1. No category

April 2006
Range and Training Land Assessment
RTLA
Technical Reference Manual:
Ecological Monitoring
On Military Lands
DRAFT
Preface
The first version of this Manual was released in 1999. This revised 2006 version contains an update
of RTLA programmatic issues as well as the following new or significantly updated sections:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Importance of Ecological Models to Monitoring Programs
Potential Monitoring Attributes and Management Applications Related to Range and
Training Land Assessment
Selecting Indicators of Resource Condition
Monitoring Protocols and Program Documentation
Data Management
Electronic Data Collection Tools
Monitoring Noxious and Invasive Plants
Monitoring Rare Plants
Soil Compaction Assessment
Unpaved Roads Condition Assessment and Relationships to Erosion and Sedimentation
Bivouac and High-Use Area Monitoring
Water Quality Monitoring
Fuels and Fire Effects Monitoring
Watershed Assessment
Aquatic Biomonitoring
The ACCESS LCTA User Interface, Structured Query Language, ACCESS LCTA Data Dictionary,
and LCTA Field Data Logger Programs sections of the 1999 manual have been published as separate
reference documents that can be downloaded from the CEMML web site at
http://www.cemml.colostate.edu/.
i
Executive Summary
Range and Training Land Assessment (RTLA) Technical Reference Manual: Ecological Monitoring
on Army Lands provides technical information for the monitoring of natural resources, with emphasis
on Army lands where training and testing activities occur. The scientific principles and monitoring
guidance provided here is fully applicable to other public and private lands. With the exception of
wildlife monitoring, the manual is intended to provide Range and Training Land Assessment (RTLA)
Coordinators and land managers with a comprehensive, single-source document for the
implementation of resource monitoring activities in support of conservation and Integrated Training
Area Management (ITAM) programs.
The manual focuses on the inventorying and monitoring of vegetation, abiotic components, and
ecological processes, and the assessment of changes or trends in resource condition resulting from
environmental factors, land management, and military training activities. The monitoring process and
program components are described with the aim of promoting sound and defensible monitoring data
to support adaptive resource management and sustainable training land use.
The manual compiles information from many sources, including federal, state and private land
management agencies, the scientific literature, and other unpublished reports. While these agencies
may differ in their objectives and approaches to land management, they share a common goal of
achieving scientifically sound and effective programs to balance land use and the sustainability of
natural resources.
Since the mid-1980s, the Army’s Range and Training Land Assessment (formerly Land Condition –
Trend Analysis or LCTA) program has provided the foundation for ecological monitoring on Army
installations. This manual incorporates much of the knowledge gained from the RTLA program, and
expands upon the scientific principles and approaches that have evolved with new programs and
objectives.
Users of this manual are encouraged to treat it not as a step-by-step guide on how to develop,
implement and manage ecological monitoring programs, but rather as a collection of informational
sources that can be chosen from and put together to fit a particular geographic location, installation,
assemblage of natural communities, populations, species, or other management concern. The manual
has been organized to follow a logical progression from how to establish monitoring goals and
objectives, to sampling and measurement techniques, followed by data collection and data
management considerations. The manual concludes with analysis techniques and procedures to
interpret and report monitoring results, with an emphasis on providing feedback for adaptive resource
management.
ii
A brief description of the chapters contained within this manual is presented below:
Chapter 1: The Origin and Development of the Range and Training Land Assessment (RTLA)
Program
This chapter discusses the history and development of RTLA and the changing scope of
the program.
Chapter 2: Introduction to Resource Monitoring
This chapter introduces the reader to resource monitoring by defining monitoring,
describing levels of monitoring, possible management applications, management and
monitoring objectives, and providing guidelines for developing a successful monitoring
program. New sections in this updated version include: Ecological Indicators, and
Written Protocols.
Chapter 3: Introduction to Sampling
This chapter introduces sampling by discussing principles of sampling and sampling
design.
Chapter 4: Measuring Vegetation Attributes and Other Indicators of Condition
This chapter defines common vegetation attributes and explains many methods for
measuring these attributes. The original LCTA design and methods is also discussed.
New sections in this updated version include: Monitoring Noxious and Invasive Plants,
Surveying and Monitoring Rare Plants, Soil Compaction Assessment, Road Condition
Assessment and Its Relation to Erosion and Sedimentation, Bivouac and High-Use Area
Monitoring, Water Quality Monitoring, Fuels and Fire Effects Monitoring, Watershed
Assessment, and Aquatic Biomonitoring.
Chapter 5: Data Management
This chapter introduces the concept of data management and also provides common data
management tasks for RTLA data.
Chapter 6: Electronic Data Collection Tools
This new chapter describes hardware, software and other considerations for collecting
data electronically in the field. Examples of several applications are provided.
Chapter 7: Data Analysis and Interpretation
This chapter discusses the statistical methods for analyzing monitoring data and for
interpreting and extrapolating the results. It also describes some common statistical
software and tools.
Each chapter contains a list of References and Appendices to support the in-chapter material.
iii
Questions or comments concerning this document can be sent to:
Center for Environmental Management of Military Lands (CEMML)
Attn: RTLA Technical Support
Dept. 1490
Colorado State University
Fort Collins, CO 80523-1490
Or
Commander
US Army Environmental Center
ATTN: SFIM-AEC-EQN
RTLA Technical Proponent
5179 Hoadley Road
Aberdeen Proving Ground, MD 21010-5401
The suggested citation for this document is:
Center for Environmental Management of Military Lands (CEMML). 2006. Range and Training Land
Assessment (RTLA) Technical Reference Manual: Ecological Monitoring on Army Lands. Prepared for
the U.S. Army Environmental Center (USAEC), Aberdeen MD by the Center for Environmental
Management of Military Lands, Colorado State University, Fort Collins CO.
This manual is intended to be a “living document” that will be periodically updated based upon comments
from field users and managers, advances in technical and scientific knowledge, and human ingenuity.
This Manual can be downloaded from the CEMML web site at http://www.cemml.colostate.edu/.
iv
Acknowledgements
RTLA Technical Reference Manual: Ecological Monitoring on Army Lands , was originally a
product of many different efforts funded by various agencies and installations of the United States
Army in support of the Integrated Training Area Management (ITAM) program. This updated version
was prepared under contract to the Army Environmental Center by the Center for Environmental
Management of Military Lands (CEMML), Colorado State University, Fort Collins, Colorado. The
authors were Chris Bern, Paul Block, Robert Brozka, William Doe, Mark Easter, David Jones, Matt
Kunze, Gary Senseman, and William Sprouse. Helpful reviews were provided by Jason Applegate
and Jimmy Harmon. The authors acknowledge the contributions made by Alan Anderson
(USACERL, Champaign, IL.) and Pat Guertin (USACERL, Champaign, IL).
v
Table of Contents
1
The Origin and Development of the Range and Training Land Assessment (RTLA)
Program _________________________________________________________________1
1.1 Introduction to the RTLA Program _____________________________________________
1.1.1
Introduction ______________________________________________________________
1.1.2
Sustainable Range Program (SRP) and Integrated Training Area Management (ITAM)___
1.1.3
History and Development of RTLA ___________________________________________
1.1.4
RTLA Process ____________________________________________________________
2
1
1
1
2
5
1.2
Use of RTLA Data in Reporting and Problem Solving ______________________________ 5
1.3
Conclusions _________________________________________________________________ 6
1.4
References __________________________________________________________________ 7
Introduction to Resource Monitoring___________________________________________8
2.1 Introduction ________________________________________________________________
2.1.1
What is Monitoring? _______________________________________________________
2.1.2
Purpose of Monitoring _____________________________________________________
2.1.3
Steps in Monitoring________________________________________________________
8
8
8
9
2.2 Importance of Conceptual Ecological Models to Monitoring________________________ 11
2.2.1
Developing Conceptual Models _____________________________________________ 17
2.3 Levels of Monitoring ________________________________________________________
2.3.1
Qualitative and Semi-quantitative Monitoring (Level 1) __________________________
2.3.2
Quantitative Monitoring (Level 2) ___________________________________________
2.3.3
Quantitative Monitoring (Level 3): Demographic, Age or Stage Class Analysis ________
18
19
19
20
2.4 Management and Monitoring Objectives________________________________________
2.4.1
Potential Attributes and Management Applications Related to RTLA ________________
2.4.2
Management Goals and Objectives___________________________________________
2.4.3
Monitoring Objectives ____________________________________________________
2.4.4
Paired Management and Monitoring Objectives_________________________________
20
21
25
27
28
2.5
Determining Benchmarks ____________________________________________________ 31
2.6 Selecting Variables to Measure ________________________________________________ 32
2.6.1
Selecting Indicators of Resource Condition ____________________________________ 32
2.7
Monitoring Intensity and Frequency ___________________________________________ 39
2.8 Written Protocols and Program Documentation__________________________________ 42
2.8.1
Elements of a Monitoring Protocol and Plan ___________________________________ 47
2.9
2.10
Summary: Guidelines for Developing a Successful Monitoring Program _____________ 49
References _______________________________________________________________ 51
vi
3
Introduction to Sampling ___________________________________________________55
3.1 Principles of Sampling _______________________________________________________
3.1.1
Why Sample? ___________________________________________________________
3.1.2
Populations and Samples___________________________________________________
3.1.3
Parameters and Sample Statistics ____________________________________________
3.1.4
Accuracy and Precision____________________________________________________
3.1.5
Sampling and Nonsampling Errors ___________________________________________
3.1.6
Hypothesis Testing Errors and Power Analysis _________________________________
55
55
55
56
56
58
58
3.2 Sampling Design ____________________________________________________________
3.2.1
Defining the Population of Interest ___________________________________________
3.2.2
Selecting the Appropriate Sampling Unit ______________________________________
3.2.3
Determining the Size and Shape of Sampling Units ______________________________
3.2.4
Determining Sample Placement (Sampling Design) ______________________________
3.2.5
Permanent vs. Temporary Plots _____________________________________________
3.2.6
Sample Size Requirements _________________________________________________
65
65
66
67
69
79
80
3.3
4
References _________________________________________________________________ 88
Measuring Vegetation Attributes and Other Indicators of Condition ________________93
4.1 Vegetation Attributes ________________________________________________________
4.1.1
Frequency ______________________________________________________________
4.1.2
Cover __________________________________________________________________
4.1.3
Density ________________________________________________________________
4.1.4
Biomass ________________________________________________________________
4.1.5
Structure _______________________________________________________________
4.1.6
Dominance or Composition ________________________________________________
93
93
94
95
96
98
98
4.2 Methods for Measuring Vegetation Attributes ___________________________________ 99
4.2.1
Frequency Methods______________________________________________________ 105
4.2.2
Cover Methods _________________________________________________________ 108
4.2.3
Density Methods ________________________________________________________ 120
4.2.4
Biomass/Production Methods ______________________________________________ 130
4.2.5
Forest and Tree Measurements _____________________________________________ 133
4.2.6
Photo Monitoring _______________________________________________________ 139
4.3 Specialized Monitoring Attributes and Approaches ______________________________
4.3.1
Soil Erosion____________________________________________________________
4.3.2
Monitoring Noxious and Invasive Plants _____________________________________
4.3.3
Surveying and Monitoring Rare Plants _______________________________________
4.3.4
Soil Compaction Assessment ______________________________________________
4.3.5
Land Uses _____________________________________________________________
4.3.6
Road Condition Assessment and Its Relation to Erosion and Sedimentation __________
4.3.7
Bivouac and High-Use Area Monitoring _____________________________________
4.3.8
Water Quality Monitoring_________________________________________________
4.3.9
Fuels and Fire Effects Monitoring __________________________________________
145
145
157
172
180
183
184
187
194
203
4.4 Overview of Original LCTA Design and Methods _______________________________ 212
4.4.1
Sampling Design and Plot Allocation ________________________________________ 212
vii
4.4.2
4.4.3
4.4.4
Data Collection Methods__________________________________________________ 213
Findings of the Independent Review Panel____________________________________ 214
Strengths and Weaknesses of the Original LCTA Approach ______________________ 215
4.5 Integrative Approaches _____________________________________________________
4.5.1
Forest Health Monitoring _________________________________________________
4.5.2
Rangeland Health _______________________________________________________
4.5.3
Watershed Assessment ___________________________________________________
4.5.4
Aquatic Biomonitoring ___________________________________________________
5
217
217
219
226
230
4.6
Monitoring Ecological Integrity on Public Lands ________________________________ 233
4.7
References ________________________________________________________________ 239
4.8
Appendix Report of 1989 LCTA Review _______________________________________ 260
Data Management ________________________________________________________268
5.1
Data Administration________________________________________________________ 268
5.2
RTLA Data Administration__________________________________________________ 268
5.3 RTLA Data Management ___________________________________________________
5.3.1
A Priori _______________________________________________________________
5.3.2
During The Collection Process _____________________________________________
5.3.3
Database Design ________________________________________________________
269
269
270
270
5.4 Database Interface Programs for RTLA _______________________________________ 271
5.4.1
LCTA/RTLA Program Manager ____________________________________________ 271
5.4.2
Access RTLA __________________________________________________________ 272
5.5
6
Electronic Data Collection Tools ____________________________________________274
6.1
7
References ________________________________________________________________ 272
Introduction ______________________________________________________________ 274
6.2 Considerations when Choosing Field Computers ________________________________
6.2.1
What Field Computer Meets Your Needs? ____________________________________
6.2.2
Major Types of Field Computers ___________________________________________
6.2.3
Cost __________________________________________________________________
275
275
275
278
6.3 GPS/GIS Software from Field to Office ________________________________________
6.3.1
Software Development and Systems Integration _______________________________
6.3.2
Software Solutions for Increased Productivity _________________________________
6.3.3
Overview of Software Features_____________________________________________
6.3.4
RTLA Monitoring Software _______________________________________________
281
281
281
281
282
6.4
Website Resources for Electronic Data Collection _______________________________ 282
6.5
References ________________________________________________________________ 282
Data Analysis and Interpretation ____________________________________________283
7.1
Introduction and General Guidance___________________________________________ 283
viii
7.2 Analyzing Monitoring Data __________________________________________________ 284
7.2.1
Overview of Statistical Applications_________________________________________ 284
7.2.2
Types of Data __________________________________________________________ 285
7.3 Confidence Intervals _______________________________________________________
7.3.1
Assumptions ___________________________________________________________
7.3.2
Calculating Confidence Intervals ___________________________________________
7.3.3
Comparing a Point Estimate to a Threshold Value ______________________________
7.3.4
Comparing Two Independent Samples _______________________________________
7.3.5
Comparing Two Non-independent Samples (permanent plots) ____________________
285
286
288
288
291
292
7.4 Statistical Tests for Monitoring Data __________________________________________
7.4.1
Caveats for Statistical Tests _______________________________________________
7.4.2
Statistical Significance and Confidence Levels ________________________________
7.4.3
Hypothesis Testing ______________________________________________________
294
294
295
296
7.5 Choosing a Statistical Procedure _____________________________________________
7.5.1
Normality Assumptions___________________________________________________
7.5.2
Frequency/Binomial Tests_________________________________________________
7.5.3
Parametric Tests ________________________________________________________
7.5.4
Non-Parametric Tests ____________________________________________________
7.5.5
Multivariate Analyses ____________________________________________________
297
297
300
300
315
320
7.6
Interpreting Results ________________________________________________________ 327
7.7 Climate Data Summarization ________________________________________________ 329
7.7.1
Sources of Climatic Data__________________________________________________ 329
7.7.2
Probability of Weekly Precipitation and Climate Diagrams _______________________ 329
7.8 Extrapolating Results_______________________________________________________ 336
7.8.1
Grouping or Pooling Data _________________________________________________ 336
7.9 Linking RTLA and Remote Sensing Data ______________________________________
7.9.1
Assess Land Condition and Trends__________________________________________
7.9.2
Classify and Ground Truth Remotely-Sensed Images ___________________________
7.9.3
Accuracy Assessment of Classified Vegetation and Imagery ______________________
338
338
338
339
7.10
Additional Analyses ______________________________________________________
7.10.1 Biodiversity Indices _____________________________________________________
7.10.2 Similarity Coefficients ___________________________________________________
7.10.3 Importance Values_______________________________________________________
343
343
346
351
7.11
Software for Statistical Analysis ____________________________________________
7.11.1 Spreadsheets and Add-ins _________________________________________________
7.11.2 Command-Line and Pseudo-Spreadsheets ____________________________________
7.11.3 Graphics Capabilities ____________________________________________________
7.11.4 Selecting a Package______________________________________________________
7.11.5 Stand-Alone Sample Size and Power Analysis Software _________________________
354
355
356
357
358
358
7.12
Data Analysis using MS Access, MS Excel, Systat, and ArcGIS __________________ 359
7.13
Guidelines for Reporting Monitoring Results _________________________________ 360
ix
7.13.1 Purpose and Types of Reports______________________________________________
7.13.2 Generic Report Organization ______________________________________________
7.13.3 Style and Format ________________________________________________________
7.13.4 Tables ________________________________________________________________
7.13.5 Graphics ______________________________________________________________
7.13.6 Brief Format ___________________________________________________________
7.13.7 Suggested Range and Training Land Assessment (RTLA) Report Outlines for Various
Audiences ____________________________________________________________________
7.13.8 Additional Guidelines ____________________________________________________
360
361
364
366
367
368
370
374
7.14
References ______________________________________________________________ 374
7.15
Appendix Statistical Reference Tables_______________________________________ 378
x
1 The Origin and Development of the Range and Training
Land Assessment (RTLA) Program
1.1 Introduction to the RTLA Program
1.1.1
Introduction
Effective management of Army lands requires information regarding initial resource conditions and
knowledge of impacts from various types of military training. Environmental degradation associated
with military training can have significant short and long-term effects on soil physical properties,
vegetation community composition and structure, soil erosion, wildlife and associated habitats, and
long-term productivity.
There are several reasons for understanding and minimizing the ecological impacts of military
training. First, the National Environmental Policy Act of 1969 (NEPA) and U.S. Army Regulation
200-2 (Department of the Army 1980) require that the Army minimize or avoid both short and longterm impacts caused by military training activities. Second, Army training lands are finite while
training demands and intensity due to technological changes and base realignments have increased.
Therefore, it is in the Army’s interest to sustain soils and vegetation resources on current training
lands to meet mission requirements for realistic training and testing. This is especially true in places
where the Army trains on and manages land belonging to non-Army federal and state agencies.
By documenting and understanding training-related impacts, excessive or irreversible damage and
associated land rehabilitation costs can be minimized. Impacts to training lands can result in off-site
problems, such as degradation of habitats, water quality, and air quality. Only by establishing
effective monitoring programs can the Army optimally manage the resources entrusted to it.
The RTLA program inventories and monitors natural resources, and manages and analyzes
information regarding resource condition and trends. Data and results are pertinent to management of
training lands from the training area to installation scales and provide input to management decisions
that promote sustained and multiple uses on military lands. The RTLA program is a long-term effort
designed to evaluate relationships between land use and condition, as well as document natural
variability over time.
1.1.2
Sustainable Range Program (SRP) and Integrated Training Area
Management (ITAM)
The Sustainable Range Program is the Army's overall approach to improving the way it designs,
manages, and uses its ranges to meet mission training responsibilities. The SRP proponent, Deputy
Chief of Staff, G-3, defines two core programs that focus on the doctrinal capability of the Army's
ranges and training lands: 1) the Range and Training Land Program (RTLP), and 2) the Integrated
Training Area Management (ITAM) program. The SRP core programs are integrated with the
facilities management, environmental management, munitions management, and safety program
functions that support the doctrinal capability to ensure the availability and accessibility of Army
1
ranges and training lands. Within the Army Test and Evaluation Command (ATEC), SRP is defined
by its test ranges and ITAM programs and is similarly integrated with the programs described in the
preceding statement.
ITAM includes RTLA (formerly known as Land Condition-Trend Analysis or LCTA), Training
Requirements Integration (TRI), Land Rehabilitation and Maintenance (LRAM), and Sustainable
Range Awareness (SRA). The ITAM Program is the Army's formal strategy for focusing on sustained
use of training and testing lands. The intent of the ITAM Program is to systematically provide a
uniform training land management capability across the total Army and to help the Army manage its
lands to ensure no net loss of training capabilities while supporting current and future training and
mission requirements. The effective integration of stewardship principles into training land and
conservation management practices ensures that Army lands remain viable to support future training,
testing, and other and mission requirements. ITAM integrates elements of operational, environmental,
master planning, and other programs that identify and assess land use alternatives.
1.1.3
History and Development of RTLA
In April 1984, the Secretary of the Army commissioned an independent panel of scientists to evaluate
natural resource management on selected military installations. Among the panel’s findings was that
the “single largest constraint to resource management is that natural resource uses are subordinate to
military-mission activities.” The panel also noted excessive soil erosion resulting primarily from the
use of tracked vehicles. Among the recommendations for improving natural resource management on
military lands were the following:
1) Develop and maintain natural resource management plans that are based on current resource
inventories and that contain clearly stated resource-use objectives, site-specific prescriptions, and
monitoring procedures designed to evaluate action strategies through integrated resource
management.
2) Require that new or more adequate natural resource inventories be completed on all military
installations and civil works projects by a prescribed time, such as 30 September 1987, with the
information to serve as the base for preparing integrated resource management plans.
Following the 1984 study, the Land Condition-Trend Analysis (LCTA) program was initiated by the
Department of Army as a top-down program emphasizing uniform data collection methodologies to
provide regional, MACOM, or national-level assessments of land condition. Development of
methodologies for measuring disturbance, vegetation and erosion was undertaken by the U.S. Army
Corps of Engineers Construction Engineering Research Lab (CERL), Environmental Division.
Implementation of the original methodology (Tazik et al. 1992) (100m permanent transects) took
place at dozens of installations across the United States and in Germany in the 1980s and 1990s. Pilot
projects were established in 1985 at Fort Carson and Fort Hood to test and validate field methods.
Among the first installations to adopt the program in 1986 were Hohenfels Training Area in
Germany, and the Piñon Canyon Maneuver Site in southeastern Colorado. On 18 August 1987, a
memorandum was issued by the Assistant Secretary of the Army for Installations and Logistics, John
W. Shannon, calling for Army-wide implementation of LCTA, and noting that the program could
potentially “place the Army in the forefront as knowledgeable and responsible land managers.” The
program was implemented on a large scale in 1989, when LCTA was established by CERL at eight
Forces Command (FORSCOM), Training and Doctrine Command (TRADOC), and Army Materiel
Command (AMC) installations. By 1992, LCTA was in place at 48 Army, National Guard, and
2
Marine Corps installations, funded primarily on a voluntary basis by individual installations using
base operations funds.
In November 1993, functional proponency for ITAM was transferred to the Training Directorate,
Office of the Deputy Chief of Staff for Operations and Plans (ODCSOPS) at HQDA. This action
served to align policy and funding for the program with Army training programs, provided official
recognition of the program, and instituted it in the Army budgeting and planning process.
The emphasis of LCTA at the outset was on data comparability and standardization, permitting
aggregation at the Major Command (MACOM) and Department of Army (DA) levels. However,
varying installation needs for environmental data, ecoregion differences among installations, and a
need for more information on a smaller scale (individual training areas on installations), generated
alternative data collection methodologies. Following initial implementation, some installations
continued with standard LCTA data collection, sometimes supplementing it with additional
monitoring to meet local needs. Other installations largely abandoned the original LCTA
methodology in favor of other methods developed by individual RTLA (then LCTA) coordinators.
Examination of results, changing program proponency, and shifts in user needs led to a re-evaluation
of programmatic and technical aspects of the program in 1996. This effort and resulting
recommendations were termed “LCTAII” (USAEC 1996a; 1996b). LCTAII redefined program
objectives to support training and range operations as a priority while continuing to provide long-term
baseline and trend information for land management needs. Installations were thus given flexibility in
using a variety of scientifically acceptable data collection methods that were appropriate to meet local
needs. HQDA representatives reviewed the requirements of the Army Training and Testing Area
Carrying Capacity (ATTACC) model. Data needed for Universal Soil Loss Equation (USLE) erosion
estimates and disturbance information were identified as data required from all installations to support
carrying capacity efforts. The ATTACC data requirement was therefore considered a “core” data
requirement. Training and Range Operations representatives developed a list of questions that LCTA
could help answer. The questions were directly related to carrying capacity/recovery, maneuverability
and trafficability, and cover and concealment. Workshop participants agreed that an ecoregion
approach for prescribing methods is less relevant than applying methods that best suit individual
installation conditions. It was also recognized that data elements required by different users may
overlap to varying degrees. With the adoption of ITAM by the training and operations community,
the modern RTLA program has evolved to a decentralized, installation-level program that focuses
first and foremost on installation needs, and may provide information to Major Commands
(MACOMs) and Headquarters, Department of the Army (HQDA) as requested.
RTLA objectives are defined by both higher-echelon commands (e.g., Army Headquarters, National
Guard Bureau, State Army National Guard Organizations) and individual installations. Army
Regulation (AR) 350-19 – The Army Sustainable Range Program (SRP) provides policy guidance for
the ITAM Program and RTLA. It incorporates guidance previously contained in AR 350-4, ITAM,
and assigns responsibilities and provides policy and guidance for: managing and operating U.S. Army
ranges and training lands to support their long-term viability and utility to meet the National defense
mission; planning, programming, funding, and executing the core programs comprising the Army’s
Sustainable Range Program, the Range and Training Land Program, and the Integrated Training Area
Management Program; integrating program functions to support sustainable ranges; assessing range
sustainability; and managing the automated and manual systems that support sustainable ranges.
In 2004, the LCTA ITAM component was renamed Range and Training Land Assessment (RTLA) to
reflect its role in training land management and training support. Relative to the early LCTA
3
Program, the RTLA Program is more focused on training support and training land management, and
may be used to support National Environmental Policy Act (NEPA) and other compliance and
planning efforts related to Army Transformation, Restationing, and Realignment. Current policies
allow installation-level land managers and Range Operations staffs to determine how they can best
collect and use resource data to support foundational/long-term and site-specific land management
decisions such as training area allocation, training area use and land rehabilitation effectiveness.
The current RTLA mission is “to inform the process of military land management to maximize the
capability and sustainability of land to meet the Army training and testing mission .” RTLA staff and
supporting data and products support Army missions by participating in land-use planning activities
and integrating environmental expertise in the initial stages of mission and stationing/restationing
planning. This is crucial for ongoing missions, Base Realignment and Closure (BRAC) actions, Army
Transformation, and Army Modularity. RTLA monitoring activities may also be included as sources
of baseline data and as mitigation in NEPA documents. By being the Range Manager or trainer’s
environmental advocate, RTLA Coordinators can provide expertise to enable training areas to be
available, in compliance with applicable regulations, and optimally managed for sustained training.
Within the context of the RTLA mission statement, programs should:
•
•
•
•
•
•
•
Provide, in an efficient manner, data, analytical capabilities, and recommendations
associated with sustained use of testing and training lands.
Provide data to support training land management and land use decisions.
Provide data to identify and monitor LRAM sites and evaluate the effectiveness of
LRAM efforts.
Provide data input to an installation's plans, such as the Integrated Natural Resources
Management Plan (INRMP), Integrated Cultural Resources Management Plan (ICRMP),
installation master plan, Range and Training Land Plan (RTLP), etc.
Provide a means for installation training land managers to measure and monitor natural
resources.
Provide methods to assess the effects and impacts of training and testing on natural
resources.
Assess the impacts of natural resources management on training and testing (e.g.,
prescribed burning, agricultural leasing, livestock grazing, etc.).
Changes to the program include a requirement for written, comprehensive protocols, and regular
reporting of results to installation staffs and MACOMs.
1.1.3.1
Original RTLA Objectives and Methodology
The development of the original LCTA methodology at CERL followed several guiding principles:
1) The protocol would provide objective, scientific data documenting existing conditions of soils,
vegetation, and wildlife through inventory and classification of Army lands.
2) The methods would be standardized and applicable across a wide range of ecological
conditions, and allow for summaries and comparisons of the data among installations.
3) Data collection would have to be repeatable.
4) The data would be able to evaluate the capability of lands to support sustained military use.
The original LCTA methodology describing detailed procedures for monitoring vegetation, soils,
small mammals, birds, and reptiles and amphibians is presented in Tazik et al. (1992). In 1989 an
independent review panel was convened to review the technical merit of the LCTA program. A report
4
issued by the panel found the methodology technically sound, and recommended some minor changes
(Cook 1989). A discussion of the original LCTA methodology is presented in Section 4.4 Overview
of Original LCTA Design and Methods.
Permanent plots were allocated using a stratified random design, using an unsupervised classification
of a satellite image and a map of the soil types. The objective was to represent all major soil and
vegetation types on an installation. Budgetary constraints determined the number of plots would be
limited by how many two crews could inventory in a season, which was generally around 200. On a
typical 100,000-acre installation, this amounted to one plot for every 500 acres, which grew to be a
rule of thumb for allocating plots.
1.1.4
RTLA Process
RTLA and resource monitoring in general should be viewed as a dynamic process, which is
influenced by, and in turn influences, a number of other program areas. Figure 1-1 illustrates RTLA
as a process designed to meet specified land management objectives. The major process components
include 1) Defining Objectives, 2) Monitoring, 3) Data Management and 4) Reporting. These
components are driven and supported by a variety of inputs and activities, and feedback loops.
The sequencing of these components is critical to a successful RTLA program. Monitoring in the
field cannot proceed without a clear idea of what questions must be answered and what objectives are
to be met. Likewise, data management cannot occur without data collected from the field and other
sources, and without an idea of how the data should be summarized.
1.2 Use of RTLA Data in Reporting and Problem Solving
Since its inception, the emphasis of RTLA has been on data collection. Prior to 2000, on most
installations RTLA data was not used extensively for reporting, problem solving, and adaptive
management. Few or no reporting requirements partially resulted in a high proportion of the RTLA
effort being expended on data collection, rather than on data management, evaluation, analysis, and
site-specific applications. As a result, data collected provided limited feedback to support adaptive
resource management, mission sustainability, and evaluation of the monitoring design and methods.
More recently, numerous installations have begun focusing RTLA efforts, applying RTLA data
successfully, and adapting objectives and methods using a systematic approach. Recent requirements
regarding protocol development and reporting will promote optimal application of RTLA data.
5
Training and Land
Management
Needs
LCTAII Process:
HQDA/MACOM Data
Requirements
ATTACC
OBJECTIVES
Evaluation and
Improvement
Installation RTLA
Program Goals and
Objectives
Technical References
and Instructional
Manuals
ITAM Learning
Module for
LCTA
HQDA/MACOM
Policy Guidance and
Resourcing
Army ITAM
Regulation
Army ITAM
PAM
Resource Management Tools:
DATA COLLECTION
Floristic, other resource surveys
Vegetation mapping
GIS Database development
Monitoring Plans and
Protocols
Technical Support Services:
Quality Assurance
Quality Control
Methods, Data Management, Data
Analysis, Software, Hardware
DATA MANAGEMENT
Data Analysis Training
(formal and informal)
Independent Evaluations of
Approaches:
Vegetation Measurements
Rapid Assessments
Integrated Approaches
Soil Erosion Models
REPORTING
Data Analysis and Synthesis
Management Recommendations
Program Evaluation
Figure 1-1. RTLA program inputs, process, and support mechanisms.
1.3 Conclusions
RTLA was initiated in the mid-1980s by the Department of Army as a top-down program
emphasizing uniform data collection methodologies to provide regional, MACOM, or national-level
assessments of land condition. With the adoption of ITAM by the Training and Operations
community, the RTLA program has evolved to a decentralized, installation-level management of
objectives to document the status and trends in natural resources, examine the relationships between
disturbance and condition, and support training and testing area land use decisions. Current policies
allow installation-level managers (ITAM, range operations, and land management staff) to determine
how they can best collect and use resource data to support short and long-term land management
decisions such as training area allocation, training area use and land rehabilitation.
A successful RTLA program provides scientifically valid baseline and long-term monitoring data.
Monitoring is a critical component of the adaptive management cycle, especially in the context of
ecosystem management (Leslie et al. 1996), but can only be successful if it is objective-based.
Limited resources dictate that qualitative methods be sometimes coupled with quantitative methods to
address short- and long-term objectives. Long-term monitoring plots, in addition to non-permanent
plots and other sampling sites, reduce the “noise” caused by annual variability and facilitates
detection of condition trends over time. This information supports stationing decisions, mission
change analysis and ecosystem management activities. It is important to note that RTLA encompasses
6
the collection and analysis of both field-collected and additional data at multiple spatial and temporal
scales.
RTLA has changed in recent years in response to needs and constraints coming from installations,
MACOMs, and HQDA, as well as changes in organizational responsibilities and funding. There is a
need for core elements that will remain important over time and flexible regardless of most policy
changes. Decision-making at the installation level is essential to ensure that site-specific issues can be
addressed effectively.
1.4 References
Cook, C.W. 1989. Report of LCTA Review. Unpublished report submitted to USACERL, December
1989. U.S. Army Construction Engineering Research Laboratory, Champaign, IL – Reproduced as
appendix in Chapter 4.
Department of the Army. 1980. Environmental Quality: Environmental Effects of Army Actions.
Army Regulation 200-2, Nov 1980.
Department of the Army. 2005. Army Regulation (AR) 350-19 – The Army Sustainable Range
Program (SRP). Washington D.C.
Leslie, M., G. K. Meffe, J. L. Hardesty, and D. L. Adams. 1996. Conserving Biodiversity on Military
Lands: A Handbook for Natural Resources Managers. The Nature Conservancy, Arlington, VA.
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (LCTA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
USAEC (U.S. Army Environmental Center). 1996a. Land Condition Trend Analysis II: Report for
Workshop held January 23-25, 1996 in Linthicum, Maryland.
USAEC (U.S. Army Environmental Center). 1996b. Land Condition Trend Analysis II: Report for
Workshop held August 5-7, 1996 in Linthicum, Maryland.
USAEC (U.S. Army Environmental Center). 1998a. Army Regulation 350-4: Integrated Training
Area Management. May 8, 1998.
USAEC (U.S. Army Environmental Center). 1998b. Department of the Army Pamphlet 350-4:
Integrated Training Area Management. Coordinating Draft. August 1998.
7
2 Introduction to Resource Monitoring
2.1
2.1.1
Introduction
What is Monitoring?
Natural resource inventorying is the process of acquiring information on resources, including the
presence, distribution, condition, and abundance of resources such as vegetation, soil, water, natural
processes, biotic communities, and natural and human-induced changes in resources (USDI National
Park Service 1992). Monitoring is the process of collecting specific information over time to assess
conditions and changes or trends in resource status and predict or detect natural or human-induced
changes in resource conditions. Indicators of resource condition are often used to evaluate the
“condition” or “health” of populations, communities, and landscapes.
Three types of monitoring can be described: implementation, effectiveness, and validation
monitoring. Implementation monitoring is used to determine if activities and projects are
implemented as designed or intended, or at all (Was the activity that was planned actually done?).
Effectiveness monitoring determines if activities or projects are effective in meeting management
objectives or established guidelines (Did the prescription work or have the desired effect?).
Validation monitoring determines whether the data, assumptions, and relationships used in
developing a plan are correct (Is there a better way to meet management objectives?). Validation
monitoring is often synonymous with research. Most of what is described in this document refers to
effectiveness monitoring.
Resource monitoring at various landscape scales (i.e. entire installation vs. training area vs. vegetation
community, etc.) is a challenging task for land managers and Integrated Training Area Management
(ITAM) staff. Monitoring efforts should specifically address management objectives articulated by
trainers, land managers, and those presented in integrated natural resource management plans.
Management objectives are often placed in categories such as vegetation condition/status (i.e., for
different communities), animal habitat, and soil erosion status. Monitoring is valuable also in
evaluating different approaches to land rehabilitation and maintenance. The development of missionoriented objectives and projects and a suggested scoping process is described in CEMML (2006).
Long-term changes or trends in resources will not be detected reliably if the examination of data is
restricted to inconsistent or short-term data. Success of long-term assessments using monitoring data
largely depend on creating and using a variety of efficient methods and designs that are both robust
and widely useful (Hinds 1984).
2.1.2
Purpose of Monitoring
It is important to distinguish between the respective purposes of inventory and monitoring activities.
Inventory activities typically precede and contribute significantly to monitoring efforts. Some
inventories involve a number of years to collect the necessary information. The primary purposes of
resource inventories are to: (1) document the occurrence, location, and current condition of physical
8
habitat and features (site conditions) and major associated biota; (2) identify locally rare or threatened
and endangered species, locating fragile or rare ecosystems and potential indicator species; and (3)
assess the full range of populations, ecosystem components, processes, and stresses (i.e., both natural
and human-caused disturbances), which form the framework for subsequent sampling during the
monitoring process (USDI National Park Service 1992). The first cycle of data collection for a
monitoring program is sometimes referred to as the “initial inventory” or “inventory year”.
The primary purposes of monitoring are to: (1) provide indicators of ecosystem health or status; (2)
define limits of normal variation (i.e. natural variability); (3) detect changes in condition, abundance,
structure; and (4) identify and understand the effects of management and land-uses. In some cases,
monitoring is used to determine compliance with environmental regulations and standards. If properly
designed, monitoring activities can provide information linking changes in resource conditions and
potential causes.
Monitoring provides a rational and objective basis for taking management actions and expands
current knowledge of ecosystem properties and processes (Spellerberg 1991). Monitoring should also
provide feedback between natural resource conditions and management objectives, an essential
component of adaptive resource management. Because inventorying and monitoring help to bridge
the gaps between land managers, training and range operations, and research activities, they require
both communication and support at a number of administrative levels to be efficient and effective.
2.1.3
Steps in Monitoring
Once management objectives have been established, a monitoring program can be designed and
reviewed. A conceptual overview of the monitoring process is presented in Figure 2-1. Monitoring
efforts must be continually evaluated to ensure that the selected attributes and indicators are sensitive
to change and the methods employed are effective.
The steps outlined in Figure 2-1 are universally applicable regardless of the scale or duration of the
project. Although monitoring is an ongoing process, periodic reporting, ideally following each data
collection period, should be performed consistently in order to provide valuable feedback regarding
management activities and the success of the monitoring program itself. Analysis involves examining
individual components to establish status and interrelationships. Data synthesis subsequently
examines the results at multiple scales over time to examine changes and trends. Analysis, synthesis,
and reporting of results should be presented in a format that is consistent with the needs of decision
makers.
9
Management plan/ objectives
Propose general monitoring objectives
Select/develop conceptual
ecological models
Evaluate available data:
- historical data
- species lists
- abundance and distribution data
- physical environment
- land uses and disturbance
- GIS data
- other sources
Identify specific monitoring
objectives and priorities
Define personnel and
budget constraints
Propose sampling
design and methods
Examine existing data or
conduct pilot sampling
Revise objectives, design,
or methods, as necessary
Analyze and evaluate
data
Does sampling design
and methods meet
monitoring objectives?
no
yes
Conduct/continue
monitoring
Analyze and
evaluate data
Report results and make
management recommendations
Assess results relative to
management objectives
Assess results relative to
monitoring objectives
Figure 2-1. Steps in the development and implementation of a monitoring program (modified from
USDI National Park Service (1992) and The Nature Conservancy (1997)).
10
2.2
Importance of Conceptual Ecological Models to Monitoring
Conceptual models are a critical tool in the development of a monitoring program. Once the primary
ecological communities of interest and other priorities are known, the process of understanding the
important elements and processes begins. Conceptual ecological models can be very helpful in
summing up what is known about a system and identifying attributes to assess changes in those
systems over time. Models should be based on focused management questions that establish the
appropriate context for the model. Although they are simplifications of the real world, models can
capture relationships that are the basis for predicting changes in our management goals or
conservation targets over time.
Conceptual ecological models can be created for a variety of conservation targets, including
individual or groups of species, vegetation community types or assemblages of community types.
The type of model constructed depends on the scientific questions asked, goals and objectives of the
project, and characteristics of the conservation targets at the site (Poiani 1999). The models developed
relate to and drive the management and monitoring goals and objectives, and support the selection of
specific attributes or indicators. In the context of a monitoring program, conceptual ecological
models 1:
•
•
•
•
•
•
•
•
•
•
•
Store important information and capture institutional knowledge
Provide users with predictive capabilities and scenario-building information
Identify priority conservation targets, processes, stressors, and threats (actual or potential)
affecting them
Help managers and scientists understand ecosystem dynamics, responses to stressors
(natural and anthropogenic), and ranges of natural variability
Identify links between states/ecosystem components, drivers, stressors, system responses,
and monitoring attributes
Facilitate selection and justification of monitoring attributes and indicators
Facilitate evaluation of monitoring data
Provide a framework for interpreting monitoring results in an adaptive management context
and prioritizing management actions
Document assumptions, knowledge, experience, and unknowns/information gaps
Are valuable communication tools for a variety of audiences.
Help identify thresholds of condition that may be difficult or impossible to reverse.
The term “model” may seem intimidating, but in fact models need not be unnecessarily complex. A
variety of terminology and frameworks for conceptual models exists, and may be confusing for the
first-time user. The stock, flow, converter, and connector framework grew out of dynamic systems
modeling and mechanistic approaches oriented toward quantitative modeling. Models using this
framework are often referred to as control or computer simulation models. Examples of control
models include population viability and extinction models, forest succession and disturbance models,
and fire simulation models.
More simplified, management-oriented conceptual ecological models are often used to help direct
management and monitoring activities and are described here. They include information about
important system components, processes affecting those components, effects of stressors, and often
specific attributes that are linked to components and stressors. The models may be represented in
tabular form (Table 2-2), simple diagrams or box and wire flowcharts (Figure 2-2 through Figure
1
Adapted from TNC (1994), Peacock (2000), and Gross (2003).
11
2-6), narratives, or a combination of the above. More generalized, non-quantitative models that
illustrate system dynamics and response are known as driver-stressor models or state and transition
models. For arid and semiarid systems, state-transition (also known as state-transition-threshold)
models are often applied (Westoby et al. 1989, Stringham et al. 2001, Bestelmeyer et al. 2003).
State-transition models are widely used by the Natural Resources Conservation Service (NRCS),
often in the context of grazing management. Under this model, plant communities or ecosystem types
are grouped into states that are distinguished from other states by large differences in plant functional
groups, ecosystem processes, and the resultant characteristics, including management requirements.
Transitions are the shifts between states caused by internal and external factors, and are generally not
easily reversed through “natural” processes. Thresholds of condition that delineate irreversible
damage may also be included in these models. Box and arrow diagrams are the most commonly-used
format for these management-oriented conceptual models.
Stressor models and state-transition models are occasionally blended to form conceptual models, and
the two terms are sometimes used interchangeably. Strengths and weaknesses of these model types
are presented in Table 2-1. Descriptions of transitions in state and transition models often incorporate
information about stressors and their effects, lessening the need for a separate narrative to explain the
model, as opposed to stressor models, which generally require a separate narrative to describe the
processes and changes at work. Regardless of the model type used, scientific references supporting
the model relationships should be included to provide rationale and lend credence to the conceptual
model. The selection of terminology may be influenced by the framework used by other land
managers and scientists in the area. The attributes or indicators identified using the model should be
measurable in some way to address the relationship among drivers, stressors, effects, states, etc.
Table 2-1. Strengths and weaknesses of state and transition and driver-stressor models. (adapted
from Gross 2004).
State and transition models
•
•
•
•
•
clear representation of alternative states
sometimes address multiple community conditions within a state
relatively simple
excellent communication with most audiences
typically contain no quantitative modeling of flows, processes, states
Driver-stressor models
•
•
•
•
•
•
provide clear link between agents of change/stressors and important attributes
simple and easy to communicate
no feedbacks
little to no quantitative modeling of flows, processes, states
frequently generalized and sometimes incomplete and inaccurate
can be too general to directly link to specific attributes/indicators
Examples of conceptual ecological models for different ecosystems are presented below. Note that
formats vary considerably. In general, diagrammatic models are accompanied by a descriptive
narrative to explain some of the relationships or processes illustrated by the graphic.
12
Table 2-2. Examples of ecosystem, drivers, stressors, ecological effects, and related attributes
relevant to conceptual ecological models.
Ecosystem
Drivers
Ecological Effects,
Responses, or Affected
Systems
Stressors or Agents of
Change
Related Attributes
Soils and
topography
• Erosion
• Disturbance/use
• Compaction
•
•
•
•
Sediment and nutrient transport
Altered soil structure
Altered runoff response
Reduced productivity
• Soil loss estimates
• Suspended sediment,
Precipitation
• Flood
• Drought
• Erosion
•
•
•
•
Fire risk
Invasion by non-native plants
Altered substrates
Habitat types/quality
• Fire frequency,
Community
dynamics
•
•
•
•
• Changes in community structure
• Habitat type conversion
• Population changes
• Presence and abundance
Land uses
• Disturbance agents, animals,
• Native community structure
• Establishment and spread of
• Multiple, and high
•
•
•
•
•
Fire
Establishment
Dispersal
Predation/mortality
Competition
vehicles, other activities
Altered fire regimes
Introduction/dispersal of
nonnative species
Habitat fragmentation
Domestic livestock grazing
Resource utilization
• Increased or decreased fire
•
•
•
•
intensity
Altered fire return interval
Change sin
seasonality/timing of fires
Prescribed burning
Presuppression and
suppression activities
•
•
•
•
invasive species
Water quality degradation
Changes in wildlife abundance,
distribution, survival
Loss of species
Loss of habitat
• Type conversion
• Nonnative species invasion
potential
• Hydrologic response and water
quality
• Soil seed bank
• Vegetation community and
habitat types
• Changes to species populations
and distributions
water quality
seasonality, severity
• Weed abundance
• Native habitat quality
of species, functional
groups, and structural
groups.
• Extent of habitat types
• Population estimates
overlap with other
driver categories
• Fire extent
• Fire frequency
• Fire intensity and/or
severity
• Water quality and runoff
parameters
• Habitat quantity and
quality
• Nonnative species
abundance and spread
13
Figure 2-2. Fire process model for Arkansas River Valley Prairie and Oak Ecosystem coarse
community types. From the TNC U.S. Fire Learning Network Workshops held in 2002 and 2003
(http://tncfire.org/documents/USfln/ARRivVal_models_revB.pdf).
Figure 2-3. Conceptual diagram of changes in a shrub-steppe community in the absence of fire
(Miller et al. 1999, modeled after Archer 1989). This conceptual ecological model was used to identify
important elements of a monitoring protocol developed for the Oregon Army National Guard (Jones
2000).
14
T1
I
Open stand of sagebrush
with productive
herbaceous perennial
understory.
II
Dense sagebrush cover.
Depleted perennial
herbaceous understory
with sagebrush seedlings
present.
T2
T6
T7
T3
IV
Dense sagebrush cover.
Abundant annuals, few
herbaceous perennials
and sagebrush seedlings
present.
T9
T12
T8
T5
T4
III
Recently burned.
Perennial herbaceous
species and sagebrush
seedlings.
VI
Repeated burns. Only
annuals with no perennial
herbaceous species or
sagebrush present.
T10
T11
V
Recently burned.
Dominated by annuals
with sagebrush seedlings
present.
T12
T12
Figure 2-4. State and transition model for sagebrush-grass ecosystem (Laycock 1991, after Westoby
et al. 1989). A generic state-and-transition diagram applicable to sagebrush-grass vegetation at
Yakima Training Center, Washington and other Columbia Basin installations is presented in Figure
2. The following description of states is modified from Laycock (1991). The boxes represent stable
states and the arrows are transitions between states. States I, II, and III represent states that may
occur in areas without nonnative annuals. State II represents the degraded state resulting from
prolonged heavy grazing, which remains dominated by sagebrush for extended periods. Fire
(transition 3), training disturbance, disease, or some other disturbance that kills adult sagebrush will
release the perennial understory from competition. State IV represents a situation that might occur
in heavily grazed or repeatedly disturbed areas where a well-adapted annual, such as cheatgrass,
replaces native perennials in the understory. Fire (transition 8) and repeated fire (transition 10) can
then convert this community into a stable state dominated by annuals (state VI). Transition 12
represents human intervention in the form of seeding to adapted perennial species such as crested
or Siberian wheatgrass. If successful seeding of native grasses and forbs takes place, transition 12
may lead to State III over time if annuals do not dominate and sagebrush is planted as well. In fact,
YTC land managers use criteria similar to these states to determine the need for seeding following
fire (i.e., state III versus state VI). Laycock notes that some transitions such as those numbered 2, 7,
and 11 are especially difficult to cross.
15
Figure 2-5. Ecological restoration model for upland ecosystems at Fort Benning, Georgia. SERDP,
Unpublished report.
16
Figure 2-6. State and transition conceptual model contained in the NRCS ecological site description
for a shallow sandy rangeland type in New Mexico.
(http://www.nm.nrcs.usda.gov/technical/fotg/section-2/esd/sd2.html).
2.2.1
Developing Conceptual Models
The process of developing conceptual models requires the thoughtful integration of knowledge and
opinion from a variety or sources, but the models may be relatively simple. Some complex systems
may be very difficult to encapsulate within a single conceptual model, necessitating the development
of submodels for important components. The key is to develop one or more conceptual models
addressing appropriate temporal and spatial scales at the right level of detail, moving from simple
models to more comprehensive models. As more is learned about the system, each model can be
refined iteratively. The suggested steps listed below build upon Maddox et al. (1999) and Gross
(2003).
1. Clearly state the goals of the conceptual model(s).
2. Identify boundaries of the system of interest.
3. Identify and gather key informational resources and data, including published and
unpublished reports and scientific papers, subject matter experts, academic and resource
management experts, and local professional knowledge and experience. It is possible that
applicable models or relevant building blocks have already been developed.
17
4. Articulate key management questions that the model(s) will support.
5. If desired, use informal or formal workshops with local or regional experts, biologists,
researchers, and land managers to develop initial frameworks that can be refined over
time.
6. Decide on the structure of the model, based on available resources, the intended audience,
and planned applications.
7. Identify key model components such as states, natural and anthropogenic stressors,
transitions, subsystems, and interactions. The driving processes of the system should be
easily discernible.
8. Describe relationships among the states in the system. This may include stressors,
ecological factors, and responses, succession, disturbance agents/factors, and other
elements
9. Identify and prioritize attributes and indicators.
10. Periodically review, revise, and refine models using in-house resources and external
reviewers and incorporate new information as it becomes available. Monitoring may
provide feedback that supports or contradicts model assumptions and key components.
Some steps will be undertaken sequentially while some will happen simultaneously.
2.3
Levels of Monitoring
Approaches to monitoring can be described as “planes” or “levels” of monitoring ranging from the
inexpensive and fast to the costly and more time-consuming. Monitoring information and
methodologies can be classified as qualitative/semi-quantitative (Level 1), quantitative (Level 2), and
demographic (individual/age/stage class analysis) (Level 3) (Menges and Gordon 1996; The Nature
Conservancy 1997). Parameters that are measured or estimated include abundance (e.g., number,
density, cover, frequency), condition (e.g., vigor, reproductive success, size, biomass, level of damage
or disease, etc.), and population or life-history structure (e.g., size class, age-class, number of
individuals or that meet specific criteria). The level of monitoring and parameters chosen will depend
largely on the management and monitoring objectives supported by the chosen approach and the
availability of resources to carry out the monitoring, including time, fiscal resources, and staff
resources. A discussion of parameter and indicator selection is presented in Section 2.6.
Well-written monitoring goals and objectives indicate which level of monitoring is initially
appropriate. Most objectives are best addressed by monitoring resources using several levels of
intensity. Monitoring a mixture of parameters within the different monitoring levels is also beneficial.
For example, quantitative measures of abundance are often combined with qualitative information
such as presence-absence records and ranked data. Efficiency of data collection can often be
improved by using diversified approaches. The addition of photographic documentation (Level 1) to
almost any other type of monitoring is almost always beneficial. Level 1 and 2 information is most
commonly used to assess resource condition. Level 3 studies are often reserved for research activities
or in the case of small populations at risk. A generalized summary of goals and analytical approaches
to different levels of monitoring is presented in Table 2-3. The discussion of monitoring levels, which
emphasizes vegetation monitoring, is based on information provided in Menges and Gordon (1996)
and The Nature Conservancy (1997).
18
Table 2-3. Levels of monitoring vegetation, with goals and statistical approach (modified from Menges
and Gordon 1996).
Level
1. distribution of populations/
communities
2. population/community size or
condition
3. demographic monitoring (individuals)
2.3.1
Goals
measure trends across populations,
hypothesize trends in size
measure trends within populations and
hypothesize mechanisms/causes
anticipate population trends and
understand mechanisms
Analysis Approaches
descriptive
trend analysis
population variability, age class analysis
Qualitative and Semi-quantitative Monitoring (Level 1)
Level 1 monitoring is used for both single or multiple locations of species, communities, or other
subjects of interest. For example, discreet populations or communities can be mapped through field
surveys or aerial photo interpretation. The occurrence, extent, or distribution of
populations/communities can be effectively monitoring using this approach. Where multiple
occurrences are monitored, the number of locations provides Level 1 information about abundance.
This approach allows for examination of changes in populations and locations, changes in area
occupied, and large changes in abundance. Using broad ocular estimation categories will only permit
coarse indicators of population changes. Increasing the number of estimation categories improves the
potential for data analysis but makes repeatability more difficult.
Level 1 condition information includes descriptions and presence of attributes or conditions at one or
more locations. For example, the presence or absence of different types of erosion, disturbance,
hydric soils, seedlings and regeneration, snags of a particular size or type, or stage of vegetation
phenology. In addition to presence/absence type data, observers can make estimates of abundance for
the condition being exhibited (e.g., 15% crown dieback).
Level 1 structure information includes presence/absence of individuals within specified age or height
classes, or coarse estimates of abundance within age or size classes. This information gives an
indication of population or community structure and recruitment.
-
Level 1 abundance information includes:
presence/absence of a population/individuals at a particular location
size of the area encompassed by the population or community
estimates of abundance using broad categories or log-scale rankings (1-10, 10-100,
100-1000)
photo monitoring
Advantages of Level 1 information include relatively low costs and rapid procedures. Communities or
populations are assessed in their entirety or by sampling a portion. Disadvantages of Level 1
information include low repeatability and precision, susceptibility to interpretation and observer bias,
inability to perform quantitative analysis, and ability to detect only relatively large changes.
2.3.2
Quantitative Monitoring (Level 2)
Quantitative monitoring consists of data collection using methods and approaches that have high
precision and minimal observer bias and subjectivity. The principal difference between Level 1 and
Level 2 monitoring is that actual measurements are made and items counted using Level 2. Examples
19
of methods include density counts, frequency frames, prism cruises, and canopy measurements.
Ocular estimates of canopy cover may be treated as either Level 1 or Level 2 information, depending
on the method used and the number of cover classes employed. However, increasing the resolution of
the estimates (e.g., estimating cover to the nearest 1% or %5) does not necessarily increase the
precision of data collection). In fact it may comparisons among observers more problematic. Cover
estimation using fairly broad classes is often analyzed using the midpoint of each class as the cover
value. The size of the sampling unit (0.5 m X 1.0 m vs. 10 m X 10 m) also influences the use of the
data as Level 1 (relatively large quadrats for descriptive purposes) or Level 2 (relatively small
quadrats for analytical purposes) data.
Population and community abundance is sampled or censused quantitatively using a number of
measurements, including the number of individuals, density of individuals, percent cover, and
frequency. These measurements can be made on individual species, groups of species, or all species
present. Temporary or permanent plots or transects are used with Level 2 information gathering.
Condition information consists of relative cover (by species or ecological group) and number or
percent exhibiting a particular condition. Population structure measurements typically involve
counting the number of individuals in size distribution or height classes (e.g., number of
inflorescences, number of insect galls, height).
Applying Level 2 methods allows for data analysis, interpretation, and prediction. Smaller changes in
condition, structure, and abundance are detectable with Level 2 methods versus Level 1 methods.
Another advantage is that the effects of management on sites, areas, and species are more readily
evaluated. From a sampling perspective, data collection is repeatable by different individuals over
time. Disadvantages include higher requirements of expertise, time, and effort. Also, no information
about individual responses or fates is collected.
2.3.3
Quantitative Monitoring (Level 3): Demographic, Age or Stage Class
Analysis
Level 3 monitoring consists of collecting information for assessing life history or demographic
parameters such as survival and mortality and age or size distribution of the population. These studies
often measure marked or mapped individuals over time, whereas Level 2 monitoring may make
similar measurements on plants but not track individuals over time. By tracking individuals,
relationships between recruitment, survivorship, and mortality can be assessed for different age and
size classes. Causality can sometimes be determined from these studies.
Data collection typically involves measuring abundance attributes and assessing condition. Other
qualitative or quantitative information is often collected to help establish relationships between site
factors and demographic responses.
Strengths of Level 3 monitoring include all those listed for Level 2 monitoring. Additional strengths
include increased change detection capability, enhanced knowledge of life histories and response to
site conditions, and ability to predict demographic changes over time. Weaknesses include those
listed for Level 2 monitoring and high cost and time requirements.
2.4
Management and Monitoring Objectives
Monitoring should be objective-based. The success of a monitoring effort is based upon its ability to
assess the success or failure of specific management objectives. Objectives must be realistic, specific,
20
measurable, and written clearly. A limited suite of attributes should be chosen for assessing changes
in the overall condition of plant communities. It is up to the individual who implements and manages
resource monitoring to articulate objectives for his or her program. If good management objectives
already exist at an installation, then the remaining task is to define monitoring objectives related to
them. Monitoring objectives may also be influenced by original program goals, need for continuity
with historic data, methods, and available resources. A good discussion of management and
monitoring objectives is presented in Elzinga et al. (1998).
2.4.1
Potential Attributes and Management Applications Related to RTLA
When considering the scope and possibilities of a monitoring program, it can be helpful to be aware
of potential applications and the data required to support them. Data collection associated with these
examples could range from qualitative to quantitative. Discussion of management and monitoring
objectives is facilitated through the use of these examples, which are based on projects on DoD
installations. These examples are by no means exhaustive, but are meant as a springboard for
discussion and program development. The examples of applications and attributes presented here are
organized by broad program goals similar 2 to those often used on installations to organize and focus
monitoring activities and projects:
1)
Document the baseline condition and trends over time in vegetation, soils and hydrologic
stability, land cover, aquatic health, ecological processes, and other indices of biotic or
abiotic integrity and provide supporting evidence to planning and compliance documents.
2)
Document and characterize military land use/training intensity, frequency, types,
patterns, and damage over time, evaluate relationships between training and resource
condition (e.g., soils, water quality, vegetation, wildlife habitat, etc., including landscape
and site-specific impacts), and provide information that may be used to model
environmental impacts of training.
3)
Characterize the quality and sustainability of training environments. These issues are
related to planning events (units) and managing land use (DPT/Range) and may overlap
with conservation concerns.
4)
Provide information for evaluating the success of ecosystem management (considerable
overlap with baseline and trend program area above).
5)
Support land rehabilitation and maintenance (LRAM) and help evaluate success of
LRAM projects.
6)
Provide information, monitoring results, and regular reports to ITAM, Training/Range,
Natural Resources, the Installation Commander, and MACOM ITAM staffs for planning,
adaptive management, and optimizing RTLA application.
Attributes and potential management applications are listed for each program goal. Efforts to support
each program goal may have both short and long-term components consisting of a combination of
2
These are not necessarily formal RTLA program goals
21
quantitative and qualitative elements. More specific objectives can be developed and listed for each
monitoring attribute (lowest hierarchical level).
Document the baseline condition and trends over time in vegetation, soils and hydrologic stability,
land cover, aquatic health, ecological processes, and other indices of biotic or abiotic integrity and
provide supporting evidence to planning and compliance documents.
•
•
•
•
Soil erosion and hydrologic stability
o Soil loss estimates and soil erosion status – point estimates and spatially distributed
modeling
o Qualitative evidence of soil loss, movement, flow, litter movement, etc.
o Surface soil stability – aggregate stability, biological crusts, litter, plant cover, etc.
o Soil compaction
o Vegetation cover – total, amount of annual vs. perennial vegetation
o Document seasonality of erosion potential
Water quality and sedimentation effects
o Suspended sediment, turbidity
o Aquatic bioassessments – benthics, other indices of stream quality
Vegetation/biotic integrity
o Amount and distribution of vegetation/community/habitat types
o Weed abundance and spread (see more below under ecosystem management)
o Plant community composition and structure
o Density, cover, frequency, area of occurrence, health/vigor
o Diversity and richness indices
o Presence and abundance of expected vegetation functional groups
Faunal species, indicators and habitat quality
o Characterization of status and effects of training on species and habitats of concern and
faunal indicators
o Songbird/neotropical migrant bird surveys
o Small mammal and herptile surveys
Document and characterize military land use/training intensity, frequency, types, patterns, and
damage over time, evaluate relationships between training and resource condition (e.g., soils, water
quality, vegetation, wildlife habitat, etc., including landscape and site-specific impacts), and provide
information that may be used to model environmental impacts of training.
•
•
•
RFMSS and other military usage data – tabular and spatial data
o Monthly and yearly totals by unit type and training area
o Seasonal usage and trends in usage over multi-year periods
o Training usage intensity of particular facilities/areas of interest
Off-road vehicle maneuver patterns and unit footprints
o GPS mapping of actual usage to determine use patterns, number of miles traveled,
and proportion of on vs. off-road travel.
o Analysis of vehicle travel patterns relative to terrain features and land
cover/vegetation.
Training/testing damage, degradation characterization and mapping
o General patterns
22
Site specific/known areas (sometimes repeatedly used areas)
Post-maneuver damage assessments
o Application of remote sensing imagery and tools to map disturbance and detect
changes over time.
Complement monitoring with research aimed at building knowledge of links between
training effects and resource conditions.
Use experimental approaches as well as “control” or “reference” sites/areas for comparison
Provide information for watershed assessments and erosion modeling
Applications using condition and usage data from other program areas
o
o
•
•
•
•
Characterize the quality and sustainability of training environments. These issues are related to
planning events (units) and managing land use (DPT/Range) and may overlap with conservation
concerns.
•
•
•
•
•
•
•
Mapping land condition status – criteria may be singular or a combined index of condition.
Minimizes harmful effects of training, helps Range personnel distribute training optimally,
and provides input to planning tool such as ATTACC
o Soil erosion
o Weed presence/risk
o Habitat degradation/quality for species or communities of concern
o Fire risk and effects of wildfire on training and scheduling
o Disturbance extent and severity
o Other measures of training land carrying capacity
Recovery requirements (# of years) for different ecological management units
o Applicable to natural and assisted land rehabilitation efforts and modeling training
sustainability vis-à-vis soil erosion and vegetation cover
Vegetation and land cover maps
o Vegetation types
o Structural types – physiognomic subclass level, can be species independent
o Acreage losses or gains in important training environments over time?
Aerial and ground concealment (from detection/observation)
o Vertical structure and tactical concealment value
Vehicle mobility and terrain analysis
o Spacing and size of trees
o Soil trafficability
o Soil conditions
o Slope steepness and topography
o Stream crossings
o Roads, trails, and accessibility
o Condition of roads and trails – driveability
Sensitive areas: slow-go and no-go areas
o Cultural, environmental, other constraints mapped for troop use.
Incorporation of multiple overlays into custom products for planning and training
Provide information for evaluating the success of ecosystem management (considerable overlap with
baseline and trend program area above).
23
•
Fire Regime and Effects
o Area burned
o Burn severity
o Fire frequency/return interval
o Vegetation and hydrologic response (pre-post burn or control vs. burned area)
o Vegetation dynamics over time
o Fuel loads
o Relationships between changes in fire regime (frequency, intensity, extent, etc.)
and natural resources conditions (vegetation types/quality, soil erosion, etc.)
•
Forest Management
o Tree/forest health indices
o Species dominance and forest structure
o other
•
Invasive Plant Management
o Mapping species distributions and areas of concentration
o Monitoring status and changes in abundance/distribution over time
o Evaluating success of management efforts
•
Habitat Conservation and Enhancement
o Total acreage in each vegetation/habitat/land cover type
o Habitat fragmentation or loss over time
o Monitoring attributes of interest (structure, composition, function, indicator
species)
•
Ground Truthing Data for Remote Sensing Applications
o Accuracy assessments of vegetation, land cover, disturbance, or other maps using
RTLA data from known locations
o Multiple approaches and applications related to this and other program areas
Support LRAM and help evaluate success of rehabilitation projects.
•
•
•
•
•
•
Identification and assessment of potential LRAM sites
Evaluation of species used in seeding mixes
Evaluation of effectiveness of structural approaches or controls
Evaluation of management success for functional groups such as perennial cover, individual
species, composition, and structure (e.g., perennial bunchgrasses, forbs, shrubs, trees, total
vegetation cover)
Assessment of desirable structural components from habitat or training perspective
Maintenance of soil erosion losses below threshold levels
Provide information, monitoring results, and regular reports to ITAM, Training/Range, Natural
Resources, the Installation Commander, and MACOM ITAM staffs for planning, adaptive
management, and optimizing RTLA application.
•
Comprehensive technical reports
24
•
•
•
•
•
•
2.4.2
Assessment or determination of “baseline”, “threshold”, or “reference” conditions using
RTLA and other data
Range and Training reports and briefs
Command briefs
MACOM ITAM report packs
Information sharing and collaboration with other local agencies and organizations
Ongoing collaboration with Natural Resources staff to optimize and complement efforts
Management Goals and Objectives
Management objectives will vary depending on the management mission of a particular organization.
For example, management objectives driven by military training (or the training community) may be
very different from those established by land management or conservation professionals.
Management objectives help to direct resource management by defining desired conditions or trends
in resource conditions. Sources of information for setting objectives include existing management
plans and environmental (e.g., NEPA) documents, ecological models, reference sites or comparison
areas, related or similar species and communities, expert opinion, and historic records and
photographs (Elzinga et al. 1998). A complete management objective, which forms the basis for one
or more monitoring objectives, should include the following components:
what will be measured – direct measurement of species/community or habitat indicator (indirect)
location or geographic area of interest – defines the limits to which results will be applied
attribute measured, e.g., size, density, cover, frequency, qualitative estimate of abundance, areal
extent
objective action - maintain, increase, or decrease
quantity/status - measurable status or degree of change for attribute – can be quantitative or
qualitative
time frame - length of time specified for management to prove effective.
Objectives can describe either a desired condition or a change relative to current conditions:
The first type can be described as target or threshold objectives. This type of objective uses a
predetermined threshold to gauge the effectiveness of management. For example: maintain the size of
population A at 450 individuals; increase the acreage of open woodland to 6000 ha; maintain the
presence of threatened species A and B at Site C. The success of meeting the objective is assessed by
comparing the current state of the measured attribute either to the desired state or an undesired state.
Presence of the undesired state should serve as an indication that management should be altered.
The second type can be described as change or trend management objectives. This type of objective
specifies a change relative to the existing situation. For example: increase perennial grass cover by
40%; decrease severe off-road disturbance by 20%; decrease frequency of weed species Y by 50%.
Trend objectives are useful when little information is available to describe a desired future condition,
or where the current status is less important than trends over time. Change detection objectives are
25
often appropriate when a significant change in management occurs and change is anticipated. If
preliminary sampling provides information about the population status, then a change objective may
be rewritten as a threshold objective.
Some of the major management concerns related to resource condition and military land uses include:
•
•
•
•
•
Off-road vehicle and other training impacts to vegetation and soils
Spread of noxious weeds or species which outcompete desirable vegetation
Changes in structural attributes of plant communities – loss of concealment resources,
wildlife habitat
Impacts of altered fire intensities/frequencies and other natural processes
Soil erosion, sedimentation, and water quality impacts
In light of these management concerns, the following are examples of general management goals
related to vegetation and soils:
1. Sustain and maintain healthy and diverse ecosystems.
2. Maintain soil stability and susceptibility to erosion at acceptable levels.
3. Maintain realistic and sustainable training environments for desired training loads.
4. Revegetate selected disturbed areas to pre-disturbance conditions.
5. Minimize the establishment and spread of undesirable non-native plants.
These general management goals should be refined so that corresponding monitoring objectives can
be developed to address specific needs. If the success or failure of the management objective cannot
be gauged, then there is no way of knowing if management activities are effective, or if management
goals are met. The following are examples of specific management objectives:
♦ Maintain the current (2005) spatial distribution and abundance, i.e., acreage, of each major
plant community from 2005-2010 (target objective).
♦ Within each community type, maintain 2003-2005 native grass, forb, shrub, and tree cover
from 2005-2010 (target objective).
♦ Within each community type, maintain 2003-2005 native grass, forb, shrub, and tree diversity
from 2005-2010 (target objective).
♦ Increase the forb diversity of woodland communities by 25% between 2005 and 2015
(change objective).
♦ In existing shrub communities, maintain current (2003-2005) densities for each shrub species
from 2005-2010 (target objective).
26
♦ Allow a decrease in the ranked abundance of Chlorogalum purpureum var. purpureum
(Purple Amole) in each of the 5 permanent macroplots at the Jones Mountain Site (Fort
Hunter Liggett, CA) of no more than 1 rank class between 2005 and 2007.
♦ Within each training area, maintain current (2003-2005) soil erosion rates from 2005-2010
(target objective).
♦ Within each community type, maintain current (2003-2005) levels of bare ground from 20052010 (target objective).
♦ Maintain the current (2004) areal extent of forest, woodland, and grassland communities from
2004-2009 (target objective).
♦ In forested areas used for bivouac, maintain an overstory tree density of at least 40 trees/ha
from 2003-2008 (target objective).
♦ For revegetation sites, increase cover of desirable plant species to within 50% of undisturbed
plant cover after three years of recovery (change objective).
♦ For revegetation sites, provide at least 20% perennial grass cover, 10% shrub cover, and 5%
perennial forb cover after 2 years of recovery (target objective).
♦ For burn sites, increase total plant canopy cover to 25% after 1 year (target objective).
♦ Maintain the current distribution and abundance of weed species X on Fort USA from 2003
to 2008 – this could be treated as quantitative or qualitative, depending on the approach that
is most feasible (target objective).
♦ Decrease the number of hectares on Fort USA infested with species Y (species is common to
abundant) to 400 ha (target objective).
♦ For sites treated for weed infestations, decrease the density of target weed species by at least
50% one year after treatment (change objective).
♦ Maintain the number of km of road shoulders with a knapweed (all species) ranked
abundance of 5 or more (target objective).
2.4.3
Monitoring Objectives
Complete management objectives form the basis for monitoring or sampling objectives. In addition to
the “what”, “where”, and “when”, a monitoring objective should specify information such as the
target level of precision (acceptable error), power, confidence level (false change error rate), and the
magnitude of change we want to detect. Without specified targets for these parameters, estimates of
population parameters might have excessively large confidence intervals or low power (e.g., only a
20% chance of detecting the magnitude of change that was desired). The necessary components of
monitoring objectives differ for target management objectives and change management objectives.
The sampling objective for target objectives is to estimate a parameter in the population, estimate a
proportion, or to estimate total population size. This estimate is then compared to the threshold value
27
specified. To accomplish this, it is necessary to specify the confidence level (i.e., how confident do
you want to be that your confidence interval will include the true value?), and the confidence interval
width (i.e., how close to the estimated mean do you want to be?).
Example:
Management objective: Decrease the density of Juniperus trees less than 10 cm in diameter in
abandoned agricultural fields to 15 trees/acre between 2005 and 2010.
Monitoring objective: Estimate the density of Juniperus trees less than 10 cm in diameter. We
want to be 90% confident that mean density is within 10% of the estimated true value.
The sampling objective for change objectives is to determine if there has been a change in a
population parameter for two or more time periods. These objectives must include the desired power
(missed change or Type II error), the acceptable false change errors rate (Type I error), and the
desired minimum detectable change (MDC) (the smallest change you are hoping to detect).
Example:
Management objective: Increase the density of flowering individuals of Tauschia hooveri
(Hoover’s desert parsley) at the Yakima Ridge site by 25% between 2005 and 2015.
Monitoring objective: We want to be 90% confident of detecting a 25% increase in mean
density with a false change error rate of 0.10. This objective specifies a power of 90%, a false
change error rate of 10%, and an MDC of 20%.
If the sampling interval is not specified in the management objective, it should be specified in the
monitoring objective (i.e., seasonally, annually, every 2 years, 5 years, etc.). The sampling interval
can be less than the timeframe specified in the management objective. For example, if a given change
is desired over a 6 year period, monitoring every 2 or 3 years may be appropriate to see if there has
been progress toward the objective.
When monitoring does not involve sampling, the management objective should provide enough
information to evaluate its success or failure. This is the case where qualitative assessments are done
for areas or where a complete census is performed. Management objectives of this type therefore do
not need to provide additional components beyond what, where, and when.
2.4.4
Paired Management and Monitoring Objectives
2.4.4.1
Examples of Training-oriented Objectives
Example 1:
Management Objective (desired status or condition - management threshold)
“Mean soil erosion status (estimated soil loss/published soil loss tolerance) shall be maintained at
levels of less than 100% for all major landcover types on the installation.”
Monitoring Objective
28
“We want to make annual estimates of erosion losses for all the major landcover types. We want to be
90% confident that the estimate is within 10% of the true value.”
Example 2:
Management Objective (desired condition)
“[Given that sagebrush is sensitive to off-road maneuvers and fire] we want to maintain mean big
sagebrush cover of at least 25% in existing shrub stands across the installation.”
Monitoring Objective
“We want to obtain an estimate of the average canopy cover of big sagebrush in existing shrub stands
every two years. We want to be 90% confident that the estimate is within 20% of the true value.”
2.4.4.2
Examples of Land Management-oriented Objectives
Example 1:
Management Objective (desired trend)
“We want to see a 50% decrease in yellow star thistle in areas sprayed with herbicides on Fort Hunter
Liggett during the next three years.”
Monitoring Objective
“We want to be 90% sure of detecting a 50% change in the density of yellow star thistle on
herbicided areas and untreated areas annually for the next 3 years. We are willing to accept a 10%
chance that we conclude a change took place when in fact there was no change.”
Example 2:
Management Objective (desired trend)
“Through the application of management activities, we want to see a 30% increase in the cover of
native warm-season grasses at the McLaughlin Cemetery Glade (Fort Leonard Wood, MO) between
2004 and 2007.”
Monitoring Objective
“We want to be 90% sure of detecting a 30% change in the aerial cover of native warm-season
grasses at the McLaughlin Cemetery Glade 3 years after the initiation of restoration activities.
Monitoring will be done annually. We are willing to accept a 10% chance that we conclude a change
took place when in fact there was no change.”
Example 3:
Management Objective (desired status)
“We want to maintain the current flowering population of Tauschia hooveri (Hoover’s desert parsley)
at the Yakima Ridge site (Yakima Training Center, WA) over the next ten years.”
29
Monitoring Objective
“We want to obtain annual estimates of the population of Hoover’s desert parsley at the Yakima
Ridge site from 2003-2013. We want to be 90% confident that the estimates are within 20% of the
estimated true mean.”
2.4.4.3
Additional Examples of Paired Management and Monitoring Objectives
1. In grassland communities, decrease the frequency of Bromus tectorum (cheatgrass) by 30% from
2005-2008.
We want to be 80% certain of detecting a 30% decrease in frequency with a false change error rate
of 0.20.
2. Maintain native grass and forb species diversity at 2003-2004 levels.
Obtain estimates of grass and forb diversity at 2 year intervals with 90% confidence intervals no
wider than ± 10% of the estimated diversity.
3. Maintain the areal extent and distribution of minimally, moderately, severely, and completely,
disturbed (e.g., off-road maneuver and assembly) areas at 2005 levels.
Estimate annually the extent (i.e., number of square meters, ha) of disturbed lands in each
disturbance category and map areas using GPS – the objective has all the information for
evaluating results.
4. In areas subject to extensive off-road maneuvers, allow a decrease in the cover of native plants of
no more than 30% relative to undisturbed conditions between 2005 and 2008 (compared to
undisturbed areas).
Be 80% confident of detecting a 30% relative decrease in native plant cover with a false-change
error of 20% (20% chance of concluding that a change took place when in fact there was no
change).
5. In forested areas used for bivouac, maintain an overstory tree density of at least 40 trees/ha from
2003-2006.
We want to be 95% certain that the estimates are within 15% of the estimated true density.
6. In existing shrub communities, maintain current (2003-2005) densities for each shrub species from
2005-2010.
We want to be 95% confident that annual density estimates are within 10% of the estimated mean
density.
7. Increase the forb diversity of woodland communities by 25% between 2005 and 2010.
We want to be 90% sure of detecting a 25% relative increase in forb diversity. We are willing to
accept a 10% chance of a false-change error.
30
8. Increase the Jones Mountain (Fort Hunter Liggett, CA) population of Chlorogalum purpureum var.
purpureum (purple amole) to 500 individuals by the year 2009.
We want to be 95% confident that the population estimate is within ± 10% of the estimated true
value. This objective applies where sampling is used. If all of the individuals in the population are
counted (census), then the monitoring objective is already specified within the management
objective (are there at least 500 individuals by 2009 – yes or no?).
9. Decrease the ranked abundance of Lythrum salicaria (purple loosestrife) in each of the four
permanent macroplots at the Ives Road Fen site by 2 rank classes between 2004 and 2006.
Estimate the ranked abundance of purple loosestrife in each macroplot – the objective has all the
information for evaluating results. Estimates could be made annually or in 2004 and 2006.
10. Do not allow erosion status estimates (estimated loss/allowable loss) for each training area to
exceed 100% in any given year from 2004-2014.
Estimate erosion status annually for each training area. We want to be 90% confident that the
estimate is within ± 20% of the estimated true value.
Additional examples of management and monitoring objectives for RTLA programs are presented in
CEMML (2006).
2.5
Determining Benchmarks
Benchmarks or management thresholds provide goals for resource management, which in turn help to
guide monitoring efforts. In the context of quantitative monitoring, benchmark conditions must be
defined for the attributes of interest, which are measured during sampling. Benchmarks are often a set
of well-defined conditions as opposed to a qualitative ranking such as “poor”, “fair”, or “good”.
However, qualitative benchmarks may be used where monitoring resources are limited or where
qualitative management objectives are specified. In some cases, monitoring data are compared to
initial conditions to gauge improvement or degradation, emphasizing trend over the attainment of
specific conditions. In fact, both approaches can be employed simultaneously without expending
additional effort or cost, the only difference being that attention must be given to defining
benchmarks.
Commonly used benchmarks include current conditions or those preceding changes in management,
pristine or near-pristine sites (reference areas), historic or presettlement conditions, desired plant
community, and projections from biotic and abiotic information – “climax” or potential vegetation.
Some of these benchmarks represent hypothetical ecological standards. Desired plant community
(DPC) is a concept that has been adopted by some land managers as a practical benchmark for
vegetation management. This concept has also been described as desired future condition (DFC).
Wagner (1989) describes the DPC concept:
“The DPC is … an expression of the site specific vegetation management
objectives instead of the more common, subjective way of stating objectives such
as changing vegetation condition from poor to fair or from fair to good. The
description of the characteristics of the DPC (species composition, production,
31
cover structure, etc.) is based on those of a real, documented community
occurring on the same or like site in another area. Therefore, vegetation
management objectives expressed as a DPC, besides being more specific and
measurable, are ultimately more realistic… The DPC is consistent with the site’s
documented capability to produce the required vegetation attributes through
management, land treatment, or a combination of the two. The DPC is a
management determination which may correspond to the existing plant
community, the potential plant community, or some intermediate community.”
Where resource conditions generally meet management objectives and changes in management are
minimal, current conditions may be chosen as a benchmark for quantitative or qualitative monitoring.
The emphasis might therefore be on maintenance of conditions versus improvement over time.
2.6
Selecting Variables to Measure
Assessing the condition of any ecosystem, be it forest, woodland, grassland, shrubland, or arid
ecosystem, is highly complex, requiring the examination of a number of factors which characterize or
contribute to the degradation or improvement of various ecological units. Changes or trends in
ecological resources can be detected in the short and long-term by monitoring resources either
directly or indirectly through indicators. The selection of variables to measure is largely determined
by program objectives and corresponding data requirements. Required data elements are simply
whatever has been requested by the responsible party or agency. Population-level monitoring can be
straightforward if the species lends itself to being measured directly. Community and landscape-level
monitoring can present the most difficulties in the selection of attributes. For this reason, indicators
are used most often at the latter scales. Whatever the variables chosen, they should be robust yet
specific enough to respond to anticipated or unknown stresses and changing conditions.
Compliance-type monitoring generally specifies the attributes to be monitored. Standardization of
data collection requirements across organizational units (e.g., MACOM requirements for installations,
range condition ratings by BLM districts) simplifies the process of determining what to measure by
specifying data needs. For example, the Army Training and Testing Area Carrying Capacity
(ATTACC) model, currently being developed by Combat Training Support Center (CTSC) for the
Office of the Deputy Chief of Staff for Operations and Plans (ODCSOPS) requires soil loss estimates
by training area in order to estimate the capacity of lands to support training. Soil erosion status is
used to help determine “land condition”. Training load and land rehabilitation costs are also required
by the ATTACC model. The ATTACC requirements and calculation of carrying capacity estimates
are presented in the ATTACC Program Handbook (CTSC 1998).
2.6.1
Selecting Indicators of Resource Condition
Assessment of condition requires that judgments are made regarding the ecological significance of the
indicator variables measured. Multiple indicators are preferable to single indicators because of the
decreased chance of false positive and false-negative signals. Qualitative measurements of indicators
may be appropriate and cost effective where large impacts are anticipated and readily apparent.
32
The importance of indicator selection cannot be over emphasized, since any long-term monitoring
project will only be as effective as the indicators chosen (Cairns et al. 1993). Once management
goals have been specified, a framework must be developed for selecting indicators and utilizing the
data that is collected. The number of potential indicators is relatively high, and the selection of
several “good” indicators is not an easy task. However, economic and ecological considerations help
to limit the number of indicators that can be measured to a fraction of those available. In some cases,
indicators may need to be developed if they are specific to an installation need (e.g., rutting, types of
training damage, etc.).
The following list of indicator characteristics was developed for applications to environmental and
water quality, but are widely applicable to terrestrial ecosystems as well (Cairns et al. 1993). The
ideal indicators are:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Biologically relevant, i.e., important in maintaining a balanced community;
Sensitive to stressors without an all or none response or extreme natural variability;
Broadly applicable to many stressors and sites;
Diagnostic of the particular stressor causing the problem;
Measurable, i.e., capable of being defined and measured using a standard procedure with
documented performance and low measurement error;
Interpretable, i.e., capable of distinguishing acceptable from unacceptable conditions in a
scientifically and legally defensible way;
Cost-effective, i.e., inexpensive to measure, providing the maximum amount of
information per unit effort;
Integrative, i.e., summarizing information from other, unmeasured indicators or variables;
Historical data are available to define “natural” variability, trends, and possibly
acceptable and unacceptable conditions;
Anticipatory, i.e., capable of providing an indication of degradation before serious harm
has occurred: early warning;
Nondestructive of the ecosystem;
Potential for continuity in measurement over time;
Of an appropriate scale to the management problem being addressed;
Not redundant with other measured indicators;
Timely, i.e., providing information quickly enough to initiate effective management
action before unacceptable damage has occurred.
Numerous research and monitoring projects have been undertaken to identify indicators that are
ecosystem appropriate and meet the criteria described above. In some cases it is appropriate to
measure the attribute of interest directly (e.g., is the acreage of Purple Loosestrife expanding?) while
in other cases the use of indicators is appropriate (e.g., are we maintaining the quality of native plant
communities in area x?). Some indicators are discrete, measurable attributes whereas other may be
represented by one or more additional attributes or may require further determination of specific
metrics.
The selection of appropriate indicators for arid and semiarid environments has received much
attention in recent years. Some work has focused on selecting site-specific indicators and attributes,
whereas other work has investigated indicators that are broadly applicable across a range of
environments.
33
The Rangeland Health Program was initiated following a report released by the Committee on
Rangeland Classification (1994). Development of Rangeland Health concepts, methods, and
applications has been spearheaded by USDI, USDA and university researchers. The program aims to
evaluate an ecological site's potential to conserve soil resources by assessing a series of indicators for
ecosystem processes and site stability. Seventeen indicators are evaluated to assess three ecosystem
attributes (soil and site stability, hydrologic function, and biotic integrity) for a given location.
Indicators include rills, water flow patterns, pedestals and terracettes, bare ground, gullies, wind scour
and depositional areas, litter movement, soil resistance to erosion, soil surface loss or degradation,
plant composition relative to infiltration, soil compaction, plant functional/structural groups, plant
mortality, litter amount, annual production, invasive plants, and reproductive capability. The
assessment techniques employed by the Rangeland Health methodology (Pellant et al. 2005) is
qualitative in nature and is not intended to be used as a monitoring tool. However, the indicators have
been shown to meet important selection criteria outlined above. A number of methods could be
applied to assess the condition of the indicators listed. The Rangeland Health framework and
approaches are further described in Section 4.5.2 Rangeland Health.
In Arches National Park, Utah, Belnap (1998) used a systematic approach to select indicators of
natural resource condition. The basic approach involved sampling a number of vegetation and soil
variables in impacted and unimpacted areas. Variables that differed significantly between compared
sites were chosen as potential indicator variables. Potential indicators were subsequently evaluated
using site-specific criteria. The required selection criteria for indicators included low impacts of
measurement, repeatability of measurements, correlation with land-use (i.e., visitor/recreation)
disturbances, and ecological relevancy. Those indicators that met the required criteria were then
evaluated for additional desirable characteristics, including: (1) quick response to land-use
disturbance and management actions, (2) minimal spatial, temporal, and climatic variability, (3) ease
of sampling, (4) large sampling window, (5) cost effectiveness, (6) short training time, (7) baseline
data available, and (8) response over a range of conditions (impacts are evident even for minimal
disturbance). The remaining indicators were then examined for ecological relevancy. A final set of
indicators was then chosen and field tested. This approach should prove applicable in situations
where indicators are required to measure both land-use impacts and response to management actions.
Several years may be required to survey habitats, develop a list of potential indicators, determine
ecological relevance, and field-test chosen indicators (Belnap 1998).
Examples of vegetation and soil indicators selected for areas in Utah (Belnap 1998) and New Mexico
(Whitford et al. 1998) include:
-
bare patch index
cover of long-lived grasses
weighted soil surface stability index
cover of plant species toxic to livestock
cover of exotic species
cover of increaser species
number of social recreational trails
soil crust index
soil compaction
soil aggregate stability
vascular plant community composition
soil surface protection index
soil biological characteristics
34
Some of these indices are measured directly some are derived from several measurements or
variables.
Additional work on rangeland indicator selection was published in 2003 by the Sustainable
Rangelands Roundtable (SRR) 3. The SRR is a collaborative partnership process jointly funded by the
USDA Forest Service, Bureau of Land Management, the U.S. Geological Survey, and Colorado State
University. The SRR mission is to promote ecological, economic, and social sustainability of
rangelands through the development and widespread use of criteria and indicators for rangeland
assessments, and by providing a forum for dialogue on rangeland sustainability. The SRR met 11
times and employed a collaborative Delphi Process 4 to produce a report of criteria and indicators for
assessing sustainable rangeland management in the United States. During the course of these
meetings, more than 100 scientists representing about 50 agencies, universities, professional societies,
NGOs, and private businesses participated in the process. A framework of five criteria and 64
indicators was developed to promote standardized, periodic rangeland inventory and monitoring at
multiple scales. The first three criteria relate to resource condition, whereas the last two pertain to
social, economic, and legal issues. Indicators for the first three criteria are presented below.
Criterion 1. Conservation and Maintenance of Soil and Water Resources of Rangelands
Soil-based Indicators
1. Area and percent of rangeland soils with significantly diminished organic matter and/or
high carbon:nitrogen (C:N) ratio.
2. Area and extent of rangelands with changes in soil aggregate stability.
3. Assessment of microbial activity in rangeland soils.
4. Area and percent of rangeland with significant change in extent of bare ground.
5. Area and percent of rangeland with accelerated soil erosion by water or wind.
Water-based Indicators
6. Percent of water bodies in rangeland areas with significant changes in natural biotic
assemblage composition.
7. Percent of surface water on rangeland areas with significant deterioration of their
chemical, physical, and biological properties from acceptable levels.
8. Changes in ground water systems.
9. Changes in the frequency and duration of surface no-flow periods in rangeland streams.
10. Percent of stream length in rangeland catchments in which stream channel geometry
significantly deviates from the natural channel geometry.
Criterion 2: Conservation and Maintenance of Plant and Animal Resources on Rangelands
11. Extent of land area in rangeland.
12. Rangeland area by plant community.
13. Number and extent of wetlands.
14. Fragmentation of rangeland and rangeland plant communities.
15. Density of roads and human structures.
3
http://sustainablerangelands.cnr.colostate.edu/
Delphi is an iterative process whereby experts answer a set of questions; individual responses are tabulated and returned to the
participants, along with summary analyses and comments. Individuals are then afforded an opportunity to revise their original
answers in response to the group feedback. The process continues until a pre-determined level of consensus is achieved.
4
35
16. Integrity in natural fire regimes on rangeland.
17. Extent and condition of riparian systems.
18. Area of infestation and presence/absence of invasive and other nonnative plant species of
concern.
19. Number and distribution of species and communities of concern.
20. Population status and geographic range of rangeland-dependent species.
Criterion 3: Maintenance of Productive Capacity on Rangelands
21. Rangeland aboveground biomass.
22. Rangeland annual productivity.
23. Percent of available rangeland grazed by livestock.
24. Number of domestic livestock on rangeland.
25. Presence and density of wildlife functional groups on rangeland.
26. Annual removal of native hay and nonforage plant materials, landscaping materials,
edible and medicinal plants, and wood products.
The National Park Service has developed a program called Vital Signs Monitoring 5 (VSM). VSM
organizes approximately 270 park units into 32 monitoring networks to conduct long-term monitoring
for key indicators of change, or “vital signs.” Vital signs are measurable, early warning signals that
indicate changes that could impair the long-term health of natural systems. Individual networks link
parks that share similar geographic and natural resource characteristics. Each network is tasked with
designing a single, integrated program to monitor both physical and biological resources, such as air
quality, water quality, soils, exotic species, and threatened and endangered species. The list of
environmental vital signs selected for monitoring the health of these resources is expected to vary
among networks, reflecting the needs and natural resources of the parks. Partnership opportunities,
ecological indicators and methodologies involved differ from one network to another, as appropriate.
Vital Signs employs a framework consisting of three levels (Level 1, Level 2, and Level 3 categories)
to which each network vital sign is assigned. For example, the vital sign 'Biological Soil Crusts' is
assigned to Level 1 = Geology and Soils, Level 2 = Soil Quality, and Level 3 = Soil Structure and
Dynamics. The vital sign 'Landbirds' is assigned to Level 1 = Biological Integrity, Level 2 = Focal
Species or Communities, and Level 3 = Birds. The three most common vital signs identified by the
initial 12 networks are exotic plant species occurrence, changes in land cover type (e.g., agricultural
to suburban), and vegetation community composition and structure.
The USDA Forest Service, Forest Health Monitoring (FHM) is a national program designed to
determine the status, changes, and trends in indicators of forest condition on an annual basis. Forest
indicators used by this program include crown condition, lichen communities, ozone injury, downed
woody debris, tree damage, vegetation diversity and structure, tree mortality, and soil condition. FHM
is further described in Section 4.5.1 Forest Health Monitoring.
The Forest Inventory and Analysis (FIA) Program is another USDA Forest Service program
(http://fia.fs.fed.us/). FIA reports on status and trends in forest area and location; in the species, size,
and health of trees; in total tree growth, mortality, and removals by harvest; in wood production and
utilization rates by various products; and in forest land ownership. FIA recently assumed
responsibility for all former Forest Health Monitoring Program (FHM) plot work on a national level.
In addition to traditional forestry and forest mensuration measurements, FIA also collects information
on indicators that were developed and initially measured by the FHM Program in the 1990s (see
5
Description of the Vital Signs Program from http://www.nature.nps.gov/protectingrestoring/IM/vitalsignsnetworks.htm
36
Section 4.5.1 Forest Health Monitoring). In 1999, responsibility for these data elements was
transferred to FIA and now are a subset of the FIA sample grid (one forest health plot for every 16
standard FIA plots). On the forest health plots, core FIA measurements (for example, height,
diameter, species) are made, as well as the forest health measurements.
Harper et al. (1996) have outlined several indicators of community quality specific to southern pine
woodlands, including: a) wiregrass dominance as an indicator of little fire suppression or soil
disturbance, b) old growth pine as an indicator of high quality sites for many TES, c) other indicator
species, and d) structural and compositional aspects. Land-use impact issues on military installation in
the Southeast include: fragmentation and land-use conversion, fire and fire suppression, alteration of
hydrology, groundcover disturbances, erosion and sedimentation, soil compaction, exotic and pest
species, and unnatural fertilization. Indicators must accordingly be applicable to these and other
management issues.
Several studies of national-level indicators have been undertaken in recent years. The National
Research Council (2000) recommended a number of ecological indicators at the national level.
Recommended indicators of ecosystem extent and status include land cover and land use. Changing
proportions of various categories are indicative of changes in these indicators. Recommended
indicators of ecological capital include total species diversity, native species diversity, soil organic
matter, and nutrient runoff. Recommended indicators of ecosystem functioning include production
capacity (chlorophyll density), net primary production, carbon storage, stream oxygen, and trophic
status of lakes. These indicators are broad, but are widely applicable and have been refined for
specific needs at regional and local scales. Additional research involving selection and evaluation of
ecological indicators has been published by The Heinz Center (2003). The State of the Nation's
Ecosystems provides up-to-date information regarding the status of data for evaluating the condition
of numerous national indicators (http://www.heinzctr.org/ecosystems/report.html). Examples of
indicators for forest, grassland, and shrubland ecosystems are presented in Table 2-4.
37
Table 2-4. System dimension and biological component indicators for forests, grasslands, and shrublands (The Heinz Center 2003).
Forests
CATEGORY
SYSTEM DIMENSIONS
BIOLOGICAL
COMPONENTS
INDICATOR(S)
DESCRIPTION
Forest Types
How is the area occupied by major forest types changing?
Forest Pattern and Fragmentation
How fragmented are U.S. forests?
At-Risk Native Species
What are the percentages of forest-dwelling species that are at different levels of risk of
extinction?
Area Covered by Non-native Plants
What percentage of the plant cover in forests is not native to the region?
Forest Age
How much of the nation's forests is young, middle-aged, or old?
Forest Disturbance: Fire, Insects, and
Disease
How many acres are affected each year by fires, insects, disease, windstorms, and ice?
Fire Frequency
Are forest fires burning much more or less frequently than in presettlement times?
Forest Community Types with Significantly
Reduced Area
How much area is occupied by forest types that have significantly declined in area since
presettlement times? Are these forest types increasing or decreasing in area at present?
Grasslands and Shrublands
SYSTEM DIMENSIONS
BIOLOGICAL
COMPONENTS
Area of Grasslands and Shrublands
How much land is covered by grasslands and shrublands?
Land Use
How are grasslands and shrublands used? How many acres are used for livestock grazing;
oil, gas, and mineral development; rural residences; intensive recreation; "protected areas";
and the Conservation Reserve Program?
Area and Size of Grassland and
Shrubland Patches
What fraction of grasslands and shrublands is found in patches of various sizes?
At-Risk Native Species
How many grassland and shrubland native species are at different levels of risk of
extinction?
Non-native Plant Cover
What percentage of grassland and shrubland plant cover is not native to the region?
Population Trends in Invasive and
Noninvasive Birds
Are invasive bird populations increasing more than other bird populations?
Fire Frequency Index
Are grassland and shrubland fires occurring more or less frequently than in presettlement
times?
38
2.7
Monitoring Intensity and Frequency
Determining the intensity (level of monitoring approach and effort) of sampling is influenced by three
principal factors: 1) program objectives, 2) funding and other resources, and 3) actual or perceived
threats or level of risk. Intensity and frequency of data collection is influenced also by program or
management objectives, including documents such as Integrated Natural Resource Management Plans
(INRMPs); or ongoing mission, restationing, or land acquisition environmental documents including
Environmental Impact Statements (EISs) and Environmental Assessments (EAs). In addition to
supporting training activities and conservation goals, monitoring is often specified as mitigation for
activities that potentially impact the environment.
The frequency or periodicity of repeated sampling should reflect the rate of change within the
attributes or indicators selected for monitoring. Moreover, the frequency should be designed to give
early warning of significant degradation. Monitoring should be frequent enough so that a population,
community, or ecosystem would not undergo extreme degradation between sampling periods
(Committee on Rangeland Classification 1994). Using this logic, more frequent sampling may be
required in areas which receive more disturbance (i.e., are more dynamic). In general, sampling more
than once per year is impractical, except perhaps for small scale investigations.
Monitoring intensity and frequency decisions may vary by plant community, sites of special interest
or concern, or disturbances levels and may change from year to year. The simplest and most
expensive monitoring plan would sample all locations every year. Other plans would survey some
plots some years and other plots all years. The costs associated with sampling may be reduced by
making alterations to the sampling design.
Several monitoring scenarios are presented to illustrate possible solutions and their respective
tradeoffs (monitoring scenarios measure change in bare ground on an installation over a series of
years):
a) Sample all plots every year (Table 2-5A). This is the best (highest confidence that the data
represents actual conditions) and the most expensive sampling protocol. The data set is
complete from one year to the next. Options for statistical analyses are the greatest (i.e., data
from all years can be used). When all plots are surveyed all years, sample size is the largest
and variance, or the difference among the samples, is the lowest. If statistically significant
differences (P<0.05) among years exists, this data set will most likely identify those
differences. The benefits of surveying all plots annually include greater assurance of
identifying and interpreting cause and effect and the ability to analyze larger subsets of data.
Analyzing subsets of data are beneficial when changes in mission occur. For example,
information may be needed for an area not described by a recognized designation. The larger
the pool of information, the greater the chance a representative subset is available. When
breaks occur between years of monitoring, interpretative and predictive ability decrease, but
so do costs.
39
A. BareGround
100
Percent
80
60
40
20
0
1989
1990
1991
1992
1993
1994
1995
Years
C. BareGround
B. BareGround
80
80
60
60
Percent
100
Percent
100
40
20
40
20
0
0
1989
1990
Years
A1
1992
Years
D. BareGround
E. BareGround
1991
1992
1993
1994
A1
1989
1995
80
80
60
60
Percent
100
Percent
100
40
20
0
1
1990
1
1991
1
1993
1
1994
A1
1995
40
20
A
1989
1
1990
2
1991
A
1992
Years
3
1993
4
1994
A
1995
0
A 1 2 3 4
1989
A 1 2 3 4
1992
Years
A 1 2 3 4
1995
Table 2-5. Effect of various sampling schemes on monitoring results. (A) All plots are monitored all
years (n = 66). (B) All plots are monitored annually for a specified number of years and then
monitored at designated intervals; in this case, every fourth year. The dashed bars indicate data not
quantitatively collected during interim periods. (C) All plots are monitored every third year (n = 66) - a
single subset of plots (n = 25) are monitored in interim years. (D) All plots are monitored periodically
(n = 66); in this case, every third year. During interim periods subsets are sampled on a rotating
basis, (sampling without replacement); in this case, over four interim periods (n = 17). (E)
Comparison of subset data to data from all plots. The letter 'A' designates all plots are monitored.
The numbers (1, 2, 3, and 4) designate the subsamples of plots.
40
b) Sample all plots for several (e.g., three) consecutive years to determine baseline conditions
and decrease sampling frequency thereafter (Table 2-5B). The intent in using this scheme is to
develop a baseline of information that identifies the response of vegetation and soils to changes in
weather, training intensity, and other natural or anthropogenic disturbances. Once the variability of
the system is known, the sampling frequency may be extended. Data patterns during subsequent
collections are explained in part by the more continuous, earlier data set. A possible sampling
sequence would be years 1, 2, 3, 4, 5, 10, 15, 20, etc. Plant communities that recover quickly from
military and/or non-military impacts would need fewer years to establish a baseline of information
and a larger spacing between subsequent surveys (Years 1, 2, 3, 6, 12, 18, 24, etc.). In contrast, an
area that recovers slowly or is heavily impacted may require a longer baseline and a shorter time
between later surveys (e.g., Years 1, 2, 3, 4, 5, 8, 11, 14, 17, 20, etc.).
This approach is less expensive than continuous monitoring and statistical analysis options are still
numerous. Breaks between surveys may cause funding problems in that requirements vary greatly
between some years. During interim years other sampling needs can be addressed such as special
projects, wildlife surveys, sensitive sites, etc.
Because plots are not monitored during interim years, the condition of the resources of interest is not
quantitatively known. Given similar levels of use and an understanding of the effects of weather on
the vegetation and the soils, land condition can be estimated if mathematical or other relationships
have been established. In the example presented (Table 2-5B), the causes of the decline of bare
ground in 1992 and the subsequent increase through 1994 would not be known. However, these
changes may have been reliably predicted if between the 1991 and 1992 field seasons training
intensity or livestock use was known to decrease, precipitation to increase, or precipitation occurred
at advantageous times. Any of these or other factors could cause an increase in vegetation or plant
litter and the decline of bare ground. The next installation-wide survey would mark current
conditions. While data may not be collected yearly, training intensity and timing, weather, and other
factors affecting vegetation should be documented.
c) Sample some plots all years and all plots during key years (Table 2-5C,D,E). Some monitoring
occurs all years; therefore a quantitative value exists. Because sample sizes are smaller, the amount of
variation is greater than when all plots are monitored. Fluctuations in funding requirements are not as
great as when there are breaks in monitoring. The cost is slightly higher than Option 2.
In this scheme all plots are surveyed periodically (e.g., every third year), and either a single subset
(Table 2-5C) or rotational subsets (Table 2-5D) are sampled during the interim years. Monitoring a
subset of plots every year (typically sampling without replacement) is sometimes referred to as a
rotating panel design. Plots surveyed during the interim years are either chosen randomly or by set
criteria (e.g., watershed A, Training Area C, Bivouacs x, y, and z, etc.). Plots monitored during
interim years provide a continuous source of information on vegetation and soil responses to weather,
training, and other land uses.
The variation within any subset is dependent on the plots in the group. When a single subset is
monitored during interim years, the variation within the group may be secondary to the information
that group provides, such as continuous monitoring of highly utilized training areas. While variation
may be higher than on the rest of the installation, site-specific information will be relevant to training
needs.
41
A rotational monitoring system strives for a random collection of installation information. In this
case, some groupings may be more variable than others may and some may be more responsive to
change (Table 2-5E). Groups 2 and 3 have more variation than Group 1 in 1989; the opposite is noted
in 1992, and a similar relationship to 1989 is illustrated in 1995. Statistically, there was no difference
among the groups (P>0.05).
Inconsistent or inadequate funding can make it difficult to meet monitoring requirements. To
minimize the effects of funding oscillations, programs should be flexible to avoid severe impairment
or being rendered useless by temporary reduction of funds. For instance, a 25 percent budget
reduction may be remedied by using smaller crew sizes, identifying vegetation to the life form versus
species level (assuming this still satisfies monitoring objectives), focusing on areas that are more
dynamic or disturbed, acquiring support from installation sources (e.g., project vehicles), or merging
other projects into the responsibilities of the field crew. Nonetheless, severe budget cuts may require
significant alterations to the monitoring plan and protocols. In some cases, quantitative monitoring
may not be practical due to the relationship between sample sizes and monitoring precision or change
detection capability. Once data are collected and analyzed, an appreciation for the minimum amount
of data necessary for a valid survey will be more apparent.
Another consideration in planning monitoring efforts is major changes in land-use patterns,
frequencies, and intensities. An example might be the mobilization of troops for active duty, or a
large scale exercise (brigade, division) that takes place for the first time or infrequently on an
installation or in a particular geographic or training area. These types of events can affect funding and
field accessibility. As training intensity can increase exponentially, so can vegetation degradation,
habitat loss, erosion rates, sedimentation, and the need for data. Qualitative data, such as qualitative
surveys and photographs, may be the only feasible methods under these circumstances.
Ultimately, training activity and intensity, vegetation and physical characteristics, and the land
management goals of an installation determine the most appropriate monitoring scheme. Ideally, an
installation will be able to survey all plots for a minimum of three consecutive years. Then, following
a review of the data, the best long-term monitoring program can be designed. No two installations
have the same training uses, vegetation, soils, or management goals. By reviewing the data and
testing the various scenarios, an installation can identify the best, least expensive alternative.
2.8
Written Protocols and Program Documentation
A monitoring protocol is an essential tool for program managers to organize, design, and evaluate
data collection and analysis efforts. Protocols can provide varying amounts of detail depending upon
their intended use. Some provide general guidelines while others provide very detailed information.
Protocols are sometimes referred to as inventory and monitoring guidelines, monitoring plans or
handbooks, or sampling protocols, but are typically more detailed than general implementation plans.
Protocols provide site-specific and detailed information about resources and management concerns;
management objectives and their corresponding monitoring objectives; monitoring methodologies
and data collection procedures; data management, quality control, and storage procedures; and data
analysis and interpretation. It may also contain information about sampling locations and plots, data
sheets, photographs, and other important documentation. The protocol can focus on the entire
installation (as a complex of communities and land uses), or specific entities such as plant
communities/habitats, land-use types, or particular species or populations of concern. Some protocols
are comprehensive, including all required components, while others may include a single component
of overall procedures (i.e., field methodology, data management, reporting, etc.). Examples of
42
protocols include documents written for the National Park Service System (USDI National Park
Service 1992; Halvorson et al. 1988), longleaf pine communities (Rudd and Sutter 1996), a Biosphere
Reserve (Shopland 1998), Army resource monitoring (Tazik et al. 1992), and camping or recreational
impact assessment (Cole 1982, 1984, and 1989; Kitchell and Connor 1984).
A monitoring protocol helps to ensure that monitoring goals are well defined and prioritized, cost
efficiency is maximized, and scientific standards and statistical rigors are appropriate to the resource
or management area (i.e., confidence levels, statistical power, and minimum detectable changes,
where quantitative methods are applied). Protocols help to provide continuity for monitoring
programs, especially where turnover rates are high and program resources are variable from year to
year. By organizing existing information and setting specific goals, protocols help program managers
to justify ongoing monitoring programs. The development of a protocol is appropriate for both
fledgling and well-established RTLA programs.
Different levels of monitoring (i.e., qualitative, semi-quantitative, quantitative) are specified that
reflect monitoring objectives and available resources. Program evaluation based on data analysis and
professional judgment, continuity with historic data, the tailoring of protocols to individual
installations, and flexibility over time are all important attributes of a protocol.
A number of agencies and organizations have developed broad-level monitoring protocols, including
the Environmental Monitoring and Assessment Program (EMAP – Environmental Protection
Agency), Forest Inventory and Analysis (FIA - Forest Service), and Natural Resources Inventory
(NRI – Natural Resources Conservation Service) http://www.nrcs.usda.gov/technical/NRI/ and
Nusser et al. 1997). These programs have been developed largely for regional and national
assessments. Additional information regarding national-level standardized approaches are discussed
in Section 4.5 Integrative Approaches and Section 4.6 Monitoring Ecological Integrity on Public
Lands.
The following sections highlight broad initiatives and site-specific protocols developed for a variety
of ecosystems. Despite the fact that some approaches are designed for state, regional, or agency-wide
summary purposes, methodologies, indicators and other program aspects are highly relevant to the
development of site-specific efforts. In some cases, modifications to the sampling design to smaller
scales may be required to implement these approaches effectively. This list is not comprehensive, and
represents information that is readily available from the Internet or other sources. Many unpublished
monitoring protocols have undoubtedly been developed by local, county, state, and private entities,
and individuals are encouraged to seek out these resources in their appropriate locales.
National Park Service 6
Excellent guidance for all aspects of developing monitoring programs and protocols is available from
the NPS Inventory and Monitoring Program web site
(http://science.nature.nps.gov/im/monitor/index.cfm). A database of protocols developed by Parks
across the U.S is available at http://science.nature.nps.gov/im/monitor/protocoldb.cfm. Additional
resources and protocols are being developed by the NPS Vital Signs Monitoring Program
(http://www.nature.nps.gov/protectingrestoring/IM/vitalsignsnetworks.htm), which is supported by
the Inventory and Monitoring Program (Figure 2-7). For each vital sign (i.e., ecological indicator),
networks develop a Protocol Development Summary document that gives a brief overview of
sampling protocols they plan to develop over the next 3-5 years to address the vital signs. The
6
Description adapted from http://science.nature.nps.gov/im/monitor/vs_framework.htm
43
Protocol Development Summaries contain a justification statement, monitoring questions and specific
objectives, and the basic approach to be followed in developing a long-term monitoring protocol for
the vital sign. By sharing their plans for vital signs and protocol development, networks can identify
additional opportunities for collaboration and development of common approaches. As an additional
tool to facilitate collaboration and coordination of monitoring planning and design, the monitoring
program is developing a database of vital signs identified by various networks, including justification
statements, monitoring questions, and specific monitoring objectives. Regional and Park-level NPS
projects are excellent sources of information for installations, as military lands share many
commonalities with Park Service Lands (e.g., highly variable in size and resources, provide refugia
for threatened and endangered species, have high-use areas surrounded by vast wildland areas,
extensive areas of rugged terrain with limited access). For example, information regarding inventory
and monitoring efforts and planning for the Mojave Network is found at http://hrcweb.lvhrc.nevada.edu/mojn/index.htm. As of 2004, twenty-two of thirty-two park networks had received
funding. Among the major Federal landholders, NPS protocol development is the closest
approximation to the bottom-up approaches espoused by the Army’s RTLA Program.
Figure 2-7. National Park Service Inventory and Monitoring Program, Vital Signs Monitoring
Networks (Source: http://science.nature.nps.gov/im/monitor/networks2.htm Jan 12 2006).
U.S. Forest Service
The Forest Service has developed and implemented a number of standardized long-term monitoring
protocols for various purposes. Forest Health Monitoring is described in detail in Section 4.5
Integrative Approaches. Apart from the national monitoring programs numerous monitoring plans
44
have been implemented on National Forests to address specific management concerns and resources.
Readers are encouraged to search for relevant monitoring information from their respective locales
and regions.
USFS Inventory and Monitoring Institute (IMI). IMI provides technical leadership and
service for agency-wide collection, management, and analysis of scientifically reliable
social and ecological information used in ecosystem management. Site includes
information regarding Local Unit Criteria & Indicators Development Project (LUCID)
(http://www.fs.fed.us/institute/imi.html)
U.S. Environmental Protection Agency
The EPA’s Environmental Monitoring and Assessment Program (EMAP) is a research program to
develop the tools necessary to monitor and assess the status and trends of national ecological
resources. EMAP's goal is to develop the scientific understanding for translating environmental
monitoring data from multiple spatial and temporal scales into assessments of current ecological
condition and forecasts of future risks to our natural resources. The program is divided among
numerous components that focus on ecological issues or geographic areas
(http://www.epa.gov/emap/).
Natural Resource Monitoring Partnership
The Natural Resource Monitoring Partnership (NRMP) is a collaborative effort by the natural
resource management community to improve monitoring efforts in order to support effective
evaluation and decision-making. Current participants include state, Canadian provincial, and federal
natural resource management agencies, nongovernmental organizations, and academic institutions
(http://biology.usgs.gov/status_trends/nrmp/MonitoringPartnership.htm).
The Partnership is led by several collaborative teams made up of volunteers from participating
agencies/organizations. However, the Partnership has no formal authority and does not exert any
control over the work of any individual or institution: it exists only to help improve the design and
implementation of monitoring programs by improving communication and coordination among
individuals and institutions engaged in natural resource monitoring.
The initial focus of the Partnership is on developing two collaborative, internet-based tools to foster
coordination and collaboration of monitoring efforts. The first is a monitoring protocol library - an
internet accessible, searchable database that provides information on monitoring protocols and
resource assessment methodologies organized to facilitate reference and use. This is a database of
protocols that will be a library or catalog developed and maintained by users. The second tool is a
monitoring "locator". The locator is an internet-based, GIS application that allows users to identify
resource monitoring projects within a particular area (e.g., State, county, Canadian Province, or other
selected geographic area). Search tools will help users find information about ongoing and historic
monitoring according to different scales, targets, and objectives.
Others
The Nature Conservancy (TNC). Many TNC documents can be viewed or downloaded from the
Conserve Online Library (http://conserveonline.org/). Several TNC products are listed below:
45
Provencher, L., A.R. Litt, and D.R. Gordon. 2000. Compilation of Methods Used by the
Longleaf Pine Restoration Project from 1994–1999 at Eglin Air Force Base, Florida.
Product to Natural Resources Division, Eglin Air Force Base, Niceville, Florida. Science
Division, The Nature Conservancy, Gainesville, Florida.
Rudd, N. and R.D. Sutter. 1998. Monitoring Protocols for the Barrens Restoration
Demonstration Project at Arnold Air Force Base, TN. The Nature Conservancy: Chapel
Hill, NC.
Rudd, N. and R.D. Sutter. 1996. Monitoring Protocols for Rare and Endangered Species
and Longleaf Pine Community Types at Fort Bragg and Camp Mackall Military
Reservations, NC. The Nature Conservancy: Chapel Hill, NC.
The Nature Conservancy of Hawaii. Measuring Conservation Actions in East Moloka‘i,
Hawai‘i Final Report. April 2004
The Nature Conservancy Landscape Conservation Networks. North American Fire
Learning Network –
The Center for Environmental Management of Military Lands at Colorado State University
(CEMML)
Several monitoring protocols developed for military lands may be available upon request.
http://www.cemml.colostate.edu/. It is anticipated that a RTLA protocol database will be
established to document the status of installation RTLA protocols and describe projects,
methods, objectives, etc.
Environment Canada
Environment Canada, Ecological Monitoring and Assessment Network (EMAN). Lists
protocols for monitoring freshwater, marine, and terrestrial ecosystems
http://www.eman-rese.ca/eman/ecotools/protocols/
There are advantages and disadvantages to using established, standardized protocols versus sitespecific, customized protocols. In most cases, the optimum approach may involve integrating
elements of standardized approaches with site specific methods to create a diverse and robust
program.
Advantages
o Standardized protocols have often been developed using a peer review process
and extensive field testing.
o Data collected using standardized protocols is compatible with other data
collected suing the same procedures on adjacent and other lands.
o Implementation can begin rapidly, as protocol development and field testing may
not be necessary.
o Support and partnership with other agencies can facilitate and accelerate program
implementation.
46
Disadvantages
o Site-specific needs and objectives may be more detailed than those in
standardized protocols.
o Sampling designs and methods may need to be modified to provide statistically
acceptable results where quantitative methods are used.
o Tailoring of standardized methods, indicators, and attributes may need
adjustment, as some data may be extraneous or cumbersome to collect.
o Intensity of data collection methods may exceed site-specific needs and available
resources.
2.8.1
Elements of a Monitoring Protocol and Plan
A comprehensive monitoring protocol or site-specific handbook should contain all relevant
information for the monitoring project. The following outline is recommended as a starting point.
These components are directly related to steps involved in the overall monitoring process (Section
2.1.3).
The detailed protocol should supplement an implementation or monitoring plan. Implementation
plans can be either short-term (one to several years) or long-term (several to many years), and should
include information about the frequency and timing of data collection, budgetary information for data
collection, data management and quality control, and data analysis, an implementation timetable
listing tasks and milestones, and other crucial program planning and support information.
47
EXECUTIVE SUMMARY
1.
INTRODUCTION
2.
LANDSCAPE AND COMMUNITY INFORMATION
2.1
SITE DESCRIPTION
2.1.1
Climate
2.1.2
Geology, Topography and Soils
2.1.3
Hydrography and Water Resources
2.2
STATUS OF RESOURCE SURVEYS, INVENTORIES, AND OTHER MONITORING EFFORTS
2.2.1
Installation survey and monitoring programs and data
2.2.2
Resource monitoring programs and approaches employed by local, regional, or national
land owners and agencies (e.g., BLM, USFS, NPS, TNC) –
2.3
DESCRIPTION OF MAJOR PLANT COMMUNITIES
3.
DESCRIPTION OF TRAINING
3.1
3.2
3.3
4.
INSTALLATION MISSION AND PRIMARY UNITS
TRAINING FACILITIES (RANGES AND TRAINING AREAS) AND LAND USE PATTERNS
TRAINING IMPACTS TO TRAINING/RANGE AREA NATURAL RESOURCES
SYSTEM DYNAMICS AND MANAGEMENT CONCERNS
4.1
4.2
MANAGEMENT-ORIENTED CONCEPTUAL ECOLOGICAL MODEL(S)
SUMMARY OF MANAGEMENT CONCERNS
5.
RTLA PROJECT DESCRIPTIONS
6.
DETAILED APPROACHES BY PROJECT
6.1
PROJECT NAME (1)
6.1.1
Status and Background
6.1.2
Attributes and Indicators
6.1.3
Evaluation of Historic Data and Efforts
6.1.4
Management and Monitoring Objectives
6.1.5
Sampling Design
6.1.6
Data Collection Protocols
6.1.7
Data Management, Analysis and Reporting
6.1.8
Monitoring Schedule and Funding Requirements
6.2
PROJECT NAME (2) – FOLLOW SAME OUTLINE FOR EACH PROJECT
6.2.1
Status and Background
6.2.2
Attributes and Indicators
6.2.3
Evaluation of Historic Data and Efforts
6.2.4
Management and Monitoring Objectives
6.2.5
Sampling Design
6.2.6
Data Collection Protocols
6.2.7
Data Management, Analysis and Reporting
6.2.8
Monitoring Schedule and Funding Requirements
7.
MONITORING SCHEDULE SUMMARY BY PROJECT
8.
RTLA DATA ANALYSIS AND PROGRAM REPORTING GUIDELINES
9.
COORDINATION AND MANAGEMENT IMPLICATIONS.
10.
PROGRAM BUDGET
11.
REFERENCES
12.
APPENDICES (AS APPROPRIATE)
48
A monitoring program should have flexible components that can be modified as more is learned about
the installation and land usage. Development of both a detailed monitoring protocol and monitoring
plan are essential to program success. Following each field season, these documents should be
reviewed and updated to reflect new information and changing program needs.
2.9
Summary: Guidelines for Developing a Successful Monitoring
Program
There are three principal difficulties that must be overcome if an ecological monitoring design is to
succeed: (1) one of the main ecological difficulties is selecting and quantifying specific biotic
conditions within the existing and continuous spatial and temporal variability (e.g., appropriate
indicators); (2) the major statistical difficulty is having enough replication in all of the different
places and types we want to examine; and (3) the cost of monitoring (Hinds 1984).
Because of these specific reasons and additional complexities associated with monitoring natural
resources, their natural variability, unexpected events, and a wide variety of possible sampling
designs, program managers are obliged to adapt as conditions change. There is, therefore, is no
universally applicable set of guidelines to ensure successful long-term monitoring of large land
parcels or landscapes. Some excellent guidance is provided by the Park Service Inventory and
Monitoring Program for developing integrated monitoring programs, including links to documents
that are relevant to protocol designs (http://science.nature.nps.gov/im/monitor/index.cfm).
The following list of attributes was adapted from Stohlgren et al. (1995) and CEMML (2006).
Attention to these issues will enhance the effectiveness of monitoring program staff, maximize the
value of long-term monitoring projects and data, and optimize the value of monitoring to adaptive
management.
(1) Develop professional knowledge. A wide range of information and knowledge is required to
successfully design and implement monitoring projects. In addition to academic knowledge, site
specific knowledge and familiarity are essential to success, including everything from disturbance
ecology to knowledge of the road and trail network. Other important aspects of professional
knowledge include familiarity with other land management and research agencies/entities, knowledge
of Department of Defense and Army natural resources policies and regulations, and knowledge
regarding military missions and training requirements.
(2) Secure long-term funding and project commitment. This responsibility lies with a number of
individuals involved with the program, including the RTLA program manager, ITAM coordinator,
and higher-level installation, MACOM, and Headquarters staff. Understand the budgeting and work
plan process, and get to know key personnel at all levels.
(3) Solicit user’s needs early in the process. Because resource monitoring on military installations is
not intended as an academic exercise, practical applications and management needs must drive both
the formulation of project goals and the visualization of products and problem-solving scenarios.
Make the program relevant by understanding related training and management issues. In the longterm, this may be the ingredient that determines the success of the project.
49
(4) Develop flexible goals. Goals must be flexible and articulated clearly, reflecting current issues and
problems yet providing a basic level of continuity. The goals should also maintain some ability to
address unanticipated future issues and concerns. By selecting sets of “core” parameters that address
primary concerns and objectives, iterative changes to program goals are not likely to affect the longterm integrity of important parameters.
(5) Refine objectives. This process involves the process of reducing general problems to specific ones,
identifying specific objectives, and setting priorities for specific inventory and monitoring data needs.
For example, priorities for data collection must be weighed against the practical constraints of
collecting limited data at many sites or more data at fewer sites.
(6) Pay attention to data management. This suite of tasks can include quality control and quality
assurance for field data, data acquisition and archive, metadata development, and statistical analyses.
(7) Be creative and experiment in the sampling design phase. There is little consensus on sampling
designs and methodologies for landscape-level studies, and too often projects have become “locked
into” designs too early in the course of the program, precluding the adoption of helpful changes.
Considerations such as plot size and shape, the parameters selected for measurement, and the
frequency, precision, and accuracy of measurements can be adapted successfully to unique settings.
However, radically changing a protocol in long-term studies may require calibrating the new methods
to the old ones.
(8) Obtain peer review of monitoring proposals and reports. Peer review helps to alleviate problems
early in the implementation of a project or even before it begins. Data analysis plans should also be
developed prior to data collection, understanding that some features of the design and methodology
may change in the initial phase of the program. Moreover, determining adequate sample sizes is an
iterative process, which requires recalculating variance and spatial replication needs as data is
collected over time.
(9) Avoid bias in selecting plot locations. An important feature of landscape-level monitoring is the
specific intent to extrapolate information from plots to landscapes (and perhaps from landscapes to
regions). Bias in site selection can cause the appearance of trend when in fact none exists (Palmer
1993). When sampling is restricted or biased, such as when sampling locations are intentionally
placed close to roads (accessibility sampling), the representativeness of the sample is uncertain (Krebs
1989). The selection of “typical” or “reference” stands is likewise biased and violates the precepts of
probability sampling that allow results to be extrapolated across the population. Program managers
can minimize bias by first defining the population of interest, and subsequently selecting random sites
from that population.
(10)
Ensure adequate spatial replication. This is one of the main obstacles to statistical
extrapolation or inference. Because of the preponderance of unique land uses and types and the
variability present within each, sample size adequacy must be addressed through creative
restructuring of the landscape. Temporal variability increases sample size requirements because it
adds a source of variability beyond sampling errors, non-sampling errors, and inherent spatial
variability. Pilot data is typically required to determine adequate sample size.
(11)
Ensure adequate temporal replication. Temporal (i.e. seasonal, yearly, or cyclic) variability
can only be addressed by evaluating several years of data. Because of resource limitations, there is
50
often a tradeoff between spatial and temporal replications. Like the determination of sample size, the
determining the frequency of sampling can be an iterative process.
(12)
Synthesize information with other studies and sources of information. Synthesis involves
examining information from different sources in order to answer questions or understand
relationships that were not evident from the individual projects. Examples of this include examining
data collected at different scales, using quantitative field data to ground-truth remotely sensed
information, and gathering different types of information that may complement or be related to the
primary data being collected. For example, climatic information is readily available for most
locations, and can help explain variability and trends in vegetation data. Training load information
organized from a variety of sources can be an important corollary to resource condition data. Water
quality and flow information may be collected by another entity within the installation, can be an
important indicator of watershed stability and vegetation condition. Experimental and retrospective
studies can provide important information regarding effects of specific and integrated stressors,
respectively, over time.
(13)
Conduct periodic program review. It is important to maintain an adaptive approach. Periodic
evaluation can help measure the success of the monitoring program (can it effectively accomplish
what was intended?). When established as a structured process, evaluation can also involve other
program players such as land managers, trainers, and independent scientists, thus promoting positive
relationships and interaction. Most importantly, the applicability of the program to training land
management and other land management concerns will determine the success of the program.
(14)
Interact and communicate effectively. Regular interaction and communication with relevant
staffs and agencies will help promote communication and help your projects stay focused on highpriority objectives. Interaction also fosters trust and respect among various parties.
(15)
Transfer information at different scales. Information transfer should be addressed at a number
of levels of understanding, including installation organizations (land management, range operations,
training, etc.), higher headquarters, professional meetings and symposia, and public awareness.
Products, reports, and presentations are all essential tools in educating and informing audiences about
results and successes. Increasingly, spatial information is communicated through the use of maps and
visual tools, which should be used extensively for all audiences. Mechanisms for writing technical
reports which address specific objectives and/or hypotheses should be established to promote timely
examination of data and information transfer.
2.10
References
Archer, S. 1989. Have southern Texas savannas been converted to woodlands in recent history?
American Midland Naturalist 134: 545-561.
Belnap, J. 1998. Choosing indicators of natural resource condition: A case study in Arches National
Park, Utah, USA. Environmental Management 22(4): 635-642.
Bestelmeyer, B.T., J.R. Brown, K.M. Havstad, R. Alexander, G Chavez, and J.E. Herrick. 2003.
Development and use of state-and-transition models for rangelands. Journal of Range Management
56:114-126.
51
Cairns, J. Jr., P.V. McCormick, and B.R. Niederlehner. 1993. A proposed framework for developing
indicators of ecosystem health. Hydrobiologia 263: 1-44.
Center for Environmental Management of Military Lands (CEMML). 2006. Handbook of Effective
Practices for Range and Training Land Assessment (RTLA) Coordinators. Prepared for the Army
Environmental Center, Aberdeen MD. Colorado State University, Fort Collins, CO.
Cole, D.N. 1982. Wilderness Campsite Impacts: Effect of Amount of Use. Research Paper INT-284.
USDA Forest Service, Intermountain Forest and Range Experiment Station, Ogden UT. 34 pp.
Cole, D.N. 1984. An Inventory of Campsites in the Flathead National Forest Portion of the Bob
Marshall Wilderness, Montana. Unpublished report on file at: USDA Forest Service, Intermountain
Research Station, Forestry Sciences Laboratory, Missoula, MT. 19 pp.
Cole, D.N. 1989. Wilderness Campsite Monitoring Methods: A Sourcebook. General Technical
Report INT-259. USDA Forest Service, Intermountain Research Station, Ogden, UT. 57 pp.
Committee on Rangeland Classification. 1994. Rangeland Health: New Methods for Classifying,
Inventorying, and Monitoring Rangelands. Board on Agriculture, National Research Council,
National Academy Press, Washington, D.C.
CTSC (Combat Training Support Center). 1998. U.S. Army Training and Testing Area Carrying
Capacity (ATTACC) Program Handbook, Version 1.1. July 1998.
Elzinga, C.L., D.W. Salzer, and J.W. Willoughby. 1998. Measuring and Monitoring Plant
Populations. BLM Technical Reference 1730-1. USDI Bureau of Land Management, National
Applied Resource Sciences Center, Denver, CO.
Gross, J. E. 2003. Developing Conceptual Models for Monitoring Programs. National Park Service
Inventory and Monitoring Program web site.
http://science.nature.nps.gov/im/monitor/docs/Conceptual_modelling.pdf
Gross, J. E. 2004. Conceptual Models for National Park Service Inventory and Monitoring Networks.
Presentation to the Southeast Region Inventory and Monitoring Meeting, January 2004.
Halvorson, W.L., S.D. Veirs, R.A. Clark, and D.B. Borgais. 1988. Terrestrial Vegetation Monitoring
Handbook. National Park Service, Channel Islands National Park, Ventura, CA.
Harper, M.G., R.A. Fisher, A.M. Trame, and C.O. Martin. 1996 – draft. Plant Community
Management Abstract: Southern Pine Woodlands. Technical Report SERDP—U.S. Army
Construction Engineering Research Laboratory, Champaign, IL.
Hinds, W.T. 1984. Towards monitoring of long-term trends in terrestrial ecosystems. Environmental
Conservation 11(1): 11-18.
Jones, D.S. 2000. Vegetation Monitoring Protocols for the Central Oregon Training Site (COTS),
Oregon Army National Guard. Center for Environmental Management of Military Lands, Colorado
State University, Fort Collins, CO. CEMML TPS 00-9.
52
Kitchell, K.P. and J. Connor. 1984. Canyonlands and Arches National Parks and Natural Bridges
National Monument Draft Recreational Impact Assessment and Monitoring Program. Unpublished
paper on file at: U.S. Department of the Interior, National Park Service, Moab, UT. 80 pp.
Krebs, C.J. 1989. Ecological Methodology. Harper and Row, New York.
Laycock, W.A. 1991. Stable states and thresholds of range condition on North American rangelands:
a viewpoint. Journal of Range Management 44(5):427-433.
Maddox, D., K. Poiani, and R. Unnasch. 1999. Evaluating Management Success: Using Ecological
Models to Ask the Right Monitoring Questions. In Sexton, W.T., A.J. Malk, R.C. Szaro, and N.C.
Johnson, Eds. 2001. Ecological Stewardship, A Common Reference for Ecosystem Management
(Volume III). Elsevier Science.
Menges, E.S. and D.R. Gordon. 1996. Three levels of monitoring intensity for rare plant species.
Natural Areas Journal 16(3): 227-237.
Miller, R.F., T. Svejcar, and J. Rose. 1999. Conversion of shrub steppe to juniper woodland. In
Monsen, S.B., R. Stevens, R.J. Tausch, R.F. Miller, and S. Goodrich, compilers. 1999. Proceedings:
Ecology and Management of Pinyon-Juniper Communities Within the Interior West; September 1518 1997, Provo, UT. USDA Forest Service Rocky Mountain Research Station, RMRS-P-000, Ogden,
UT.
National Research Council. 2000. Ecological Indicators for the Nation. National Academy Press,
Washington, D.C.
Nusser, S.M. and J.J. Goebel. 1997. The national resources inventory: A long-term multi-resource
monitoring programme. Environmental and Ecological Statistics 4(3): 181- 204
Palmer, M.W. 1993. Potential biases in site and species selection for ecological monitoring.
Environmental Monitoring and Assessment 26: 277-282
Peacock, G. 2000. Utilizing State and Transition Models to Assist in the Decision Making Process.
Paper presented at the State and Transition Modeling Workshop, Logan, Utah. Sponsored by the
USDA Natural Resources Conservation Service Grazing Lands Technology Institute.
Pellant, M., D. Shaver, D.A. Pyke, and J.E. Herrick. 2005. Interpreting Indicators of Rangeland
Health, Version 4. TR-1734-6. USDI Bureau of Land Management, Denver, CO.
Poiani, K. 1999. Conceptual Ecological Models and Site Conservation Planning: A Brief Overview.
The Nature Conservancy.
Rudd N. and R. Sutter. 1996. Monitoring Protocols for Rare and Endangered Species and Longleaf
Pine Community Types at Fort Bragg and Camp MacKall Military Reservations, North Carolina.
Report for DOD. The Nature Conservancy, Southeast Regional Office, Chapel Hill, NC.
Shopland, J. 1996. Designing a Comprehensive Monitoring Program at El Triunfo Biosphere
Reserve. The Nature Conservancy, Conservation Science Web Page.
53
Spellerberg, I.F. 1991. Monitoring Ecological Change. Cambridge University Press, Cambridge. 334
pp.
Stohlgren, T.J., D. Binkey, T.T. Veblen, and W.L. Baker. 1995. Attributes of reliable long-term
landscape-scale studies: Malpractice insurance for landscape ecologists. Environmental Monitoring
and Assessment 36: 1-25.
Stringham, T.K., W.C. Kreuger, and P.L. Shaver. 2001. States, transitions, and thresholds: Further
refinement for rangeland applications. Oregon State University Agricultural Experiment Station,
Special Report 1024, Corvallis, Oregon.
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (LCTA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
The Heinz Center. 2003. The State of the Nations Ecosystems. Available from Cambridge University
Press or electronically from http://www.heinzctr.org/ecosystems/report.html
The Nature Conservancy (TNC). 1994. Ecosystem Models: A Template for Conservation Action. The
Nature Conservancy, Arlington, VA.
The Nature Conservancy (TNC). 1997. Vegetation Monitoring in a Management Context Workbook. Workshop coordinated by The Nature Conservancy and co-sponsored by the U.S. Forest
Service, held in Polson, MT, September 1997.
USDI National Park Service. 1992. Natural Resources Inventory and Monitoring Guideline: NPS-75.
National Park Service, Washington D.C. 37 pp.
Wagner, R.E. 1989. History and Development of Site and Condition Criteria in the Bureau of Land
Management. Pages 35-48 in W.K. Lauenroth and W.A. Laycock (eds.), Secondary Succession and
the Evaluation of Rangeland Condition. Westview Press, Boulder, CO.
Westoby, M., B. Walker, and I. Noy-Meir. 1989. Opportunistic management for rangelands not at
equilibrium. Journal of Range Management 42(4):266-274.
Whitford, W.G., A.G. De Soyza, J.W. Van Zee, J.E. Herrick, and K.M. Havstad. 1998. Vegetation,
soil, and animal indicators of rangeland health. Environmental Monitoring and Assessment 51(1/2):
179-200.
54
3 Introduction to Sampling
3.1
3.1.1
Principles of Sampling
Why Sample?
Sampling involves taking measurements on a subset of a population to make inferences about that
population. Efficient sampling can provide precise population estimates at a reasonable cost.
However, sampling is not always a necessary or appropriate component of a monitoring program.
Alternatives to sampling include performing a census of all individuals within a population or using
qualitative techniques that are not intended to represent the entire population of interest.
Sampling is necessary when information about the entire population (or community, or ecosystem) is
desired, but where census or qualitative techniques are impractical or do not meet objectives.
Sampling provides not only an estimate of the population or attribute of interest, but a measure of the
variability of the estimate, which can be interpreted as the quality of the estimate.
3.1.2
Populations and Samples
Defining the statistical population being studied is an important elementary step in any sampling
design. The statistical population is also referred to as the “sampling universe”. The statistical
population may or may not be a biological population, and it may in fact be subdivided by artificial or
administrative boundaries, or may be a defined subset of a larger population. Statistical populations
can be defined both narrowly and broadly. Examples of clearly specified populations include: the
white-tailed deer population of Fort Leonard Wood, Missouri; the population of blue oak (Quercus
douglasii) seedlings on Fort Hunter Liggett, California; and mature longleaf pine stands on Fort
Benning, Georgia. Difficult to define populations are often related to issues of spatial scale and the
fact that biological populations change over time (Krebs 1989). Examples of poorly defined
populations include: populations of long-lived biennial or perennial plants that have large
underground storage organs and local populations appear to shrink or grow from year to year; the
number of seeds in a seed bank, which is highly dependent on inputs, storage, and mortality; or the
population of fish species in a stream reach.
Sampling units are the individuals or objects that constitute a population. Commonly used sampling
units are transects (lines), quadrats (plots), points, or individual plants. A sampling unit is a distinct,
discrete member of a population that can be analyzed by grouping or as a whole. It is any quantity
(size, percentage, height, etc.) used as a measurement and is representative of the population (Bonham
1989). The definition of sampling unit is related to the measurement that is chosen. For example, if
we are interested in shrub height, then individual shrubs are the sampling units. A specified number
of individuals might be selected randomly, constituting a sample of X number of plants. If mean
density of plants or canopy cover is the attribute of interest, then the sampling unit would be the plot,
frame, line, or other space where the counting or measurement takes place. In short, the sampling unit
is the object that is measured. Examples include: trees with DBH> 10 cm, canopy cover of a species
within a frame, or the interception of a leaf at a point.
55
A sample is made up of a number of observations (n) made for the measurement of interest using the
sampling units defined. Therefore, in a statistical sense, the word “sample” refers to a set of
observations or measurements.
If it is desired to make statistical inferences about a widespread biological population (e.g., old field
communities at Camp Ripley, Minnesota), then that widespread population should be sampled. The
sample should aim to be as representative as possible of the type as a whole. Sometimes the entire
population of interest is inaccessible (e.g., impact area, dud area, outside installation boundary, etc.)
so that only a portion of the true area of interest can be sampled. The challenge of defining the
statistical population and subsequently relating it to the biological population of interest is pervasive
in quantitative sampling, and less so with qualitative approaches.
3.1.3
Parameters and Sample Statistics
Several measures help to describe or characterize a population. Population parameters include
measures of central tendency (e.g., mean, median, mode) and measures of dispersion (e.g., standard
deviation, range). Measures that describe or characterize a population are called parameters. Because
it is often impossible to calculate parameters, they are estimated by using random sampling. An
estimate of a population parameter is called a statistic. The true population parameters are generally
unknown (accuracy of estimates are therefore also unknown) so sample statistics are used to estimate
them. This process is referred to as extrapolation or inference from the sample data. In statistical
equations, population parameters are represented by Greek letters and sample statistics by Latin
letters.
Sample statistics will vary from sample to sample for samples taken from the same population.
Statistics also change as the population being sampled changes. The best statistics are unbiased, have
a high level of precision, and are consistent (i.e., they become better estimators of the parameters as
sample size increases) (Zar 1996).
3.1.4
Accuracy and Precision
Accuracy is the closeness of a measurement to the actual or true value of the variable measured.
Precision is the closeness to each other of repeated measurements of the same quantity. The best
measurements and designs produce results that are both accurate and precise. Because the accuracy of
estimates is often unknown, precision is often used as a surrogate to indicate the relative benefits of
different methods or approaches. However, high precision or repeatability does not explicitly infer
high accuracy. When samples are collected, measures of variability are calculated to provide a
measure of the precision of the sample. Biased sampling procedures may produce inaccurate
estimates.
3.1.4.1
Measures of Precision
Commonly used measures of precision include the standard deviation, standard error, confidence
intervals, and the coefficient of variation. The sample variance is the basis for all of these measures.
The standard deviation (SD) is the positive square root of the variance, and therefore has the same
units as the original measurements:
56
(∑ x )
−
2
s=
∑x
2
i
i
n −1
n
where:
x = the value of an individual sample
n = total number of samples (collections of observations)
The standard error of the mean (SEM or SE) is the standard deviation of a number of separate mean
values. The SE can be used to compare a number of samples drawn from the same population.
Optimal sampling design and data collection procedures have low standard errors.
SE =
s
n
where:
s = standard deviation
n = sample size
The standard error is minimized by reducing the standard deviation (s) or by increasing the sample
size (n). Increasing n often results in a lower standard deviation. Standard errors are used directly,
when reporting results in tabular or graphic form (e.g., mean ± 1 SE), and indirectly, in the
calculation of confidence intervals. Confidence intervals provide an estimate of precision about a
sample mean. More specifically, confidence intervals specify the likelihood (i.e. confidence) that the
interval contains the true value (see Section 7.3 Confidence Intervals for a more detailed discussion).
The coefficient of variation (COV) of a sample is a unitless, relative measure that provides an
estimate of the sample variability relative to the sample mean. It is referred to as a relative measure of
dispersion. This measurement may be calculated only for ratio-scale or proportional data. The COV is
defined as:
COV =
s
x
where:
s = standard deviation
x = sample mean
The COV is frequently multiplied by 100 to be expressed as a percentage. The COV is especially
useful when comparing the results of different survey design and data collection methods. The
approach, which produces the lowest COV, is desirable from a statistical standpoint. Table 3-1
illustrates the use of coefficient of variation to compare several field methods for estimating mean
tree diameter at breast height (DBH). From the standpoint of precision, the caliper method has the
lowest COV, followed by the tape, Biltmore stick, and ocular estimation. COV should not be the sole
basis for selecting a method; issues such as nonsampling errors (e.g., bias), and cost should also be
considered.
57
Table 3-1. Comparison of three methods using the coefficient of variation.
Method
Diameter tape
Calipers
Biltmore stick
Ocular – no equipment
3.1.5
DBH (inches)
12.9
12.6
13.3
13.7
Standard
Deviation
2.4
2.1
2.6
3.5
Coefficient of
Variation
.19
.17
.20
.26
Sampling and Nonsampling Errors
It is essential to minimize errors in a monitoring project. Errors associated with monitoring can be
separated into sampling errors and nonsampling errors. Sampling errors arise by chance, and may be
caused by applying an inappropriate method for a given situation. Some error, for example, may be
inherent in the measurement being applied. Sampling errors occur when the sample does not reflect
the true population. Sampling errors are often systematic and can be measured.
Nonsampling errors are sampling mistakes which cannot be measured. However, they can be
minimized. Examples of nonsampling errors include: (1) using biased selection rules, such as
selecting “representative samples” by subjectively locating sample units or substituting sampling
units that are easier to measure (e.g., proximity to roads, convenience sampling); (2) using sampling
units in which it is impossible to accurately count or estimate the attribute; (3) sloppy field work and
data management; (4) transcription and recording errors; and (5) incorrect or inconsistent species
identification (Bureau of Land Management 1996).
Nonsampling errors can be minimized by designing studies appropriately. For example, smaller
nonsampling errors (higher precision) are associated with point intercept and line interception
techniques compared to quadrat cover estimation (Floyd and Anderson 1987, Kennedy and Addison
1987). When measuring density, quadrat sizes should be sized so that observers are not counting an
excessively high number of plants. Training of field personnel is essential, especially when different
personnel will be collecting data over time. Training and periodic field “testing” help to maximize
consistency in measurement and estimation. Lastly, field forms should be designed to be
straightforward and economical, allowing for easy interpretation by transcribers. Data entered should
be checked against field forms for correctness.
3.1.6
Hypothesis Testing Errors and Power Analysis
A common goal in ecological monitoring is to determine whether the magnitude of a variable is
increasing or decreasing or whether or not a change in a resource has taken place. Power analysis is a
statistical technique for determining the ability of a monitoring program to detect change in a
resource. In other words, it is the probability that a change will be detected when a change has really
occurred. Portions of this section are excerpted with permission from Anderson et al. (1996).
Power analysis is a statistical technique useful in quantifying the ability of a monitoring program to
detect change in the monitored resource. A number of authors recommend an increased use of power
analysis techniques in the design and analysis stages of controlled studies and monitoring programs
(e.g., Toft and Shea 1983, Rotenberry and Wiens 1985, Peterman 1990a, Peterman 1990b). Few
applications of power analysis techniques are documented using vegetation and disturbance
monitoring data from DOD installations. However, investigators have successfully applied these
58
techniques to RTLA wildlife monitoring (Rice et al. 1995, Rice and Demarais 1995, Hayden and
Tazik 1993) and RTLA line transect data (Mitchell et al. 1994, Brady et al. 1995, Anderson et al.
1996). The use of power analysis techniques in these studies has proven useful in evaluating current
data collection methods and providing insight into the effects of modifications to those
methodologies.
Power analysis techniques are not commonly used with RTLA data, in part because installation
personnel are unaware of the consequences of Type-II errors, are unfamiliar with the procedures to
conduct power analysis, or are unaware that the results of power analysis can strengthen statistical
inferences made from monitoring data. In fact, power analysis techniques require only limited data
such as those currently available with most original and some modified RTLA data sets. Power
analysis calculations are relatively simple to carry out and are easily interpreted.
Poorly designed monitoring programs may not detect changes when in fact they have occurred.
Therefore, management decisions should take into consideration both the power of the analysis and
the detection of change/no change in the resource. Applications of power analysis to RTLA data are
discussed in detail by Anderson et al. (1996).
3.1.6.1
Power Analysis
A tool commonly used in the statistical analysis of data is the test of a hypothesis or test of
significance (Snedecor and Cochrane 1980). The hypothesis under test is usually referred to as the
null hypothesis (Ho) and is tested against the alternative hypothesis (Ha or H1). For each hypothesis,
the data are examined to see if the sample results support the hypothesis. The null hypothesis for
many monitoring programs is that no change has occurred in the monitored resource. The alternative
hypothesis is that a change has occurred.
Two types of errors are associated with any statistical test (Table 3-2). Type-I error (α) is the
probability of rejecting the null hypothesis when the null hypothesis is true. Type-II error (β) is the
probability of failing to reject the null hypothesis when the null hypothesis is false. If a resource
manager interprets the output of a monitoring system as indicating that a biologically important
change has occurred, the conclusion should prompt some management action or response. If a real
change has occurred, the correct decision was made. If no real change has occurred, the manager is
probably reacting to inherent variability in the process monitored; a false change (or Type-I) error
would have been made and the manager would have taken actions that were not required. If a
manager interprets the output of a monitoring system as indicating that no change has taken place, no
action will be indicated. If, in reality, no change has taken place, this action would be the correct
decision. However, if there really were a change that the monitoring system missed, a missed-change
(or Type-II) error would have been made. Missed-change errors mean that a change, usually
detrimental, was missed and that remedial actions will be delayed until a time when they may be more
expensive or less effective. A practical example illustrating the potential consequences of committing
false-change and missed-change errors is presented in Table 3-3.
59
Table 3-2: Statistical decision and error probability table.
“True” Condition
No change has taken
There has been a real
place
change
Statistical Decision
Monitoring detects a change
(reject Ho)
False-Change Error
(Type I) α
No Error
(Power) 1-β
Monitoring detects no change
(fail to reject Ho)
No Error
(1-α)
Missed-Change Error
(Type II) β
Table 3-3. Consequences associated with erroneous decisions resulting from Type I and Type II
errors when testing the null hypothesis of no change (from Tanke and Bonham 1985).
Type I Error:
It is erroneously concluded that a
change took place – Ho rejected
when true
Belief about range trend:
True status of resource:
Management decision:
Consequences:
short-term
long-term
Trend up
Trend static
Increase stocking
Trend down
Trend static
Decrease stocking
Decrease in condition
Decreased forage production
Loss in AUMs
Implications to ranch
stability
Type II Error:
It is erroneously concluded that
conditions are static – Ho accepted
when false
Belief about range trend:
True status of resource:
Management decision:
Consequences:
short-term
long-term
Trend static
Trend down
Maintain stocking
Trend static
Trend up
Maintain stocking
Further decrease in condition
Decreased forage production
Loss in AUMs
Implications to ranch
stability
Statistical power (1-β) is the probability that a particular test will reject the null hypothesis at a
particular level (α) when the null hypothesis is false. For monitoring programs, this is the probability
that a change will be detected when a change has really occurred. A common misunderstanding of
statistics often leads resource managers to interpret a failure to reject the null hypothesis to mean that
the null hypothesis is true. Whether the null hypothesis can be considered true depends on the power
of the test. If a monitoring program has high power and a change in a resource has not been detected,
a manager can conclude that no change has occurred in the resource. If the monitoring program has
low power and a change has not been detected, the manager cannot conclude that a change has or has
not occurred.
60
Failure to employ power analysis may result in the development and continuation of monitoring
programs that are incapable of meeting monitoring objectives or the misinterpretation of results from
existing monitoring programs. As a result, increased use of power analysis in both the design and
analysis stages of monitoring programs is called for (Toft and Shea 1983, Rotenberry and Wiens
1985, Hayes 1987, Peterman 1990a, Peterman 1990b). Hayes (1987) reviewed the toxicology
literature and found high power in only 19 of the 668 reports that failed to reject the null hypothesis.
In many cases conclusions were made as if the null hypothesis was proven to be true. However, only
in those studies with high power should the null hypothesis have been accepted as true. In the studies
with low power, the results should have been interpreted as inconclusive. Examinations of statistical
power associated with studies reported in specific journals and representing many topic areas have
shown similar results (Cohen 1977, Reed and Blaustein 1995, Forbes 1990).
3.1.6.2
Applying Power Analysis
Power analysis is applied principally in two forms (Peterman 1990a). Prior power analysis is used in
the design stage to determine the appropriate sample size required to yield a specified power
(Peterman 1990b, Rotenberry and Wiens 1985, Toft and Shea 1983). Post hoc power analysis is used
after data have been collected to determine the minimum detectable change or effect size (MDC or
MDES) for an existing survey (Rotenberry and Wiens 1985). The two approaches differ only in the
data required and the parameters solved for in the equation.
The use of prior power analysis is an important consideration when implementing a new monitoring
program at an installation. Although the best sample size is generally the largest sample size, at some
point the rate of increase in precision and power diminishes with increasing sample size (Green
1979). The question of concern with limited funds and manpower is not what is the best sample size?
but rather how many samples are required to meet management objectives? The original sample
allocation protocol used for RTLA was based on the population size (land area) rather than the
population variance (Diersing et al. 1992). As a result, recommended sampling intensity protocols
may not be optimal because budgetary and logistic constraints are usually the primary factors
dictating the magnitude of change that can be detected. Power analysis techniques using RTLA data
from similar installations or preliminary surveys could be used to estimate desired sampling intensity
based on ecological similarities.
Post hoc power analysis is a useful tool for existing monitoring programs. The use of power analysis
allows natural resource staff to determine the level of minimum detectable change. Only by knowing
the MDC for important variables can installation managers determine if the monitoring program is
fulfilling management objectives. This recommendation applies wherever quantitative data is
collected and change detection is an objective.
3.1.6.3
Statistical Tests and Power Analysis
The determination of statistical significance and the estimation of the probability of error in the
statistical conclusion are made within the framework of a particular statistical test. As such, the
statistical test is one factor that determines the statistical power (Lipsey 1990). Numerous statistical
tests are applicable to monitoring data analysis. Power equations for many of these tests are available.
Population change over time and associated power can be estimated with two sample tests and paired
tests using individual years (Cohen 1977). The same tests using the means of blocks of years before
and after an event also can be used to make the tests less sensitive to random annual environmental
variation (Cohen 1977). Gerrodette (1987, 1991) provides power equations for regression tests for
linear and exponential change. Green (1989) provides power equations for multivariate tests.
61
Bernstein and Zalinski (1983) provide power equations for monitoring programs that also employ
control plots. Although a number of power analysis models are available for use, Kendall et al. (1992)
concluded that power estimates using data from the first and last years (two sample data) is a
reasonable and robust procedure and is a good indicator of power, even for other trend tests.
For permanent plots such as those used in the original RTLA survey design, paired plot comparisons
(t-test) between two years are selected to determine the power of sampling protocols. This type of test
was selected because paired tests are appropriate for repeated measurements associated with
permanent sample plots (Snedecor and Cochrane 1980). This type of analysis requires only two years
of data. As such, it is applicable to the majority of installations currently implementing RTLA and
other monitoring programs. This applicability is especially true for data summaries that are available
from long-term survey data collected only every three to five years (Tazik et al. 1992, Price et al.
1995). A requirement of only two years of data may encourage installation personnel to employ the
techniques early in the implementation process, when the results are most useful. The power
associated with paired plot comparisons is more easily calculated than other methods so installation
personnel may be more likely to make use of the technique during data analysis. This general type of
analysis is applicable to many questions of interest to installation personnel.
The null and alternative hypotheses associated with paired plot comparisons are presented below. The
null hypothesis is that there is no change in the monitored resource. The alternative hypothesis is that
a change has occurred in the monitored resource.
The null hypothesis
H o : μ1 = μ 2
The alternative hypothesis
H a : μ1 ≠ μ 2
where:
μ1 = first year mean
μ2 = second year mean.
The power equations for paired plot comparisons are shown in Equation 1 and Equation 2 (Green
1979). Equation 1 calculates the number of plots required to detect a specified effect size with
specified values of α and β and an estimated variance for the attribute of interest. These equations
represent the a priori use of statistical power analysis techniques. Equation 3 calculates the minimal
detectable effect size for specified values of α, β, sample size, and estimated variable variance, and
represents post hoc use of statistical power analysis.
Equation 1: Power equation to estimate sample size when using a paired or “resampling of sites”
approach (prior).
n=
( tα + t β ) 2 S 2
Δ2
62
Equation 2: Power equation to estimate sample size when using a two sample or “reallocation of
sites” approach (prior).
n=
2( t α + t β ) 2 S 2
Δ2
Equation 3: Power equation to estimate minimum detectable change or effect size (post hoc).
MDC =
2
tα + t β ) s
(
2
n
where:
n
α
β
MDC
tα
tβ
Δ
s2
=
=
=
=
=
=
=
=
sample size
Type-I error level
Type-II error level
minimum detectable change
student t value associated with α for infinite degrees of freedom
student t value associated with β for infinite degrees of freedom
absolute effect size
variance of the differences between measurements.
Sometimes the tabular value Z is substituted for t values (Table 3-4). The Z values are obtained from a
table of normal deviates, and are not affected by n, the sample size. It is appropriate to use t values for
the specified α for n = infinity (Zar 1996). The t or Z value for β in power analysis is always
evaluated one-tailed (Green 1989).
Table 3-4. Z values for power analysis.
Type II Error
(β)
0.40
0.20
0.10
0.05
3.1.6.4
Power
(1-β)
0.60
0.80
0.90
0.95
Zβ
0.25
0.84
1.28
1.64
Biological Significance and Minimum Detectable Change
Statistical significance is a statement about the magnitude of a variable without regard for the
importance of the value. Biological or ecological significance is a statement about the magnitude of a
value of a variable based on management considerations. Biological significance is related to
statistical significance by considering the stability, power, and robustness of the survey methods
employed in the monitoring program. Biological significance is more important than statistical
significance when drawing a conclusion from sample data (Yoccoz 1991).
63
MDC in power analysis is the degree of change one wants to detect by the test. The choice of MDC
should be based on an understanding of the biology of the system and the economic and
implementation constraints associated with the survey. MDC is the smallest effect size that can be
detected for a given sampling intensity and specified error levels. Determining the MDC helps ensure
that statistical significance more closely corresponds to biological significance. If the MDC of a
survey is larger than the effect size that would be considered biologically significant, the study design
is considered to be weak (Cohen 1977). In weak study designs, small but biologically significant
changes in the resource may not be detected. If the MDC of a survey is smaller than the effect size
that would be considered biologically significant, the study design is considered to be strong. In
strong study designs, biologically significant changes in the resource should be detected. Without
specifying the minimum detectable change associated with a test, land managers are not provided the
information necessary to judge the strength of the available evidence (Peterman 1990b).
MDC is reported as either absolute (MDC or MDES) or relative (RMDC or RMDES). Absolute effect
size is the change that can be detected regardless of the abundance of the variable being measured.
The ability to detect an absolute change of 10 in the population implies that the protocols will detect a
change of 10 when the mean is 10 and a change of 10 when the mean is 20. Relative effect size
implies that the effect size will depend on the mean value of the variable being monitored. The ability
to detect a 25 percent change in the population implies that the protocols will detect a change of 2.5
when the mean is 10 or a change of 5 when the mean is 20. The choice of reporting format is
important and depends on the abundance of the variable being reported.
3.1.6.5
Achieving Statistical Power
The equations in the previous section illustrate that power is a function of the standard deviation,
sample size, minimum detectable change, and Type 1 error rate (α). By rearranging the equation, we
see that minimum detectable change is a function of the standard deviation, sample size, power, and
α.
There are several ways to increase statistical power (Elzinga et al. 1998):
a) reduce the standard deviation – this is accomplished by minimizing errors and implementing
an optimal sampling design. Ways to reduce the standard deviation of a sample include
optimizing sample unit size, shape, and number of observation per sample. Sample
stratification can be helpful in minimizing sample variability. However, some gains in
homogeneity may be offset by reductions in sample size.
b) increase the sample size – The increase in power results because sampling distributions
become narrower (i.e., a larger proportion of the estimates are closer to the true population
mean). This has less of an effect than reducing standard deviation where sample sizes are
already fairly large (>10).
c) increase the acceptable level of false-change errors (α)
d) increase the desired MDC – a sampling design is more likely to detect a true large difference
than a true small difference. As the size of the difference increases, there is a corresponding
increase in power to detect the difference.
The effects of these factors on statistical power are illustrated graphically in Elzinga et al. (1998).
64
3.1.6.6
Minimum Detectable Effect Sizes for Selected α and β Error Rates
Careful thought should be given to the consequences of both Type-I and Type-II statistical errors and
the appropriate rates of errors that are accepted. A Type-I error means that a management practice
such as site rehabilitation may be implemented where it is not necessary. A Type-II error means that a
necessary management practice may not be implemented because the problem is not detected. The
more stringent the standard set for Type-I error, the more likely a Type-II error will occur for a given
sampling intensity.
Determination of appropriate α and β levels should be based on the relative cost of committing TypeI and Type-II errors and on criteria external to the data (Cohen 1977, Toft and Shea 1983, Rotenberry
and Wiens 1985, Green 1989). In some circumstances the ecological/management consequences of
wrongly concluded change in a variable when none has occurred (Type-I error) may be equivalent to
the consequences of failing to detect change (Type-II error). Under those conditions, the errors should
be treated equally in the analysis of data. In natural resources management, Type-II errors often are
considered more costly than Type-I errors (Peterman 1990a, Thompson and Schwalbach 1995).
Setting β lower than α implies that the cost of Type-II errors are higher than the cost of Type-I errors
(Toft and Shea 1983). The relative costs will determine the acceptable error levels for each type of
error and are likely to be installation and management-objective specific. For example, the cost of an
extensive rehabilitation program may outweigh the costs of increased monitoring efforts required to
detect a problem early when rehabilitation may be less expensive. The cost of not modifying training
levels and rotating training areas when needed may or may not exceed the cost of modifying training
prematurely and implementing other rehabilitation programs.
3.1.6.7
Effect of One-tailed and Two-tailed Tests on MDC
The use of one-tailed (directional) tests can increase the efficiency of a study by reducing the required
sample size or decreasing the minimum detectable effect size for an existing study. One-tailed tests
are used only when there is reason to expect results in one direction (Snedecor and Cochrane 1980;
Sokal and Rohlf 1981). When analyzing RTLA data, there are many instances in which results in only
one direction may be expected and are of concern. Military impacts frequently result in decreased
vegetation cover, increased soil exposure, and increases in introduced species (Jones and Bagley
1998a, 1998b, Severinghaus et al. 1979, Shaw and Diersing 1990, Thurow et al. 1993, Trumbull et al.
1994). One-tailed tests may be justified when testing for the effects of increased training.
3.2
3.2.1
Sampling Design
Defining the Population of Interest
A population consists of the complete assemblage of units about which inferences are to be made.
The population of interest, or target population, can be difficult to evaluate, even when defined very
specifically. The boundaries of the target population must be defined in a concrete manner to
facilitate sampling within it. However, characteristics of the study area such as irregular boundaries or
very large areas may require redefining the target population (The Nature Conservancy 1997). In
some cases, the population boundaries may be redefined using macroplots that fit within the
population of interest, thus defining the majority of the population as the sampled population.
Macroplots are advantageous because they can be permanently marked and random plot coordinates
can be located for sampling within their boundaries. Macroplot approaches, however, are limited in
65
that statistical inferences can only be made about the area within the macroplot. For very large
populations (or areas), a number of samples or macroplots can be randomly positioned within the area
to provide an unbiased estimate. If plots are placed subjectively within the population of interest, then
inferences can only be made to the area encompassed by the macroplots. This is because bias is
introduced in the placement of the sampling area. In some cases, a macroplot may be placed randomly
within an area considered “representative” of the population of interest. In doing so, some bias in the
positioning of the macroplot is reduced, but inferences can still only be made to the area considered
“representative”, not to the entire population.
Plant communities at the landscape level can also be considered populations of interest. Community
types may cover very large geographic areas. To minimize variability and organize the landscape into
smaller populations of interest, the landscape can be stratified in various ways (see Section 3.2.4.3).
This is one of the major differences between sampling populations and sampling communities. Many
of the same attributes sampled in the context of communities are sampled for populations. However,
additional attributes such as community structure and diversity are examined at the community level,
combining information for a number of species to assess the more complex community units.
3.2.2
Selecting the Appropriate Sampling Unit
The attribute chosen for measurement will determine the type of sampling unit selected. The most
common attributes are cover, density, and frequency. Vegetation attributes and associated
measurement techniques are discussed in Section 4. A summary of sampling units associated with
vegetation attributes is presented in Table 3-5.
Table 3-5. Sampling units for measuring different vegetation attributes.
Attribute
Cover
Sampling Unit
points along lines
(point quadrats)
individual points (point
quadrats)
quadrats
transects (lines)
Notes
points are often associated with lines to facilitate placement
and replication (for permanent plots)
the points are the sampling units when they are randomly
placed (use only for non-permanent design)
quadrats can be individual or associated with lines, where
cover estimates are made for each quadrat
the transect is the sampling unit for the line interception
method
Density
quadrats/belt transects
transects (distance
methods)
plants counted by quadrat
individual plants are measured in association with a transect or
pacing method
Frequency
quadrats (most
common)
presence-absence data is summarized either by transect or as a
group of random quadrat placements
Biomass
quadrats
Because biomass involves clipping/removal, quadrat
approaches are used. In some cases individual plants are the
sampling unit.
Other
attributes
individual plants
plant heights, weights, diameters, basal areas, etc. are
measured for individual plants.
66
3.2.3
Determining the Size and Shape of Sampling Units
The optimum sampling unit size and shape is the one that produces the highest precision, usually
expressed as the standard error or coefficient of variation, at the lowest cost. Sometimes a
compromise involves accepting broader confidence intervals to stay within a fixed budget.
Considerations for sampling unit size and shape vary by the attribute measured. Some general
concepts apply regardless of whether quadrats are being used for frequency, density, or cover
measurements. Sampling unit configuration and size are affected by characteristics of the population
being sampled (density, distribution, and size of individual plants) and sampling considerations such
as the objective of the study and travel and setup time in relation to sampling time (Elzinga et al.
1998).
Number of Species
In community sampling it is often important to represent all species as fully as possible. Where the
emphasis is on sampling recurring communities or ecological units, it is helpful to determine the socalled “minimum area” of the community (Mueller-Dombois and Ellenberg 1974). The minimum area
is defined as the smallest area on which the species composition of the community in question is
adequately represented. A good idea of the area required provides guidance on the size and number of
quadrats. Individual quadrat sizes are adjusted primarily to the convenience of assessing the
quantitative attributes selected. Determining the minimal area can be difficult in communities that are
relatively heterogeneous or fragmented. The minimal area is determined by initially sampling a small
area, for example 0.5 x 0.5 m (0.25 m2), and recording all species that occur within this small area.
Then the sample area is enlarged to twice the size, then to four times, eight times, etc. The
additionally occurring species are listed separately for each enlarged area. The sample area is
increased until very few species are added to the list with increasing area sampled (Mueller-Dombois
and Ellenberg 1974). The results are referred to as a species-area curve, which illustrates graphically
the increase in species documented as area sampled increases (Figure 3-1). In the example presented,
the number of species continues to increase even at the largest area sampled, but the rate of increase
in the number of species levels out considerably at about 6 – 8 m2.
30
25
20
15
10
5
0
0
10
20
30
40
50
60
70
2
Area Sampled (m )
Figure 3-1. Species-area curve.
Depending on the species of interest (e.g., dominant species vs. functional groups vs. rare species vs.
all/most species) the minimal area concept may have limited applicability. There is generally no
advantage to increasing the sample size if it yields no additional species. The minimal area approach
often is used to determine relevé sizes for the Braun-Blanquet method.
67
Density – For measuring density, quadrat size must be related to the size and spacing of the
individuals. Counting excessive numbers of individuals per species cannot be done accurately in large
plots unless the plots are subdivided into smaller units or the individuals are somehow marked after
they are counted. Personal judgment is therefore a limiting factor. Recommended maximum sizes for
density quadrats are 10 x 10 m for the tree layer, 4 x 4 m quadrats for all woody undergrowth up to 3
m in height, and 1 x 1 m quadrats for the herbaceous layer (Oosting 1956). As quadrat size becomes
smaller, the greater the boundary in relation to the area, thus increasing the chance of boundary plants
and decisions regarding “in” or “out” individuals. Because the number of species increases with the
amount of area sampled, it follows that if one of the objectives is to document densities for all species
present, then many small plots will be required if small quadrats are used. If large quadrats are used,
species may be well documented but among-quadrat variation may be high, depending on the
distribution of the plants.
The shape of density quadrats has a significant effect on the precision of the count. Because most
vegetation has a clumped or aggregated distribution, rectangular shapes are more efficient than square
or circular shapes (Figure 3-2). This is because long, skinny plots tend to capture variability,
clumping, or banding (i.e., patterns) in the vegetation by virtue of their shape. This statement is
generally true because the longer quadrats tend to include at least some individuals. In a clumped
distribution, a square or circular plot is more likely to either include a large number of individuals, or
none at all. By minimizing the number of very low (or zero) counts and very high counts, precision
increases because variation between quadrats decreases. To take advantage of this relationship,
quadrats should be at least as long or longer than the distance between vegetation clumps. When the
population is truly random or has a uniform distribution, the shape of the quadrat has little effect on
density estimates.
Figure 3-2. Types of spatial distributions of individuals in a population.
Frequency – The size and shape of the quadrat influence frequency results. The optimum frame size
for each species will produce frequency values between 20 and 80%. This range of values allows for
68
both increases in frequency at the upper end of the scale, and decreases in frequencies at the lower
end of the scale. Frequencies greater than 95% and less than 5% can result in heavily skewed data
distributions. If the plot unit is too small, the likelihood of a species being recorded is small, and even
the more common species will receive a low frequency value. If the sampling unit is too large,
frequency results may be similar for species that have different distributions in a community.
Moreover, because individuals of a species may be aggregated, frequencies determined using square,
rectangular, or circular plots of the same area may differ. The use of nested frequency (i.e. 5cm x
5cm, 20cm x 20cm, 60cm x 60cm) quadrats helps to avoid problems with selecting a one-size fits all
frame.
The selection of an appropriate plot size is a subjective decision, based primarily on the size and
spacing of individual of a species. Using the minimal area concept is problematic for determining the
size and number of quadrats to sample because the minimal area plots are nested, resulting in a
frequency of 100% for all species within the area (Bonham 1989). Sequential sampling with nested
frames provides information on the appropriate frame size for different species and the number of
frames required to achieve the desired precision. For each species, the frequency can be calculated
after each plot placement (# frames present/# frames sampled). Sample adequacy can be achieved
where the frequency no longer changes significantly with each additional observation.
Cover - Vegetation cover can be estimated using points, lines, or quadrats. Considerations for making
visual estimates of cover in quadrats are similar to those for density. Estimating cover in quadrats is
done typically using a transect for sampling unit location; rectangular quadrats minimize variability
and generally document species composition better than circular or square plots. Limitations to
quadrat sizes are discussed in Section 4.2 Methods for Measuring Vegetation Attributes. The principal
limiting factor in cover estimation in quadrats is that estimation becomes more difficult and less
repeatable as quadrat size increases. Considerations for cover quadrat size and shape are discussed
further in Section 4.2. A discussion and comparison of lines, points, and quadrats also is presented in
Section 4.2.
3.2.4
Determining Sample Placement (Sampling Design)
Sample locations for the designs described below can be allocated within the population of interest
using GIS. The use of all designs is facilitated by the ease of locating sample locations using Global
Positioning Systems (GPS). For non-permanent plots, GPS can be used to determine the random
sample locations while avoiding any subjective bias by the observer. For permanently samples, GPS
navigation can be used to find the proximity of the plot whose exact location can be determined using
maps, landmarks, measurements, and permanent field markers. For systematically arranged samples,
the use of GPS minimizes bias while ensuring proper dispersion of sample locations. Pre-determined
sampling locations are typically uploaded to a GPS as waypoint files before going to the field.
3.2.4.1
Simple Random Sampling
In simple random sampling, every possible combination of sample units has an equal and independent
chance of being selected. If the probabilities of selecting different sampling units vary, then the
assumptions required for most statistical analyses will be violated. During the sampling process, the
selection of a particular unit (or location) should in no way be influenced by the other units that have
been selected or will be selected, i.e., the selection of a unit should be independent of the selection of
other units. Subjectively placing sampling units in locations known to be typical, homogeneous,
representative, or undisturbed should be avoided. If such placement is used, then sample data cannot
69
be extrapolated to larger populations of interest. A graphic representation of systematic sampling (2
possible variations), simple random sampling, and stratified random sampling is presented in Figure
3-3.
systematic grid 1
systematic grid 2
A: 7 plots
B: 5 plots
C: 4 plots
simple random sample
stratified random sample
Figure 3-3. Four possible distributions of 16 sample plots (without replacement) in a population
composed of 256 square plots.
Sample units can be selected with or without replacement. Sampling with replacement allows samples
to overlap partially or completely with other samples. In sampling without replacement, a particular
unit (e.g., plot, quadrat, transect) is allowed to appear in the sample only once. Most resource
sampling is without replacement (Avery and Burkhart 1995).
For simple random samples, the estimate of the population mean is given by:
x=
where
∑x
i
n
x = population mean
xi = observed value of x in sample I
n = sample size
70
The variance of the sample measurements is calculated as:
s
2
∑(x
=
− x)2
1
n −1
The sample standard deviation is the positive square root of the variance.
The standard error of the mean, SE, (when sampling with replacement or for infinite populations) is
calculated as:
SE =
s
n
The expression presented above for standard error assumes that the sample is drawn from an infinite
population. That is to say, the sum total of the area sampled is an insignificant portion of the total
possible area sampled. If the population sampled is finite, and the sample comprises at least 5% of the
area, then the finite population correction (FPC) factor should be applied. For a sample n of a given
size, drawn from a population of N possible placements of sampling units, the fraction of the area
sampled is:
FPC =
1−
size of area sampled
total size of population
or
n ⎞
⎛
⎜1 −
⎟
N ⎠
⎝
The standard error is corrected by multiplying it by the FPC. The FPC slightly reduces the calculated
standard error; its effect becomes larger as the proportion of the population sampled increases. If the
fraction of the area is less than 5% or plotless methods are used, the FPC is considered negligible.
Simple random sampling does not ensure that every portion of the area of interest is sampled
uniformly (i.e., samples are distributed evenly within the population). Poor distribution or
interspersion of sampling units may result in a sample that is not representative of the population.
Poor interspersion can result from random clumping of units or where the sampling units are all
located along one or several transects. As Figure 3-4A illustrates, several areas within the population
of interest, notably the center and far right regions, have not received any samples.
3.2.4.2
Restricted Random Sampling
Restricted random sampling, a form of simple random sampling, promotes a more uniform
distribution of samples within the area sampled (Figure 3-4B). A comparison of Figure 3-4A and
Figure 3-4B illustrates how restricted random sampling can improve the “spread” of the samples.
Both allocations consist of 25 sample points, however, the restricted allocation significantly reduces
the likelihood that random zones are undersampled or not sampled at all. Restricted random sampling
can also be applied within stratified random sampling. For example, a strata such as training areas can
be used to restrict the allocations of samples in strata within each training area. We could specify that
each training area be allocated a minimum of 20 samples, which subsequently are allocated in a
random stratified manner within each training area. In addition to improving the distribution of
71
samples, this method helps to ensure that minimum sample size criteria are met for each training area.
Analysis of samples is the same as for a simple random sample.
A
+
+
+
+
+
+
+
+
+ +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
B
+
+
+
+ +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Figure 3-4. (A) Distribution of 25 random sample locations within the population of interest (less
uniform distribution), (B) Distribution of 25 restricted random samples within the population of
interest (more uniform distribution).
There are several methods for simple random sampling. One selection technique consists of drawing
random intersection points in a coordinate system based on x and y grid coordinates. This can be done
either manually or by using coordinates generated by computer programs, including GIS. Coordinates
that fall outside of the area of interest are rejected. This approach works well for small sampling units
(rectangular or round) but can be problematic when the sampling unit is a long line or rectangle,
where a portion of the unit may fall outside of the population even though the starting point does not.
If sampling units that extend outside of the population are rejected outright, then the sample will be
biased against locations near the outer boundary (i.e., boundary areas will have a lower probability of
being selected). If sampling unit locations are shifted inwards so that units fall just inside the
boundary, then sampling units will have a higher probability of being near the edge of the population
(The Nature Conservancy 1997). Overlapping sampling units is also a potential problem using the
grid coordinate approach.
The grid cell method eliminates some of the problems associated with the grid coordinate system. In
this method, the population area is overlaid with a conceptual grid, where the grid cell size is
equivalent to the size of each sampling unit (The Nature Conservancy 1997). For example, is the
sampling unit is a 2m X 5m (10 m2) quadrat and the population is approximately 1ha (10,000m2) in
size, then the population is divided into approximately 1,000 possible quadrat positions. Sampling is
without replacement, so none of the units overlap. To allocate plots, random points along the x and y
axes are selected as the corners (usually lower left) of each unit. If any pairs of coordinates are
repeated, the second pair is rejected and another pair is selected at random.
72
3.2.4.3
Stratified Random Sampling
Sample stratification helps to create divisions so that samples can be analyzed with other samples that
are more alike than if simple random sampling were used (i.e., all plots aggregated). Simple random
samples are taken within each stratum. Advantages of stratification include lower sample variances
and better distribution of samples across the sample space. The main disadvantage of stratified
sampling is that sample sizes can become small, especially when multiple strata are employed.
Landscapes are stratified into discrete categories for various management purposes. Classification
systems are developed as tools for communication, for aggregation of information into logical units,
for interpretation, and for the extrapolation of information among units with similar properties
(Leonard et al. 1988). Characteristics used to delineate strata include remote sensing imagery
(satellite, high altitude, low altitude), vegetation and ecological communities, land-use areas or types,
and soil taxonomic categories.
Strata should be defined by characteristics that will remain relatively constant over time. The type and
number of strata chosen will depend on the intended uses of the stratification. Classification schemes
often set arbitrary or subjective boundaries based on interpretation constraints and management
needs, even though it is understood that continua or gradients are present for most classification
attributes; discrete natural boundaries are not common. Strata should consist of units that are
functionally or ecologically sensible. Additional strata such as administrative divisions (e.g., training
areas) may be incorporated in response to management needs. These strata, which can have little or
no ecological meaning, may influence the spatial distribution of samples by creating a more uniform
sampling distribution.
Allocation of samples among strata can be equal, in proportion to stratum size (area), in proportion to
density/abundance, or in proportion to the amount of stratum variability (optimum allocation).
If samples are located randomly within each stratum, then the principles and procedures of simple
random sampling apply for each of the strata or subpopulations,. However, more complex formulae
are required to estimate means for the whole population. Formulae and examples are provided by
Krebs (1989) and Platts et al. (1987).
Stratum weight = Wh =
where
Nh
N
Nh = size of stratum h (or number of possible sample units in stratum h)
N = size of entire statistical population
Stratum weights are proportions and must add up to 1.0. The means and variances are then calculated
for each stratum using the simple random sampling equations presented in Section 3.2.4.1. Stratum
sizes are used as weights to produce a weighted mean. The overall mean per sampling unit for the
entire population is estimated as follows (Cochran 1977):
L
x ST =
where
x ST
∑N
h
xh
h −1
N
= stratified population mean per sampling unit
73
Nh
h
xh
= size of stratum h
= stratum number (1,2,3…,L)
= observed mean for stratum h
N
= total population size =
∑N
h
Using the sample data provided in Table 3-6, we have:
x ST =
(15000)(58) + (10000)( 47) + ( 20000)(76) + ( 4000)( 24) + 1500(65)
= 60.86 %
50500
The same results are produced by summing the strata weights multiplied by the mean cover.
Table 3-6. Stratified random sampling on a Midwestern installation. The number of plots per stratum
was determined haphazardly.
Stratum
Size (Nh)
(ha)
Stratum
Weight (Wh)
Sample
Size (nh)
woodland
mixed forest
grassland
coniferous forest
bottomland
15,000
10,000
20,000
4,000
1,500
0.297
0.198
0.396
0.079
0.030
54
31
72
14
9
TOTALS
50,500
1.0
180
Stratum
Mean perennial
Variance of
herbaceous cover/ cover (Nh2)
sampling unit (xh) (%)
58
47
76
24
65
454.29
119.32
178.73
328.68
420.46
The variance of the stratified mean is calculated as:
L
⎡W 2 s 2 ⎤
Variance of ( x ST ) = ∑ ⎢ h h ⎥
h =1 ⎣ n h
⎦
where
Wh
sh2
nh
= stratum weight
= observed variance of stratum h
= sample size in stratum h
The variance of the stratified means depends only on the sizes of the variances within each stratum
(i.e., not between strata) and the sample size. Therefore, the selection of homogeneous strata and
higher sample sizes will decrease the size of the weighted mean.
For the herbaceous cover data in Table 3-6, the variance of the stratified mean is calculated as:
⎡ (.297) 2 ( 454.29) ⎤ ⎡ (.198) 2 (119.32) ⎤ ⎡ (.396) 2 (178.73) ⎤ ⎡ (.079) 2 (328.68) ⎤ ⎡ (.030) 2 ( 420.46) ⎤
Variance of ( xST ) = ⎢
⎥+⎢
⎥+⎢
⎥+⎢
⎥+⎢
⎥
54
31
72
14
9
⎣
⎦ ⎣
⎦ ⎣
⎦ ⎣
⎦ ⎣
⎦
Variance of ( xST ) = 1.47
74
The standard error of the stratified mean is the square root of its variance (Krebs 1989). Note that the
variance of the stratified mean cannot be calculated unless there are at least two samples in each
stratum.
Standard error (s) of ( xST ) = variance of xST = 1.56 = 1.21
The confidence interval for the stratified mean is computed as:
x ST ± ts x ST
The number of degrees of freedom for t is approximated by (n1-1) + (n2-1) + (n3-1) + … + (nL-1).
Thus for the population mean, the 90% confidence limits are
62.72 ± (1.96)(1.21)
= 60.35 to 65.09
3.2.4.3.1
Allocation of Samples
When planning a stratified sampling design the number of sampling units to measure in each stratum
must be determined. There are two alternative procedures for distributing field plots among strata:
proportional and optimal allocation. In some situations, it may be desirable and necessary to
“reallocate” plots based on evaluation of sample sizes needed for each stratum. For example, existing
plots can continue to be used and additional plots can be added to strata that have inadequate sample
sizes. The required sample sizes can only be estimated after initial or pilot data is available to provide
an estimate of variability (see Section 3.2.6). According to Krebs (1989), for most situations the
standard error of the mean (SE) will have the following pattern in accordance with the allocation
procedure:
SE (optimal allocation) ≤ SE (proportional allocation) ≤ SE (random allocation)
Regardless of the method of allocation, the principles of stratified random sampling produce a simple
random sample within each stratum.
3.2.4.3.1.1 Proportional Allocation
This approach to stratified sampling allocates samples to strata on the basis of a constant sampling
fraction in each stratum. Most often referred to as allocation in proportion to size, proportional
allocation distributes field plots in proportion to the area of each type. For example, for the example
presented in Table 3-6, the number of plots in each stratum would be calculated as follows if the total
number of plots is known:
⎛N ⎞
n h = ⎜ h ⎟n
⎝ N ⎠
where
nh = the number of plots for a stratum
Nh = the size of a stratum
N = the size of all strata combined
n = the total number of plots to be sampled
75
Assuming that 180 plots are to be sampled:
Table 3-7. Proportional allocation of samples.
Stratum
Stratum
Size (Nh)
woodland
mixed forest
grassland
coniferous forest
bottomland
TOTALS
15,000
10,000
20,000
4,000
1,500
50,500
Calculation
Number of
Samples
(15000/50500)*180
(10000/50500)*180
(20000/50500)*180
(4000/50500)*180
(1500/50500)*180
53
36
71
14
5
180
One disadvantage of proportional allocation is that large areas receive more sample plots than small
ones, irrespective of the variation in the attribute measured. This limitation also applies to simple
random and systematic sampling. However, when strata can be defined and their areas determined,
proportional allocation should be superior to a similar, non-stratified sample.
3.2.4.3.1.2 Allocation in Proportion to Variance (Optimal Allocation)
With this procedure, the sample plots are allocated in a manner that results in the smallest possible
standard error of the mean possible with a fixed number of samples. Determining the number of plots
to assign to each stratum requires the calculation of the product of the area and the standard deviation
for each type (Table 3-8). Because optimum allocation uses the variance term to distribute samples,
the variable must be chosen carefully. Required sample sizes and proportional allocation can be
calculated for a number of important variables, and if feasible, the highest required sample size could
be chosen. Based on the available number of samples (180) and the standard deviation of each
stratum, the number of plots can be estimated.
Table 3-8. Preliminary calculation for optimal allocation.
Stratum
woodland
mixed forest
grassland
coniferous forest
bottomland
TOTALS
Area (ha)
St. Dev. of
cover (Nh2)
15,000
10,000
20,000
4,000
1,500
50,500
21
11
13
18
21
84
Area X
Variance
(Nhsh)
319710
109236
267376
72518
30758
799598
The following equation calculates the estimated sample size for each stratum:
⎛ N h sh
n h = n⎜
⎜∑N s
h h
⎝
where
⎞
⎟
⎟
⎠
n = the total sample size (either fixed, as in this case at 180, or estimated)
and the other terms are defined above.
76
⎛ 319710 ⎞
woodland : ⎜
⎟ * 180 = 72 plots
⎝ 799598 ⎠
⎛ 109236 ⎞
mixed forest : ⎜
⎟ * 180 = 25 plots
⎝ 799598 ⎠
⎛ 267376 ⎞
grassland : ⎜
⎟ * 180 = 60 plots
⎝ 799598 ⎠
⎛ 72518 ⎞
coniferous forest : ⎜
⎟ * 180 = 16 plots
⎝ 799598 ⎠
⎛ 30758 ⎞
bottomland : ⎜
⎟ * 180 = 7 plots
⎝ 799598 ⎠
total samples = 180
Note that in this case, optimum allocation produced a different distribution of samples among the
strata relative to the proportional allocation. Compared to the proportional allocation, more plots were
allocated to the mixed forest and grassland communities, which have relatively high acreages but low
standard deviations. Optimum allocation can also integrate the cost of sampling if the costs of
sampling differ among the sampling strata (Krebs 1989).
3.2.4.4
Systematic Sampling with a Random Start
Systematic sampling is commonly used for assessing forest and range conditions because the sample
units are easily located on the ground and they seem more uniformly distributed and therefore more
representative. To be considered a random sampling method, the starting point must be chosen
randomly. The regular placement of points or quadrats for cover estimation or frames along a transect
area examples of systematic sampling. Using a baseline to locate perpendicular transects at regular
intervals is another example. Point-based cover and frequency data collected along multiple transects
can be analyzed using two techniques. The first technique is to combine all quadrat data into a single
sample (binomial data treated as an overall proportion), i.e., treat the sample as if the quadrats were
the sampling unit distributed as a simple random sample. The second technique is to calculate
frequency or cover values for each transect, thus treating the transect as the sampling unit. While
systematic sampling provides good interspersion of samples, precision can be low if the pattern of the
sampling units intersects with a naturally-occurring pattern or cyclic variation in the population.
3.2.4.5
Other Methods
Several other methods are more complex in terms of both design and data analysis. They are used
sometimes when subsampling characteristics of individual plants. A brief overview of each method is
presented.
3.2.4.5.1
Cluster Sampling
Cluster sampling should not be confused with cluster analysis, which is a classification technique.
Cluster sampling is a method of selecting a sample when the individual elements cannot be identified
in advance. Groups or clusters of the elements (e.g., leaves on individual plants) are selected
randomly, and every element in each cluster is measured. For example, a manager is interested in the
number of cavities in snags along a riparian corridor. However, there are too many snags along the
corridor to make measuring all of them feasible. Therefore, the study area could be divided into 1 ha
77
plots. Each plot can contain a different number of snags. The 1 ha area becomes the cluster. A random
sample of the clusters is taken and cavities are measured on all trees within each cluster.
Cluster sampling is convenient and inexpensive with regard to travel costs. Maximum efficiency is
gained when elements within a cluster are close to one another geographically (Platts et al. 1987). It is
most efficient when different clusters are similar to each other and variability is low within clusters. If
each cluster contains large numbers of the element of interest, two-stage sampling is probably more
efficient (The Nature Conservancy 1997). Disadvantages of cluster sampling include a tendency
toward a higher variance among elements sampled and fairly complex analysis computations.
Adaptive cluster sampling uses the same initial framework, but then increases sampling if certain
conditionas are met. Before sampling, criteria are established that must be met for sampling to occur.
The process begins by selecting random sampling locations (simple or systematic). Whenever an
observed value at your sampling locations satisfies your minimum (or absolute) criteria, additional
units are added to the sample from the neighborhood of the original unit. If the criteria are satisfied in
the additional units, data is collected and still more units can be added. For adaptive cluster sampling,
computation of means and variances follows a different set of equations to account for the biased
nature of the sampling design.
3.2.4.5.2
Two-stage Sampling
Two-stage or multi-stage sampling is used where clusters have so many elements that it is prohibitive
to measure all elements in the cluster. Two-stage sampling is also applied where all of the elements in
a cluster are so nearly the same that to sample all of them would provide little additional information.
It would therefore make sense to sample only a portion of the elements within the cluster. Like cluster
sampling, groups of elements are first identified. First a random sample of the groups is selected
(primary sampling units). Then, a second random sample of elements (secondary sampling units) is
taken within each group. For example, in the example used in cluster sampling, if the number of
snags is too extensive within each 1 ha plot, then subsampling the snags in each plot makes sense.
The primary sampling unit is the 1 ha plot, and the secondary sampling unit is the snag.
Compared to simple random sampling, two-stage sampling can be less expensive because it is easier
to sample many secondary units in a group than to sample the same number of units spread out across
the population. However, statistical precision may be lower with two-stage sampling. Statistical
precision may be improved by using two-stage adaptive sampling techniques. In adaptive sampling,
the secondary unit is a “seed cell”. If the value within the seed cell meets a selection criterion, the
adjacent cells are sampled until no new cells are added to the network or the boundary of the adjacent
primary unit is met. In two-stage sampling, calculations are relatively complex because variances
must be calculated for the estimates at both stages. Formulae and examples are provided by Platts et
al. (1987) and Krebs (1989).
3.2.4.5.3
Double Sampling
Double sampling, often used in rangeland studies of plant production, uses the measurement or
estimation of one variable to estimate another. When the variable of interest is difficult and/or
expensive to measure, it is measured only in a small fraction of the sampling units. Because precision
would be expected to be low if only a few measurements are made, an auxiliary variable that is much
easier to measure is estimated in a much larger number of sampling units. Regression techniques are
then used to define the relationship between the two sets of measurements.
78
For example, because it is slow and expensive to clip, dry, and weight biomass in many sampling
units, observers often estimate biomass ocularly. The unit clipped can be plots, branches, whole
plants, or portions of plants. Bonham (1989) describes the process in estimating plant biomass using
double sampling. In this method a number of randomized observations are taken by visual estimation
of biomass weight, and, in addition, biomass is clipped at a small number of sampling units taken at
random from a large sample of visually estimated units. The data collected consists therefore of a
large sample that contains visual estimates, and within the large sample, a small sample that contains
clipped weights of biomass in addition to the visual estimates. The small dataset is used to calculate a
regression equation which represents the relationship between the clipped weight y to the estimated
weight observed by visual estimation x. Estimates of x for the large sample are then used to obtain
predicted biomass weights y.
If the auxiliary variable is measured and highly correlated with the variable of interest, double
sampling is more efficient in estimating the variable than direct measurement. The principal
disadvantage is that data analysis and sample size determinations are significantly more complex than
those required for simple random sampling.
3.2.5
Permanent vs. Temporary Plots
The selection of permanent or temporary sampling units will influence the precision of estimates that
are collected over time as well as the method of data analysis. Temporary plots are plots that are
randomly located each time sampling occurs. The data from one time period to the next is considered
independent, because the data collected in an given year has no influence on the value that is recorded
in subsequent years. When sampling units are permanent, the same plots are surveyed during
subsequent data collection. Most long-term monitoring studies make use of permanent plots for both
range and forest studies. Continuous forest inventory (CFI) and Long-Term Ecological Research
(LTER) plots are examples of permanent plots used for long-term monitoring.
3.2.5.1
Advantages of Permanent Plots
The principal advantage of permanent plots is that plant counts between time periods are to some
degree correlated with one another, thereby reducing the spatial heterogeneity relative to using
temporary plots. For most plant species, especially large and/or long-lived species, permanent plots
provide greater precision with the same number of quadrats or the same precision at smaller sample
sizes relative to temporary plots. Because of this reason and the fact that the statistical tests for
detecting changes over time are more powerful for permanent plots, the number of required sampling
units is reduced compared to temporary plots. The higher the correlation between time periods, the
more advantageous the use of permanent plots. If high mortality or short-lived species dominate, then
temporary plots may be more appropriate.
3.2.5.2
Disadvantages of Permanent Plots
Disadvantages of permanent plots include additional time required to relocate plot markers and
establish plots, failure to relocate plots, sampling impacts associated with repeated visits to a site,
need to transport, install, and map marker locations, and impact of plot markers on wildlife and
aesthetics (too many markers are unsightly).
79
3.2.6
Sample Size Requirements
Sample size adequacy should relate directly to monitoring objectives. It is an essential aspect of both
planning and evaluating monitoring efforts. Especially in quantitative sampling, the ability to meet
requirements of precision and minimum detectable change is crucial in justifying program costs and
making sound management recommendations. Sample size equations can use pilot data, historical
data, or professional opinion to estimate some of the required parameters. Precision increases with
sample size. However, precision is highly dependent on sample variability. Therefore, the optimal
design will have the lowest achievable variability with a sample size that provides acceptable overall
precision and change detection ability.
For information about computer programs for determining sample size and statistical power, see
Chapter 11 Software for Statistical Analysis.
3.2.6.1
Estimating Sample Size Using Sequential Sampling
The number quadrats per transect can be determined through relatively quick pilot sampling. In pilot
sampling, a number of different frame sizes can be evaluated simultaneously if the data are collected
in the appropriate format. For example, a quadrat for cover or density measurement can be 1 x 3 m in
size, and can be subdivided into smaller units during each placement (e.g., 1 x 2 m, 1 x 1 m, 0.5 x 3
m, 0.5 x 2 m, 0.5 x 1 m, 0.25 x 1 m, etc.). At each placement, estimates of cover or density tallies are
recorded for each frame size. Usually, a predetermined number of observations or samples are
collected and subsequently evaluated.
The number and size of frames per sample (for cover and density measures) is usually driven by the
relationship between area surveyed and number of species found, the running mean, and the running
standard deviation, which can be calculated in the field. This is also referred to as sequential
sampling. A pilot study for sequential sampling may consist of randomly positioning a number of
sampling units of different sizes and shapes (you don’t need actual frames - you can use sticks/poles
to form 2 sides of a frame) within the area to be sampled. Then choose the size and shape that yields
the smallest coefficient of variation (COV), where the coefficient of variation is the sample standard
deviation divided by the sample mean. The sampling design with the lowest COV is the most efficient
(to help determine sample size). The process consists of collecting data for one or more sampling unit
sizes and shapes. As each additional quadrat is sampled, the mean and standard deviation of the
sample are recalculated. This is repeated after each quadrat is sampled. The data should be
summarized and graphed in a random order to eliminate any patterns in the data. The sample size is
plotted graphically against the running mean and standard deviation. This can be done by hand or
using computer software. An estimate of minimum sample size is indicated where the running mean
and standard deviation level out (Figure 3-5). Elzinga et al (1998) point out that a Y-axis scale that is
too broad due to initial means and standard deviations may lead to a false impression that a leveling
out is occurring when it is not. Extreme values and zeros can confound this process and possibly
indicate that alternate sampling unit sizes and shapes may need to be changed.
For the same amount of area sampled, long, skinny quadrats (e.g., 0.5m x 2m) tend to pick up more
species of plants and are representative of the variability than square plots (e.g., 1m x 1m). The linear
configuration can also remedy problems associated with clumping by individuals/species that results
in some quadrats having high counts and other having small counts or no data. The larger the quadrat
becomes, the more difficult it becomes to estimate canopy cover.
80
3.2.6.2
Calculating Sample Size Using Formulas
If data variability within a type is high, then detecting change in the resource can be difficult. On
most installations, more plots may be necessary to minimize the variability and show statistically
significant changes or, alternatively, to confirm that conditions are relatively static. Some installations
have either pilot or historic data that can be used to calculate sample size by stratum (e.g., vegetationsoils-land use categories). Different equations exist for sample size determination, based on
temporary or permanent plots and density/cover estimates or frequency. Platts et al. (1987) present
sample size equations for estimating a single population mean or a single population total with a
confidence interval around the mean or total for simple random or stratified random sampling
designs.
It is important to note that sample size calculation is an iterative process. Sample standard deviations
will be relatively large with small sample sizes, and will generally become smaller with each
additional sample added. Therefore, sample size requirements should be recalculated after several
additional samples are collected to see their effect on sample size requirements. Estimated sample
sizes are always rounded up to the next whole number.
Corrections for sampling finite populations, where more than 5% of the population is sampled, are not
addressed here because it is rare in community monitoring to sample such a large proportion of the
population (or area) of interest.
Computer programs for calculating sample sizes and statistical power are presented in Chapter 7 Data
Analysis and Interpretation.
81
n Value Running St. Dev.
Mean
1
5
5.00
2
9
7.00
2.83
3
2
5.33
3.51
4
0
4.00
3.92
5
4
4.00
3.39
6
2
3.67
3.14
7
5
3.86
2.91
8
4
3.88
2.70
9
3
3.78
2.54
10
4
3.80
2.39
11
0
3.45
2.54
12
8
3.83
2.76
13
8
4.15
2.88
14
1
3.93
2.89
15
6
4.07
2.84
16
7
4.25
2.84
17
4
4.24
2.75
18
7
4.39
2.75
19
9
4.63
2.87
20
2
4.50
2.86
21
9
4.71
2.95
22
7
4.82
2.92
23
8
4.96
2.93
24
6
5.00
2.87
25
8
5.12
2.88
8
n Value Running St. Dev.
Mean
26
4
5.08
2.83
27
0
4.89
2.94
28
7
4.96
2.91
29
8
5.07
2.91
30
8
5.17
2.91
31
9
5.29
2.95
32
5
5.28
2.90
33
3
5.21
2.88
34
4
5.18
2.84
35
5
5.17
2.80
36
6
5.19
2.77
37
6
5.22
2.73
38
1
5.11
2.78
39
8
5.18
2.78
40
7
5.23
2.76
41
4
5.20
2.73
42
7
5.24
2.71
43
2
5.16
2.72
44
6
5.18
2.70
45
5
5.18
2.67
46
8
5.24
2.67
47
3
5.19
2.66
48
3
5.15
2.65
49
7
5.18
2.64
50
8
5.24
2.64
Mean
Std dev.
7
Running Mean
6
5
4
3
2
1
0
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49
Number of Samples
Figure 3-5. Example of sequential sampling graph using pilot data.
82
3.2.6.2.1
Sample Size Required to Estimate a Population Mean
The approach to estimate sample size as described by Cochran (1977) is appropriate for simple
random sampling designs. RTLA data is often treated as simple random sample data because sample
locations were randomly allocated in proportion to area, and the locations were thus unbiased.
The following equation calculates sample size adequacy for an installation as a whole or a particular
management or ecological unit:
n=
t 2s2
( Ex ) 2
where:
n
t
s2
E
= the number of samples required
= t distribution value for a given level of confidence (using infinite degrees of
freedom); 2-tail value provides a plus/minus confidence interval
= an estimate of the sample variance for a specific vegetation category
= level of precision desired for the estimate of the mean
_
x
= the sample mean for a specific vegetation attribute
Elzinga et al. (1998) recommend adjusting the sample sizes calculated with this equation using
procedures developed by Kupper and Hafner (1989).
Example
Monitoring Objective: Estimate the mean density of trees having a dbh ≥10 cm to within +/-10% of
the estimated true value at a 95% level of confidence.
Sample Data:
Based on pilot data, the mean density of trees with dbh ≥10 cm = 35
Standard deviation = 11
and based on the monitoring objective above:
the desired confidence level is 95%. α = 0.05, so tα = 1.96 (2-tailed)
the allowable error = 10% x (the estimated mean) = (0.10) x (35) = 3.5 trees
Therefore,
n=
1.96 2 × 112
= 37.95 or 38 samples
3.5 2
The sample size necessary to be 95% confident that the estimate of the population mean is within
10% of the true mean = 38 plots.
83
3.2.6.2.2
Sample Size Required to Estimate a Proportion
The equation for calculating the required sample size for estimating a proportion is:
n=
t 2 pq
d2
where:
n
t
p
q
d
= the number of samples required
= t distribution value for a given level of confidence; 2-tail value provides a
plus/minus confidence interval
= an estimate of the proportion
= 1-p
= the permitted error (proportion expressed in decimal form) – this is the level of
precision specified.
Example
Monitoring Objective: Estimate frequency of species J within 10% of the estimated true value, at the
90% level of confidence.
Sample Data:
Pilot sampling using 200 random quadrats produced a frequency (p) of 35% or 0.35.
Therefore, if p = 0.35, then q = 1-p = 0.65.
and based on the monitoring objective above:
the desired confidence level is 90%. α = 0.10, so tα = 1.64 (2-tailed)
the permitted error (d) = 10% = 0.10. Note that d is the absolute value of the allowable error, and
not a percentage of the mean (i.e., not 0.1 x 0.35).
Therefore,
1.64 2 × ( 0.35 × 0.65 )
= 61.19 or 62 quadrats
n=
0.10 2
The necessary sample size to be 90% confident that the frequency is within 10% of the estimated true
frequency = 62 quadrats.
3.2.6.2.3
Sample Size Required to Detect Differences Between Two Proportions
(temporary sampling units)
The equation for calculating the required sample size to detect a difference between two proportions
is:
84
n=
( Z α + Z β ) 2 ( p1 q1 + p 2 q 2 )
( p 2 − p1 ) 2
where:
n
Zα
Zβ
p
q
p1
q1
p2
q2
= the number of samples required
= Z value for the Type I error rate (2-tailed)
= Z value for the Type II error rate (1-tailed)
= an estimate of the proportion expressed as a decimal. A conservative estimate
is 0.50 if no other estimate is available.
= 1-p
= the value of the proportion for the first sample
= 1- p1
= the value of the proportion for the second sample
= 1- p2
Values of the Z-distribution for various Type II error rates.
Type II Error
Power
(β)
(1-β)
0.40
0.60
0.20
0.80
0.10
0.90
0.05
0.95
Zβ
0.25
0.84
1.28
1.64
Example
Monitoring Objective: We want to be 90% confident of detecting a 25% decrease in the absolute
frequency of knapweed 1 year after herbicides are applied. We accept a 10% chance of making a
Type II error.
Sample Data:
The current (pre-herbicide) frequency of knapweed is estimated to be 72% = p1 = 0.72
q1 = 1-p1 = 1-0.72 = 0.28
and based on the monitoring objective above:
We are interested in detecting a change of 25% so p2 is assigned the value (0.72 – 0.25) = 0.47.
therefore, q2 = 1-p2 = 0.53
Type I error = α = 0.10, so Zα = 1.64
Type II error = β = 0.10, so Zβ = 1.28
85
Therefore,
n=
(1.64 + 1.28) 2 ((0.72)(0.28) + (0.47)(0.53))
= 61.5 or 62 samples
(0.47 − 0.72) 2
The estimated sample size to be 90% certain of detecting a 25% change in frequency with a falsechange error rate of 10% (power of 90%), is 62 plots/quadrats.
3.2.6.2.4
Sample Size Required to Detect Differences between Two Means (temporary or
non-permanent plots)
If sites are randomly allocated for each sample, then the approach is different than if the samples are
permanent. For example, if we want to test the hypothesis that there is no change in an attribute
between before and after-impact times, and a new set of samples is chosen for the second survey, the
following equation is used to estimate the sample size required to detect a specified absolute change
between the two means, for given Type I and Type II error rates:
n=
where:
2(tα + t β ) 2 S 2
Δ2
Δ = detectable change
n = sample size
α = Type-I error level
β = Type-II error level
tα = Student t value associated with α*
tβ = Student t value associated with β
S2 = Sample variance from initial sample
*Note: t-values for sample sizes of infinity are approximately equal to, and are substituted for Zvalues.
Example
Monitoring Objective: At the 90% confidence level, detect a 15% change in the mean density of
individuals of species X between 2000 and 2005, with a Type II (missed-change) error rate of 10%.
Sample Data:
Pilot sampling in 2000 provided the following estimates used in the calculation –
mean density = 44 plants/plot
standard deviation = 14 plants
and based on the monitoring objective above:
Type I error = α = 0.10, so tα = 1.64 (2-tailed)
Type II error = β = 0.10, so tβ = 1.28 (β is always 1-tailed)
desired MDC = (1995 mean) x 15% = (44) x (0.15) = 6.6 plants/plot
86
Therefore,
n=
2(tα + t β ) 2 S 2
Δ2
=
2(1.64 + 1.28) 2 × 14 2
= 76.7 or 77 samples
6.6 2
The estimated sample size to be 90% certain of detecting a 15% difference between 1999 and 2004,
and with a false-change error rate of 10% (power of 90%), is 77 plots.
3.2.6.2.5
Sample Size Required to Detect Differences Between Two Means (paired or
permanent plots)
Calculating the minimum detectable effect size using power analysis can be done after two years of
sampling. Alternately, for a specified minimum level of change detection, the required sample size
can be determined.
The equation below estimates the sample size for a given absolute change and specified Type I and
Type II errors:
n=
where
(tα + t β ) 2 S 2
Δ2
Δ = detectable change
n = sample size
α = Type-I error level
β = Type-II error level
tα = Student t value associated with α*
tβ = Student t value associated with β
S2 = variance of the differences between measurements
*Note: t-values for sample sizes of infinity are approximately equal to, and are substituted for Zvalues.
Example
Monitoring Objective: I want to be 90% confident of detecting a 20% change in mean oak cover. I am
willing to accept a 20% chance of making a Type II error.
Sample Data:
Ten permanent plots were sampled in 1999 and 2004. Oak cover was measured on each plot. Results
are presented below.
1999 mean = 14.10% oak cover
2004 mean = 17.80% oak cover
mean paired difference = 4.10
standard deviation of the paired difference = 6.64
n = 10 plots sampled each year
87
and based on the monitoring objective above:
Type I error = α = 0.10, so tα = 1.64 (2-tailed)
Type II error = β = 0.20, so tβ = 0.84 (β is always 1-tailed)
desired MDC = 20% = (the larger of the 2 means) x 20% = (17.80) x (.20) = 3.56 (absolute MDC)
Therefore,
n=
(tα + t β ) 2 S 2
Δ2
=
(1.64 + 0.84) 2 × 6.64 2
= 21.4 or 22 samples
3.56 2
The estimated sample size to be 90% certain of detecting a 3.56% absolute difference in cover
between 1999 and 2004, and with a false-change error rate of 20% (power of 80%), is 22 plots.
Spreadsheets can be developed to calculate power, MDC, and required sample sizes. Table 3-9
illustrates how a power analysis spreadsheet might be organized.
Table 3-9. Example of a spreadsheet for evaluating power and sample size requirements for change
detection monitoring.
t(alphas) are 2-tailed and t (betas) are 1-tailed
For a 1-tailed test, need to multiply alpha by 2 and then determine critical t value
alpha =
beta =
0.10
0.20
t(alpha) =
t(beta) =
1.64
0.84
enter
desired
value
VEGTYPE
variable
BLUE OAK total cover
SAVANNA native cover
introd cover
peren cover
bare ground
disturbance
oak cover
3.3
1994
mean
76.60
36.70
56.00
23.80
23.60
3.00
14.10
1995
mean
85.70
48.90
67.20
31.30
22.80
1.80
17.80
ABS MDC size Relative
n
Observed
with current n, MDC (%
Desired required
change
power of 0.8
change
Desired
MDC
for
(paired sd (paired and alpha of from time
MDC
(Relative to desired
n
diff)
difference)
0.1
A)
(Absolute) 1st year)
MDC
10
10
10
10
10
10
10
9.10
15.20
12.00
8.50
9.20
2.80
4.10
6.45
10.92
6.09
8.45
6.39
3.82
6.64
5.1
8.6
4.8
6.6
5.0
3.0
5.2
6.6
23.3
8.5
27.8
21.2
100.0
36.9
15.3
11.0
16.8
7.1
7.1
6.0
4.2
20
30
30
30
30
200
30
2
7
1
9
6
3
16
References
Anderson, A.B., P.J. Guertin, and D.L. Price. 1996. Land Condition-Trend Analysis Data: Power
Analysis. Technical Report 97/05. U.S. Army Construction Engineering Research Laboratories
(USACERL), Champaign IL.
Avery, T.E. and H.E. Burkhart. 1995. Forest Measurements, 3rd Ed. McGraw-Hill Publishing
Company. New York.
88
Bernstein, B.B. and J. Zalinski. 1983. An optimal sampling design and power test for environmental
scientists. Journal of Environmental Management 16: 35-43.
Bonham, C.D. 1989. Measurements for Terrestrial Vegetation. John Wiley, New York.
Brady, W.W., J.E. Mitchell, C.D. Bonham, and J.W. Cook. 1995. Assessing the power of the pointline transect to monitor changes in plant basal cover. Journal of Range Management 48: 187-190.
Bureau of Land Management. 1996. Sampling Vegetation Attributes: Interagency Technical
Reference. BLM National Applied Resource Sciences Center, BLM/RS/ST-96/002+1730. Supersedes
BLM Technical Reference 4400-4, Trend Studies, dated May 1995. 163 pp.
Cochran, W.G. 1977. Sampling Techniques, 3rd ed. John Wiley and Sons, New York.
Cohen, J. 1977. Statistical Power Analysis for the Behavioral Sciences. Academic Press, N.Y.
Diersing, V.E., R.B. Shaw, and D.J. Tazik. 1992. U.S. Army Land Condition-Trend Analysis (LCTA)
program. Environmental Management 16: 405-414.
Elzinga, C.L., D.W. Salzer, and J.W. Willoughby. 1998. Measuring and Monitoring Plant
Populations. USDI Bureau of Land Management, in partnership with Alderspring Ecological
Consulting, The Nature Conservancy of Oregon, and the Bureau of Land Management California
State Office. BLM Technical Reference 1730-1. BLM National Applied Resource Sciences Center,
Denver, CO.
Floyd, D.A. and J.E. Anderson. 1987. A comparison of three methods for estimating plant cover.
Journal of Ecology 75: 221-228.
Forbes, L.J. 1990. A note on statistical power. The Auk 107: 438-453.
Gerrodette, T. 1987. A power analysis for predicting trends. Ecology 66: 1364-1372.
Gerrodette, T. 1991. Models for power of predicting trends – a reply to Link and Hatfield. Ecology
72: 1889-1892.
Green, R.H. 1979. Sampling Design and Statistical Methods for Environmental Biologists. John
Wiley and Sons, New York. 257 pp.
Green, R.H. 1989. Power analysis and practical strategies for environmental monitoring.
Environmental Research 50: 195-205.
Hayden, T.J. and D.J. Tazik. 1993. Integrated Natural Resources Monitoring on Army Lands and Its
Application to Conservation of Neotropical Birds, In Status and Management of Neotropical
Migratory Birds. General Technical Report RM-229. USDA Forest Service.
Hayes, J.P. 1987. The positive approach to negative results in toxicology studies. Ecotoxicological
and Environmental Safety 14: 73-77.
89
Jones, D.S. and C.F. Bagley. 1998a. Tracked Vehicle Impacts on Plant Community Characteristics:
Yakima Training Center (1994-1997 study period). TPS 98-7, Center for Ecological Management of
Military Lands, Colorado State University, Fort Collins, CO.
Jones, D.S. and C.F. Bagley. 1998b. Tracked Vehicle Impacts on Plant Community Characteristics:
Orchard Training Area (1995-1997 study period). TPS 98-9, Center for Ecological Management of
Military Lands, Colorado State University, Fort Collins, CO.
Kendall, K.C., L.H. Metzger, D.A. Patterson, and B.M. Steele. 1992. Power of grizzly sign surveys to
monitor population trends. Ecological Applications 2(4): 422-430.
Kennedy, K.A., and P.A. Addison. 1987. Some considerations for the use of visual estimates of plant
cover in biomonitoring. Journal of Ecology 75: 151-157.
Krebs, C.J. 1989. Ecological Methodology. Harper Collins, New York.
Kupper, L.L. and K.B. Hafner. 1989. How appropriate are popular sample size formulas? The
American Statistician 43: 101-105.
Leonard, S.G., R.L. Miles, and P.T. Tueller. 1988. Vegetation-Soil Relationships on Arid and
Semiarid Rangelands. In P.T. Tueller (ed.), Vegetation Science Applications for Rangeland Analysis
and Management. Kluwer Academic Publishers, London.
Lipsey, M.W. 1990. Design Sensitivity: Statistical Power for Experimental Research. Sage
Publications, Newbury Park, CA.
Mitchell, J.E., W.W. Brady, and C.D. Bonham. 1994. Robustness of the Point-Line Method for
Monitoring Basal Cover. Research Note RM-528. USDA Forest Service Rocky Mountain Forest and
Range Experiment Station.
Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and Methods of Vegetation Ecology. John
Wiley, New York. 547 pp.
Oosting, H.J. 1956. The Study of Plant Communities: An Introduction to Plant Ecology, 2nd Ed. W.H.
Freeman and Co., San Francisco. 440 pp.
Peterman, R.M. 1990a. Statistical power analysis can improve fisheries research and management.
Canadian Journal of Fisheries and Aquatic Sciences 47: 2-15.
Peterman, R.M. 1990b. The importance of reporting statistical power: the forest decline and acid
deposition example. Ecology 71: 2024-2027.
Platts, W.S., C. Armour, G.D. Booth, M. Bryant, J.L. Bufford, P. Cuplin, S. Jensen, G.W.
Lienkaemper, G.W. Minshall, S.B. Monsen, R.L. Nelson, J.R. Sedell, and J.S. Tuhy. 1987. Methods
for Evaluating Riparian Habitats with Applications to Management. General Technical Report INT221. USDA Forest Service. 17 pp.
Price, L.D., A.B. Anderson, W. Whitworth, and P.J. Guertin. 1995. Land Condition-Trend Analysis
Data: Preliminary Data Applications. USACERL Technical Report TR N-95/39/ADA300753.
90
Reed, J.M. and A.R. Blaustein. 1995. Assessment of ‘nondeclining’ amphibian populations using
power analysis. Conservation Biology 9: 1299-1300.
Rice, C.G. and S. Demarais. 1995. An Analysis of the LCTA Methods for Inventory and Monitoring
of Birds and Small Mammals on Army Lands in the Southwest United States. Final Report to
USACERL, Texas Tech. University.
Rice, C.G., S. Demarais, and R.W. Hansen. 1995. Statistical Power for Evaluating Monitoring
Methods and Analysis for the Army’s Lands Condition-Trend Analysis Program. In F. Dallmier,
(ed)., Proceedings of the SI/MAB International Symposium: Measuring and Monitoring Forest
Biological Diversity. Smithsonian University Press, Washington, D.C.
Rotenberry, J.T. and J.A. Wiens. 1985. Statistical power analysis and community-wide patterns.
American Naturalist 125: 164-168.
Severinghaus, W.D., R.E. Riggins, and W.D. Goran. 1979. Effects of Tracked Vehicle Activity on
Terrestrial Mammals, Birds, and Vegetation at Fort Knox, Kentucky. USACERL SR N77/ADA073782, Champaign, IL.
Shaw, R.B. V.E. Diersing. 1990. Tracked vehicle impacts on vegetation at the Pinon Canyon
Maneuver Site, Colorado. Journal of Environmental Quality 19: 234-243.
Snedecor, G.W. and W.G. Cochrane. 1980. Statistical Methods. Seventh Edition. Iowa State
University Press, Ames, Iowa.
Sokal, R.R. and F.J. Rohlf. 1981. Biometry, 2nd ed. Freeman Press, San Francisco.
Tanke, W.C. and C.D. Bonham. 1985. Use of power curves to monitor range trend. Journal of Range
Management 38(5): 428-431.
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (LCTA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
The Nature Conservancy (TNC). 1997. Vegetation Monitoring in a Management Context Workbook. Workshop coordinated by The Nature Conservancy and co-sponsored by the U.S. Forest
Service, held in Polson, MT, September 1997.
Thompson, F.R. and M.J. Schwalbach. 1995. Analysis of Sample Size, Count Time, and Plot Size
from an Avian Point-Count Survey on Hoosier National Forest, Indiana. In Monitoring Bird
Populations by Point Counts. General Technical Report PSW-GTR-149. USDA Forest Service
Thurow, T.L., S.D. Warren, and D.H. Carlson. 1993. Tracked vehicle traffic effects on the hydrologic
characteristics of Central Texas rangeland. Transactions of the American Society of Agricultural
Engineers 36: 1645-1650.
Toft, C.A. and P.J. Shea. 1983. Detecting community-wide patterns: estimating power strengthens
statistical inference. American Naturalist 122: 618-625.
91
Trumbull, V.L., P.C. Dubois, R.J. Brozka, and R. Guyette. 1994. Military camping impacts on
vegetation and soils of the Ozark plateau. Journal of Environmental Management 40: 329-339.
Yoccoz, N.G. 1991. Use, overuse, and misuse of significance tests in evolutionary biology and
ecology. Bulletin of the Ecological Society of America 72: 106-111.
Zar, J.H. 1996. Biostatistical Analysis, 3rd ed. Prentice Hall, Upper Saddle River, New Jersey.
92
4 Measuring Vegetation Attributes and Other Indicators of
Condition
4.1
Vegetation Attributes
The most important measurable attributes in vegetation sampling are frequency, cover, and
abundance. Other measurable characteristics include height, stem diameter, biomass, and structural
criteria such as twig diameter, leaf size, or bark thickness. Frequency, cover, and abundance are most
commonly used for descriptive purposes, whereas attributes which measure physiological parameters
such as productivity measures (e.g., litter production, seed production, annual diameter increment),
leaf-area index, and transpiration rates are measured primarily for experimental or research purposes.
Other descriptive characteristics such as composition, indices of biological diversity, and vertical
structure are derived from measured attributes.
4.1.1
Frequency
4.1.1.1
Description
Frequency is a rapid and precise method for monitoring vegetation. Frequency is the number of times
a species is present in a given number of sampling units. Species presence can be defined as rooted
within the sample unit or present in the vertical projection above the sample unit (i.e., cover). It is
usually expressed as a percentage. No counting is performed; only species presence is recorded.
Frequency is a useful index for examining changes in vegetation over time and comparing different
plant communities. Frequency can be estimated using small plots, points, or lines, which may be
distributed randomly or systematically. Nested frequency uses several sampling units nested within
one another. Using this approach provides for appropriate frame sizes for both abundant and less
common species. By reducing the frequency quadrat to a point, presence/ absence data for the more
abundant species can be recorded. This method, known as the “point quadrat” method, is used to
measure cover.
4.1.1.2
Advantages and Limitations
Estimates of frequency are sensitive to the size and shape of the sample unit, as well as the abundance
and distribution (i.e., pattern) of plants. The appropriate plot size and shape depends on the
distribution, size, and number of the plant species. The determination of quadrat dimensions requires
an iterative approach using pilot data or field experience for the specific community and species in
question. Quadrats or nested quadrats are the most commonly used sampling units.
To detect changes in species frequency, frequencies should be between 20% and 80% optimally, but
frequencies between 10% and 90% (Bureau of Land Management 1996) or 5% and 95% (Bonham
1989) are acceptable. To obtain frequency values within these limits for both dominant and less
common species, nested frequency frames of different sizes are used. Frequency comparisons are
only made only among quadrats of the same size and shape.
93
Interpreting changes in frequency is difficult because it is not obvious whether canopy cover, density,
and/or pattern of distribution have changed. Thus, frequency is not directly related to other vegetation
variables specified in typical management objectives. Plant frequency alone may not be a sufficient
basis for making land management decisions.
Frequency measurements are rapid and highly repeatable. Rooted frequency is less sensitive to
seasonal or yearly variations due to environmental factors and stressors compared to canopy
frequency.
Frequency is sensitive to variations in seedling establishment. High seedling mortality may lead to
large fluctuations in frequency data for a particular species. Recording seedlings as a separate
category remedies this problem.
4.1.1.3
Appropriate Uses
For detecting changes in vegetation, frequency is probably the most cost effective method providing
statistically defensible results (Hironaka 1985). It also may provide a cost efficient early warning of
undesirable changes in key or indicator species (West 1985a). West (1985a) warns against using
frequency data to calculate percent composition because of the sensitivity of frequency to quadrat size
and shape and pattern of plant distribution.
Frequency has been used for both rangeland and forest monitoring (emphasis on herbaceous
understory in forest environments). However, plant frequency alone may not be a sufficient basis for
making land management decisions. This is primarily because frequency is not directly related to
more commonly applied vegetation attributes.
4.1.2
Cover
4.1.2.1
Description
Cover generally refers to the percentage of ground surface covered by vegetation. This definition
applies to both estimation and measurement methods for cover. Cover can also be expressed as an
absolute quantity (i.e. square meters, hectares, etc.). Cover is considered to be one of the most
important measures of ecological significance, and is probably the most commonly measured
attribute. It is sometimes correlated with density and/or biomass, but not consistently in all
communities or sites. It also provides a measure for species that cannot be measured effectively using
density or biomass.
Vegetation cover is the total cover of vegetation on the study site or within the sample area. The
maximum amount of total cover is 100%. Basal cover or area is the percentage of the basal portion of
vegetation in contact with the ground. For trees it is measured through the diameter (using area = πr2),
usually at 1.5 m above the ground. Foliar, foliage, or aerial cover is the area of ground covered by the
vertical projection of the aerial portion of plants. Canopy cover is the area of the ground covered by
the vertical projection of the outer perimeter of the foliar cover, including small openings or gaps in
the canopy. When cover is measured by species or guilds (functional groups), the total canopy or
foliar cover may exceed 100% due to overlap of plants at sample locations. Ground cover is the cover
of plants, litter, rocks, and gravel on a site. The objectives of the monitoring design should define and
specify the type of cover measured.
94
All methods for estimating cover involve the coverage of a plant or vegetation in a quadrat.
Approaches for measuring cover are generally two and three dimensional. Points, lines, quadrats
(microplots or macroplots), and plotless techniques are used to measure cover.
4.1.2.2
Advantages and Limitations
Canopy cover and foliar cover are the most sensitive to climatic and biotic effects which may produce
misleading interpretations. This is especially true for herbaceous vegetation, and less so for woody
plants. For trend comparisons in herbaceous communities, basal cover is considered to be the most
stable type of cover because it varies less due to climatic fluctuations and grazing pressure compared
to aboveground cover.
Ground cover is often used as an indicator of soil or watershed stability. Comparisons between sites
should be made carefully because of differing ecological potentials.
Point-based methods are considered to be the most objective ways to measure cover. Quadrat-based
methods require ocular estimation by the observer and are therefore more prone to observer bias.
Ocular estimation using cover classes (especially broad classes) is not sensitive to small changes in
cover. The desired size of detectable change will influence the selection of cover method.
4.1.2.3
Appropriate Uses
Point-based methods are used primarily to measure herbaceous vegetation where there is little to no
obstruction by shrubs or other woody plants. This is generally also true for small cover quadrats,
which are inappropriate for measuring tree cover. Large quadrats or area-based cover estimates
provide a rapid and useful method to collect information for community classification, vegetation
mapping, and understanding relationships between vegetation and the site. Line transects are good for
low herbaceous species, basal cover of sod-forming grasses and bunchgrasses, and canopy cover of
shrubs, but are also applicable to tree canopies in prairies or savannas.
4.1.3
Density
4.1.3.1
Description
Density is the number of individuals or other sampling attribute per unit area. It is commonly used
because it is easily obtained and interpreted. Moreover, by actually counting plants (or stems,
inflorescences, etc.), observer bias is minimized, as long as individuals are readily distinguishable
from one another. Density has been used extensively in measuring trees and shrubs, and less so for
bunchgrasses and forbs. Comparisons should only be made among similar life forms and sizes of
plants (e.g., shrubs and shrubs, not shrubs and annual grasses). Despite the difficulties in measuring
some perennial herbaceous communities, density can be a good metric because it is less variable from
year to year than measures of cover or biomass. Quadrat approaches are typically used for estimating
densities in herbaceous communities, whereas plotless methods are often used for woody species and
trees. Homogeneous stands may be best sampled using square quadrats; long, thin quadrats or belt
transects are suggested for heterogeneous communities.
Variable plot or plotless techniques are often used to determine the density of trees and shrubs.
Methods include the point-centered quarter (PCQ), wandering quarter, line transect, random pairs,
95
nearest neighbor, and closest individual. Using plotless techniques may improve the accuracy of the
estimate because boundary errors are eliminated.
4.1.3.2
Advantages and Limitations
A significant limitation in measuring density is the recognition of individuals. Trees and single
stemmed shrubs and herbs are easy to count, although counting can be slow and tedious for abundant
herbs. Cases where the individual plant is difficult or impossible to identify due to vegetative
reproduction include spreading or creeping shrubs (e.g., krummholz vegetation), clonal or motteforming woody species (e.g., some species of Cornus, Rubus, Quercus, Populus), sod-forming
grasses (e.g., Bouteloua gracilis), and rhizomatous grasses and forbs. Some of these difficulties can
be overcome by establishing rules of thumb regarding the definition of what constitutes an individual
(e.g., minimum specified distance between individuals). The density of rhizomatous or stoloniferous
plants can be determined by counting stems instead of individuals. The density of seedlings and
mature plants should be recorded separately to minimize false indicators of positive or negative
trends.
Boundary decisions can influence results. Small and elongated quadrats have a higher proportion of
boundary to area compared to larger, square quadrats. However, rectangular and “long skinny plots”
may capture more variability in the data and species presence.
The different density methods are designed to handle both random and clumped distributions of
plants.
4.1.3.3
Appropriate Uses
Density measures are appropriate for species where individuals can be consistently identified. The use
of height classes may be important to distinguish among densities of different cohorts or size classes.
Density is useful when comparing similar functional groups or life forms; is the number of
individuals of a group or specific species increasing or decreasing? When plants are widely dispersed,
plotless methods may be the most efficient method for measuring density.
4.1.4
Biomass
4.1.4.1
Description
Production is one approach to assessing the importance or role of an individual species in a plant
community. Aboveground production during the peak growth stage is commonly used as an indicator
of site condition or forage availability. Production can be assessed relative to a benchmark value for a
particular site or species, or can be examined relative to other species or sites. Biomass estimates are
often provided in soil surveys to assist land managers in estimating carrying capacity and site
potential.
Herbage weight or biomass is an important attribute for rangeland and grassland vegetation because it
supports all secondary and higher consumer groups, either directly or indirectly (Pieper 1988). A
number of methods have been developed to estimate plant biomass. Estimation techniques (plant or
plot basis), often employ a double sampling procedure to improve on ocular estimates. Harvest
methods are destructive sampling procedures that are most often used in research studies to measure
the response of vegetation to different treatments. Tree and shrub biomass is usually estimated from
96
dimensional analysis, which uses measurements such as crown diameter, crown area, height, and
basal diameter to predict biomass from regression equations. Biomass of shrubs is usually measured
in terms of total biomass and forage for animals. Analysis by size class and site is usually necessary
to establish significant relationships between tree and shrub size and biomass (Bonham 1989). While
biomass may come closest to capturing the three-dimensional structure and dominance of species
relative to other measurements such as cover, density, and frequency, it is constantly changing over
time and is difficult to measure efficiently (Pieper 1988).
Terminology related to production is most often associated with vegetation biomass, and includes
(Bureau of Land Management 1996):
Gross primary production is the total amount of organic material produced, both above and below
ground, often for a one-year period
Productivity is the amount of production for a specified time period and area, expressed as a rate
(e.g., kg/ha/yr)
Biomass is the total weight of living organisms in the ecosystem, including plants and animals
Standing crop or phytomass is the amount of plant biomass present above ground at any given point
in time
Browse is the portion of woody plant biomass accessible to herbivores
4.1.4.2
Advantages and Limitations
Because below-ground production and animal biomass are difficult to measure, aboveground standing
crop is the attribute most often chosen for trend studies in rangelands and grasslands where biomass is
measured. Peak standing crop is generally measured at the end of the growing season. Because
different species reach peak standing crop at different times, measuring mixed communities can be
problematic (Bureau of Land Management 1996). Measurement errors generally increase with the
diversity of plant species and growth patterns due to difficulties in distinguishing individual plants
and species. Where grazing is significant, exclosures are required for estimating “undisturbed”
conditions.
Collecting production data is time and labor intensive. In some cases, relationships between cover and
production can be established so that cover data can be used to estimate production.
4.1.4.3
Appropriate Uses
Biomass measurements are appropriate in grasslands and rangeland environments. In extremely arid
environments where production is low, production estimates may be difficult to make because of
fluctuations in climate and biotic influences. Biomass data can be used to estimate carrying capacity,
range condition, and the dominance of species in a plant community.
The nature of the vegetation is an important consideration in the selection and type of sampling unit.
For extensive surveys, estimation techniques used alone or in conjunction with double-sampling
procedures may be most appropriate (Pieper 1988). Research projects will probably continue to use
direct harvest techniques despite their disadvantages
97
4.1.5
Structure
4.1.5.1
Description
Structure refers to how vegetation occupies a three-dimensional space, and can be related to height,
abundance, and vertical distribution. Vegetation is measured either by individual heights or using
predetermined height classes. In this way, information such as species composition and abundance
(e.g., cover) is collected about the different vertical layers. Structure information can be summarized
from data that includes plant heights, or from methods specifically designed for wildlife concealment
such as the cover board and Robel pole methods.
4.1.5.2
Advantages and Limitations
Methods used to collect wildlife habitat information are relatively efficient, allowing for numerous
samples to be collected in the sampling area. Visual obstruction methods have little observer bias, and
are therefore highly repeatable. Methods that rely on cover estimates require more training to reduce
observer bias. Structure data are usually complemented by other data to examine trends (Bureau of
Land Management 1996).
4.1.5.3
Appropriate Uses
Application of structure information includes evaluation of plant communities in terms of wildlife
habitat, and assessing the usefulness of different communities to military training. Training
applications include assessing vehicle mobility and concealment. Structure can be examined for
grasslands as well as communities with shrubs and trees.
4.1.6
Dominance or Composition
4.1.6.1
Description
Composition is the proportion of various plant species in relation to the total for a given area, and
may be expressed as relative cover (including basal area of trees), relative density, relative weight,
etc. (Bureau of Land Management 1996). “Composition of vegetation” typically refers to a list of
plant species found in a specific area or vegetation type. However, a simple species list does not
convey as much information as information about the proportion of each species in a plant
community. “Composition”, therefore, should not be confused with species composition or richness,.
To calculate composition, the value of the measurement for each species or group of species is
divided by the total value for the entire population. Most monitoring methods, including all of the
methods described in Section 4.2 (except frequency, cover board and robel pole methods) provide
information for calculating composition.
4.1.6.2
Advantages and Limitations
Most methods can be used to calculate composition. Issues of repeatability and precision are related
directly to the attribute selected and the method used. Changes in composition will vary by the
attribute selected. For example, cover estimates may vary annually due to climatic variability while
estimates of basal area are relatively stable.
98
4.1.6.3
Appropriate Uses
Composition is used mainly in the description of ecological sites and to evaluate community
condition. Like cover, composition can be very sensitive to biotic and edaphic influences (Bonham
1989).
4.2
Methods for Measuring Vegetation Attributes
Evaluating monitoring methods involves numerous considerations (Figure 4-1). Design and methods
decisions are probably most complex or difficult when monitoring communities (composition and
dynamics), followed by multispecies groups (functional groups or guilds), populations (structure and
dynamics), and individuals (physiology and behavior). Probable costs of monitoring follow the same
scale, from relatively high to relatively low (Hinds 1984). Photographic monitoring is recommended
as a complement of most monitoring methods. It provides an efficient and inexpensive way to
document conditions over time.
In many cases, well-defined, widely-accepted methods already exist and should be used (Davis 1989,
USEPA 1987). Frequency of sampling, sample size, plot sizes, and other details should be determined
by pilot studies or evaluation of existing data. When accepted methods are chosen, additional
evaluation and refinement of procedures may improve the cost effectiveness over the long term
(Hinds 1984). Silsbee and Peterson (1991), in the context of monitoring in the National Park System,
offer the following general guidelines when selecting a monitoring method:
“Simplicity and economy are important at every stage of program development, because
the program will have to be sustained for many years through changes in personnel and in
funding priorities (Garton 1984, Halvorson 1984, Quinn and van Riper 1990). Simplicity
and economy must not come at the expense of methodological rigor and statistical
validity, however. From the outset, thought must be given to how the data will be
processed, analyzed, and archived (Hinds 1984, Jones 1986). Techniques should be as
insensitive as possible to differences between observers, because substantial personnel
turnover is likely and because field personnel may not always be highly trained (Davis
1989). New techniques must be calibrated with old techniques whenever methods are
changed…(Lund 1983).”
Consistency is essential where data is to be compared among installations. Differences in
methodology or terminology can lead to difficulties or misleading results when comparing
installations. Where comparability is important among installations or regions, using widely-accepted
methods and communication with other installations may minimize these difficulties (Silsbee and
Peterson 1991). Some commonly-used vegetation monitoring methods are presented in Table 4-1.
99
DEFINE OBJECTIVE
SPECIFIC AREA OR TAXA
DETERMINED BY OBJECTIVE?
yes
qualitative
no
SELECT ATTRIBUTES
TO MONITOR/
MEASURE
SELECT AREA OR
TAXA
quantitative
VARIANCE AMONG
PLOTS KNOWN?
no
PRELIMINARY DATA
AVAILABLE?
no
COLLECT PILOT DATA
FROM MULTIPLE
PLOTS
yes
yes
COLLECT
PRELIMINARY DATA TO
EVALUATE COST AND
EFFECTIVENESS
ANALYZE DATA FOR
VARIANCE ESTIMATES
PRECISION AND
POWER ACCEPTABLE?
DOCUMENT FINDINGS
no
yes
RECONSIDER
ECOLOGICAL DESIGN
COST PER UNIT PRECISION
AND POWER ACCEPTABLE?
RESULTS
ACCEPTABLE?
yes
no
ACCEPTABLE
MONITORING METHOD
no
Figure 4-1. Considerations for evaluating ecological monitoring methods (modified from Hinds
1984).
100
Table 4-1. A comparison of vegetation data collection methods employed on Army Installations (modified from USAEC 1996).
PointIntercept (original LCTA1 )
Description
Intended
Application
Other
Potential
Uses
Metal pins are lowered at regular intervals
along a fixed length transect (30m, 50m,
100m, 200m, etc). The contact at ground
surface is recorded, i.e., basal area by
species, litter, rock, etc. Canopy intercepts
are recorded by species and sometimes by
height.. Pin frames having several to many
pins are sometimes used. The Original
LCTA method uses a 100m transect, a 1m
metal rod (1/4” diam.) is placed every 1m. A
telescoping range pole is used to measure
intercepts above 1m.
Ground and canopy cover (frequency)
Frequency, species composition, vegetation
condition and trend
Strengths
Objective, rapid and easily taught.
Few surveyor decisions.
Limitations
Canopy intercepts are sensitive to wind and
surveyor differences. Not recommended for
cover less than 5% or greater than 35%.
Field sampling design may not be
appropriate for all vegetation types in U.S.
Difficult to pick up minor species unless a
high number of points are used.
Requires a large number of points to meet
sample size adequacy. Focus is on
dominant plant species. Use of sharpened
pins and pin guide helps to minimize
observer bias.
Bonham 1989; Diersing et al. 1992; Tazik et
al. 1992.; Mitchell et al. 1994
Additional
Comments
References
1
Line-Intercept
Daubenmire
Cover Quadrats/
Nested Cover Quadrats
A measuring tape is stretched
between two stakes or points and can
be of any length. The intercept
distance is recorded for each
plant/species that intercepts the line.
The accumulated length for any
species divided by the length of the
transect (tape) multiplied by 100 is
expressed as percent cover for that
species.
A 20cm x 50cm quadrat is used to estimate
percent ground cover. Canopy cover is
visually estimated as a vertical projection of
a polygon drawn around the extremities of
each plant. The projections are summed
and recorded in a corresponding cover
class. Six cover classes are used and
converted to class midpoints for data
analysis.
Initially a small quadrat (i.e. 10 x 10 cm) is
designated and visual estimates are made
of the percentage of the quadrat occupied
or covered by a vertical projection to the
ground. The area is progressively doubled
to twice the area, 4 times, 8 times, etc. The
smaller quadrat is nested within the larger
quadrat. Cover is estimated relative to the
quadrat size with each enlargement.
Canopy cover
Canopy cover
Canopy cover
Frequency, basal diameter of species,
density (Bonham 1989, p. 177),
species composition, vegetation
condition and trend
Canopy cover of different vertical
strata can easily be estimated.
Equally adaptable to small and large
areas. Basal diameter of grasses is
commonly employed for monitoring.
Not suitable for dense intermingled
herbaceous species. This method is
only appropriate for species with a
relatively large basal area (i.e.
bunchgrasses, shrubs) or small gaps
in the canopy.
Frequency, species composition,
vegetation condition and trend
Frequency, species composition, vegetation
condition and trend
Not intended for plants greater than 1m in
height. Estimates are subject to surveyor
bias and requires training to standardize
observer estimates.
Cover estimates requires extensive training
and repeated comparisons with measured
data or between surveyors. Difficult to
estimate cover in large quadrats.
Suitable in sparse vegetation where
the plants are distinct and shrub
communities, such as the Western
U.S.
Data are summarized using the midpoints
for the cover class; this results in low
precision over time.
In general, cover estimates are desirable
when individuals cannot be distinguished.
Canfield 1941; Bonham 1989.
Daubenmire 1959.
Bonham 1989.
Suitable to estimate cover for small shrubs, Cover estimates are relative to quadrat size.
rhizomatous grasses and bunchgrasses.
Cover classes enable repeatable results
among surveyors.
Land Condition-Trend Analysis
101
Table 4-1. Methods matrix cont.
Braun-Blanquet
Density Quadrats (including
belt transect)
Modification of the Modified
Step-Point
Description
A surveyor deliberately and carefully
selects a non-random sample location. A
detailed description of the sample location
is made, which may include information on
slope, aspect, soil depth and type. A
species list is made. Sites are classified by
grouping locations that have a number of
species in common. Groups are usually
arranged by a computer program and
provide a classification from associations.
The number of individual species
are counted in each quadrat.
Density is expressed as the number
of individuals per unit area.
The nearest species to the point in a
forward direction, 180° arc along a
transect is recorded. Occurrences for
each species are summed and
divided by the total number of points.
Intended
Application
Classification of plant communities
Density
Frequency
Other
Potential Uses
Canopy cover and abundance estimated
by seven classes, species composition
Species composition, vegetation
condition and trend
Species composition
Strengths
Detailed description of the area.
Species composition, frequency,
monitor population dynamics,
including TES
Provides a direct count of
individuals per unit area.
Quick estimate of species
composition. Reduces surveyor bias
from pin placement.
Limitations
Subjective and dependent upon the
experience and knowledge of the
vegetation type by the surveyor.
Quadrat size is important. If the
quadrat is too small the variation will
be high and conversely a large
quadrat will be too time consuming.
Unable to conduct statistical analysis
for change over time because of the
bias. Tend to over estimate frequency
if a species is recorded at every point.
Additional
Comments
Quadrat size should be no larger than 10m
x 10m when numerous species are
present for cover estimates. May not
detect a change because of decreased
precision associated with cover classes.
Vegetation condition is based
exclusively on vegetation
presence/absence and not
environmental parameters such as
soil stability.
References
Braun-Blanquet 1965; Shimwell 1971,
Mueller-Dombois and Ellenberg 1974.
Surveyor defines what constitutes
an individual. Defining an individual
can often be difficult for some
species, such as sod-forming
grasses and clonal shrubs. The
LCTA belt transect is a form of a
density quadrat for estimating
density of woody species.
Curtis and McIntosh 1950.
Simple to obtain, rapid and objective. Nested
frequency is an effective technique to estimate
frequency for a number of species. Reduces
surveyor bias compared to other techniques.
Frequency is dependent on the spatial
distribution of the species and plant size.
Desire frequency estimates between 20% 80% within a given quadrat size to best detect
changes.
Provides information on the distribution of the
species.
Owensby 1973.
Nested Frequency
To determine frame size, a small area (i.e. 10
x 10 cm) is designated and all species present
are listed. The area is progressively increased
to twice its original size, 4 times, 8 times, etc.
and with each enlargement any new species
encountered are listed. Quadrats are located
side by side, with a smaller quadrat located
within a large quadrat. Frequency is the
percentage of quadrats in which a species is
recorded. The optimum frame size for each
species should provide a frequency value
between 0.2 and 0.8.
Frequency - used to determine vegetation
condition and trend.
Winward and Martinez 1983; Curtis and
McIntosh 1950.
102
Table 4-1. Methods matrix cont.
Forage/Biomass Clipping or
Harvest Method
Densiometer
Photographic Monitoring
Description
A concave or convex spherical
densiometer is held at elbow height. A
grid is etched on the surface of the
spherical densiometer and the grid
intersections (points) are tallied where
the canopy is open. Four estimates at
the cardinal directions are averaged for
each sampling point.
A random transect location and azimuth is established. At
a random starting point and at specified intervals, record
weights by clipping and weighing all vegetative matter for
each species occurring in the quadrat. Bag samples and
save for air-dry weighing. Oven-dry samples at 60 °C for
24 hrs. to determine air-dry weight. Pilot studies are
necessary to determine the number of samples and the
number and size of quadrats to comprise a sample.
Intended
Application
Canopy cover (usually for forest)
Permanent locations are described and marked. Identical
photographic scenes are taken over time. A 28mm or 35mm
lens is typically used.
Photo points (qualitative) are general view photographs taken
from a permanent reference point to portray dominant
vegetation and site conditions. Reference points are noted and
marked to ensure replication over time.
Photo plots (qualitative or quantitative) are close-up pictures
designed to show soil surface and vegetation characteristics.
2
Frame size is typically 1m . These are analyzed using a slide
projector or using digital means.
Visual estimates of changes in vegetation cover, structure,
and soil surface conditions. Useful for documenting changes
over time
Effective visual tools synthesize site information. Inexpensive,
repeatable. Should be coupled with field observations or
quantitative measurements when possible.
Photo plots are more time intensive than photo points, and
require analysis in the office.
May provide continuity with historic data, provides data
used in assigning “range condition”,
Other Potential
Uses
Strengths
Most effective in stands of trees taller
than 10m.
Limitations
Weather conditions may impact the
accuracy of the cover estimates, such
as bright sun reflecting in the mirror
and wind moving the overstory foliage.
Additional
Comments
Tend to over estimate canopy cover if
the understory vegetation is greater
than 1m.
Lemmon 1956, Vora 1988.
References
Important to conduct photo monitoring during the same
season(s) each year.
Bureau of Land Management 1996, Borman 1995, Magill 1989
Measuring peak standing crop (above-ground annual
production) of each plant species. Best suited for
grasslands and desert shrublands
Species composition by weight
Seasonal and annual fluctuations in climate can influence
plant biomass. Sampling can be time consuming. May be
difficult to separate current year growth from previous year.
Large numbers of quadrats may be necessary to detect
changes.
Requires drying oven and scales. Permanent transects or
quadrats are not recommended due to the destructive
nature of sampling.
Bonham 1989, Bureau of Land Management 1996
103
Table 4-1. Methods matrix cont.
Prism or Point Sampling
Description
Intended
Application
Other Potential
Uses
Strengths
Limitations
Additional
Comments
References
The surveyor pivots a glass prism or
other angle device 360° over the
sampling point. All tree stems that are
not completely offset when viewed are
tallied. The diameter at breast height
(DBH) of each tree is often measured.
The number of “in” trees is multiplied by
the basal area factor (BAF) of the device
to estimate basal area per acre. Tree
density by size class can also be
calculated.
Basal areas, volumes or numbers of
trees per unit area can be computed
from the tallied trees.
Ability to estimate timber parameters for
large area. Cost-efficient and precise
estimate of volumes.
Sighting difficulties in dense stands;
angle gauge would be more suitable.
Does not provide a good estimate of
stand structure, such as regeneration
(trees/ac).
Also referred to as probability
proportional to size (PPS) sampling.
Larger trees have a higher probability of
selection. Requires training in selecting
the appropriate prism and correcting for
slope.
Avery and Burkhart 1995.
Visual Obstruction (Robel Pole) and Cover
Board Methods
Visual Obstruction: Establish the number of vertical
cover classes and height limits based on objectives.
Select random points along a transect. Two obstruction
measurements are made at each point by determining
the highest band totally or partially visible and recording
the height.
Cover Board: A profile or density board is used to
estimate the vertical area of a board covered by
vegetation from a specified distance away. Transects
o
are often used. Four measurements (offset 90 ) are
recorded at each observation point.
Vertical cover, vegetation structure - as index of wildlife
cover or concealment
Visual Obstruction - most effective in upland and
riparian areas where perennial grasses, forbs, and
shrubs <1 m tall dominate.
Cover Board - applicable to wide variety of vegetation
types, especially those where significant changes are
anticipated (i.e., woody riparian vegetation).
Visual obstruction (Robel pole) - estimation of
production.
Measurements are simple, quick and accurate. Training
required in laying out transects and determining cover
classes/percent cover.
Visual obstruction: infrequent application in a wide
variety of rangeland ecosystems.
Bureau of Land Management 1996, Mitchell and
Hughes 1985, Nudds 1997.
104
Methods described in Section 4.2 include methods for measuring frequency, cover, density,
production/biomass, methods developed for forest measurements, and photographic monitoring.
4.2.1
Frequency Methods
4.2.1.1
General description
Quadrat frequency and nested quadrat frequency are the most common frequency methods. Quadrats
may be located using measured distances or by pacing. Plotless methods such as the step point and
point intercept were developed primarily for estimating cover, but are sometimes used to estimate
species frequency and composition. Cover or density quadrat data collected by species is easily
converted to frequency data because only presence/absence information is required at each sample
location. Frequency frames sometimes incorporate cover estimation using the corners or tine
extensions of the frame as point intercepts for basal, canopy and ground cover. Vegetation attributes
measured using this method include frequency, basal and foliar cover, and regeneration information
(if seedling information is recorded separately.
Quadrats are typically placed along transects, either at regular intervals or by pacing. Within a study
area, transects can be placed systematically with a random start, or located randomly.
4.2.1.2
Applicability
Frequency is used primarily in grasslands, shrub-grasslands, and for measuring the herbaceous layers
of forest and woodland environments.
4.2.1.3
Advantages and Limitations
Frequency is simple to measure, rapid, and objective. Surveyor bias is minimized because only
presence-absence data of rooted plants is recorded. The major sources of observer error are quadrat
placement, species identification, and boundary errors. These non-sampling (i.e., observer) errors can
be minimized through training of field personnel.
Frequency results are a function of quadrat size and shape. For example, an increase in frame size
may result in different frequency values, especially for species with intermediate frequency values
(Mueller-Dombois and Ellenberg 1974). Because of the patchy or aggregated distribution of many
plant species, a rectangular frame may produce different frequencies than a square frame of the same
area. Therefore, once a frame size and configuration are used for a particular community or series of
communities, comparisons are only meaningful among communities or over time if the same size and
shape of frame is used.
Frequency is related to density, but only where the distribution of plants is regular or random. For
example, an abundant species may have low frequency values because individuals are concentrated in
a small area, whereas the same population would have high frequency values if individuals were
distributed evenly across the site. Frequency is therefore a better indicator of dispersion or
distribution than density for commonly occurring species.
Frequency data can be obtained using several different-sized quadrats in the form of a nested frame.
When a species is recorded within a plot (or frame), it simultaneously occurs in all of the larger plots.
This increases the efficiency of data collection, especially for common species. The use of nested
105
plots facilitates having proper plot sized for different species, i.e., where frequency values fall
optimally between 20% and 80%. If only one frequency frame size is used, it is likely that many
species will fall outside optimum frequency values. It may not be possible to find a nested frequency
frame design that produces desired frequency values for all species of interest. For example, rare
species may have frequencies of <2% even in the largest nested frame.
4.2.1.4
•
•
•
•
•
•
•
4.2.1.5
Equipment
Equipment for establishing photo point
Plot location and site description form
Frequency or nested frequency form
Frequency frame
Materials for establishing and marking permanent transect (metal stakes or other markers,
compass, hammer, random azimuths)
Metric measuring tape – length depends on transect length (25 m, 50 m, 100 m, etc.)
List of plant species codes
Training
Plant species identification is the most important issue. Observers must understand how to judge
whether a plant is inside or outside a frame boundary. If cover data is collected, observers must be
familiar with the cover categories and how to read cover information using the sharpened tines
attached to the frame.
4.2.1.6
Establishing Studies and Plot Layout
Collect pilot data from the area (single community, across multiple communities) to determine the
number of samples (e.g., transects) and size and shape of quadrats required for a statistically valid
sample. If total species composition is desired, develop a species-area curve as well (see Chapter 3).
The quadrat size will depend on the characteristics of the vegetation sampled. Nested frequency is
often used when it is desirable to include most of the species present in the sampling design. If
frequencies are less than 10% or greater than 90%, it may be difficult to detect a change in those
species during subsequent sampling. For example, if the frequency of a common species is 95% in the
smallest frame size, then there is little margin for detecting a positive trend in that species in the
future.
Use the same frame size and shape throughout the life of the study. If the frequency for a species
approaches 0% or 100%, a different frame size should be used for that species; nested frames help
resolve this problem. Nested frames consist of two or more frame sizes.
Plots can be established using baseline, macroplot, or linear study designs. Because locating random
points can be time-consuming, locating the quadrats along line transects or baselines is often
preferred. Based on frequencies of common species, estimated variance, and the cost of sampling,
Bonham (1989) estimated than 25 plots randomly located on 25 randomly located transects should
give satisfactory results where the plant community is homogeneous. This estimate was based on
pooling the frequencies for all transects vs. analyzing the data by transect (see Section 4.2.1.8).
A measuring tape is stretched between the base and end-point stakes. The tape may be attached to
intermediate stakes (e.g., 25 m, 50 m, 75 m) to ensure the tape remains straight and the same line is
106
measured every time the plot is measured. The baseline technique for plot layout is often used where
the boundary of the study area or community is well defined. The linear technique is sometimes used
where larger scale studies are performed in communities spread across a larger landscape. If the
baseline technique is used, transects should be placed far enough apart that there is no overlap among
quadrats from adjacent transects.
4.2.1.7
Sampling Process
A transect is read by placing the frame against the transect at a specified starting point. The same part
of the frame should be aligned with the observation point every time (e.g., left rear corner, right rear
corner, etc.). Place the frame at specified intervals along the transect until the end of the transect is
reached or the predetermined number of frames has been recorded. Pacing can be used to measure the
interval between quadrats.
Data is collected for all species. A tally is made by species for the number of frames where the
species occurred. Woody species seedling frequencies are recorded separately. When using nested
frequency, the smallest frame is examined first and a tally is made. If a species is recorded in the
smallest frame, then it is automatically present in all larger frames. Field data sheets for nested
frequency are designed to record the smallest frame size where a species occurred, using codes for the
different frame sizes.
4.2.1.8
Data Summary and Analysis
Frequency data can be summarized by individual transect or pooled for all transects. For each frame
size, the frequency value for each species is calculated by dividing the number of quadrats occupied
by the total number of quadrats surveyed. This provides a frequency value for each species by frame
size for each transect. To calculate the total frequency for all transects, sum all occurrences by frame
size for all of the transects, then divide this number by the total number of quadrats sampled on all
transects. Multiplying by 100 converts frequency to a percentage.
Summary data for individual transects are sometimes treated as normally distributed data and area
analyzed using summary statistics (mean, standard deviation, etc.). Pooled frequency proportions are
analyzed according to the binomial distribution.
To determine if frequency of a species changes significantly over time:
transect – t test or paired t-test, ANOVA or repeated measures ANOVA
pooled – Chi square, McNemar’s test, Cochran’s Q
To determine if frequency of a species is different in community A or treatment A vs. B.
transect – t-test, ANOVA
pooled – chi square, binomial confidence interval
107
4.2.2
Cover Methods
4.2.2.1
Line Interception
4.2.2.1.1
General Description
Line interception is sometimes referred to as the line transect method. A measuring tape is stretched
between two stakes or points. The intercept distance is recorded by species for each plant that
intercepts the line by noting the starting and ending point of the interception, usually to the nearest 10
cm. The line is usually placed on or near the ground. The accumulated length for any species divided
by the length of the transect (tape) multiplied by 100 is expressed as percent cover for that species.
Cover of grasses and forbs can be measured on the line at ground level (i.e., basal area) or measured
at crowns, while half shrubs, shrubs, and trees are measured at the intercept of the perimeter of the
crown. Canfield (1941) describes the use of line interception to estimate cover of grassland and shrub
communities. Where plant crowns overlap vertically, cover can be measured for each height class, as
defined by the observer. Line interception is sometimes used in conjunction with point intercept
methods along a line transect. The length of the tape can be of any length (e.g., 30m, 50m, 100m,
200m), and depends on the variation and abundance of species of interest in the sample area.
4.2.2.1.2
Applicability
This method is designed for measuring grasses and herbaceous plants, shrubs, and trees that have a
solid crown cover or a relatively large basal area (Bonham 1989). It is ideally suited for semiarid
bunchgrass-shrub types, but can also be applied in woodland or savanna types where tree crown
edges are distinct and easily projected to the ground. The method can be applied to plants with
relatively solid (i.e., continuous) crown cover or relatively large basal areas (Mueller-Dombois and
Ellenberg 1974).
4.2.2.1.3
Advantages and Limitations
Line intercept can be adapted to a variety of types and densities of vegetation but it is best suited for
measuring shrub cover. For example, it can be used to estimate the basal cover of grasses or the
canopy cover of shrubs by species or for all species combined in an efficient and repeatable way.
Plant canopy outlines taller than about 15 m are difficult to assess accurately, and may require
sighting devices (Mueller-Dombois and Ellenberg 1974). The accuracy of the method therefore
depends largely on the accuracy of the vertical projection and the determination of the boundaries of
the crown itself (see Section 4.2.2.1.7). Grasses and forbs with sparse crowns and small basal areas,
such as many annual grasses, as well as very dense grasslands, are difficult to measure using this
method. Groundcover components such as litter and small rocks cannot be measured using point
interception.
4.2.2.1.4
•
•
•
•
•
Equipment
Equipment for establishing photo point
Plot location and site description form
Line intercept form
Materials for establishing and marking permanent transect (metal stakes or other markers,
compass, hammer, random azimuths)
Metric measuring tape – length depends on transect length (25 m, 50 m, 100 m, etc.)
108
•
4.2.2.1.5
List of plant species codes
Training
Training needs are minimal. Training should address plot establishment and layout, how to judge
canopy continuity and make measurements for the different plant species, life forms, and groundcover
categories (if collected), and species identification.
4.2.2.1.6
Establishing Studies/Plots
A transect is established using a random starting point and azimuth. A tape is stretched between the
stakes as close to the ground as possible. Data should be collected on several pilot plots within
different ecological types to determine the length and number of transects necessary to meet study
objectives for the key species. Longer transects will be required where vegetation is sparse.
Permanent transects require permanent beginning and end stakes, and are recommended for change
detection/trend analysis.
4.2.2.1.7
Sampling Process
From the starting point, walk down the tape, measuring the intercept length of each plant or species.
Determining the continuity of canopy cover is an important consideration to ensure repeatability and
precision in data collection. The most practical unit of measurement id the whole plant, including
gaps in the canopy. If a plant has branches that reach across the line with gaps in between, the gaps
can be excluded from the measurement for greater precision. However, this can be difficult to assess.
Small gaps in the canopy that are probably not ecologically significant can be ignored. The
assessment of crown cover is essentially similar to the approached used by Daubenmire (1959), who
envisioned the plant crown as extending above the area of influence of the plant. If the canopy is
significantly broken with gaps that exceed rules of thumb for “solid” crown cover, the cover of
individual components should be summed and recorded as one entry. This also facilitates the
counting of individuals if estimates of density are desired (Bonham 1989). If written guidelines are
adhered to when making measurements, precision among observers should be relatively high.
Canfield (1941) recommended a 15 m line for areas with a cover of 5-15% (the species of interest)
and a 30 m line where the cover is less than 5%. Canfield suggested that the optimum length of a
transect is one that can be measured by a two-person team in approximately 15 minutes.
4.2.2.1.8
Data Analysis
The sampling unit is the transect. Percent cover for each species is the sum of all interception lengths
for that species divided by the length of the transect and multiplied by 100. Total vegetation cover is
the sum of the cover for all species, and may be greater than 100 percent if overlapping plants are
measured. Relative composition or cover is calculated by dividing the percent cover of each species
by the total vegetation cover.
Data analysis guidelines follow those for permanent (non-independent) and temporary (independent)
plots. For permanent plots with no more than two years of data, use a paired t test or Wilcoxon rank
test. For more than two years of data, use a repeated measures ANOVA.
For temporary plots a two-sample t test or MannWhitney U can be used for comparing 2 samples;
ANOVA or Kruskal Wallis is appropriate for comparisons involving more than 2 samples or years.
109
4.2.2.2
4.2.2.2.1
Point Intercept
General Description
The point intercept method consists of using a pin, series of pins connected in a frame, or a sighting
device along a transect to estimate plant cover, usually by species. Estimates of ground cover (litter,
dead wood, bare ground, rock, etc.) and plant basal cover can be made by recording the bottom
intercept at ground level. Pin frames contain a fixed number of pins, often 10, separated by a fixed
distance (e.g., 10cm) and placed on a frame. The point intercept method is sometimes referred to as
the point-quadrat method because the point has dimensions associated with area (i.e., usually 1-2 mm
in diameter). The premise of this method is that if an infinite number of points are placed in a twodimensional area, exact cover of a plant can be determined by counting the number of points
contacting the plant (Bonham 1989). If periodic remeasurement of the same vegetation is the
objective, then transects with evenly-spaced points are recommended (Mueller-Dombois and
Ellenberg 1974). The method provides percent ground cover by composition, the percent cover that
each species contributes to the area, the relative frequency of species, and percent composition
(number of points for a species divided by the number of points for all species). Where ocular
samplers are used, only the topmost intercept can be recorded. This is also true for optical samplers
that are used for vertical projections; only presence or absence of canopy (i.e., the lowest
interception) can be recorded.
4.2.2.2.2
Applicability
The point intercept principle has wide applicability for measuring both low herbaceous cover and tree
cover. Most sighting devices and pins/point frames are designed so the observed looks downward on
the vegetation from above or from the side. Point sampling using pins (narrow metal rods) is best
suited to herbaceous and low shrub communities such as grasslands, prairie, and desert shrublands.
Sampling devices that allow upward sighting can be used to estimate the canopy cover of large shrubs
and trees. A telescopic sighting with crosshairs has been used to sample forest vegetation and tree
canopy cover. A telescoping range pole has also been used to estimate canopy intercepts by height for
heights up to 8.5m in forested communities (Tazik et al. 1992). Because the diameter of the pin can
significantly affect the amount of cover recorded, the results are only relative to surveys done with
pins of the same size (Goodall 1952). Estimates of aerial cover from point-intercept should not be
compared with other cover methods which estimate cover based on the area of influence of the
canopy (i.e. estimating cover of the perimeter including small gaps), and thus ignore openings in the
canopy.
4.2.2.2.3
Advantages and Limitations
Point interception measurements are highly repeatable and produce more precise estimates than
ocular estimation methods such as cover quadrats. Point intercept is probably the best method for
measuring ground cover and dominant species cover (Bureau of Land Management 1996). Relative to
other cover methods of equal reliability and objectivity, point intercept has been shown to be more
efficient while attaining the same level of precision (Crocker and Tiver 1948, Floyd and Anderson
1987).
When pin frames are used a much larger number of points must be read to achieve the same level of
precision that is achieved by points that are more widely spaced. A general rule of thumb is that
single pin measurements require one-third as many points as required when group of pins are used
(Bonham 1989). This is due to the fact that where plants are clumped or large, the same plant is
110
intercepted using the point frame than when using more widely-spaced single points. This can result
in overestimation of cover for those plants.
One limitations of point interception is the difficulty in recording the presence and cover of
uncommon and minor species without sampling an inordinate number of points. Less common plants
are often overlooked because a small amount of area (three-dimensional space) is covered by point
intercept surveys. For this reason, point intercept data is inadequate for documenting species richness
or diversity. Windy conditions can make surveys difficult because of the movement of plant parts.
4.2.2.2.4
•
•
•
•
•
•
•
4.2.2.2.5
Equipment
Plot location and site description form
Cover data form – aerial intercepts and ground cover for each point surveyed
Single pin (often 100cm long), pin frame, or sighting device
Tripod for mounting sighting device or pin (recommended)
Materials for establishing and marking permanent transect (metal stakes or other markers,
compass, hammer, random azimuths)
Metric measuring tape – length depends on transect length (25 m, 50 m, 100 m, etc.)
List of plant species codes
Training
Point intercept requires a minimal amount of training. Protocols for interception interpretation should
be established and demonstrations performed. Understanding of ground cover categories and their
correct application is essential.
4.2.2.2.6
Establishing Studies/Plots
Study sites should be selected within a single plant community or ecological type. Transect starting
points and azimuths should be randomly selected within the study area. For permanent transects, the
starting and end points should be marked using permanent markers. Photographs from the beginning
of the transect are recommended to help illustrate documented findings. Points are most often
measured along a transect at regular intervals. When using systematically-located single points, the
distance between pins depends on plant distribution patterns, the distance between plants, and the size
of individual plants. Pilot data should be collected to determine the number of points necessary to
collect a statistically valid sample.
4.2.2.2.7
Sampling Process
The first sampling location is determined randomly; the same distance to the first sampling point can
be used for all transects sampled. If permanent transects are used, the same starting point and
sampling interval should be used each time sampling occurs. When using ocular samplers, the
protocol is very straightforward. Looking down toward the ground, the cover category or species in
the crosshairs is recorded (groundcover will be seen if it is obscured by any vegetation above it). The
same procedure is followed when using ocular devices to estimate canopy coverage (presenceabsence only). In pin sampling, the pin is lowered vertically, preferable using a guide. When using
pins (or telescoping poles) a point can be read in several ways, depending on the desired information
and the amount of time available at each sampling location. For example, will intercepts be recorded
using height categories/canopy layers? Will all interceptions be recorded regardless of height? If total
canopy cover is the minimum requirement, then the presence/absence of cover is recorded regardless
111
of species or height. For some devices, it may be necessary to move upper layers out of the way in
order to record hits in multiple height categories.
If permanent transects (paired samples) are used, transect starting and end points should be
permanently marked. In the sampling periods that follow, a tape is stretched between the permanent
markers and the same locations on the transect are sampled. If the data are to be analyzed as
independent samples (non-permanent transects), then transect locations do not need to be permanently
marked. Point sampling locations can be determined through pacing or by laying out a tape between
two temporary stakes and making measurements at regular intervals. It is important, especially when
using independent samples, to ensure that placements of the points are objective and unbiased.
Looking at the horizon when placing the point (independent samples) can help to avoid bias in point
placement (Bureau of Land Management 1996).
4.2.2.2.8
Data Analysis
Data can be summarized as percent cover by species, relative cover or composition (percent cover of
a species divided by the total cover of all species, or density/relative density of intercepts by species,
if multiple intercepts are recorded for each location.
Data analysis procedures depend on whether the transects are permanent or temporary.
For permanent transects, the transect or the point frame is considered the sampling unit: To test for
differences or changes in mean cover (i.e. total cover, cover by species, etc.) over time, use a paired t
test or a nonparametric test. Repeated measures analysis of variance is used to test for significant
changes for more than two sampling periods. A two-sample t test (two groups) or analysis of variance
(> two groups) can be used to examine differences among groups of plots at a given point in time.
For non-permanent transects, the sampling unit can be either the transect, the pin frame, or individual
points. If individual points are used, the cover of each species in the study area is calculated as the
sum. of all the hits for that species divided by the total number of points surveyed, multiplied by 100.
If frames are used, cover by species is calculated for each frame, and the values are subsequently
averaged to get a mean value for the study area. To test for differences in average cover between
sampling periods when pins and pin frames are used, use a two-sample t test or the Mann Whitney U
test (comparing two years of data), or the ANOVA or Kruskall-Wallis test (three or more years of
data).
Total vegetation cover is the sum of all coverages for all species. If multiple hits for species are
recorded at each point, the total vegetation coverage may exceed 100%.
4.2.2.2.9
Step-Point Method – A Variation of Point Intercept
The step point method is a variation of point interception. It involves pacing transects and recording
ground, basal, and aerial plant cover at fixed intervals by placing a recording pin in a notch in the toe
of the boot and lowering the pin to the ground. This method is relatively simple and easy to use. It is
suitable for measuring major characteristics of ground and vegetation cover in relatively large areas,
and allows collection of a large number of samples within a relatively short time. A limitation of the
step-point method is that there can be extreme variation in the data collected and among examiners,
especially when sample sizes are small. Data analysis procedures are the same as those for point
intercept.
112
4.2.2.3
Visual Estimates in Quadrats/Plots
4.2.2.3.1
General Description
For measuring plant cover, point intercept and line interception involve little or no judgment by the
observer, and are therefore regarded as quantitative cover methods, the results being generally
reproducible by other observers. Ocular cover estimation methods are considered semi-quantitative
methods because the method of data collection has qualitative aspects yet the data can be used for
mathematical computations and are suitable for statistical analyses (Bonham 1989).
In this method, ranges of cover values are specified and each category is assigned a value. Methods
that are altered to allow estimation of plant cover to the nearest 1-2% or even 5% risk having
problems with observer bias and low repeatability both among observers and with the same observer
over time.
4.2.2.3.2
Advantages and Limitations
Cover estimation using quadrats is relatively simple and rapid. Cover estimation can become
laborious when plant diversity is high, thus reducing the efficiency of the method. More species are
recorded than when sampling using linear techniques because more area is being sampled. Therefore,
cover quadrats are advantageous if species richness or diversity is one of the monitoring objectives.
By using broad cover classes, observer error in assigning coverage classes is minimized. As the
number of cover classes used increases (i.e., resolution of the estimated cover increases) the
repeatability of the method is reduced (observer error increases), especially in species-rich herbaceous
communities (Bonham 1989). Because the midpoints of some cover classes are far apart, summarized
results may differ significantly if different cover classes are assigned to similar cover conditions.
When broad cover classes are used, this approach is poorly suited to change detection, as cover values
will have to change by 1 to 2 cover classes before a significant change may be detected.
4.2.2.3.3
•
•
•
•
•
•
4.2.2.3.4
Equipment
Plot location and site description form
Cover class form by species – includes an entry for each species for each quadrat.
Metal stakes and hammer for establishing transect
Tape of desired length (e.g., 25, 50, 100m)
Sampling quadrat – size varies
Compass
Training
Species identification, training in establishing transects objectively, and consistent estimation of
coverage classes are essential abilities.
4.2.2.3.5
Establishing Studies/Plots
Transects with a random starting point and azimuth is the most commonly used approach for
establishing cover quadrats. Pilot sampling may be necessary to determine the number of quadrats per
transect and the number of transects to represent a given ecological community or site. A tape is
stretched in a straight line between metal stakes. The number of frames can vary; 50 frames placed
adjacent to the transect and spaced at a fixed distance are commonly sampled.
113
4.2.2.3.6
Sampling Process
4.2.2.3.6.1 Cover Classes
Cover estimation in plots is typically based on a cover class system. The methods which use broad
classes are less susceptible to consistent human error. A higher resolution of scale toward the lower
end allows better estimation of less abundant species (Bonham 1989). The different systems are
similar in most respects. The systems presented here use from 6 to 12 coverage classes (Table 4-2).
Table 4-2. Published cover class systems for ranges of cover.
# of
Classes
Daubenmire
Braun- Blanquet
# of plants
1
2
0-5%
3
4
5
6
7
8
9
10
11
12
5-25%
25-50%
50-75%
75-95%
95-100%
sparse or v.
sparse (+)
plentiful
v. numerous
any number
any number
any number
area
occupied by
species
very small
1-5%
(small)
or 5-25%
25-50%
50-75%
75-100%
Greater Yellowstone
Area Noxious Weeds
Cover Classes
Bailey
and
Poulton
Domin-Krajina
# of plants
EcoData
0-1 (tr.)
0-1 (tr.)
solitary (+)
area
occupied
by species
insignif.
1-5%
1-5%
seldom
insignif.
1-5% (c)
5-25%
25-100%
5-25%
25-50%
50-75%
75-95%
95-100%
v. scattered
scattered
any number
any number
any number
any number
any number
any number
any number
<1%
1-5%
5-10%
10-25%
25-33%
33-50%
50-75%
75-99%
100%
5-15%
15-25%
25-35%
35-45%
45-55%
55-65%
65-75%
75-85%
85-95%
95-100%
0-1% (+)
The Daubenmire (1959) and Braun-Blanquet (1965) cover scales are the most widely used among
those presented in Table 4-2. However, the other scales presented are applicable, straightforward, and
advantageous in some cases. The Daubenmire cover scale was modified by Bailey and Poulton (1968)
by splitting the 0-5% class into two classes. The Domin-Krajina (Krajina 1933) and Ecodata (Jensen
et al. 1994) scales both have relatively high resolution classes on the low end of the cover scale. The
EcoData scale has the highest resolution in the middle and upper level cover classes compared to
other scales, and as a result may have lower repeatability over time in those classes. The Greater
Yellowstone Area Noxious Weed class system was developed specifically to support monitoring and
mapping of invasive plants (NAMWA 2002). The structure of its classes reflects the management
implications of different levels of infestation.
The Daubenmire scale is used for estimating only cover. The method is applicable to a wide range of
herbaceous and shrubby vegetation, but may become problematic when vegetation is greater than 1 m
in height. Daubenmire found that a frame size of 0.1 m2 plot with 20 cm X 50 cm inside dimensions
was very satisfactory for cover estimation. He found that as frame size increased beyond these
dimensions, it became difficult for the observer to see the entire area at once, leading to significant
errors in cover estimation. The metal frame is painted to divide the area of the frame into quarters and
subsequently into smaller divisions to create visual reference markers equal to 5, 25, 50, 75, and 95%
of the frame. Tests by Daubenmire in Washington and Oregon showed that at least 40 plots (20 cm x
50 cm) placed approximately one plot diameter apart provided satisfactory results for most species
(Daubenmire 1959). A species-area curve and sequential pilot sampling (i.e., constructing a running
114
mean and standard deviation for each species) can be used to determine optimum sample sizes for a
particular location.
The Braun-Blanquet scale gives a combined estimate of abundance and cover. The Braun-Blanquet
scale, which can be used with a number of sampling designs and quadrat sizes, is not synonymous
with the established Braun-Blanquet method (Braun-Blanquet 1965). The Braun-Blanquet method is a
European-designed system for describing and classifying plant communities and ecosystems. The
process involved delineating a “uniform” plant community, referred to as a “stand”, and then making
a detailed description of it, referred to as a relevé. The average height and the percentage cover of
each layer of vegetation in the stand is recorded. The percentage cover is taken as an estimate of the
space covered by a vertical projection of all the above-ground parts of the plant. Because most stands
are too large to estimate coverage accurately for every species, a small and representative area in the
middle of the stand is subjectively chosen
The size of the sampling unit used to construct the relevé is determined using the minimal area
concept. An initial quadrat size (1 m2 in grasslands and 4-6m2 in woodlands) is surveyed for all plant
species present. The size of the quadrat is progressively doubled and surveyed for new species until
the number of new species added per increase of area becomes small. Relevé size varies by vegetation
community, often ranging from 5 m x 10m to 20 m x 20 m or larger. Each species within the quadrat
is given a cover-abundance and sociability rating. Any additional species which occur outside the
sample area and within the stand are noted and an “x” is used as the cover abundance code. This
coverage information is considered a substitute for a description of the whole stand. Additional soil,
physiographic, and structural information is collected and noted. The species lists and coverage
ratings are organized by plot into matrixes which are used to examine associations and produce
descriptions for each association. The true Braun-Blanquet method described above uses a subjective
(i.e., biased) approach to selecting a quadrat to represent the stand. This cannot be considered
sampling because the site selection is not random. Moreover, as quadrat size becomes larger and
numerous species are present, ocular estimation is difficult to do consistently. The decreased
precision associated with relatively few cover classes makes it difficult to detect changes in cover
when using this approach. However, this is an excellent method for describing communities,
developing species lists, and distinguishing among different plant associations (Shimwell 1971,
Mueller-Dombois and Ellenberg 1974).
4.2.2.3.6.2 Collecting Data
The frame is placed along the tape at specified intervals. The canopy cover of each species is
estimated and assigned the appropriate cover class for each frame. Total vegetation cover, bare
ground, litter ground cover, etc. can also be estimated. Each species is considered separately. An
imaginary line is drawn around the foliage of each species and projected to the ground. This is
sometimes referred to as the area of influence of the plant or species, and is considered its canopy
coverage. Canopy estimates are for all plants extending into the quadrat regardless of whether they
are rooted inside or outside the quadrat. Overlapping canopy cover species is considered in the
estimate of canopy coverage; total cover can therefore exceed 100 percent. Where applicable, small
divisions should be marked on the edges of the frame to assist the observer in estimating coverage.
4.2.2.3.7
Data Analysis
The midpoints of cover classes can be used for data summary and statistical analysis. Using class
midpoints assumes that the actual cover values are symmetrically distributed around each midpoint. If
this assumption is violated, results will not be representative of field conditions. Tests can be done to
assess changes in cover of particular species or ground cover categories. Using the same rationale
115
provided for frequency quadrats (4.2.1) the sampling unit can be either the quadrat, if the quadrats are
far enough apart to be considered independent, or the transect. If the quadrats are independent and the
transects are permanent, either a paired t-test or Wilcoxon signed rank test is appropriate. For three or
more years of data, the repeated measures ANOVA is appropriate.
4.2.2.4
Points vs. Lines vs. Plots
One of the principle challenges in cover estimation is getting satisfactory estimates for the minor
species. The problem of getting an adequate sample (high precision and low standard deviation)
increases geometrically as plot size becomes smaller. Reducing the plot size to a few mm in diameter
(point interception/point frequency) typifies the problem. While Daubenmire (1959) found that
approximately 40 to 50 plots were adequate to represent conditions, Whitman and Siggeirsson (1954)
found that 1,400 points or more are required to reduce sampling error below 10% for even the major
taxa of North Dakota grasslands. A similar study of grasslands in southern Alberta by Johnston
(1956) found 1,200-4,320 points necessary to estimate the basal area of major species within 10% of
the true mean. Point intercept and line transect methods are sometimes combined for estimating plant
cover in short vegetation types.
Floyd and Anderson (1987) found that point interception achieved about the same degree of precision
as line interception in one third less sample time. They concluded that in general, point interception is
the most efficient compared with line interception and cover quadrat estimates where estimates for
dominant and common species in a community are needed. Similar results for point vs. line
efficiencies were reported by Heady et al. (1959). The length of the transect and the number and
spacing between points are important considerations. It is usually more efficient to record more
transects with fewer points per transect than fewer transect with more points per transect (Floyd and
Anderson 1987). The length of transect required can be greatly influenced by the uniformity of
vegetation plant size, and morphology. The decision regarding this sampling tradeoff depends on the
difference in time required to set up a transect (or relocate a permanent transect) compared to the time
required to read additional points (Bonham 1989).
If plant cover is defined as the vertical projection of the outlines of undisturbed foliage, then line
interception data should be about the same as cover quadrat data, providing both samples are
adequate. Based on studies of line interception requirements by Canfield (1941) and Proudfoot
(1942), Daubenmire (1959) compared estimates of canopy cover from 40-50 (20 cm x 50 cm ) plots
with cover estimates from 350 m of line interception in a scattered sagebrush shrubland and found
that estimates rarely exceeded 2.6% absolute cover. Hanley (1978) compared line interception and
quadrat estimation methods (Daubenmire cover scale) for four densities of big sagebrush in the Great
Basin. His results indicated that the methods provide comparable estimates. Where high levels of
precision and confidence are required, the line interception method was found to be preferable.
However, the Daubenmire frames were preferable where lower levels of confidence and precision are
acceptable. The major advantages of plot techniques over line interception are: (1) ground surface
cover data can be estimated along with vegetation cover, (2) the method can be more efficient if many
lines need to be laid out for line interception, and (3) more life forms (and species) are considered
(Daubenmire 1959). The principal advantage of the line interception technique is that vegetation is
measured directly as opposed to being visually estimated (Canfield 1941).
116
4.2.2.5
Spherical Densiometer (Forest Canopy Measurement)
4.2.2.5.1
General Description
Canopy cover refers to the proportion of an area covered by the vertical projection of plant crowns to
the ground surface. Techniques for estimating forest overstory cover include ocular estimation, the
moosehorn, point intercept, line interception, photographic techniques, and the spherical densiometer.
The spherical densiometer is widely used to estimate canopy cover by counting dots on a mirror grid.
4.2.2.5.2
Applicability
The spherical densiometer is intended for use in forests where understories are sparse or below 1 m
tall. It is most effective where trees are at least 10 m tall.
4.2.2.5.3
Advantages and Limitations
This method is easy to use, rapid, and relatively repeatable. Weather may affect the accuracy of cover
estimates. Effects include sunlight reflecting off the mirror and canopy movement due to wind.
Vora (1988) found that estimates of ponderosa pine canopy cover did not differ significantly between
the spherical densiometer method and ocular estimation methods. He recommends ocular methods
where understory vegetation is tall or clumped as opposed to randomly distributed, or where available
field time limits sample size.
In a comparison of methods for estimating forest canopy cover, Vales and Bunnell (1988) found
significant differences among observers when using the spherical densiometer. Ganey and Block
(1994) found that differences among methods were greater at lower canopy densities, and that
observer estimates using the spherical densiometer became more pronounced as canopy closure
increased. For these reasons it may be preferable to use a single observer when using this method.
4.2.2.5.4
•
•
•
•
4.2.2.5.5
Equipment
Plot location and site description form
Canopy cover field data class form – four entries per sample (observation) point
Metal stakes and hammer for establishing transect
Either a convex or concave spherical densiometer can be used. Both types consist of a
polished chrome mirror 2 ½ inches in diameter and having the curvature of a 6-inch sphere.
The mirrors are mounted in small wooden boxes with hinged lids. A circular spirit level is
mounted to the box. Cross-shaped, circular grids are either etched upon the mirror or
mounted between the mirror and the observer. Densiometers typically have a grid of twenty
four ¼-incg squares. Each square has four equi-spaced dots.
Training
Training in estimating canopy cover is essential. Standardization among observers is important to
avoid non-sampling errors.
117
4.2.2.5.6
Establishing Studies/Plots
Transects are the usual method for arranging observations at points. A random starting point and
azimuth are chosen and the transect is either permanently marked or a temporary plot is established.
Transect length and number of observation points is largely determined by uniformity of the canopy.
A homogeneous canopy will require fewer sample points than one having large or scattered openings.
Pilot sampling is required to determine the minimum number of observations and transects required to
satisfy desired precision and confidence levels. Observations can be made at points arranged in a
number of ways. Periodic measurements along transects are often used; observations are at fixed
intervals. Intervals should be spaced far enough apart to avoid canopy overlap between sample points.
4.2.2.5.7
Sampling Process
At each observation point, the observer holds the densiometer at elbow level, manually leveling the
device using the integral spirit level. The 24 squares with four imaginary equi-spaced dots are
scanned. Dots that fall where open canopy is seen in the mirror grid are counted, subtracted from the
96 possible dots, and divided by 96 to give a proportion mean crown completeness (MCC) above
each point.
4.2.2.5.8
Data Analysis
Cover estimates are analyzed as continuous data. Point data is averaged for the sample (e.g., transect).
4.2.2.6
Cover Board
4.2.2.6.1
General Description
A profile or cover board is used to estimate the vertical area and extent of vegetation cover on a board
from a specified distance. The technique is designed to evaluate changes in vertical cover and
structure over time. Photo plots (close-up) and photo points (general) should be used with this method
to provide additional evidence of changes and to characterize ground cover conditions.
4.2.2.6.2
Applicability
The cover board method is applicable to a wide range of habitat types to evaluate the amount of
screening cover available to wildlife species (Mitchell and Hughes 1995), especially where significant
changes are anticipated.. It has been used to examine relationships between cover and habitat use for
birds in the Great Plains (Jones 1968, Robel et al. 1970, Guthery et al. 1981), rodents (Rosenzweig
and Winakur 1969), and deer (Tanner et al. 1978, Griffith and Youtie 1988).
4.2.2.6.3
Advantages and Limitations
This technique is rapid and easy to duplicate. The size of the board (Bureau of Land Management
1996) and the height of the increments (Guthery et al. 1981) can be modified to meet the purpose of
the study.
4.2.2.6.4
•
•
•
Equipment
Study location and site information forms
Cover board field data sheet
Cover board and 1 m sighting rod
118
•
•
4.2.2.6.5
Metal stakes and hammer
Compass
Training
Training is required to ensure that crews can properly lay out transects, make cover board estimates
consistently, and identify plant species.
4.2.2.6.6
Establishing Studies/Plots
If screening cover is being estimated for only one or a few similar species, transects should be located
in habitat that is typical for those species. Data can be collected using either a linear approach or a
center point approach. The linear or transect approach is used most often and is described here.
Transects can be permanently marked or temporary. Pacing can be used to locate observation points
where plots are temporary.
Sample points are located by following a transect and taking cover readings at intervals (sampling
stations) along the transect. Sampling stations can be randomly or systematically located; systematic
locations are more efficient. To prevent overlap among sampling stations, both transects and sample
stations should be at least 20 m apart. The board is approximately 2 to 2.5 m long and 10 to 30 cm
wide. The board is divided into 0.5 m sections which are painted with alternating white and black or
orange. Mitchell and Hughes (1995) recommend using a 1 m rod to standardize the height of the
observer’s eye for all observations. A bottom spike or hinged arm attached to the back of the board
can be used to anchor the cover board in place so that a single observer can take readings.
4.2.2.6.7
Sampling Process
At each observation point, four readings (i.e., placements) are made using random azimuths. The
profile (cover) board is placed at the first sample point 15 m from the station. For each segment of the
cover board, the cover class is estimated and recorded. Cover classes such as those developed by
Daubenmire (1959) or Bailey and Poulton (1968) are often used (see Section 4.2.2.3.6.1). Either total
cover for the board or cover of each increment (vertical strata) can be estimated. If increments cover
is taken, estimate the increment at ground level first, then proceed to the subsequent strata.
4.2.2.6.8
Data Analysis
The average cover is calculated as a whole and for each vertical layer by using the midpoints for the
cover class system used.
average percent horizontal cover for the stand =
sum of readings
total number of readings
average percent horizontal cover for each height increment =
∑ cover readings for increment
total number of readings for increment
Permanent plots are suggested where trend analysis is the objective.
119
4.2.3
Density Methods
Density, a measure of abundance, is defined as the number of individuals in a given unit of area.
Density is often used to describe the vegetation of plant communities. Several potential problems
associated with density measurements include defining what constitutes an individual plant, selecting
a quadrat size and shape with associated boundary error, and estimating density using plotless
methods (Bonham 1989).
4.2.3.1
Density Quadrats
4.2.3.1.1
General Description
The density of individuals is estimated by counting the number of individuals within plots. Plot sizes
and dimensions vary widely depending on the sampling objectives and other considerations. Before
beginning sampling using density quadrats, the following should be considered: (1) distribution of the
plants, (2) size and shape of the plot, and (3) the number of observations needed to obtain an estimate
that satisfies the sampling requirements.
4.2.3.1.2
Applicability
This method is applicable to a wide variety of vegetation communities and is suitable for counting
trees, shrubs, grasses, and forbs.
4.2.3.1.3
Advantages and Limitations
Advantages
Because density is recorded on a per area basis, comparisons among different sites can be made even
if different quadrat sizes were used.
The density of mature perennial plants is not affected as much by annual environmental variation as
attributes such as cover or production.
Density can provide useful information on seedling abundance, survival, and mortality. However,
overall density may appear unchanged even though significant mortality and regeneration has taken
place. Using stage or age classes can provide more information to help interpret changes in the
population.
Depending on the life form, density sampling can be rapid and repeatable.
120
Limitations
Some difficulties arise when trying to count individuals. Trees and single-stemmed annuals rarely
present problems but other life forms can more problematic. It can be very difficult to impossible to
distinguish individuals among spreading shrubs, stoloniferous or rhizomatous plants, and clonal
shrubs and some trees. Bunch grasses and caespitose or single-stemmed herbs are usually countable,
especially where plant outlines are distinct. In cases where individuals are indistinguishable on a
consistent basis, the decision must be made whether one can count individuals or just parts of
individuals, such as the number of stems or shoots (Mueller-Dombois and Ellenberg 1974). As long
as a definition or rule of thumb is applied consistently, counting plant parts can provide an accurate
count.
Accurately counting individuals at quadrat boundaries is important to minimize errors of accuracy
and precision. Boundary decisions become more common with smaller quadrats because of the
increase in boundary length in relation to the quadrat area. Again, definitions or rules of thumb for
what constitutes an “in” plant should be developed and applied consistently.
In dense populations, sampling can be tedious and time-consuming, thus raining the risk of nonsampling (observer) errors.
A single quadrat shape cannot efficiently and adequately sample all species or life-forms of interest.
Density estimations are sometimes limited to a few species for this reason.
4.2.3.1.4
•
•
•
•
•
4.2.3.1.5
Equipment
Study location and site information form
Density field data form
Measurement equipment will depend on sampling design and plot layout: large quadrats
require tapes that are tied or anchored at the corners; large circular plots may require a center
pin and a single tape or length of cord, small circular or rectangular plots can be constructed
from metal, wood, or other materials.
Metal stakes and hammer for marking corners or transect location (if permanent quadrats or
transects are used)
Compass
Training
Plant identification must be consistent among observers. Where individuals of certain species are
difficult to distinguish, rules of thumb should be developed and applied consistently. The same is true
for making boundary decisions.
4.2.3.1.6
Establishing Studies/Plots
Pilot studies should be conducted to determine the most efficient quadrat size and shape, the number
of quadrats, and the number of samples (observation points or transects) that meet study requirements
(see Chapter 3). As a general rule, long and thin quadrats work better than circles, squares, or short
and wide quadrats (Krebs 1989). This is because more species are included in long, narrow quadrats
because of the tendency for vegetation to be clumped (Bonham 1989). Larger quadrats are generally
recommended for use in sparse vegetation to minimize the number of counts that are very small or
zero.
121
A strip quadrat, sometimes referred to as a “transect” or “belt transect”, is a relatively long, thin
quadrat. The strip usually varies in width between 2 and 10 m and can extend in length from tens of
meters to hundreds of meters. Placing small quadrats at predetermined intervals along a transect is a
modification of the strip transect method. When a strip transect is used, there is no estimate of the
variance of the density for that transect.
4.2.3.1.7
Sampling Process
Quadrats can be established using random locations within the study area, associating quadrats with
randomly placed transects, or by establishing a systematic grid for sampling. For relatively small
study areas, random pairs of coordinates (x, y) can be selected, where the point represents the lower
left corner of the quadrat. Quadrats are placed at the locations of the random coordinates. The number
of individuals (or other counting unit) of each species of interest is counted and the results recorded
on the field data form. Sampling continues in this manner until the required number of quadrats has
been read. Where transects are used to determine plot placement, starting points are determined
randomly and the point is marked. The transect is run out a predetermined length using a random
azimuth. Quadrats are placed along the transect at specified intervals and counts are made for each
quadrat.
4.2.3.1.8
Data Analysis
Densities can be calculated on a quadrat basis (3.5 plants/2 m2) or plot basis (35 plants/20 m2), and
subsequently expanded to the desired area units (e.g., 17,500 plants/hectare). The average density per
quadrat for each size/age class for each species is calculated by dividing the total number of plants
counted in the sample for each size/age class by the number of quadrats in the sample. The process is
done separately for each species, depending on objectives. The estimated total density of the
macroplot (population) is calculated by multiplying the average density per quadrat by the total
number of possible (non-overlapping) quadrats in the macroplot.
Confidence intervals can be constructed for density estimates. The type of comparisons performed
will depend on whether the plots are permanent or temporary.
4.2.3.2
Distance Methods
Distance measurements, or plotless sampling, utilize spacing between plants to estimate density.
Plotless sampling often is used in forest inventories, but also is used were individual plants are
discrete and not clonal. These techniques are "based on the concept that the number of plants/unit
area can be estimated from the average distance between two plants or between a point and a plant"
(Bonham 1989). Distance measurements can save time and may improve accuracy because there is no
boundary effect (Mueller-Dombois and Ellenberg 1974; Greig-Smith 1983; Bonham 1989).
To determine density:
1) Sum all distances in the samples and divide by the total the number of samples to obtain mean
distance:
Mean Distance = Sum of Distances / Number of Samples
2) Square the Mean Distance to calculate Mean Area:
122
Mean Area/Plant = (Mean Distance)2
3) Divide the mean area by the unit area to express the density, in this case hectares:
Density = Mean Area2 / 10000 m2
Density can be expressed as the density of a single species or the relative density of a species to
all species sampled.
4.2.3.2.1
Applicability
Distance measures are best used when plants are randomly distributed. Commonly used during forest
inventories, distance measures can be used to estimate shrub densities, forbs, and bunch grasses
(Laycock 1965). When used with bunchgrasses in Idaho, Laycock (1965) noted that plants must be
distinct and that individual species must be sampled separately, because grasses tend to have several
species grouped together.
4.2.3.2.2
•
•
•
•
Equipment
Metric measuring tape
Compass
List of plant codes
Caliper or meter stick for measuring diameter for Diameter at Breast Height (DBH) (optional)
4.2.3.2.2.1 Closest Individual
Closest individual method measures from a sampling point to the nearest plant of the life form being
measured (i.e., shrubs or trees, or shrubs and trees). The method is continued from the next random
sampling point to the next nearest plant (Figure 4-2A). To be truly random would require all plants to
be labeled and then a group of plants chosen as starting points (Mueller-Dombois and Ellenberg
1974), which would negate the need for sampling. Instead criteria are set prior to sampling for how
each sampling point is to be located (e.g., starting points are identified at 10-m intervals along a
transect. Approximately 50 samples/site are necessary.
4.2.3.2.2.2 Nearest Neighbor
The nearest neighbor method begins by measuring the distance from the sampling point to the nearest
plant and then measures between the first individual to its nearest neighbor (Figure 4-2B).
4.2.3.2.2.3 Random Pairs
The random pairs method begins at the sampling point to the nearest individual plant. A
perpendicular line to the line is drawn between the sampling point to the nearest individual. This can
be done by standing at the sampling point and extending one's arms out to either side. The plant
closest to the first individual on the other side of the perpendicular line is the second individual of the
pair (Figure 4-2C).
123
4.2.3.2.2.4 Point-Center Quarter
Point-center quarter begins at the sampling point. In each quadrat the distance between the sampling
point and the nearest individual plant is measured. A minimum of 20 sampling points is suggested.
Often the diameter at breast height of trees is also recorded, allowing for the estimation of relative
dominance as well. When species information is recorded along with DBH, point-center data can be
used to calculate 1) density, 2) relative density, 3) relative dominance, 4) frequency, and 5) relative
frequency. The method is simple, rapid, and works well when individuals are randomly dispersed
(Figure 4-2D). The point-center quarter method is often favored over the other measures because it
uses fewer sampling points and reaches minimum sample sizes more quickly.
4.2.3.2.2.5 Wandering Quarter
Wandering quarter begins at a sampling point. An angle of direction is determined (e.g., 0o or due
north). Visualizing two 45o partitions to either side of the angle of direction, proceed to the closest
plant. At this plant, again visualize two 45o angles that flank the original angle of direction. Within
the resulting arc, locate the nearest plant, measuring the distance. Continue throughout the stand for a
minimum of 25 observations (Figure 4-3).
124
A.
B.
D.
C.
50
40
30
20
10
50
40
20
30
CEMO2
JUMO
PIED
10
RHTR
0
RICE
Figure 4-2. Distance methods illustrated. A. Closest individual, B. Nearest neighbor, C. Random
pairs, D. Point-center quarter in a pinyon-juniper woodland (Pinus edulis-Juniperus monosperma).
Mountain mahogany = CEMO2 (Cercocarpus montanus), oneseeded juniper = JUMO (Juniperus
monosperma), pinyon pine = PIED (Pinus edulis), skunkbush sumac = RHTR (Rhus trilobata), and
wax currant = RICE (Ribes cereum).
125
E.
50
40
30
20
10
50
40
20
30
CEMO2
JUMO
PIED
10
RHTR
0
RICE
Figure 4-3. Wandering quarter method. Execution of wandering quarter in a pinyon-juniper woodland
(Pinus edulis-Juniperus monosperma). Mountain mahogany = CEMO2 (Cercocarpus montanus),
oneseeded juniper = JUMO (Juniperus monosperma), pinyon pine = PIED (Pinus edulis), skunkbush
sumac = RHTR (Rhus trilobata), and wax currant = RICE (Ribes cereum).
126
4.2.3.2.3
Example - Point-Center Quarter Method
An example of point-center quarter sampling and data summarization is presented in Figure 4-4,
Table 4-3, and Table 4-4. Data from any of the other distance sampling measures would be treated in
a similar fashion for the estimation of density with the addition of the correction factor to the mean
distance.
100
90
80
70
60
50
40
30
20
10
80
70
60
50
CEMO2
JUMO
40
30
20
PIED
RHTR
RICE
10
0
Figure 4-4. Use of point-center quarter in a pinyon-juniper stand (Pinus edulis-Juniperus
monosperma). Five parallel transects, 20-m apart were surveyed using the point-center quarter
method. Starting points for each sample were 20-m apart along the transect lines. The nearest
individual to the starting point was measured within each quarter. Mountain mahogany = CEMO2
(Cercocarpus montanus), oneseeded juniper = JUMO (Juniperus monosperma), pinyon pine = PIED
(Pinus edulis), skunkbush sumac = RHTR (Rhus trilobata), and wax currant = RICE (Ribes cereum).
127
Table 4-3. Point-center quarter distance measurements and calculations for mean distance, mean
area/plant, and density.
Quarter
VegID
Distance (m)
Quarter
VegID
Distance (m)
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
JUMO
PIED
PIED
RHTR
CEMO2
CEMO2
PIED
RICE
JUMO
PIED
PIED
RHTR
JUMO
JUMO
PIED
RHTR
PIED
PIED
PIED
RHTR
PIED
RHTR
RHTR
RICE
JUMO
JUMO
JUMO
JUMO
JUMO
JUMO
PIED
RICE
PIED
PIED
RHTR
RICE
PIED
PIED
RICE
RICE
4
3.4
3.7
4.5
4
5
5
5.3
8
2
5
2.5
5
7
4
2.5
4
3
2.7
2
6
3.5
3.5
3
1
2
2
3.5
4.5
3.5
3
2.5
3.7
7
5
6
5
6.5
3
4
11
11
11
11
12
12
12
12
13
13
13
13
14
14
14
14
15
15
15
15
16
16
16
16
17
17
17
17
18
18
18
18
19
19
19
19
20
20
20
20
JUMO
PIED
RHTR
RHTR
JUMO
JUMO
JUMO
JUMO
JUMO
JUMO
PIED
RICE
PIED
PIED
RHTR
RHTR
JUMO
PIED
RICE
RICE
PIED
PIED
PIED
RHTR
JUMO
PIED
RHTR
RHTR
PIED
RHTR
RHTR
RHTR
CEMO2
PIED
RHTR
RHTR
JUMO
JUMO
PIED
PIED
7.3
6.5
6
5
2
2
3
1.5
7
4
3
2.3
3.2
6.7
4.5
5.4
7
4
4.5
2.8
2.7
3.7
2.5
2
5
3.8
4.5
3.2
2
4
2.5
6
4.5
5
2.5
3.5
4.5
4
9
3
a
b
c
d
326.4
80
4.08
16.6464
600.7305
Sum of Distances
Measures
Mean Distance
(a/b)
Mean Area/Plant
c2
Density
10000/d
m
m
m2
hectare
128
Table 4-4. The relative and actual density of woody species in a pinyon-juniper stand (Pinus edulisJuniperus monosperma). Relative Density = # of Plants of Species A/Total # for all species.
VegID
CEMO2
JUMO
PIED
RHTR
RICE
Total
#
a
3
21
28
19
9
80
Calculated Relative Density
(a/80)*100
3.75
26.25
35.00
23.75
11.25
Actual Relative Density
5.65
26.74
29.76
20.53
18.08
Table 4-5. The frequency and relative frequency of woody species in a pinyon-juniper stand (Pinus
edulis-Juniperus monosperma). Frequency = # of Points Species A occurred/Total # of Points
Sampled.
VegID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Grand2
Observations
CEMO2 JUMO
1
2
1
2
4
2
1
4
2
1
1
1
3
2
21
Sum of Observations
Frequency
VegID CEMO2 JUMO
10
55
PIED
RHTR
2
1
2
1
3
1
1
1
2
2
1
1
2
1
3
1
1
1
2
28
PIED
90
Sum Frequency for All Species
Relative Frequency
4
22
36
RICE
1
1
1
1
2
1
1
1
1
2
2
1
2
2
1
2
3
2
19
9
80
= (# points with Species A/Total # of Points Sampled) * 100
RHTR RICE
60
35
e.g., (2/20)*100
250
= (Frequency of Species A / Sum Frequency of All Species) * 100
24
14
e.g., (10/250)*100
Results -- Actual density of individuals in the survey area was 632/hectare. The estimated density of
plants using the point-center quarter method was 642/hectare (8000 m2 was surveyed with 535 plants.
This value was corrected for a hectare). The ranking of species by relative density was the same for
calculated and actual numbers; however, the difference between point-center quarter measure and the
absolute count varied up to 38% for the species Ribes cereum (RICE).
129
4.2.3.2.4
Limitations and Modifications of Distance Methods
Distance measures work on the assumption that the distribution of species in an area is random;
however, non-random dispersion of individuals occurs in most life forms (Mueller-Dombois and
Ellenberg 1974, Smith 1980). Not having a random distribution generally does not produce a
significant error when point-center quarter or wandering quarter methods are used, unless clumping
of individuals is obvious. For the other measures noted, lack of randomness can affect results. In spite
of this, other calculations descriptive of the area can be used, such as relative density and relative
frequency. With little additional effort, DBH can be collected on tree species and relative dominance
and cover determined.
Correction factors are suggested for the distance measures described with the exception of the pointcenter quarter method, which does not need a correction factor. Based on empirical testing, the
correction factor for the nearest neighbor method is 1.67 times the mean distance, 2.0 for closest
individual, and 0.8 for random pairs (Cottam cited in Mueller-Dombois and Ellenberg 1974). Data
from all methods are typically comparable when the correction factors are used (Cottam and Curtis
1956).
4.2.4
Biomass/Production Methods
A difference in time exists between biomass and production. Biomass is the weight of vegetation at a
point in time (t), whereas production includes the dimension of time (t, t+1, t+2, etc.) (Holecheck et
al. 1989). Production, therefore, is a rate expressed as the change in biomass between two or more
sampling intervals. While there is a difference between the terms, there is no difference in the
methods of collecting data. Either can include or be limited to above and below ground materials.
Above ground materials can be further defined to include or exclude litter. For the purposes of
describing the methods the term 'standing biomass' will be used; thereby, limiting discussion to above
ground plant materials.
Standing biomass measures are used for estimating forage for wildlife and livestock and fuel load.
Standing biomass is another descriptor of species contribution/importance within and between plant
communities, physical distinct areas, different land uses, etc. Standing biomass data can be collected
by area, species, or life forms (e.g., grasses, forbs, shrubs, etc). Vegetation is collected from any of a
series of rings from an area appropriate for the vegetation (e.g., 0.1 m2, 0.25 m2, 0.5 m2). Generally, a
circular frame is preferred to limit edge effect. Frames are tossed over shoulders or placed along
transect lines to achieve a random sample.
Biomass data are frequently compared between controlled, or excluded areas, and areas under some
type of utilization.
4.2.4.1
Applicability
Production and standing biomass measures are primarily used in grasslands. The methods are used to
estimate forage, herbaceous species contribution to community types, and productivity.
130
4.2.4.2
Advantages and Limitations
Methods span a wide-range of harvesting levels. The more time consuming methods are the most
accurate; however, with practice, field crews can visually estimate standing biomass consistently.
Harvesting and estimating by species is laborious, and often species are grouped by into functional
groups (annual grass, perennial grass, annual forb, etc).
4.2.4.3
Equipment
•
•
•
•
•
•
4.2.4.4
Clippers
Paper bags
Marking pens
Spring scale
Drying oven
Scale (to 0.1 or 0.01g)
Harvest
Plant biomass can be measured directly by clipping vegetation rooted in the quadrat to ground level
or to a specific height, and weighing. Weights determined in the field represent the wet mass of plant
materials. To minimize variation among plant materials due to environmental conditions (e.g.,
humidity, soil moisture, etc.), materials are oven-dried following field collection and dry weights
determined.
Direct harvest is the most time consuming and labor expensive measure of standing biomass, but it
requires little training and the information is accurate. Materials can be sorted (e.g., by species, life
form, palatability, etc.) in the field or in the lab. The data are reported as grams/unit area (e.g., g/m2)
or grams/species or life form.
There are various degrees of harvest. Plants can be clipped to ground level and potentially killed or
clipped to a set height to estimate rate of recovery. When plants are clipped to ground level,
subsequent sampling takes place some distance from the initial harvest location.
Collected plant materials are placed in paper bags and labeled with collection site and treatment
information. Materials in their bags are oven-dried at 80o C until they reach a constant weight (i.e., no
additional weight loss with additional drying time). Before weighing, the materials are allowed to
return to room temperature. In humid areas, materials should be weighed as soon as cooled. Plant
material is weighed in the bag, the weight is recorded, the bag is emptied, and the empty bag is
weighed. The difference between the weight of the full bag and the empty bag is the weight of the
material. An alternative method is to weigh 10 or more bags and assume the average bag weight is the
weight for all bags. The number of decimal places attributed to the weight of a sample should reflect
the level of precision and accuracy of all of the previous steps and the size of the smallest sample. If
samples weigh 100 g or more there is no need to include decimal points; Samples less than 10 g are
more accurately reported using 1-2 significant figures.
4.2.4.5
Double Weight Sampling
Double weight sampling is efficient and can approach the accuracy of direct harvest with sufficient
training (Reich et al. 1993). Sampling units (quadrat, species, life form) weights are estimated by
sight. A portion of all plots sampled are clipped, materials bagged and weighed (wet weight with field
scale) following ocular estimates. A regression of estimated and actual dry weights is made.
131
Estimated weight values for materials from unclipped plots are then determined. An adequate number
of clipped plots and training are necessary.
4.2.4.6
Dry-Weight Rank
Dry-weight rank sampling is a non-destructive method for estimating biomass. Quadrats (50 to 100)
are placed randomly throughout the sampling area. The size is not critical as long as the most
common species occur in most quadrats (Bonham 1989). The three most abundant species are ranked
1(most abundant), 2, and 3. The first, second, third ranked species are then multiplied by an
empirically developed factor; 70.19, 21.08, or 8.73, respectively (Bonham 1989). The products of
rank and multiplier are then added by species to give dry weight percent.
4.2.4.7
Estimated biomass
Plant material weight is estimated for a plot. The main types of vegetation (life forms) are then
identified and the weight of the plant material is apportioned (e.g., 150g; 25% forbs, 25% half shrubs,
and 50% grasses). As in double weight sampling, a subset of plots are clipped, bagged, and weighed
to correct estimates.
4.2.4.8
Robel Pole
4.2.4.8.1
General Description
This method provides estimates of standing plant biomass, and has been used primarily for
determining the quality of wildlife habitat. in terms of height and vertical cover. Robel pole readings
can be correlated to forage production or standing crop by clipping and weighing forage on a sample
after making the Robel pole readings. The pole readings provide an estimate of total standing
biomass, which includes both the current year’s production as well as any material from previous
years. The Robel method can be used to monitor vertical cover, production, and structure.
4.2.4.8.2
Applicability
The Robel pole method is most effective in upland and riparian areas where dominant vegetation
consists of perennial grasses, forbs, and shrubs less than about 1.2 m in height (Bureau of Land
Management 1996).
4.2.4.8.3
Advantages and Limitations
This method is rapid, relatively simple, and easy to apply. The equipment is inexpensive, lightweight,
and easy to construct.
4.2.4.8.4
•
•
•
•
•
•
Equipment
Study location and site information form
Robel pole field data sheet
Robel pole
Cover class definitions for the area or plant community
Metal stakes and hammer
Compass
132
4.2.4.8.5
Training
Data accuracy is depends on consistent training and observer skill. The primary proficiency tasks are
locating and laying out transects, determining cover classes, and reading the Robel pole.
4.2.4.8.6
Establishing Studies/Plots
Sites to be sampled should be selected and located on a map of the study area prior to data collection.
Bureau of Land Management (1996) recommends a minimum of 50 observation points per transect.
The pole is approximately 1 1/8” in diameter and 48 inches long, painted with alternating bands of
white and gray colors. A 157-inch (4 m) cord is attached to the pole at a height of 1 m to standardize
the distance of the measurements. White bands are numbered in black. Based on monitoring
objectives, the number of vertical cover classes and the height limits for each should be established.
For example, 2-foot height limits have been used to create a 4-class system, where the fourth class is
represented where obstruction height exceeds 4 feet.
4.2.4.8.7
Sampling Process
The technique works best with a two-person field crew, but a spike can be attached to the bottom of
the pole so that it can be pushed into the ground, allowing a single observer to conduct the sampling.
The field crew travels along a transect and takes two measurements from each observation point,
perpendicular to the transect. One person holds the Robel pole at the observation point and the second
person holds the end of the cord perpendicular to the transect. The observation measurement is made
by determining the highest 1-inch band totally or partially visible and recording the height on the
field data form. Readings are taken at specified intervals until the transect is complete.
4.2.4.8.8
Data Analysis
The measurements are totaled for each observation point and for the transect. The sum is divided by
the total number of observations to yield the average visual obstruction. The average cover value can
be used to determine if management or monitoring objectives were met.
Data from the Robel pole can be correlated with standing crop or forage production. The relationship
is established by clipping and weighing the vegetation within a specified quadrat frame directly in
front of the Robel pole after a reading is made. Approximately 25 clipped quadrats may be necessary
to establish a good correlation between Robel pole readings and standing crop (Bureau of Land
Management 1996).
If plots are clipped, temporary plots are recommended. The appropriate non-paired test should be
used. ANOVA is appropriate for comparing more than two sampling periods.
4.2.5
Forest and Tree Measurements
4.2.5.1
Introduction
Most forest measurement approaches are designed to assess timber resources, but many methods can
provide information that is applicable to wildlife, watershed, recreation, and military training
concerns. This section will provide a brief overview of approaches and methods that relate to tree,
stand, and forest attributes that are useful in forest monitoring and management. Material related to
tree form, volumes, and growth and yield models, which provide information for timber management,
133
will not be addressed. For more information see Beers and Miller (1973), Wenger (1984), and Avery
and Burkhart (1995). While most forest inventory procedures are based on the “stand” concept, other
strata within the landscape can be successfully used. Because forest stands by definition tend to be
uniform in terms of species composition, age, structure, and productivity, they are relatively
homogeneous. This is advantageous from a tree inventory perspective but may not be so when other
vegetation or ecological attributes are the principal attributes measured.
4.2.5.2
Attributes of Interest
Tree densities are provided indirectly by a number of the methods discussed. In order for the diameter
or height of a tree to be measured, that tree must be within the sampling area or somehow qualify.
Densities can be expressed for all trees meeting certain minimum criteria such as diameter, height,
crown class, and/or species.
The most frequent tree measurement made by foresters is diameter at breast height (DBH). DBH is
defined as the average stem diameter, outside bark, at a point 1.3 m (4.5 feet) above the ground.
Direst measurements of DBH are usually made with a diameter tape, tree caliper, or Biltmore stick.
With a diameter tape, tree circumference is automatically converted to a diameter. For circular trees, a
steel diameter tape is the most consistent method for measuring DBH. When irregular or elliptical
stems are encountered, two caliper readings made at right angles provide a more accurate
measurement compared to the diameter tape. Specific guidelines have been developed for measuring
the DBH of irregular, forked, and leaning trees, as well as those on slopes. Tree diameters are
frequently measured to the nearest 0.1 inch. Measurements can be grouped into diameter classes (e.g.,
1 inch or 2 inch classes). If 1 inch classes are used, the numbered class forms the midpoint of the
class. For example, the class limits for 9-inch trees are 8.6 to 9.5 inches.
Another attribute of interest is tree height. Tree height, in conjunction with tree density, is an
important descriptor of community structure and habitat. Many types of height-measuring devices
have been developed. The most common types used are the clinometer and the Abney level. When
using these instruments, the observer stands at a fixed horizontal distance from the tree (usually 50,
66, or 100 feet). Tangents of the angles to the top and base of the tree are multiplied by horizontal
distance to derive the tree height. Instruments used generally yield height readings directly in feet or
meters at fixed horizontal distances from the tree.
Tree age or class is often measured because it provides additional information about stand structure
and age distribution. Trees can be classified initially according to their relative dominance or crown
levels. For example, the following four-class system is presented by Avery and Burkhart (1995):
1) Dominant. Tree with crown extending above the general level of the crown cover and
receiving full light from above and partly from the side; larger than the average trees
in the stand, with crowns well developed but possibly crowded on the sides.
2) Codominant. Trees with crowns forming the general level of the crown cover and
receiving full light from above, but comparatively little from the sides; usually with
medium-sized crowns more or less crowded on the sides.
3) Intermediate. Trees shorter than those in the two preceding classes, but with crowns
either below or extending into the crown cover formed by codominant and dominant
trees, receiving little direct light from above and none from the sides; usually with
small crowns considerably crowded on the sides.
4) Overtopped. Trees with crowns entirely below the general level of the crown cover,
receiving no direct light either from above or from the sides.
134
Trees in different crown classes may be of varying ages or the same age. Tree ages are measured
using counts of growth rings from tree cross sections at ground level, or by using an increment borer
to estimate the age of standing trees. The age of the tree is estimated by counting the growth rings and
then adding the estimated number of years required for the tree to reach the height of the boring.
4.2.5.3
Sampling Designs
A number of different sampling designs can be used to inventory or monitor timber resources.
Systematic, simple random, and stratified random designs are discussed in Chapter 3. Timber
inventory estimates require that the size of the forest area must be known and an unbiased sample of
trees must be measured. The choice of a particular inventory method is determined by relative costs,
size and density of trees, the size of the area to be sampled, desired precision, the resources available
for field work, and the available time. Except for the case where a complete tally of trees is justified
(a census), trees will be sampled within desired areas. This process is referred to by foresters as
“cruising”. As with other types of sampling, considerations for developing an efficient sampling
design include plot size and shape, number of samples, and the sampling design used (e.g.,
systematic, random, etc.). Two methods for selecting sample trees will be covered here: (1)
probability proportional to frequency, and (2) probability in proportion to size, or point sampling.
4.2.5.3.1
Strip or Plot Sampling - Probability Proportional to Frequency
With probability proportional to frequency, the likelihood of selecting trees of any given size is
determined by the frequency with which that tree size occurs within the stand. This technique is
implemented in the field by strip cruising or fixed area plot sampling. Within the strips and plots that
are defined on the ground, individual trees are tallied and characteristics such as species, height,
DBH, etc. is assessed. The sample tallies are then expanded to the desired per unit area basis by
applying an appropriate expansion or blow up factor.
With the strip cruising system, strips are established at equally spaced intervals, such as 5, 10, or 20
chains (1 chain = 66 feet). The sample strip is usually between ½ chain and 2 chains wide, depending
on tree density. Strips often are placed in north-south or east-west directions and should crosstopographic gradients to the greatest extent possible. Strip cruises are designed to sample a
predetermined percentage of the forest area. Strip sampling may cover between 2 and 10 percent of
the forest area. Accurate determination of strip lengths and center lines requires careful measurement
by two individuals. Tallies are summed for all strips combined and the values are multiplied by the
stand blow up factor to generate an estimate for the forest area.
Advantages of strip sampling include:
1) Sampling is continuous, and little time is wasted traveling between strips
2) In comparison with plot sampling of the same intensity, strips generally have fewer
borderline trees because the perimeter to area ration is smaller
3) With a crew of two, there is less safety risk in remote or hazardous areas
Disadvantages of strip sampling include:
1) Errors occur due to inaccurate estimation of strip width, partly because the cruiser is walking
while tallying
2) Tree heights may be underestimated unless they are checked at a good distance from the trees
135
3) Windfall and brush can be a significant hindrance compared to plot sampling
4) It is difficult to do spot checks of results because the strip is rarely marked on the ground
(Avery and Burkhart 1995).
The line-plot system of cruising is the most traditional approach for forest inventory. It consists of a
systematic tally of trees on a number of fixed area plots arranged in a rectangular or square grid. A
random starting point for the first compass line is chosen and subsequent lines are placed at equal
spacing, usually along an edge of the forest area. Plots are usually circular, but squares and rectangles
may also be used. 1/4 or 1/5-acre plots are commonly used in sawtimber-sized forests; smaller plots
are used for sapling and poletimber stands. Regeneration counts may use plots as small as 1/1000
acre. English and metric plot radii for a variety of plot sizes are presented in Table 4-6. Systematic
line-plot inventories are often planned on a percent cruise basis. However, when attempting to meet
desired standards of precision, the number of samples required should be determined best by using
sample size equations based on desired precision and confidence.
Table 4-6. Radii for various sizes of circular plots.
Plot size
(acres)
1
1/2
1/4
1/5
1/10
1/20
1/25
1/40
1/50
1/100
1/500
1/1000
Plot radius Plot radius
(ft.)
(m)
117.8
83.3
58.9
52.7
37.2
26.3
23.5
18.6
16.7
11.8
5.3
3.7
35.90
25.39
17.95
16.06
11.34
8.02
7.16
5.67
5.09
3.60
1.62
1.13
Systematic grids are preferred over random plot locations because of the ease of location of plot
centers. Line-point sampling can be done with crews of 1 to 3 persons. Lines are determined using a
compass. Intervals between plots are measured using a chain, tape, or by pacing. For circular plots,
the radius should be measured out in four directions. Trees and tallied for the plot using a data sheet
that is divided into sections for each plot.
Advantages of plot sampling include:
1) A one-person crew is efficient
2) Cruisers are not obstructed by brush and windfalls because tallying is not done while
following a compass line
3) A pause at each plot center allows more time for checking stem dimensions and borderline
trees
4) Tree tallies are separate for each plot, permitting summaries and variances of data by timber
types, stand sizes, or area condition classes (Avery and Burkhart 1995).
136
4.2.5.3.2
Point Sampling - Probability in Proportion to Size
Point sampling is a method which selects tally trees based on their size rather than their frequency.
Sample points, arranged in a similar fashion to plot centers, are located within the area of interest and
a prism or angle gauge is used to select “in” trees. Trees close enough to the observation point to
completely fill in the fixed sighting angle are tallied; stems which are too small or far away are not
counted. The resulting tally is used to calculate basal areas, volumes, or numbers of trees per unit
area. The probability of tallying a given tree depends on its cross-sectional area and the sighting angle
used. More trees are tallied as the angle becomes smaller. Point sampling does not require direct
measurement of trees or plot areas (Avery and Burkhart 1995), except where information about tree
diameter distribution is desired. A basal area factor (BAF), determined by the angle of the gauge or
prism chosen, is used to convert tree tallies to basal area per unit area. BAFs commonly used range
from BAF 5 for use in light-density pole stands to BAF 40 or 60 in large western timber where trees
are scattered. The BAF is usually chosen to provide a tally of 5-12 trees per point.
Commonly used instruments for conducting point sampling are stick-type angle gauges, the Panama
angle gauge, wedge prisms, and the Spiegel relascope. Each instrument will have a specified BAF. To
tally trees with stick-type gauges, all trees larger than the defined angle are counted. The Panama
gauge is used in a similar fashion. To tally trees with the wedge prism, the prism is held vertically
over the sample point at a right angle to the line of sight. Tree stems not completely offset when
viewed through the prism are counted; others are not tallied. Borderline trees can either be alternately
counted or not counted, or the distance from the sample point to the center of the tree and the tree
diameter can be measured and the plot radius factor for the tree diameter can be applied. The
Relascope is a more complex, expensive instrument that has many features and can compensate
automatically for slope steepness when sampling points. Sample data by diameter class and height is
presented in Table 4-7.
Table 4-7. Tally of stems by height class from a point cruise of 25 points.
dbh (in.)
Height (# of logs)
1
2
3
total
8
10
12
14
Total
16
20
36
28
68
54
20
170
12
28
30
6
76
20
24
12
56
The following equation calculates basal area per acre:
Basal Area (BA) per acre =
total trees tallied
xBAF
number of points
With a BAF of 10 and 170 trees tallied at 25 points, the estimated total basal area per acre is 170/25 X
10 = 68 square feet per acre.
Because each diameter class of tree has a different imaginary plot zone, the per-acre conversion factor
varies from class to class. The number of trees per acre for any diameter class is:
137
Trees per acre =
(number of trees tallied)x(per acre conversion factor )
total number of points
where the per-acre conversion factor for BAF 10 (conversion factors for different BAFs are presented
in forestry manuals and handbooks) is:
43,560
π (dbh × 2.75)2
Tree densities are calculated from the sample data in Table 4-7 using BAF 10 per-acre conversion
factors from published tables:
8 - in. class =
28(28.65)
= 32 trees/acre
25
10 - in. class =
68(18.35)
= 13 trees/acre
25
12 - in. class =
54(12.74 )
= 27 trees/acre
25
14 - in. class =
20(9.35)
= 7 trees/acre
25
Advantages of point sampling include:
1) It is not necessary to measure or establish a fixed plot boundary, thus increasing efficiency
2) Basal area and volume per acre are derived without direct measurement of stems
Disadvantages of point sampling include:
1) Heavy underbrush reduces sighting visibility and cruising efficiency
2) Because of the relatively small size of sampling units, carelessness and errors in tree tallies
can produce significant errors in stand estimates relative to plot sampling
3) As with strip and line-point cruising, slope compensation can cause problems leading to large
errors unless proper procedures are followed
4) Edge effects can be potentially serious, especially in small or narrow stands (Avery and
Burkhart 1995).
138
4.2.6
Photo Monitoring
4.2.6.1
General Description
Photo monitoring is a simple and effective means to monitor resource conditions and document
changes over time. Photo monitoring can be an integral part of any monitoring project. Before
beginning photo monitoring, the purpose or objective of the project should be defined, as well as the
study site location and the “what, when, and how” of monitoring (Borman 1995; Hall 2001). Photo
monitoring sites can include RTLA plot locations, riparian areas, upland areas with native species,
weedy areas, burned areas, revegetation areas, or other areas of interest. Homogeneous areas will
require fewer survey points than heterogeneous areas. Both landscape (photo point) and close-up
(photo plot) photos can be taken at each photo point. Photo monitoring is most effective in
documenting large changes over time, but is also valuable to illustrate lack of change and gross
vegetation structure.
4.2.6.2
Photo Points
Photo points represent the broad view of a study site. Photo points can be used to: (1) relocate study
sites (the sites themselves as well as approach photos), (2) provide a visual record of general
conditions along transects or at plots, (3) monitor changes in site conditions (e.g., plant structure,
weed presence and abundance, and disturbance), (4) estimate population size (Elzinga et al. 1998),
and (5) estimate vegetation density and height (Van Horn and Van Horn 1996).
When subsequent pictures are taken at photo points, the same camera position and angle used during
the initial photo are applied. The camera location and photo point should be permanently marked and
include the same area and landmarks as shown in the initial photos. Accessibility is important when
selecting a photo point location. Photo points can be located in random or subjective locations, and
may or may not be associated with other data collection efforts. A high density of photo points does
not make this method more quantitative, but may capture changes more effectively.
An identification label can be included in the foreground of each photograph. The label can be
prepared on paper or other re-usable medium with a non-white background (gray or light blue paper
works well, white dry-erase boards produce glare and are often illegible). Include the date, site
reference number and/or name, and location on each label. Observations should be recorded when
photos are taken. Observations can include information about vegetation composition, vegetation
phenology and vigor, soil conditions, wildlife observations, precipitation events and effects,
recreational use, other forms of disturbance, and other pertinent information (University of California
Cooperative Extension 1994). Some digital cameras allow audio recordings to be made, which require
office time to transcribe.
4.2.6.3
Photo Plots
Photo plots can be used as qualitative or quantitative records of resource condition over time. Closeup photos can show soil surface characteristics and the amount and type of ground cover. Photo plots
can also be used to estimate plant cover and plant species densities.
Photo plots are usually permanently located using standard size frames (de Becker and Mahler 1986;
Schwegman 1986; Foster et al. 1991). A 3 by 3 foot or 1 by 1 meter frame is typically used, but
different sizes and shapes can also be used depending on monitoring objectives. As frame size
becomes larger, greater camera heights are required to capture the whole area clearly. Frames can be
139
made of PVC pipe, steel or aluminum rods, or other material. Temporary, removable rods can be
placed across the frame to delineate a grid to facilitate counting or cover estimation. Permanent photo
plot corners can be marked with angle iron or rebar, which can be spray painted a bright color to
facilitate relocation. The location of each photo plot should be described and mapped using azimuths
and distances from locatable features. Approach photos may be helpful in relocation. An
identification label should be included in each photo plot which includes the date, plot identifier, and
location. The camera point should not change over time, and preferably be located on the north side
of the photo plot so that repeat pictures can be taken at any time during the day without casting a
shadow across the plot (Bureau of Land Management 1996).
Photo plots can function as a density sampling unit when individual plants can be clearly identified in
a frame. This approach may be helpful when field time is limited, but may have limited accuracy.
(Elzinga et al. 1998). Photo plots can also be used to estimate cover by overlaying a line grid or dot
grid on the photograph and counting intersections or dots for each species of interest. However, small
or rare species may be underestimated or missed altogether. Ocular estimates can also be used with
the photos. Analysis of photos can be done using prints, projected slides, and digital displays.
4.2.6.4
Equipment Needed
The following equipment is needed for photo monitoring at photo points or photo plots:
•
•
•
•
•
•
•
•
•
4.2.6.5
35 mm camera with 28 or 35 mm focal length lens or digital camera.
Field notebook or personal digital assistant (PDA) with plot/point locations, maps, azimuths,
location photos, and initial photos.
Compass and/or GPS unit.
100-ft (30-m) (or appropriate length) measuring tape.
Paper or other re-usable medium for photo labeling board.
Frame to delineate photo plots.
Metal fence posts (and post driver), steel bars/stakes (preferably concrete reinforced rebar), or
other permanent markers for photo points or plots. Stamped metal fence posts are
recommended over T-bars because of their flexibility, which allows them to remain in place
if disturbed and bend if driven over. Stamped metal fence posts are also inexpensive and
easier to carry and drive into the ground (Hall 2001).
Camera tripod.
Extra memory card(s) for digital camera.
Photo Storage and Management
A durable field notebook should be kept for a photo monitoring project that includes maps and other
helpful information such as compass direction, camera location, and other notes. A personal digital
assistant (PDA) can also be used to store and manage this information when revisiting sites. A second
notebook for the office can include all photos, plot locations, maps, azimuths, and observation notes.
A filing system described by Hall (2001) includes photographs to illustrate camera locations and
photo points, and the photo monitoring layout and location. Hall (2001) recommends filing
everything for each study (e.g., maps, data, filing system forms, slides, prints, and digital memory
cards) in a separate expandable file. Files are organized by geographic location of the study site and
by date.
Photos should be developed and cataloged as soon as possible. Identifying information (e.g., site
name, date, photo number) should be written on the back of prints, slide frames, or as part of the file
140
name for digital photos. A print should be made from all negatives or digital files for reference. Slides
and prints should be stored in archival quality (i.e., non-acid, non-PVC) sheets for protection.
Digital photos should be backed-up on a hard drive as well as other media such as CDs. Digital
photos should be accompanied by a file such as a readme.txt or spreadsheet to identify and describe
each image (Hall 2001). Alternatively, digital or scanned images can be stored and managed in
Microsoft (MS) Access or an ArcGIS geodatabase.
4.2.6.5.1
Storing Digital Images in Microsoft Access
MS Access supports the storing of Object Linking and Embedding (OLE) objects in tables. An OLE
object can be a MS Excel spreadsheet, a MS Word document, graphics, sounds, or other binary data
and can be linked to or embedded in a MS Access table. To use this feature in a table a data field must
first be created in the table with a data type of OLE object.
The procedures for embedding an object in a MS Access table are listed below:
1. Move to the record in which you want to insert the object and click the OLE Object field in a MS
Access database table.
2. On the Insert menu, click Object.
3. In the Insert Object dialog box, click Create from file, and then specify a path to the file. If you
don't know the path, click Browse.
4. Select the Display as Icon check box if you want the object to appear as an icon instead of as the
object itself. Displaying an object as an icon can be helpful when an object contains supplementary
information that doesn't have to be displayed. Displaying an object as an icon also uses significantly
less disk space.
5. Click OK.
If the application that the object is being copied from supports OLE drag-and-drop editing, then,
instead of using the Object command, you can drag the file directly from the Windows Explorer to
the OLE Object field in a MS Access database.
The procedures for linking an object in a MS Access table are listed below:
1. Move to the record in which you want to link the object, and click the OLE Object field in a MS
Access database table.
2. On the Insert menu, click Object.
3. In the Insert Object dialog box, click Create from File, and then specify a path to the file. If
you don't know the path, click Browse.
4. Select the Link check box.
5. Select the Display as Icon check box if you want the object to appear as the icon for the
application where the object was created instead of as the object itself. Displaying an object as an
icon can be helpful when an object contains supplementary information that doesn't have to be
displayed. Displaying an object as an icon also uses significantly less disk space.
6. Click OK.
When objects are linked they are not stored directly in the database, only a reference to the file
location is stored. When changes are made to the file they will be reflected when opened in the
database.
141
To view an OLE object double click the OLE object field, the application which supports that object
will open showing the object.
Although it is possible to store OLE objects, such as images, within the database itself, it is not
generally the preferred way. For more information on this and alternative techniques refer to the
resources below.
For more information on handling images in forms and databases:
http://www.mvps.org/access/forms/frm0030.htm
Image Control may cause a GPF when browsing between records:
http://www.mvps.org/access/bugs/bugs0044.htm
How to display images from a folder in a form, a report, or a data access page:
http://support.microsoft.com/default.aspx?scid=kb;en-us;285820
ACC2000: How to Load OLE Objects from a Folder into a Table:
http://support.microsoft.com/default.aspx?scid=kb;en-us;198466
4.2.6.5.2
Storing Digital Images in an ArcGIS Geodatabase
In ESRI ArcGIS 9, digital photos can be stored as raster attributes in geodatabase feature classes and
stand-alone tables (ESRI 2004). A variety of image formats can be used (e.g., JPEG, BMP, TIF,
MrSID). The ESRI website (http://www.esri.com/) and ArcGIS 9 documentation contain additional
technical resources.
4.2.6.6
Guidelines for Photo Monitoring
The success of photo monitoring will depend on the level of replication, quality of the photos, and
how well the system meets the monitoring goals and objectives. The following are guidelines for an
effective photo monitoring program. Additional guidelines are provided by Todd (1982), Brewer and
Berrier (1984), Borman (1995), and Hall (2001).
1. Film Cameras: A quality 35 mm camera is essential (Hall 2001). Photos with good color and
resolution are needed to accurately assess change. The camera and film influence the color
and resolution of photos (Hall 2001). Point-and-shoot automatic and single lens reflex (SLR)
are the two types of cameras available for film use. SLR cameras produce superior photos
because of the enhanced photographic composition and fine-tuned adjustment capability. A
28 or 35 mm focal length is recommended (Borman 1995). Elzinga et al. (1998) recommend
lens sizes of 28-75 mm for photo plots and 50-200 mm for photo points. Lenses with wide
minimum aperture settings (f1.6-f1.8) allow for better photographs in low-light conditions.
Fixed focal length lenses are superior because the actual focal length is known. Zoom lenses
may increase camera versatility for different photographic formats, but are usually less sharp
than fixed lenses and can be problematic because lenses must be precisely set to reproduce
the original image (Hall 2001). If a zoom lens is used, it is best to use the minimum or
maximum focal length to ensure the same focal length each time. The slowest shutter speed
(but generally not less than 1/60 of a second) and the smallest aperture setting help maximize
the depth of field so that more objects in the foreground and background are in focus.
142
The three techniques available for recording images from film cameras include color slides,
color prints, and black-and-white prints. Slides are often preferable to prints because of the
high photographic quality, ease of storage, and usefulness in presentations. Prints can be
made from slides for subsequent field work or inclusion in reports; however, quality may
degrade. High quality scans of both prints and slides facilitate their incorporation into written
reports.
Slower films (ASA 25-100) produce sharper, less grainy images, but faster films (ASA 200400) may be needed for low light, shady, early morning, or evening photographs. All pictures
should be in color, regardless of whether they are a primary or secondary monitoring tool
(Bureau of Land Management 1996). When using film, Hall (2001) recommends using both
color and black and white because color film fades with time while black and white does not.
2. Digital Cameras: The number of pixels influences the color and resolution of photos taken
by a digital camera (Hall 2001). Digital cameras are categorized in terms of pixel count,
which is the number of individual pixels that go into making each image. Typical pixel
counts vary from 1 million (1 Megapixel, or 1MP) to around 14 MP), and most cameras have
between 2MP and 6MP. Cameras with higher pixel counts will produce higher quality photos
than lower resolution cameras. However, file sizes increase exponentially with increasing
pixel resolution.
Most digital cameras have both optical and digital zoom. Optical zoom works similar to a
zoom lens on a film camera; the lens changes focal length and magnification as it is zoomed.
Digital zoom simply crops the image to a smaller size, and then enlarges the cropped portion
to fill the frame. Digital zoom results in a significant loss of image quality.
A variety of different memory cards are available for saving images from a digital camera.
Compact flash cards are widely used, and vary in memory from 32 to 512 MB. When
purchasing a memory card, look for cards that can save images at 32x or better to reduce
storage time. The ‘x’ refers to the 150 kilobytes saved per second on the compact flash card.
A 32x card should be capable of saving 4,800,000 bytes/second, or 4.8 MB/sec. Images from
a digital camera can be saved in a variety of formats including JPEG, BMP, and TIF.
Software such as MS Photo Editor and ThumbsPlus
(http://www.cerious.com/featuresv7.shtml) can be used to reformat and modify digital
images.
If photographs are to be used for legal purposes, some special considerations apply. Film
cameras may be a better choice (e.g., endangered species, cultural sites, etc) since they are
less prone to alteration. For digital photography, software is available that writes digital
image parameters at the time of downloading. At a later date, the photo can be “tested” and
determined as “untouched.” In lieu of such software, Levy-Sachs et al. (2004) recommend to:
•
•
•
Keep the source media intact that show the original directory with dates, file size,
and file names. This may require keeping the original floppy or compact flash card.
Make files read-only.
Rename files if they have been modified; do not overwrite the original file.
3. Timing considerations: The timing of photography depends on what is being monitored, and
timing should be consistent from year to year. Timing will depend on management goals and
objectives and knowledge about when to identify responses to management. Seasonal
143
conditions and phenological stages will vary, and the stage of growth may be more important
than the exact date. Periods before and after treatment and at the end of a growing season
may also be important to document.
4. Taking photographs: General-view photographs document a scene and do not contain a size
control board to focus and orient the camera for re-photography. Therefore, the photo will
require some of the horizon to aid in relocation. If necessary, pairs of photographs can be
taken and joined to illustrate the desired frame. This is called stitching in digital photographic
software.
4.2.6.7
4.2.6.7.1
Additional Photographic Approaches
Repeat Photography
Repeat photography uses a pair or sequence of photos to document the effects of climate, succession,
management, and other variables on range and forest lands over time. Repeat photography is useful
for evaluating historic changes when no other data exist. Repeat photography is applicable for
evaluating short- and long-term effects of tree and shrub growth and colonization, weed invasion,
recovery from fire, grazing, recreational uses, off-road vehicle impacts, or other agents of
disturbance. Suggestions for relocating and re-photographing previously photographed sites, and for
interpreting the changes are provided by Dingus (1982), Dutton and Bunting (1981), Gruell (1980,
1983), Humphrey (1987), Johnson (1987), Magill (1989), and McGinnies et al. (1991).
4.2.6.7.2
Videography
Video photography can be used to document conditions in the same ways as photo points or photo
plots. Videography provides a “big picture” overview that allows for additional perspectives. Still
pictures can be derived from frames of video footage. However, video footage is not easily
reproduced because the camera can be difficult to mount in the same position as the initial shot. Field
time is reduced considerably, but the effort required for processing and analyzing the images is
extensive. Using video to sample vegetation may be difficult because of problems associated with low
resolution, making species identification and the detection of small individuals difficult (Leonard and
Clark 1993). Video footage has become a popular component of computer interface tools for
displaying data associated with sampling locations.
4.2.6.7.3
Panoramic Photography
Digital cameras allow the creation of panoramic images that provide up to a 360° view of the photo
point. This is accomplished by taking a series of photos that overlap by a third to a half. The subjects
should be at a distance to reduce distortion. The camera is kept horizontal by using a tripod. The
photographs are stitched together using software such as Ulead Cool 360
(http://www.ulead.com/cool360/runme.htm), Pixtra PanoStitcher (http://www.pixtra.com/), or other
camera software, and can be viewed as an Apple QuickTime file (http://www.apple.com/quicktime/).
144
4.3
4.3.1
Specialized Monitoring Attributes and Approaches
Soil Erosion
Estimates of soil erosion and deposition from water or wind can be measured directly or derived
using established erosion prediction models. The decision whether to use direct measurements or
models depends on several variables including monitoring objectives, site conditions, data accuracy
and precision needs, cost, and time requirements. A successful erosion monitoring program will most
likely use a combination of quantitative field measurements, qualitative indicators, and erosion
prediction models. Qualitative indicators of erosion such as evidence of overland flow, rills, gullies,
and sediment deposition are described in Section 4.5.2 Rangeland Health.
4.3.1.1
4.3.1.1.1
Measuring Soil Erosion by Water
Plot-Scale Soil Erosion
Measuring erosion at the small plot-scale allows for a close examination of underlying physical
processes and the dominant controls on these processes. The dominant erosion processes acting at the
plot-scale include rainsplash and sheetwash. Plots must be several meters long before rill erosion
becomes important (Mutchler et al. 1994).
Rainfall simulators are commonly used to measure soil erosion at the plot-scale. Rainfall simulators
have the ability to create controlled and reproducible artificial rainfall, which eliminates the spatial
and temporal variability of natural storms (Meyer 1994). Use of a rainfall simulator can expedite data
collection and allow comparison of erosion rates from plots with different site characteristics and
management activities. Several different types and sizes of rainfall simulators are available depending
on site characteristics, plot size, and monitoring objectives (Meyer 1994).
A typical rainfall simulation involves applying of a known amount of rainfall onto a bounded plot for
a specified period of time. Runoff and sediment are routed to a trough or outlet at the bottom of the
plot and collected at equal time intervals throughout the simulation. Suspended sediment
concentration is determined for each runoff sample and net erosion is calculated using the sediment
concentration data and the runoff hydrograph (Porterfield 1979). Independent variables measured for
analysis can include percent vegetation cover, slope, aspect, and physical soil properties such as soil
texture and soil moisture.
Rainfall simulators have several disadvantages which may limit their application in erosion
monitoring. Operation of rainfall simulators can be very time and labor intensive, often requiring
several people to conduct a single simulation. Since a large volume of water is needed, simulations
are usually limited to sites adjacent to roads or areas accessible to vehicles. Results from rainfall
simulations are not easily extrapolated to larger areas such as watersheds due to changes in erosion
processes and deposition with increasing spatial scale (Bunte and MacDonald 1999). These
disadvantages make rainfall simulators best suited for relative comparisons of erosion rates between
sites.
145
4.3.1.1.2
Hillslope-Scale Soil Erosion
The dominant erosion processes acting at the hillslope-scale include rainsplash, sheetwash, and rill
erosion. Hillslopes can take a variety of shapes, including convex, concave, and planar, and do not
contain first-order stream channels. Hillslopes can vary widely in area, ranging anywhere from 15 m2
to upwards of 10,000 m2.
Erosion at the hillslope-scale can be measured with large rainfall simulators, but the easiest and most
economical method is to use silt fences. A silt fence is a synthetic geotextile fabric that is woven to
provide structural integrity with small openings (0.3 to 0.8 mm) that pass water but not sediment
(Robichaud and Brown 2002). Silt fences can be installed on planar hillslopes or unchanneled swales
using wooden stakes or rebar. A second fence can be installed below the first fence to evaluate trap
efficiency. Periodically, or after a rain event, the sediment in the fence is collected and weighed in the
field. A soil sample is taken during each cleanout to determine water content for calculating sediment
dry weight. The contributing area above each silt fence should be determined with a GPS so that
erosion rates per unit area can be calculated. An inexpensive tipping bucket rain gage with a data
logger can be used to relate rainfall amounts and intensities to the measured hillslope erosion rates.
The technical reference document Silt Fences: An Economical Technique for Measuring Hillslope
Soil Erosion (Robichaud and Brown 2002) provides further details on equipment vendors,
installation procedures, statistical design, and analysis methods for silt fences.
The low cost and high trap efficiency of silt fences make them suitable for erosion monitoring on
military lands. Potential applications include:
•
•
•
•
Estimation of natural or background erosion rates
Validation of erosion prediction model results
Evaluation of the effectiveness of LRAM treatments (i.e., seeding)
Evaluation of the effects of different military land use activities (i.e., vehicle tracking,
bivouac)
• Evaluation of the effects of other management activities (i.e., roads, fire, timber harvest)
4.3.1.1.3
Watershed-Scale Soil Erosion (Sediment Yield)
4.3.1.1.3.1 Instream Sediment Monitoring
Measuring sediment yields at the watershed-scale requires both continuous streamflow measurements
and suspended sediment and/or bedload samples. Streamflow measurement techniques are described
in Section 4.3.8 Water Quality Monitoring. Suspended sediment refers to the portion of sediment load
suspended in the water column, while bedload is the material that rolls along the streambed
(MacDonald et al. 1991).
Suspended sediment concentrations are determined gravimetrically (Stednick 1991). A known
volume of water is passed through a 0.45 µm filter and weighed, dried at 105°C for 24 hr, and
weighed again to determine concentration in mg/L. Suspended sediment samples are collected using
hand-held, cable-operated, or automatic pumping samplers (Edwards and Glysson 1988). Suspended
sediment concentrations can show considerable spatial and temporal variability, requiring careful
attention to sampling location and frequency (Bunte and MacDonald 1999). In general, a higher
sampling frequency is needed during periods of high runoff. Easier-to-measure attributes such as
turbidity may be used to predict suspended sediment concentrations and facilitate high-frequency
sampling. Turbidity threshold sampling is an automated procedure for using turbidity to govern
suspended sediment sample collection during a runoff event (Lewis 1996). The equipment consists of
146
a programmable data logger, a turbidimeter mounted in the stream, a pumping sampler, and a stagemeasuring device.
Once sufficient samples and streamflow measurements have been collected at a site, sediment
concentrations can be reasonably predicted for a range of streamflow conditions using a sediment
rating curve (Colby 1956). Separate sediment rating curves can be constructed for individual storm
events, different seasons (i.e., summer or winter), or different stages of the runoff hydrograph (i.e.,
rising or falling limb). Sediment yields are determined by multiplying streamflow times suspended
sediment concentration for each time interval, and summing the products for the time period of
interest (i.e., storm event, daily, yearly, etc.) (Porterfield 1972).
Bedload is an important component of the overall sediment yield from a watershed, but the
measurement of bedload is difficult (MacDonald et al. 1991). Sampling devices placed on or near the
streambed may disturb the rate of bedload movement. Bedload transport is also subject to extreme
spatial and temporal variability, making it difficult to obtain representative samples for a given
interval of time (Edwards and Glysson 1988; Bunte and MacDonald 1999). Portable bedload
samplers such as the Helley-Smith are the most commonly used devices for the direct measurement of
bedload. The Helley-Smith consists of an expanding nozzle, sample mesh bag, and frame (Edwards
and Glysson 1988). The sampler is placed on the streambed for a specified period of time, and the
sediment caught is dried and weighed to determine a transport rate in mass per unit stream width
(MacDonald et al. 1991). Other measurement devices such as continuously recording bedload traps
and tracers are reviewed in Bunte and MacDonald (1999).
4.3.1.1.3.2 Sediment Detention Basin Surveys
Sediment detention basin surveys offer another method to directly estimate sediment yields from
small watersheds. Surveys of detention basins are best suited for estimating sediment yields on longer
time scales (i.e., annually), as it may be impractical to conduct surveys after individual storm events
(Walling 1994). Sediment yield estimates from detention basins will include both suspended load and
bedload.
A number of different survey methods are available depending on size of the basin (ASCE 1975;
Walling 1994). For small basins, electronic total stations can be used to survey the elevation of
sediment above the base of the basin. Alternatively, the accumulated sediment can be excavated and
weighed. Larger basins may require the use of survey boats with automated hydrographic data
collection systems. ASCE (1975) elaborates on field measurement techniques to determine sediment
volumes in reservoirs of different sizes. Calculation of sediment volume requires survey data from the
original basin bottom. Core samples are necessary to determine bulk density in order to calculate
sediment mass. Regardless of which method is used, the trap efficiency of the detention basin must be
known in order to obtain reliable estimates.
4.3.1.1.3.3 Sediment Delivery Ratios
A sediment delivery ratio (SDR) can be used to indirectly estimate sediment yield at the watershedscale. A SDR is the ratio of net sediment yield divided by the total potential erosion in the watershed
(Novotny and Olem 1994). The total potential erosion in the watershed can be determined from
measured hillslope erosion rates or by using models such as USLE or RUSLE. The magnitude of the
SDR will depend on several factors including the location and size of the sediment sources, relief,
slope, drainage pattern, vegetation cover, land use, and soil type (Renfro 1975; Walling 1994). Most
SDR equations are a function of watershed area, and the SDR decreases as watershed area increases
due to continual sediment deposition. Other variables may include relief ratio and main channel slope.
Equations for SDRs have been developed for several regions of the U.S. and are presented in Walling
147
(1994). Bunte and MacDonald (1999) suggest that a SDR should only be used in the specific area
where it was developed, and should be based on multiple years of data. SDRs can be calibrated and
validated using measured or modeled hillslope erosion rates and measurements from instream
monitoring or sediment detention basin surveys.
4.3.1.2
Modeling Soil Erosion by Water
4.3.1.2.1
USLE, RUSLE, and WEPP
The most common method used for estimating soil erosion by water on military lands is the Universal
Soil Loss Equation (USLE) (Wischmeier et al. 1978). The USLE equation is:
A = R x K x LS x C x P
where
A
R
K
LS
C
P
= annual soil loss from sheet and rill erosion in tons/acre
= rainfall erosivity factor
= soil erodibility factor
= slope length and steepness factor
= cover and management factor
= support practice factor
Estimating soil loss with the USLE requires information on site and climatic factors which are
generally fixed over time (i.e., slope length and steepness, soil erodibility, and rainfall erosivity),
factors which may change over time such as soil surface cover/configuration and vegetation cover
(cover and management factor), and conservation practices designed to mitigate soil erosion losses.
Site, soil, and climatic characteristics can usually be obtained from site measurements and published
soil surveys. If no conservation practices are implemented or those in place remain constant over
time, then the cover and management factor (C factor) is the one variable that has the greatest
influence on how soil loss estimates change from year to year.
All factors, except the cover and management factor and slope length factor, can be obtained from
published sources, typically a Natural Resources Conservation Service (NRCS) county soil survey.
The following subfactors are required by the USLE to calculate the C factor for a particular site:
•
•
•
percent plant canopy cover
percent plant, litter and rock on soil surface (ground cover)
average minimum canopy height (drip height)
The Revised Universal Soil Loss Equation (RUSLE) essentially replaces the USLE as the preferred
empirical equation for computing average gross erosion rates across landscapes.
Like the USLE, the soil loss computed by the RUSLE is the amount of soil lost from a landscape
profile described by the user. A landscape profile is defined by a slope length, which is the length
from the origin of overland flow to the point where the flow reaches a major flow concentration or a
major area of deposition. The computed soil loss is an average erosion rate for the landscape profile.
Erosion can vary widely even on a uniform slope, depending on slope position and configuration of
the slope profile. Neither the USLE nor RUSLE estimates deposition or the amount of sediment
148
leaving a field or watershed; only soil movement at a particular site is estimated. The Modified USLE
(MUSLE) applies an empirically-derived runoff coefficient from individual storm events to predict
sediment yield from a watershed. However, its application is limited and coefficients must be derived
for individual watersheds.
The RUSLE maintains the same basic six-factor structure as the original USLE. Each factor has been
either updated with recent information, or new factor relationships have been derived based on
modern erosion theory and data. Like the USLE, RUSLE does not explicitly consider runoff or the
individual erosion processes of detachment, transport, and deposition. All of the equations used to
derive the factors have been modified and enhanced to account for a variety of field conditions.
Procedures for computing each factor from basic data have also been developed for cases where
published values are not readily available. Finally, the procedures and factor values have been
computerized into a variety of automated programs and routines to assist the user with selecting
parameter values and performing the RUSLE computations (Renard et al. 1997).
Major changes to the USLE incorporated into RUSLE include:
R factor: new and improved isoerodent maps and erodibility index (EI) distributions for some
areas
K factor: time-variant soil erodibility which reflects freeze-thaw in some geographic areas
LS factor: new equations to account for slope length and steepness
C factor: additional sub-factors for evaluating the cover and management factor for cropland and
rangeland
P factor: new conservation practice values for cropland and rangeland
Data requirements for calculating the C factor include:
1. Effective root mass in top 4” of soil (lb./acre) or estimate of annual site production potential
(lb./acre) to generate a root mass value for a given plant community
2. Percent canopy
3. Average fall height (ft)
4. Surface roughness value (index of average micro-elevation): values generally range from 0.3 to
1.5
5. Percent ground cover (rock + litter, excluding plant basal cover)
The improvements made by RUSLE make the USLE-based erosion prediction model generally more
applicable across a wider range of landscapes, both geographically and scale-wise. The effect of
RUSLE on USLE soil erosion estimates will vary by location. Although a total generalization cannot
be made, RUSLE estimates of gross soil erosion have been found in some cases to be less than those
computed using USLE (Lal et al. 1994; Jones et al. 1996). The automated capabilities using the
RUSLE software facilitate data input and parameter selection. The software is available through
USDA-ARS or can be downloaded from the RUSLE website
(http://fargo.nserl.purdue.edu/rusle2_dataweb/RUSLE2_Index.htm).
Original RTLA methods were designed in part to collect information necessary to calculate soil
erosion potential based on the USLE (Tazik et al. 1992). The following data is not currently collected
or part of the original RTLA methodology; additional field data collection (FC) or information
gathering (IG) is required to calculate RUSLE estimates for RTLA plots.
R Factor
• consult
updated R value maps or consult with local NRCS staff (IG).
149
K Factor
• no
additional data required.
LS Factor
• choose
C Factor
• estimate
appropriate LS equation based on topography, land use, rill/interrill
ratios (FC).
effective root mass in top 4” of soil (lb./acre) or annual site production
potential (lb./acre) to generate a root mass value for a given plant community [IG
(e.g., soil survey) and/or FC]. A potential productivity condition rating could be
developed which adjusts for site condition.
surface roughness value (index of micro-elevation in inches): generally ranges
from 0.3 to 1.5 (FC).
•
surface cover function B-value code: represents the relative effectiveness of
surface cover for reducing soil loss. The choice of B value is based on the ratio of
rill/interrill erosion under bare soil conditions (IG).
•
P Factor
•
collect information on conservation practices in place (FC).
The USLE and RUSLE are reliable tools for land managers because they are relatively easy to use,
widely applied, and generally accepted by the natural resources community. A conservative approach
is to use USLE or RUSLE to look at trends in erosion estimates at particular locations. The user
would thus have an indication of relative changes in soil loss at a particular site. The absolute values
of the estimates thus become less important as the emphasis shifts to trends of degradation or
improvement. However, this type of approach may constrain the spatial extrapolation of site
estimates. The number of sample sites used for erosion estimation should be validated statistically in
order to allow for spatial extrapolation with a high level of confidence.
The Water Erosion Prediction Project (WEPP) represents the culmination of several decades of
research, field studies and model development by the U.S Department of Agriculture and its
subordinate organizations, primarily the Agricultural Research Service (ARS) and the Natural
Resources Conservation Service (NRCS), formerly the Soil Conservation Service (SCS). WEPP
model development began in 1985 and was fully documented and implemented in 1995. WEPP is
intended to replace the USLE/ MUSLE/RUSLE models and expand the capabilities for erosion
prediction in a variety of landscapes and settings. WEPP also contains features and components that
have been derived from or similar to other accepted erosion models, to include USLE, CREAMS,
EPIC, SWRRB, ANSWERS and AGNPS. The model can be applied to crop land, pasture land, range
land, forested land, and lands disturbed by construction and mining. The model has had only limited
testing on military lands but several studies are ongoing.
4.3.1.2.2
Erosion Status (ES) and Erodibility Index (EI)
Soil type and climate largely determine the ability of lands to remain productive over the long term
despite soil erosion losses. Therefore, the concept of erosion status was developed to standardize
losses relative to allowable or “tolerable” soil losses. Soil loss tolerance (T) values ranging from 1 to
5, represented in tons/acre/year, are published for most soils. Erosion status is the estimated erosion
rate (A) calculated by USLE/RUSLE divided by T.
150
An erodibility index (EI), also referred to as a potential erosion map, can be a valuable tool for
planning and land management. The required data are largely fixed for the area being examined.
Climatic, soil, and topographic characteristics remain largely unchanged over time, and no data
regarding plant cover and soil surface attributes or condition are necessary for the calculation.
Estimates can be calculated for specific sites, slopes, or locales, or for larger areas using GIS on a cell
by cell basis. The calculation is very similar to that for erosion status (ES), but eliminates C (cover
management factor):
EI = (R * K * LS) / T
The results of the EI calculation are unitless, and represent an index of potential erodibility. The EI
represents the inherent erosion potential of the land when no vegetation is present and accounts for
allowable losses (T) among soil series. An EI can be used as a relative measure of erodibility, and can
help to identify areas that, if disturbed, are more susceptible to long-term damage. The methodology
for ES and EI is described by Warren et al. (1989). Published values for R, T, and K are used most
often. A map of LS values can be generated using various GIS software tools, or collected in field for
small field-scale or hillslope applications. Despite some of the difficulties in obtaining quality data
layers for all of the required elements, especially LS, EI estimates are still useful in examining
relative erosion potential at the landscape scale. Additional sources of LS values should be
investigated to improve the accuracy of the LS values used to develop the EI map.
4.3.1.2.3
Selecting a Model
This section is based on The Soil Erosion Model Guide for Military Lands: Analysis of Erosion
Models for Natural and Cultural Resources Applications (Doe et al. 1999), prepared as part of an
ongoing initiative within the Department of Defense (DoD) to evaluate soil erosion prediction
methods and technologies for application on military lands. The overall goal of this initiative is “to
determine the best methods to predict soil erosion by wind and water on DoD installations over
applicable spatial and temporal scales as a function of both human and natural activities”.
(USACERL 1997) recently examined existing technologies and needs for evaluating soil erosion on
military lands.
Doe et al. (1999) evaluated twenty-four soil erosion models which were identified as having potential
for use by military land managers. The models were reviewed and evaluated against a set of criteria
which provide the users with guidance on which models may best support their intended applications
related to soil erosion prediction, planning and mitigation. It also discusses the linkage of models to
geographic information systems (GIS) and user interfaces to facilitate data input and analysis.
Most soil erosion models are developed for specific types of land use activities or application. As
their use becomes more universal and accepted in practice, they may be applied to other types of soil
erosion issues or problems, albeit with appropriate caution and validation. In developing or selecting
a specific soil erosion model for application, model developers and potential users must ask the
following questions:
1) Who will use the model?
2) What equipment (computer hardware, software, etc.) is available to run the model and analyze
its results?
3) What is the users’ level of technical competency with models?
151
4) What is the most common scale (space and time) of the application?
5) What level of accuracy and precision is required or desired?
6) How are the results to be interpreted and used?
7) What are the sources and availability of data to support the models?
A classification of 24 erosion models, based on mathematical formulation, spatial structure, temporal
structure, and scale of application is presented in Table 4-8. Reference links to websites that support
each model are presented in Table 4-9. Some of the criteria presented in Table 4-8 are fundamental in
the selection of a model (e.g., spatial scale of application) whereas others are largely academic
preferences (mathematical formulation) which affect predictions. It is unlikely that any single model
can perfectly fit or accomplish all of the applications intended for complex land management
decision- making. The best solution may be to identify a suite of erosion models that can be best
applied to specific questions or land management problems (Doe et al. 1999). Model selection may
differ according to the scale of the problem or from one climatic regime to another.
152
Table 4-8. Erosion model classification matrix (reprinted from Doe et al. 1999).
APEX
ALMANAC
RUSLE
AGNPS
MUSLE
USPED
CREAMS
SWRRB
SPUR
SWAT
x
x
x
x
x
x
x
HUMUS
GLEAMS
CASC2D
MULTSED
ARMSED
WEPP profile
watershed
SIMWE
ANSWERS
KINEROS
EUROSEM
SHE
SEMMED
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
landscape/regional
continuous
single event
distributed parameter
x
Scale of Application
watershed
x
x
x
x
x
x
x
x
EPIC
Structure
(Temporal)
field scale
USLE
Structure (Spatial)
lumped parameter
MODEL NAME
physical/process -based
empirical
Mathematical
Formulation
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
153
Table 4-9. Website addresses for soil erosion models (updated from Doe et al. 1999).
MODEL NAME
WEB REFERENCE
AGNPS
Http://dino.wiz.uni-kassel.de/model_db/mdb/agnps.html
http://www.ars.usda.gov/Research/docs.htm?docid=5199
Http://pasture.ecn.purdue.edu/~aggrass/models/agnps/Index.html
http://www.ars.usda.gov/Research/docs.htm?docid=9760
Http://pasture.ecn.purdue.edu/~aggrass/models/answers/
http://www.ars.usda.gov/Research/docs.htm?docid=9792
http://gcmd.nasa.gov/records/CASC2D.html
http://www.engr.colostate.edu/~pierre/ce_old/Projects/CASC2DRosalia/Index.htm
Http://dino.wiz.uni-kassel.de/model_db/mdb/creams.html
http://eco.wiz.uni-kassel.de/model_db/mdb/epic.html
http://www.ars.usda.gov/Research/docs.htm?docid=9791
http://www.silsoe.cranfield.ac.uk/nsri/research/erosion/eurosem.htm
http://www.ars.usda.gov/Research/docs.htm?docid=9797
http://www.wiz.uni-kassel.de/model_db/mdb/gleams.html
http://eco.wiz.uni-kassel.de/model_db/mdb/humus.html
http://gcmd.nasa.gov/records/HUMUS.html
Http://Dino.wiz.uni-kassel.de/model_db/mdb/kineros.html
http://www.iwr.msu.edu/rusle/
http://www.ott.wrcc.osmre.gov/library/hbmanual/rusle.htm
http://fargo.nserl.purdue.edu/rusle2_dataweb/RUSLE2_Index.htm
Http://www.frw.ruu.nl/fg/demon.html
http://skagit.meas.ncsu.edu/~helena/gmslab/reports/CerlErosionTutorial/de
nix/Examples/simulation_of_land_use_impact.htm
http://topsoil.nserl.purdue.edu/nserlweb/isco99/pdf/ISCOdisc/SustainingT
heGlobalFarm/P081-Mitasova.pdf
http://eco.wiz.uni-kassel.de/model_db/mdb/spur.html
http://www.ars.usda.gov/Research/docs.htm?docid=9793
Http://www.brc.tamus.edu/swat/
Http://www.brc.tamus.edu/swatgrass/index.html
http://eco.wiz.uni-kassel.de/model_db/mdb/swrrbwq.html
http://topsoil.nserl.purdue.edu/usle/index.html
http://pasture.ecn.purdue.edu/~engelb/agen526/USPED.html
http://skagit.meas.ncsu.edu/~helena/gmslab/denix/denix.html
http://topsoil.nserl.purdue.edu/nserlweb/weppmain/wepp.html
ALMANAC
ANSWERS
APEX
CASC2D
CREAMS
EPIC
EUROSEM
GLEAMS
HUMUS
KINEROS
RUSLE
SEMMED
SIMWE
SPUR
SWAT
SWRRB
USLE
USPED
WEPP
Graphical User
Interfaces
MOSES
WMS
http://pasture.ecn.purdue.edu/~meyerc/MOSES/
http://www.bossintl.com/html/wms_overview.html
http://www.ems-i.com/WMS/WMS_Overview/wms_overview.html
While the technology and modeling competency of land managers will undoubtedly improve with
future developments and experience, the applications will most likely remain constant. These
applications can be classified as either (1) predictive or (2) evaluative. The predictive applications
154
will be used for both short and long-term planning horizons to provide land managers and land
users with an understanding of how military activities may impact soil erosion on their lands
internally, as well as consequent trans-boundary impacts. These impacts may include downstream
sedimentation and degradation of water quality in streams or air pollution caused by windtransported materials from an installation site to an off-site community.
Riggins and Schmitt (1994) describe various applications of predictive and evaluative soil erosion
modeling. Modeling can be used in the predictive sense by military land managers to:
•
•
•
•
•
•
•
Calculate the erosion thresholds for a specific watershed, training/testing area or
installation-wide
Calculate expected long-term average annual soil loss for a given parcel of land
Calculate expected soil loss for an interval (monthly, seasonal or training rotation)
Calculate expected soil loss from a single storm (rainfall-runoff) event or single
military exercise
Compute sediment yield, either annually or for a single event, from a watershed
Determine the locations within a watershed or training/testing area that are most
sensitive (from a soil erosion perspective) to specific military activities
Examine potential responses in soil erosion resulting from changes in land use or
climatic change
Modeling can be used in an evaluative sense to:
•
•
•
Measure and compare the effects of implementing soil erosion mitigation practices
Monitor and evaluate watershed stability and ecological health over time
Test and evaluate data collection methods and instrumentation
Doe et al. (1999) concluded that several models have high potential for solving the unique erosion
problems found on military lands. While established empirical models, such as the Revised
Universal Soil Loss Equation (RUSLE), continue to have useful applications for some purposes,
the study recommends that several of the new generation of physically-based, distributed
parameter models have the greatest potential for use by DoD land managers. In particular, the
following models are recommended for use depending upon the specific application requirements:
♦
♦
♦
♦
♦
4.3.1.3
AGNPS
CASC2Dd
SIMWE
SWAT
WEPP
Modeling Soil Erosion by Wind
The wind erosion estimation method most often used, the Wind Erosion Equation (WEQ) with
various modifications, was developed for agricultural land applications by Woodruff and
Siddoway (1965). The most important factors affecting wind erosion, especially in arid and semiarid areas are soil aggregation, soil moisture, soil surface, vegetative cover, large, open expanses,
and strong winds (Skidmore 1994). The basic wind erosion formulation is presented below:
155
E = f(I, K, C, L, V)
E = potential average annual soil loss
I = erodibility index
K = soil ridge-roughness factor
C = climatic factor
L = unsheltered median travel distance of wind across a field
V = equivalent vegetative cover
The Revised Wind Erosion Equation (RWEQ) was initially developed by the Agricultural
Research Service (ARS) beginning in 1991 using advanced technology from the Wind Erosion
Prediction System (WEPS). WEPS is a daily wind simulation model applicable to field-scale
applications. A description of WEPS is provided by Hagen (1991) and Hagen et al. (1995). WEPS
was developed primarily for cropland applications, and is currently being enhanced for application
to disturbed lands (Hagen 1997). The basic form of the RWEQ is:
Average Soil Loss = f(weather, soils, crop, management)
The following subfactors are applied within each factor:
WEATHER
wind velocity above threshold
wind direction
wind preponderance
air density
average air temperatures
solar radiation
days with snow cover
rainfall amount
rainfall erosive intensity (EI)
number of rain days
irrigation (amount, rate, number of days)
SOILS
soil erodible fraction
soil wetness
oriented roughness
random roughness
soil roughness decay
soil crust factor
surface rock cover
CROPS
flat residues (percent soil cover)
standing residues
crop canopy
residue decomposition
156
Application of wind erosion prediction technologies and models to military land management are
limited. Recent experimental approaches to evaluate the effects of simulated off-road maneuvers
on wind erosion potential have been undertaken at the Orchard Training Area, Idaho (Grantham
2000) and Fort Bliss, Texas (Marston 1986).
4.3.2 Monitoring Noxious and Invasive Plants
4.3.2.1 Background
Federal agencies are directed under Executive Order (EO) 13112 to prevent the introduction,
provide control, and minimize the economic, ecologic, and human health impacts of invasive
species (Clinton 1999). In response to EO 13112, Army Policy Guidance for Management and
Control of Invasive Species was distributed in June 2001. The memorandum states: “Invasive
species can be a threat to natural resources, impact local economies, and present problems for the
military mission.” Invasive species include plants, animals, and other organisms (e.g., microbes).
In the context of this document, the discussion of monitoring is limited to plants. The requirements
for implementing invasive species management on military land is identified in the U.S. Army
Environmental Program Requirements under the Sikes Act (natural resources stewardship), the
Endangered Species Act (protection and management of listed species and critical habitat), the
Clean Water Act (effects of invasive species on erosion control and wetlands), and other
documents. According to the Sikes Act, installations are required to monitor invasive species
populations, track the presence and status of invasive species over time to determine when control
measures are necessary, and evaluate the effectiveness of prevention, control/eradication, and
restoration measures.
In this document, noxious, invasive and other non-native plant species are referred to collectively
as “invasives” or non-indigenous plant species (NIS). The establishment and spread of NIS is a
primary management concern on military and other public lands, and can have significant
ecological and economic impacts. Noxious plants are those that have been identified by state and
Federal agencies as having the capability to pose serious threats, primarily to agriculture and
wildlife. Many non-indigenous (i.e., non-native, alien, exotic) plant species exist on military
installations. Most are non-invasive and do not compete excessively with native plants. Others are
considered invasive by land managers or are classified as noxious by government agencies, and
are therefore prime candidates for monitoring and management efforts.
Invasive species alter ecosystems by changing fire regimes, degrading wildlife habitat, displacing
native species (including threatened and endangered species), altering soil properties and
processes such as nutrient status and soil erosion, changing vegetation structure (e.g., ability to
penetrate vegetation, presence/absence of physiognomic groups), and adversely affecting native
biodiversity. NIS can also adversely impact military operations, reduce military carrying capacity,
compromise long-term sustainability of training lands, diminish training realism, and restrict
training land availability. For example, Scot’s Broom (Citysus scoparius), a dense woody shrub
that invades open grassland training areas in the Pacific Northwest, changes the dynamics of lineof-sight, mounted, and dismounted training. Cheatgrass (Bromus tectorum) has altered fuel loads
and fire regimes in the West and Intermountain West, affecting both erosion and restrictions on
live-fire training. Yellow Star Thistle (Centaurea solstitialis) forms impenetrable thickets to 2m
tall in central California, reducing training realism and making mounted and dismounted
157
maneuvers difficult. Westbrook and Ramos (2005) documents the impacts of invasive species
using twelve case studies of Army, Navy, Air Force, and Marine Corps installations.
Information regarding the management of NIS is widely available in the literature, on the Internet,
and from local, state, federal, and non-governmental agencies. Guidance for Non-Native Invasive
Plant Species on Army Lands: Western United States (USACE 2003a) and Guidance for NonNative Invasive Plant Species on Army Lands: Eastern United States (USACE 2003b) are good
starting points for understanding NIS management on Army lands, and contain lists (one eastern,
one western) of primary and secondary species of concern. The species lists and accompanying
management abstracts were developed from federal and state data, expert opinions, and invasive
qualities of species. The Primary lists contain species with wide distributions, strong negative
ecological impacts, and that are currently being actively controlled by land managers. The
Secondary lists contain species that are increasing at such a rate that they will soon have a similar
impact as Primary species.
Awareness of potential invasive species establishment is the first line of defense. Actions that
prevent the introductions of invasive species (e.g., use of wash racks, inspection of clothing, etc.)
are often the easiest and most cost effective way to combat these species. Understanding and
anticipating pathways facilitates early detection and rapid response, which are critical to
effectively managing invasive species. Monitoring is an integral part of NIS management efforts,
and must be scientifically valid to be useful in adaptive management.
4.3.2.2 Standardized Approaches and Data Elements
The US Forest Service, the Bureau of Land Management, the National Park Service and other
federal and state agencies have developed methods and guidance for monitoring non-native
invasive plant species. Some programs share common methods and data elements, and most
incorporate information about the location, spatial distribution, and abundance of invasive plants.
Many invasives monitoring efforts focus on mapping and inventory. With maps or inventory
information, a strategy focused on removing new and isolated infestations and containing the
principle infestation can be developed. Once contained, the size of the infestation is reduced,
working from the outside in (NAMWA 2002). It is important to remember that each procedure
that has been developed is designed to meet specific requirements. Some are designed for use by
professionals, volunteers, landowners, or a combination thereof. In many cases, the requirements
may encompass core data only. Additional data collection (additional attributes or detail) may be
necessary to meet local monitoring needs.
Mapping standards developed by the North American Weed Management Association (NAMWA)
have been incorporated into invasive species inventory and monitoring protocols by the US Forest
Service (USDA Forest Service 2002a), the National Park Service, the US Fish and Wildlife
Service, and other public and private organizations. The standards were designed to be compatible
with most existing invasive species inventories, and to facilitate information sharing across
ownership and management boundaries. In most cases, NAMWA data are combined with
additional site-specific data when inventorying and mapping nonnative plant species. The
NAMWA system includes the following attributes:
1. Date
2. Examiner
3. Plant Name
158
4. Common Name (optional)
5. Plant Codes - Use of codes from the USDA PLANTS database is highly encouraged.
6. Infested Area (with units) – area of land containing a single weed species. Area of
land containing one weed species. An infested area of land is defined by drawing a
line around the actual perimeter of the infestation as defined by the canopy cover of
the plants, excluding areas not infested. Areas containing only occasional weed plants
per acre do not equal one acre infested. Generally, the smallest area of infestation
mapped will be 1/10th (.10) of an acre or 0.04 hectares but some users record infested
areas as small as .01 acre or 0.004 hectares. It is highly recommended that only a
single weed species be entered for each infested area (NAMWA 2002).
7. Gross Area (with units, optional) – This field is intended to show general location and
population information. Like Infested Area, it is the area of land occupied by a weed
species. Unlike Infested Area, the area is defined by drawing a line around the general
perimeter of the infestation, not the canopy cover of the plants. The gross area may
contain significant parcels of land that are not occupied by weeds. Gross area is used
in describing large infestations. When a value is entered for gross area, the assumption
is that the area within the perimeter of the weed population (area perimeter) is an
estimate or the product of calculating the area within a described perimeter. It is not a
measured value. If a value for Gross Area is entered, a value for Infested Area must
still be entered. The value for Infested Area is derived from estimating the actual or
percentage of land occupied by weed plants (NAMWA 2002).
8. Canopy Cover – Canopy is estimated as a percent of the ground covered by foliage of
a particular weed species. A variety of methods are acceptable; ocular estimates using
cover classes are most commonly used. Cover will be recorded as a numeric value. If
the inventory procedure includes the use of cover classes such as Greater Yellowstone
Area Cover Class System (<1%, 1-5%, 5-25%, 25-199%), 10 point cover classes, or
Daubenmire codes, the mid point of the cover class will be entered as the cover value.
9. Ownership
10. Source of the Data
11. Country
12. State or Province
13. County or Municipality
14. Hydrologic Unit Code – required for aquatic species only
15. NAMWA optional data fields include:
• Location – legal, Latitude and Longitude, UTMs
• Quad Number
• Quad Name
• Area Surveyed
The NAMWA standards are designed to provide a minimum set of data, and the collection
additional biotic, abiotic (i.e., site), and resource management or land use information may also be
beneficial. Additional attributes might include the following:
•
•
•
•
•
Topographic position (slope, aspect, landscape position)
Bare ground
Distance to features such as roads, streams, livestock water sources, etc.
Invasive growth stage (e.g., seedling, bolt, bud, flower, seed set, mature
(Roberts et al. 1999)
Plant community seral stage
159
•
•
•
•
•
Disturbance evidence, extent, and severity (military disturbance, fire, flood,
small mammals, wind throw, mortality from insects/disease, etc.)
Land use activities (e.g., grazing, other management)
Dominant plant community type
Plant community structure (presence/abundance of structural life forms, etc.)
Management or other administrative unit
Montana State University Extension has produced a booklet that describes a protocol for weed
mapping with GPS units. The free booklet contains detailed information on statewide mapping
procedures, using and submitting data, and data recording and management, including computer
mapping and GPS. Minimum NAMWA standards are included as well as some additional fields.
A metadata form for recording weed data collection is included. The protocol was revised in 2001
and can be downloaded at http://www.montana.edu/places/mtweeds. The protocol uses the Greater
Yellowstone Ecosystem Cover Class System (<1% cover, 1-5% cover, 5-25% cover 25-100%
cover) and accommodates both hand drawn and GPS maps. The protocol is presently being
applied outside Montana as well.
Data collection for these systems may consist of GPS-based data recorders or PDA systems with
GPS accessories. In some cases, mapping may be done manually using aerial photographs and
color coding of each species. Data is subsequently downloaded to databases linked to GIS. In this
way, data fields can be easily checked for quality, queried, displayed, and shared. See Section 6
Electronic Data Collection Tools for further guidance.
The mapping systems described above provide important information for management, but have
limitations as well. The primary limitations of the mapping systems include:
1)
2)
3)
4)
The use of broad cover class systems render the data insensitive to
relatively small changes in plant cover abundance.
Locations surveyed are for the most part subjectively located, often biased
to roads and riparian corridors.
The determination of infestation occurrence and mapping boundaries may
vary widely among observers.
The expertise of observers varies widely, influencing the accuracy of area
and cover estimates and plant identification accuracy.
In addition to these observations, Stohlgren et al. (2002) recommend the use of standard circular
plots at every fifth to tenth polygon mapped, and the addition of at least five random samples
within major strata to capture large-scale patterns in a statistically valid way. Stohlgren et al.
(2002) also recommend the adoption of quality control and quality assurance procedures to
evaluate observer bias and accuracy.
4.3.2.3 Setting Priorities
The priority to detect and control or eradicate individual plant species is a function of regulatory
status, ecosystem threats, species biology, potential economic/agronomic effects, cost of control,
impact on training activities, and other considerations. Many species are not classified as noxious
or invasive but may be considered serious threats by land managers. These species may be
widespread or found in discrete locations. Incipient NIS populations or incidences are those that
have recently established and are typically located in small patches or locales. It is crucial to
monitor these incipient or early infestations, as small patches of some species can spread rapidly
160
and become expensive and difficult to eradicate or control (Forcella and Harvey 1988; Stein and
Flack 1996). Small or widely scattered infestations are difficult to detect, especially on large
installations. Monitoring for both widespread and incipient NIS is therefore crucial to minimize
effects of NIS and reduce management costs and long-term ecological impacts on native flora and
fauna species and communities. Once a list of known exotic plant species has been obtained for a
given area, each individual species can then be ranked relative to the other species. Species which
pose an immediate threat to natural areas can then be targeted for control efforts, while species
which have small potential impacts are given a lower priority for management. In general, species
that have low invasiveness and a low priority for management and monitoring: 1) have stable or
decreasing populations, 2) primarily colonize disturbed areas and do not readily spread into intact
native vegetation, and 3) will be controlled or eliminated with natural succession or reestablishment of natural processes (especially restoration of fire or hydrologic regime) (Sutter
1997). Several ranking systems have been developed to help prioritize threats from invasives in an
objective manner.
The Handbook for Ranking Exotic Plants for Management and Control (Hiebert and Stubbendieck
1993) was developed to provide land managers with a tool to effectively evaluate the potential
impacts of known exotic plant species. The advantage of using this approach is that managers can
objectively evaluate different management strategies based on information obtained from literature
and field surveys. This approach encourages managers to consider the full range of the potential
impacts for their management decisions (Hiebert and Stubbendieck 1993). The benefits of
managing specific exotic plants can be weighed against the potential costs of different
management actions. The ranking system provides a sound justification for management plans,
and can also provide justification for future program authorization and funding (Hiebert and
Stubbendieck 1993). The ranking system uses numerical ratings in an outline format to evaluate
the current and potential ecological impacts and distributions of species in the areas of concern.
The ranking system also evaluates different control options for a given species. Information for the
ranking system can be obtained from both literature reviews and field surveys.
The Alien Plants Ranking System (APRS), based on the Hiebert and Stubbendieck (1993) system,
is a cooperative effort by the National Park Service, Northern Arizona University, Ripon College,
University of Minnesota, and the U.S. Geological Survey
(http://www.npwrc.usgs.gov/resource/2000/aprs/aprs.htm). APRS is a computer program that
helps managers prioritize decisions concerning invasive non-native plants. It is an analytic tool
that can be used to: 1) separate harmless non-native plants from invasive, harmful non-native
plants at a site; 2) identify weeds that could impact a site in the future; and 3) address control of
each weed at a site so that control costs can be weighted against the weeds' impact. APRS was
developed and first tested in grassland and prairie parks in the central United States. However, the
program has been robustly designed to apply to other ecoregions (APRS Implementation Team
2000).
Another well-documented system for assessing threats is the Invasive Species Assessment
Protocol (Morse et al. 2004). This protocol was developed cooperatively by NatureServe, The
Nature Conservancy and the U.S. National Park Service as a tool for assessing, categorizing, and
listing non-native invasive vascular plants according to their impact on native species and natural
biodiversity in a large geographical area such as a nation, state, province, or ecological region.
This protocol is designed to make the process of assessing and listing invasive plants objective and
systematic, and to incorporate scientific documentation of the information used to determine each
species’ rank (Morse et al. 2004). The protocol is used to assign each species an Invasive Species
Impact Rank (I-Rank) of High, Medium, Low, or Insignificant to categorize its negative impact on
161
natural biodiversity within that region. The protocol includes 20 questions grouped into four
sections: Ecological Impact, Current Distribution and Abundance, Trend in Distribution and
Abundance, and Management Difficulty (Morse et al. 2004).
Regardless of the system used to prioritize the threat and management needs for invasives,
decisions regarding management and monitoring of NIS ultimately lie with individual land
managers.
4.3.2.4
Types of Monitoring Activities
Monitoring activities undertaken to document invasive plants can be grouped into five main types.
The following program areas could be part of an RLTA weed monitoring strategy 7 and form the
basis for detailed monitoring objectives:
1) Occurrence - Document the presence/absence of particular invasives in a defined area (e.g.,
installation-wide, specific communities or training areas, road segments, etc.) and record the
arrival of additional NIS over time.
2) Abundance - Estimate the current abundance (area occupied, density, cover, etc.) of
invasives of interest within specific areas or sample plots.
3) Expansion and Trends - Assess changes in the abundance or expansion of certain invasives
over time and identify which habitats are colonized by particular invasives. Weed
populations can spread over time in upland and wetland environments, along roads and
trails, and in response to various disturbance. Separate infestations can grow together to
form a single large infestation. An infestation can split to form several populations where
only one previously existed. Finally, the size and shape of an infestation and the abundance
of invasives within it can change over time.
4) Biology and Ecological Responses - Document NIS biology such as longevity, seedling
survival, seed production, response to disturbance, and herbivory. This may also include
evaluating relationships between changes in the abundance of invasives and ecosystem
properties/changes or land use disturbance (e.g., altered fuel loads, soil disturbance
severity).
5) Management Effectiveness - Assess the effects of control measures on the abundance of
invasives and native species. Compare different weed control measures in terms of cost,
effectiveness, or impact on native flora and fauna.
In addition to the mapping systems described above, a number of other approaches have been used
to monitor invasives. For example, Haber (1997) describes some examples of projects applied to
specific needs:
1. Linear highway surveys and trail monitoring. Roadside surveys can be conducted using a
continuous recording approach or by sampling transects radiating out from the road at
predetermined intervals or locations. These projects proved to be an effective use of volunteer
labor. Although it concentrates on only a portion of the landscape, this type of survey can be
effective to detect high priority infestations in high-probability areas. This can be an important
component of an early warning and invasive species eradication program, especially in areas
with high ecological value.
7
Adapted from Ainsworth (1999) and Haber (1997).
162
2. Population expansion. Various methods can be used to document the rate of expansion of a
single plant, patch (typically clonal) or population of an invasive species. The approach varies
with the species being monitored (form, longevity, reproductive strategy. Examples are
provided for monitoring:
i. Basal area expansion of a perennial plant
ii. Basal area expansion of a clump or distinct patch
iii. Rate of propagation of a floating aquatic species
iv. Expansion of herbaceous plants (annuals or biennials)
3. Monitoring impacts of invasives on native vegetation. These examples provide a basic
framework for monitoring activities that can be modified to suit individual interests or needs
of the region. Projects can be undertaken solely on species that are presently known to be
invasive in natural habitats or a broader focus can be maintained by monitoring other exotics
whose invasive potential have not been clearly, or at all, established.
4.3.2.5
Developing Objectives
Management objectives provide the rationale and direction for an invasive species management
program, including monitoring efforts. Objectives can be refined as more information becomes
available. Management objectives should identify the species (individual or collectively), where
the species of interest are located (e.g., installation-wide, in a single training area or watershed,
other discrete locations), a specific attribute to measure or estimate (e.g., area occupied, density,
cover, frequency, etc.), the desired trend over time (e.g., maintain or decrease and by how much),
and the time frame needed for the management to prove affective. How a species is monitored is
dependent on the monitoring objectives (see Chapter 2). Objectives are typically classified as
target/threshold objectives (e.g., an estimate of population size) or change/trend objectives (e.g.,
intended to document a change or a specified size over time). Monitoring objectives often include
a target level of precision, the desired magnitude of change, and the amount of acceptable error.
This process supports adaptive management and bolsters sustainable training by providing
defensible data.
If the sampling interval is not specified in the management objective, it should be specified in the
monitoring objective (i.e., seasonally, annually, every 2 years, 5 years, etc.). The sampling interval
can be less than the timeframe specified in the management objective. For example, if a given
change is desired over a 6 year period, monitoring every 2 or 3 years may be appropriate to see if
there has been progress toward the objective. When monitoring does not involve sampling, the
management objective should provide enough information to evaluate its success or failure. This is
the case where qualitative assessments are done for areas or where a complete census is
performed. Management objectives of this type therefore do not need to provide additional
components beyond what, where, and when. Examples of management and monitoring/sampling
objectives for several invasives management goals are presented below.
Management Objective: Maintain a frequency of Chinese lespedeza (Lespedeza cuneata) of less
than 5% in Training Area B5 for the next 5 years.
Monitoring Objective: Estimate the frequency of Chinese lespedeza annually within 10 percent of
the true mean with 90 percent confidence.
Management Objective: Maintain the current extent of Lehmann lovegrass (Eragrostis
lehmanniana) infestations on East Range for the next 5 years. (infestations are identified using
specific criteria that can be developed locally)
163
Monitoring Objective: Map the all infestations of Lehman lovegrass every 2-3 years (or at a
minimum this year and five years from now) using specified mapping guidelines.
Management Objective: Decrease the occurrence (frequency) of invasive species along roads in
Training Area Alpha by 30% over the next five years (2004-2008).
Monitoring Objective: Estimate the frequency of invasives with a precision of 85% and 90%
confidence.
Management Objective: Maintain (or decrease) the current distribution and areal extent of target
noxious and invasive plant species (see list) on Fort Harrison from 2005 to 2009. This objective
could be treated as quantitative or qualitative, depending on the approach that is most feasible
(target objective).
Monitoring Objective: Map all occurrences of target noxious and invasive plant species on Fort
Harrison in 2005 and 2009. For each species, compare the number of occurrences and total
acreage over time to assess whether or not the management objective has been achieved.
Management Objective: For sites treated for infestations, decrease the cover of Cogon grass
(Imperata cylindrical) by at least 75% one year after treatment (change objective).
Monitoring Objective: We want to be 90% sure of detecting a 75% change in the cover of Cogon
grass in treated areas. We are willing to accept a 20% chance that we conclude a change took
place when in fact there was no change.
Management Objective: Maintain the number of km of road shoulders with a knapweed (all
species) ranked abundance of 5 or more (target objective).
Monitoring Objective: Perform ranked abundance surveys for knapweed species on all roads
annually.
Management Objective: Obtain a 50% decrease in yellow star thistle in areas sprayed with
herbicides on Fort Hunter Liggett during the next three years.
Monitoring Objective: We want to be 90% sure of detecting a 50% change in the density of yellow
star thistle on treated and untreated areas annually for the next 3 years. We are willing to accept a
10% chance that we conclude a change took place when in fact there was no change.
Management Objective: In grassland communities, decrease the frequency of Bromus tectorum
(cheatgrass) by 30% from 2005-2010.
Monitoring Objective: We want to be 80% certain of detecting a 30% decrease in cheatgrass
frequency with a false change error rate of 0.20.
Management Objective: Decrease the ranked abundance of Lythrum salicaria (purple loosestrife)
in each of the four permanent macroplots at the Ives Road Fen site by 2 rank classes between 2006
and 2008.
Monitoring Objective: Estimate the ranked abundance of purple loosestrife in each macroplot. In
this case, the objective has all the information for evaluating results. Estimates could be made
annually or in 2006 and 2008.
4.3.2.6 Invasive Species Monitoring Guidelines
Weed monitoring projects may have considerations beyond those applied to monitoring vegetation
in general. However, the same principles of sampling and setting up a monitoring plan for plant
164
communities or populations apply. In some cases, monitoring objectives for NIS may be satisfied
by existing monitoring efforts. As with general monitoring, management and monitoring
objectives drive most sampling and inventory efforts, ecological models are incorporated into the
project, sampling designs and methods are tailored to monitoring objectives and species
characteristics, results are evaluated relative to monitoring objectives, and the process supports
adaptive resource management. A wide variety of methods may be appropriate, ranging from
qualitative to quantitative approaches as well as demographic studies. The selection of methods
will depend on the management and monitoring objectives and other considerations such as the
spatial distribution of NIS (clumped/patchy, random, regular, associated with roads or other
features, etc.), the number of species that will be monitored, the size of the areas being considered,
species biology, the level of detail desired, and the availability of resources and staffing. The
timing of surveys will also be influenced by the visibility and phenology of the species of interest.
The primary differences between general plant monitoring and invasives monitoring are that there
is more emphasis on mapping invasives than native vegetation and ranking invasive species for
management and monitoring is highly emphasized. Moreover, the information gained is directly
related to management of invasives. A generalized approach for initiating monitoring and mapping
of invasive plants is presented in Figure 4-5. A broad range of information may be needed for each
invasive species present, including data on each species’ range, abundance, life history
characteristics, habitat preferences, spread rate, impacts on other species, value, presence of
natural control agents, and response to control actions (Erich 1997).
The following general guidelines for invasives inventory, monitoring, and mapping are adapted
from NPS (2002):
•
Set clear goals and objectives.
•
Integrate mapping, inventory, and data collection and management with other
vegetation management activities to the greatest extent possible.
•
Prior to conducting field inventories, collect all available existing information about
invasives currently or potentially occurring on the installation or surrounding lands. A
list of known and potential invasives should be used when mapping. This information
is essential to the integration of ecological models with protocol development.
•
For mapping and monitoring purposes, distinguish between species that are widely
distributed and well established versus those occurring in discrete locations.
•
If it is not feasible or desirable to address all invasive species, develop a priority list of
target species and or areas. Additional considerations when selecting species include
the ability to see the plants in the field, their growth habit, plant phenology,
distribution in restricted habitats, and other considerations.
•
Base the scale and intensity of inventory and monitoring efforts on the required level
of accuracy (e.g., mapping and inventory) and precision (sampling estimates). All
quantitative sampling should be designed to meet precision or change detection
requirements of monitoring objectives.
165
Figure 4-5. Considerations in initiating an effective weed inventory and mapping program.
(Pamela Benjamin (2002), in Inventory and Monitoring for Invasive Plant Guidelines (NPS 2002)).
•
Focus efforts on new invasives and species that will be most difficult to control if not
managed early. Invasive species that pose the greatest threat to achieving management
goals and those that can be easily controlled are the highest priority species for
management (CNAP et al. 2000). When the species is widespread, often the decision
is to control the species, since the probability of eradication is unlikely (Hobbs and
Humphries 1995) (Figure 4-6).
•
Prioritize areas having high natural or cultural resource values. These areas may
include sensitive or high quality communities, habitats supporting TES, riparian and
wetland habitats, and other areas. Other high priority areas may include those that
have a high potential for invasion, such as roadsides, riparian corridors, and disturbed
areas.
•
Plan a systematic survey to include as much of the installation as possible.
166
•
Installations less than 1,000 ac (400 ha) in size should be inventoried with a high level
of accuracy. Larger installations can be inventoried less intensively, although
intensive surveys should be undertaken in areas where invasives are concentrated,
known invasive corridors, and areas with high natural or cultural values.
•
When possible, structure inventory protocols to ensure they are applicable and
appropriate for future monitoring efforts or repeated measures. The development of
detailed and well-documented protocols is one way to ensure the value of current and
future efforts.
•
Develop systems for documenting areas that are free of invasives as well as those
where invasives are present. The areas included in all surveys should documented for
each mapping or monitoring effort.
•
Design adaptive inventory and monitoring procedures to accommodate patterns of
plant invasions (e.g., types of sites or areas).
•
If possible, expand invasive mapping, inventory, and monitoring outside installation
boundaries. Invasive plants are often well established on adjacent private and public
lands. Information gathered outside the installation boundary may be helpful in
illustrating the relative success of installation natural resources management and
stewardship. It may also identify sources of invasives and promote proactive measures
to reduce their impact on the installation.
Quarantine
Eradication
Species
Abundance
Invasive
Species
Abundance
Documenting and detecting new invasives as they arrive in an area is a critical part of any
monitoring program. Early detection and management can greatly reduce control or eradication
costs and influence the prioritization of invasives (Figure 4-6). Some invasive populations remain
small for a number of years (lag stage), after which an episodic or one-time event (e.g., flood, new
pollinator to an area, above average rainfall or cooler/warmer temperatures than average, etc.) can
promote a rapid population spread. Populations of other species more or less grow continually and
are not noticed until they are widespread (Hobbs and Humphries 1995).
Control
Effective control unlikely
Invasion
Time
Cost
Cost
Figure 4-6. Shift in invasive species abundance and priorities for control. Adapted from Hobbs
and Humphries (1995).
167
Some sampling design considerations for the early stages of plant invasions are presented in Table
4-10.
Table 4-10. Species distribution and sampling design considerations in the initial stages of plant
invasion (Thomas et al. 2002).
Management
Objective
Monitoring
Objective
Invasive Species Distribution
Detect & control
invasive exotics
before establishment
Early detection
of new invasions
Species is known to occur regionally or on
adjacent lands, but has not yet been
confirmed within park. If present at all,
distribution is sparse, limited in extent and
may vary from sparse individuals to dense
patches
Locate and control
establishing
populations
Location of
establishing
populations
Species has been found in small, localized
patches. Finding and controlling patches
might prevent large-scale invasion.
Distribution is somewhat limited in extent and
may vary in intensity from sparse individuals
to dense patches
Target limited control
resources toward
highest priorities
Map invasion
front; identify
areas for
targeted
management.
Species is established and abundant in some
areas. Targeted control to protect sensitive
areas, prevent further invasion, and reduce
the rate of spread along invasion fronts. In
some areas it may no longer be feasible to
locate and control individual patches.
Sampling Design
Considerations
Rare event.
Emphasis on extensive searching,
attention to boundaries/corridors,
rapid assessment
Moderately rare event.
Grid-based or stratified random
sampling with higher selection
probabilities in areas likely to be
invaded
Moderately frequent to common
event.
Variable sampling intensity to
locate and map invasion fronts.
Monitoring the status and trends of known populations or infestations is another component of
most invasives management programs. Even with limited management, knowledge regarding the
distribution, extent, abundance, and ecological influence of invasive plants can help land managers
plan for future projects and document current conditions. Notable trends in invasives and native
desirable species may signal natural ecosystem recovery or degradation. Monitoring status and
trends in populations is sometimes combined with efforts to monitor the effects of management.
For example, monitoring can provide managers with information about the relative success of
different control measures and help weigh the costs of different methods according to their
effectiveness. In some cases, experimental designs employing replications and control may be the
most effective way to document differences in the effectiveness of control methods. However,
despite some drawbacks, less rigorous and less expensive designs employing before and after
measurements may provide adequate data for evaluating the effectiveness of different control
methods. The information gained from these projects help managers adjust their efforts
accordingly.
4.3.2.7
Additional Considerations for Non-Native Invasive Species Monitoring
Phenology of non-native invasive species – Many of the same considerations for threatened and
endangered species monitoring apply to NIS monitoring. The time of year for monitoring should
correspond to the period where the species of interest is at its peak numbers and is visible to
observers. In some cases, this period may be before most native vegetation has developed new
growth (e.g., counting basal rosettes in early spring) or later in the season after native vegetation
has senesced (e.g., counting flower stalks in late summer). If one of the monitoring objectives is to
168
evaluate effects of one of more NIS on native species or communities, then monitoring should take
place during the mid to late portion of the growing season. Once the best period for sampling or
inventory is determined, surveys in future years should be conducted during the same
phenological period.
Precision and Change Detection Capability – Needs regarding the precision and change
detection capability of weed monitoring programs may differ from those for monitoring native
plant communities. In many cases, it is crucial to know the occurrence or relative abundance or
patch size of NIS to guide control efforts. Thereafter, it can be important to know the rate of
expansion for particular invasive plants. This information can be used by natural resource
managers to prioritize species and focus control efforts. However, the ability to detect a 5% vs.
15% change in a particular population or subpopulation is probably not important. The importance
of statistically powerful sampling designs may surface when evaluating the effects of NIS on
native species and communities (Ainsworth 1999).
Selecting a Sampling Universe – The selection of inventory and monitoring areas and the
positioning of sampling sites (i.e., plots, transects, quadrats, etc.) will influence the extrapolation
of results. For example, sampling restricted areas such as roadsides for Dalmatian toadflax
(Linaria dalmatica) may document high densities but are not representative of conditions further
away from roads. Sampling large areas, especially in rugged terrain, is a challenging prospect and
may or may not be appropriate. The focus of the effort, ecology of NIS, and ecological threat level
will help determine the choice of restricted versus large-scale sampling areas. Unless an inventory
is being done, whereby all individuals or populations are documented, samples should be placed in
an unbiased manner using random or systematic procedures. The important point is that the results
apply only to areas having an equal probability of being sampled.
Monitoring Effects of Biocontrol Agents – Biological control (biocontrol) agents are
increasingly used to help manage NIS. In the context of NIS management, biocontrol most often
consists of the introduction of predator insects from regions where the plant species originated.
These species may consume or weaken aboveground and belowground plant tissue, interrupt
important life-cycle elements, or reduce seed production or viability. Monitoring for the effects of
these agents on target and nontarget species can be difficult. The most straightforward assessment
involves recording weed density, seed production, growth, or some similar measure of abundance
or vigor, release a biological control agent, and repeat the surveys periodically to gauge the
effectiveness. Performing this procedure at a number of introduction sites is preferred. This
approach, however, cannot account for environmental factors that might affect all sites and NIS
abundance/vigor. A more advanced approach is to compare sites where biocontrol agents have
been released with similar nearby sites where the agent has not been introduced (Ainsworth 1999).
These efforts may combine vegetation measurements, insect abundance information, and insect
damage data. Specifics of monitoring biocontrol and other control efforts are described in Farrell
and Lonsdale (1997) and Wilson and Randall (2003).
4.3.2.8
Sampling Design Examples
The following are examples of sampling designs that can be used to monitoring NIS.
169
Locating Points along a Road
using ARCMAP
Points can be added to a roads data layer. Segments
of line can be divided into +/- equal units using
ARCMap.
Make a copy of the roads layer, and name it
Working. Use Editor and select the Working layer.
Click on a segment of a line with the edit tool
.
Click on the Editor’s drop down arrow and select
Divide. To make somewhat equal segments, choose
the second option (Place points separated by every
units). Note that at the x/y coordinates are at the
center of the segments.
Use the Select by Attribute to locate all the segments
with an ObjectID = 0, the road segments have values
greater than 0. Save the selected segments to a new
file, SamplePts.
Select a point symbol to identify the sampling
points.
Chinese lespedeza
Firing Points 707 & 708
In the area of Firing Points 707 & 708 an invasive species,
Chinese lespedeza (Lespedeza cuneata), was noticed. As a
potential fire-enhancing species, one of the installation’s invasive
species management objectives is to limit Lespedeza to no more
than 5% frequency in any area. The monitoring objective is to be
90% confident of detecting the percent frequency in a defined
area.
A sample size of 20 sites with 200 quadrats (2 dm x 1 m) was
determined to sample the area adequately. Two 50 m by 2 dm wide
line transects were established at each point and the
presence/absence of the species was noted every 5 m for a distance
of 1 m.
Sampling is conducted yearly in early summer, prior to flowering.
If more than a 5% frequency is noted, the area is sprayed.
170
How much detail do you need? In TA Bravo, there was no significant change in the area
of invasive species between 2003 and 2004, either for all species combined or for individual
species. These data do not detail the abundance (e.g., density, cover, etc.) of the invasive
species at each site, but additional abundance information may not have been a requirement.
Figure B suggests that London rocket is moving down a wash and that a new species,
common thistle is now present. Figure C suggests that there is little change in the location
and extent of the three species, although the patch boundaries have shifted somewhat. The
“x”s in Figure A indicate survey locations where invasive species were not encountered.
171
4.3.2.9 Web Resources for Monitoring Noxious and Invasive Plants
Center for Invasive Plant Management. Provides state-specific information.
http://www.weedcenter.org/
Global Invasive Species Database. Invasive Species Specialist Group. http://www.issg.org/
Hawaiian Ecosystems at Risk Project (HEAR). http://www.hear.org/
Invasivespecies.gov - USDA. A gateway to Federal and State invasive activities and programs and
species links. http://www.invasivespecies.gov/
National Biological Information Infrastructure, Invasive Species Information Node. USGS.
Species profiles, plants and animals. http://invasivespecies.nbii.gov/
North American Weed Management Association (NAMWA). Provides education, regulatory
direction, professional improvement, and environmental awareness regarding the degrading
impacts of exotic, invasive noxious weeds. http://www.nawma.org/
National Park Service, Invasive Species Monitoring Resources
http://science.nature.nps.gov/im/monitor/TechnicalGuidance.cfm#TGInvasives
Invasive.org - Joint project by USDA Forest Service and USDA APHIS PPQ and The University
of Georgia. Archive of photographs of invasive and exotic species and other resources.
http://www.invasive.org/
TNC Invasive Species Initiative. Invasive species documents and photographs
http://tncweeds.ucdavis.edu/esadocs.html
TNC Eastern Invasives Management Network
http://tncweeds.ucdavis.edu/networks/eastern/eastern.html
USDA APHIS. List of Federal noxious weeds.
http://www.aphis.usda.gov/ppq/weeds/noxiousweedlist.pdf
4.3.3
Surveying and Monitoring Rare Plants
4.3.3.1
Introduction
A rare plant species is any native vascular or non-vascular plant that exists in low numbers or in a
very restricted range, giving it regulatory status for a heightened level of concern by federal 8, state
8
USFWS Terminology from http://endangered.fws.gov/glossary.pdf
Endangered species An animal or plant species in danger of extinction throughout all or a significant portion of its range.
Threatened species An animal or plant species likely to become endangered within the foreseeable future throughout all or a
significant portion of its range.
Candidate species (candidate) A plant or animal species for which FWS or NOAA Fisheries has on file sufficient information on
biological vulnerability and threats to support a proposal to list as endangered or threatened.
172
(e.g., threatened, endangered, sensitive species), and Natural Heritage (state or global rank)
programs, or other special management status by land managers. Species at risk or of concern and
Federal Candidate species are often given special management consideration despite their not
having a legal protection status. Plant species of interest will be referred to as ‘rare species” in this
section.
Some of the contributing causes of species rarity and endangerment include habitat loss and
degradation, disease, competition for resources by introduced and native species, collecting of
specimens, development, herbivory by nonnative species, and alteration of natural processes such
as fire, flooding, and other disturbances. Rare plant surveys are conducted to determine the
presence and location of rare plant species and rare vegetation communities on a particular site.
The need for surveys is often related to proposed construction or disturbance due to changes in the
military mission.
Rare plant monitoring consists primarily of four types of activities (White and Bratton 1984;
Lancaster 2000):
1)
2)
3)
4)
Designation and formulation of list of rare species and prioritization of efforts.
Establishment of a record of historic locations and information from previous surveys.
Rare plant surveys consisting of field verification of sites and additional field surveys
for new sites/populations, which may also include population estimates and mapping
of the spatial extent of populations. Menges and Gordon (1996) refer to mapping of
locations and their spatial extents as Level 1 rare plant monitoring.
Monitoring status and trends in populations over time. Monitoring is generally
divided into periodic inventory studies (i.e., census), sampling studies, and
demographic studies (Palmer 1987). Menges and Gordon (1996) refer to quantitative
assessment of abundance or condition as Level 2 monitoring, and demographic
monitoring as Level 3 monitoring.
The discussion here focuses on rare plant surveys and monitoring activities. Important information
regarding monitoring rare plants may be found in management and recovery plans developed by
land managers or the U.S. Fish and Wildlife Service. Because many species may be at risk at a
particular location, and because monitoring time and resources are limited, not all species can be
monitored intensively. A multilevel or hierarchical approach is therefore often used. The decision
to perform surveys (with mapping extents – Level 1 monitoring), population monitoring (Level 2
monitoring), or demographic studies (Level 3 monitoring) is directly related to the urgency of
management and available resources, and should be closely linked with management goals and
objectives for a particular site (Menges and Gordon 1996). Decisions to apply more intensive
monitoring (e.g., quantitative population inventory and sampling, demographic studies) types
should be based on the results of less intensive efforts. Demographic studies can be expensive and
time-consuming and are relatively uncommon except where necessary or required by compliance
needs. Decisions regarding monitoring approaches should also consider the management
implications of results. Intensive demographic studies should not be undertaken when the
Listed species A species, subspecies, or distinct population segment that has been added to the Federal list of endangered and
threatened wildlife and plants.
Species of Concern An informal term referring to a species that might be in need of conservation action. This may range from a need
for periodic monitoring of populations and threats to the species and its habitat, to the necessity for listing as threatened or endangered.
Such species receive no legal protection and use of the term does not necessarily imply that a species will eventually be proposed for
listing. Synonymous with “species at risk”.
173
likelihood of management action is low to none (Palmer 1987). As rare plant populations decline,
monitoring information is increasingly important for their conservation and is considered the
minimum requirement for their management (Collins et al. 2001).
4.3.3.2
Rare Plant Surveys
Rare plant surveys are the foundation for documenting the presence of species and monitoring
changes in their distribution and abundance over time. Although rare plant surveys may confirm
the presence of a rare species (Lancaster 2000), unless a site is small and extremely degraded
(WNHP 2004), such surveys are seldom comprehensive enough to rule out the presence of a
species from the area of interest. This is because sparsely populated or dormant plants may be
overlooked, and some unlikely habitats may be ignored. The more often an area is surveyed
without finding a rare plant population, the less likely a dormant or overlooked population is
present (Robson 1998).
Many military installations have several plant taxa of special interest (e.g., rare, sensitive, or
protected taxa). DoD land managers are required to conduct threatened and endangered plant
surveys under the Endangered Species Act to monitor, protect, and manage rare species and their
habitats and minimize the potential impacts of military training. The size and distribution of plant
populations are determined during surveys, allowing biologists to make informed management
recommendations for the area and the preservation of the species. As new rare plant populations
are discovered, the information is incorporated into a GIS database.
Conservation Data Centers (CDCs) and associated Natural Heritage Programs, established in all
50 states, are important repositories for the collection, storage, and management of rare plant data,
and are thus important sources of information when planning rare plant surveys (Robson 1998).
Following surveys, any new or updated information should be provided to the appropriate CDC.
The U.S. Forest Service and the Bureau of Land Management maintain sensitive plant lists that
include plants without legal protection as well as those that are monitored to minimize the need for
listing. In addition, regional offices of the U.S. Fish & Wildlife Service Endangered Species
Program, U.S. Forest Service, and BLM also maintain information on plant taxa of special
interest. State-level Natural Heritage, Nature Conservancy, and native plant societies also maintain
databases regarding the occurrences of rare species.
4.3.3.2.1
Personnel
It is crucial to find a well qualified plant surveyor with a strong background in plant taxonomy
(bachelors degree or higher in botany or similar field). The following additional qualifications are
highly desirable:
• Ability to identify most or all the plant species found in the area of interest and use
taxonomic keys to identify unfamiliar taxa, and to recognize certain difficult-to-identify
taxa requiring assistance from professional taxonomists.
• Knowledge of local native flora, including potential rare species, the ecology of the
area, and unique plant communities or assemblages (Robson 1998; Lancaster 2000;
WNHP 2004).
• Familiarity with state and federal laws and regulations regarding rare plant protection,
permitting, and environmental assessment (Robson 1998). It is illegal to collect
federally protected taxa without appropriate permit. Taxa may also be protected under
state law. For example, California has strict laws and penalties for illegal collection;
Arizona has strict laws regulating collection of cacti and other economically important
taxa; and Colorado has a daily limit for collection of the state flower (CEMML 1996).
174
•
•
Field skills, including a good understanding of maps, navigation, use of GPS, and other
tools to map rare plants (WNHP 2004).
Demonstrated ability to prepare detailed technical reports (Robson 1998).
4.3.3.2.2
Field Preparation
Prior to entering the field a list of potential rare species occurrences should be generated, based on
similar range and habitats within the study area. Natural Heritage Program and CDC databases,
distribution maps, local and regional botanists and other experts, reports and literature, and plant
field guides for the area are used develop the list (Robson 1998; Lancaster 2000; WNHP 2004).
Plant lists for the surrounding counties or parishes can also be used to compile a list of taxa that
might occur in the survey area. A lack of historic information or recent data could simply mean
that no one has visited the area before and the likelihood of other rare plants appearing in the study
area should be assessed (Robson 1998).
The species scientific name, habitat preferences, and optimal time for sampling should be included
in the lists (WNHP 2004) along with a description, illustration, and/or photograph (when
possible). In addition to the species lists, known locations of rare plants, plant communities, and
their associated habitats should be mapped on topographic maps and aerial photos (< 1:30,000
scale, preferably in color) and taken into the field (Robson 1998; Lancaster 2000). In addition,
potentially developed or disturbed sites should be mapped, along with all available information on
the physical environment (e.g. topography, geology, soils) (Lancaster 2000).
The surveyor should study herbarium specimens and flora guides, noting important distinguishing
features of potential rare plant species in the area (Lancaster 2000). Data from published sources
and herbarium labels can be useful for determining common habitats, flowering dates (important
for determining ideal sample dates), associated species (Robson 1998), and other biological
characteristics (e.g. annual, perennial, reproductive strategies, saprophyte, etc.). When flowering
date information is not available, phenological information should be used for the potential species
to select data collection dates (Lancaster 2000). Familiarity with the project location, size and type
of disturbances, and possible alternative study area locations will help the surveyor gather
information and plan field work (Robson 1998).
4.3.3.2.3
Survey Design
A rare plant survey should adequately cover the study area including each plant community,
uncommon plant associations, and important features (e.g., riparian areas, geologic features)
potentially supporting rare plants (Lancaster 2000). Surveys should be conducted at least twice
during the growing season at the peak of target species presence (when definitive characteristics
are most visible). All previously documented rare species occurrences should be revisited
(Lancaster 2000; WNHP 2004). A more extensive plant survey to explore for undocumented taxa
or occurrences should include seasonal considerations to survey for warm and cool season
perennials, summer and winter annuals, and ephemeral habitats (e.g., intermittent streams, vernal
pools) and events (e.g., snow melt). Surveying the study area under a variety of conditions and at
different dates or phenological stages helps to account for climatic fluctuations (e.g., droughts
affecting annual production), lack of perennial flowering (decreasing the ability to identify
species) or above ground growth in subterranean perennials, and the effects of grazing and pests
(e.g., insects, disease), which can reduce vigor and make plants less noticeable. Further surveys
may involve more intensive searching within defined areas or searching as many areas as possible
within a larger study area (Lancaster 2000).
175
Each representative plant community within a study area should be examined thoroughly. This is
important because occasionally rare plants are found in habitats where they are not expected. In
areas less than a square mile in size (640 acres or 259 ha or one Section), it may be possible to
visit all plant communities in a day, depending on terrain and density of vegetation (Robson
1998). The surveys should be designed so that each habitat is visited at the most optimal time for
detecting potential species, which usually requires more than one visit per habitat (WNHP 2004).
In circumstances when the area is extremely large, efforts should focus on unusual, rare, or other
habitats where rare species are most likely to be found. Because rare plant populations are often
scattered and small, surveys focus on areas and features likely to have rare plants (e.g.,
microhabitats, seasonal water patterns, edge of transition zones between habitats, and certain plant
associations) (Robson 1998; Lancaster 2000; WNHP 2004). All potential areas and features with
rare plant species should be searched. When conducting floristic surveys for rare plants the goal is
not to identify all plant species in the given area, but instead to identify plant species encountered
to the extent necessary to determine their level of rarity (Lancaster 2000). Quantitative survey
techniques are designed for analysis at the vegetation community level, and thus are inappropriate
for rare plant surveys since they are biased toward dominant species and have narrow coverage of
the study area (Robson 1998; Lancaster 2000).
The choice of search technique is influenced by topography and vegetation. Two common search
methods (meander and patterned searches) are described by Lancaster (2000). A patterned or
systematic survey minimizes overlap and maximizes the area surveyed (Robson 1998; Lancaster
2000). By walking a series of roughly parallel transects in a search unit, maximum coverage of the
area is achieved. Vegetation cover, plant density, the size and species of plant, and its visibility
through the vegetation influence the spacing of the search transects (Robson 1998; Lancaster
2000). In a meander search, each species is recorded during random (i.e., surveyor-determined)
walks through a habitat type or plant association. The search is terminated when no more new
species are observed. Meander searches work well in rugged terrain or in irregularly shaped areas,
but may oversample an area (i.e., sample the same ground more than once) and can be biased to
areas with easier walking. Patterned searches can be advantageous because they maximize the
coverage of a search while minimizing overlap and are less biased towards easy-walking terrain.
However, patterned searches may overlook unique habitats when accessibility to an area is poor.
Lancaster (2000) recommends using meander searches to locate boundaries of biotic attributes and
then using transects in a pattern search for complete coverage of the areas of interest. For a more
detailed discussion on search methods see Nelson (1984; 1987).
4.3.3.2.4
Data Collection
Documenting results of rare species occurrences is important for monitoring of species population
status and for mitigation planning, and may be the most important step in rare plant surveying
(Robson 1998). It is also important for understanding rare plant species biology, habitat
preferences, and phenology. At the site of each occurrence, a variety of information is collected,
including the location (GPS coordinates plus other descriptive information), date, surveyor name,
collection number if specimen collected, rare taxa present, the size of the population (counted or
estimated), associated species, topographic and site information, and other relevant information.
The perimeter of the population may also be mapped manually on a photograph or map or using
GPS. The locations of individuals or populations are sometimes marked in the field to facilitate
monitoring and mitigation from anticipated land uses (Robson 1998). The distribution within and
outside the study area should be noted.
176
Voucher specimens should be collected unless the respective species population is extremely low
(sample must be < 4% of remaining species population). A minimum amount of plant material
diagnostic of the rare plant species should be sampled when populations are small (Lancaster
2000). Robson (1998) states that a voucher specimen should only be collected if the local rare
plant population has more than 20 individuals and it should only consist of enough of the plant to
confirm its identity (such as a branch with several leaves and a few flowers and/or fruits). For
some species (i.e. mustard family) the entire above ground portion of the plant can amount to a
single branch, in which case efforts should be made to leave the root system intact. When annuals
only have one branch, they should only be picked after most of the seeds have fallen from the
plant (Robson 1998). Collecting voucher specimens often requires a permit. Brayshaw (1996)
provides guidance for proper mounting and labeling of voucher specimens. Photographs are an
effective alternative when collecting is not possible. As with collections, diagnostic features of the
plant should be emphasized along with the associated habitat of each species (Robson 1998;
Lancaster 2000). Robson (1998) recommends that the following characteristics be recorded to
accompany each photograph when a voucher specimen is not collected: measurements of leaf,
flower and/or fruit width and length; overall plant height; color, shape and surface characteristics
of the stems, leaves and reproductive structures; and type of inflorescence.
Designated rare plant report forms should be used to document rare plant occurrences and
pertinent information including precise location (using legal land description, latitude and
longitude or UTM coordinates, directions for relocation, and photographs), habitat, plant
community, associated species, aspect, slope, elevation, relative abundance, soil type and texture,
drainage descriptions, date; and information on plant phenology, vigor, size or age classes
(Robson 1998; Lancaster 2000). In addition, factors potentially affecting the plants, such as
moisture conditions, competition, insect pests, current land use practices, grazing pressure, fire,
and any other threats should be noted. These forms should accompany each voucher specimen and
should be sent to a publicly accessible herbarium. If a voucher specimen is not obtained, copies of
the photographs and morphological data collected should be sent to the herbarium. In addition,
copies of the forms and photographs should be sent to the appropriate Conservation Data Center
noting the location of plant material vouchers collected and associated collection numbers. The
reference information used to key each plant should also be provided. A taxonomic expert should
be used to verify proper identification of rare species that are difficult to identify and to review
and confirm all specimens and photographs/descriptions (Robson 1998; Lancaster 2000). The
name and contact information of the collector and the name of the botanist who reviewed the
specimen/photographs should be included in all documentation.
177
4.3.3.2.5
Reporting and Documentation
A rare plant survey report must provide adequate information for the receiving agency or reader to
assess the quality and extent of the survey (Lancaster 2000). It should include a brief description
of the project including justification for the study and survey design, pre-survey information from
existing databases, survey results, dates of survey, photographs, percentage of area covered and
amount of time spent surveying, maps showing rare plant populations, copies of rare plant
occurrence forms, and notes on limitations of the survey or confidence in the results (Robson
1998; Lancaster 2000). Mitigation suggestions should be made if any rare plants would be
impacted by proposed or future projects (Robson 1998). The name and qualifications of the
surveyor should be provided in the report, along with a signature sheet for all contributing
botanists to assure quality control and full disclosure of results (Robson 1998; Lancaster 2000).
4.3.3.3
Rare Plant Monitoring
Rare plant monitoring consists of collecting and analyzing [typically quantitative] data to
document the condition of a population or community over time. The goals of these efforts may
include detecting changes at a site or evaluating the effects of management actions (Palmer 1987).
Various combinations of qualitative and quantitative measurements may be nested within a single
monitoring program (Menges and Gordon 1996).
4.3.3.3.1
Census or Inventory
An inventory is a count of all of the individuals in a population, and can be repeated periodically
to assess changes in the population. On the surface, a rare species inventory would appear the
easiest approach, and would give an indication of the stability of adults in the population.
However, seedling and other inconspicuous life stages may not be well documented, and estimates
may be confounded by climatic extremes (e.g., drought), periodic dormancy or difficulties
associated with annual or biennial species (Palmer 1987).
4.3.3.3.2
Sampling Populations and Communities
Sampling can be more time-consuming than an inventory, but it provides a means to subsample
large or extensive populations and collect additional data (Palmer 1987). Sampling uses repeatable
designs and methods to make estimates regarding population and community attributes. Attributes
of interest include the number and spatial extent of patches or populations; vegetation
characteristics such as population size, cover, density, and size of individuals; and reproductive
information such as the number of reproducing individuals, average seed production, the number
of flowers per plant, and the number of flowering individuals. Information about population
structure can be collected or inferred from measurements, but the fate of individual plants cannot
be known. For example, the frequency distribution of plant sizes present can be indicative of the
age structure of the population. Community sampling differs from population sampling in that all
plant forms or species may be recorded versus just the rare species. Sampling is sometimes
combined with inventory efforts.
Sampling rare plants follows the same principles and methods used for monitoring plant
communities and populations in general. Decisions regarding permanent or temporary sampling
units, size and shape of sampling units, number of samples, and the selection of attributes to
collect determine the accuracy and precision of results. Populations are sampled using permanent
or nonpermanent grid systems, belt transects, quadrats, or well-defined unmarked areas (Menges
and Gordon 1996). In general, the population is considered the fundamental unit of analysis.
Individuals should be counted, if feasible, although the number of rooted stems may be counted or
percent cover estimated if individuals are difficult to distinguish or an alternate measure is desired
178
for management purposes. For plant cover and structure, the tradeoffs between vegetation
measurement (e.g., point intercept, line interception) and ocular estimation should be considered
in relation to specific monitoring objectives (Menges and Gordon 1996).
Rare plant sampling typically takes place within predefined areas and only one to several species
are recorded. Special consideration must be given to sampling designs for rare species when
populations are small, since small populations (extent or number of individuals) can create
problems with plot placement, replications, and randomization (Travis and Sutter 1986). Projects
designed to evaluate the effects of management on a rare plant populations most often apply
community sampling approaches, which provide more information than single species sampling
(Palmer 1987).
There are few published protocols for monitoring rare plants, as most programs are tailored to
local needs and are species-specific. The Fire Effects Monitoring and Inventory Protocol includes
a protocol for sampling rare species (USDI Geological Survey 2004). The FIREMON Rare
Species (RS) method is used to assess changes in uncommon, perennial plant species when other
monitoring methods are not effective. This method monitors individual plants and statistically
quantifies changes in plant survivorship, growth, and reproduction over time. Plants are spatially
located using distance along and from a permanent baseline and individual plants are marked
using a permanent tag. Data are collected for status (living or dead), stage (seedling, nonreproductive, or reproductive), size (height and diameter), and reproductive effort (number of
flowers and fruits). This method is primarily used for threatened and endangered species and
uncommon grass, forb, shrub, and tree species of special interest. Local or regional land managers
may be the best source of information for existing protocols for quantitative monitoring of rare
plants.
4.3.3.3.3
Demographic Studies
Demographic studies are the most intensive types of sampling, involving searching, mapping
and/or marking, data collection, and often multiple trips to a site to monitor seedlings,
reproductive plant parts, and other attributes. For declining populations of threatened, endangered,
or locally imperiled species, monitoring the population alone may not provide the information
necessary for effective management. Demographic studies are employed to understand the biology
of a species and provide information about reproduction, recruitment, survivorship, age structure,
and changes in the population over time. Questions related to demographic studies include (Lesica
1987):
1)
2)
3)
4)
5)
At what stage(s) in the plant’s life are individuals being lost?
What factors are causing reductions in recruitment?
At what age does reproduction begin?
How does reproductive output vary from year to year and over the lifetime of
an individual?
Is recruitment adequate to replace individuals that have been lost?
The two major types of demographic monitoring are observational and experimental.
Observational studies are descriptive with no effects applied by the observer. In experimental
studies, the aim is to assess the effects of treatments applied by the investigator or land managers
(Travis and Sutter 1987). Demographic monitoring is also described in Chapter 2. In order for
demographic monitoring to be justified, the results should lead to a management action to assist
species recovery. In many cases, data on population demographics is necessary to identify the
179
underlying causes of population trends, but is too costly to collect. Additional discussion of
specific techniques and obstacles to collecting demographic data are presented in Owen and
Rosentreter (1992).
4.3.3.3.4
General Guidelines for Rare Plant Monitoring
1)
2)
3)
4)
5)
6)
7)
8)
9)
4.3.4
Develop a detailed protocol containing the rationale, methods, and analysis
procedures to be employed.
Use a progressive approach, beginning with mapping and occurrence
documentation and using more intensive monitoring as needed to support project
goals.
Plan data analysis in advance to ensure sampling and designs will satisfy
monitoring objectives. Have sampling designs and methods reviewed by a
qualified biologist or statistician.
Employ random sampling to ensure that population estimates are unbiased when a
census is not practical.
Permanent plots for Level 2 monitoring have some advantages over non-permanent
plots, but reductions in spatial variability may be offset by increased efficiency of
nonpermanent plots. Recent advances in GPS facilitates the relocation of
approximate sites within defined areas, greatly improving sampling efficiency and
allowing for more intensive data collection at few locations or more extensive
sampling over larger areas.
Perform pilot sampling to determine the number of samples required to achieve the
desired level of precision.
Only monitor high priority species.
Incorporate species biology and life history aspects into sampling designs (e.g.,
species phenology, visibility, dormancy, flowering period, etc.)
Summarize and analyze results relative to monitoring objectives and evaluate
sampling designs and methods accordingly.
Soil Compaction Assessment
Soil compaction can be a major consequence of vehicle traffic and foot trampling in military
training areas (Halvorson et al. 2001; Halvorson et al. 2003). Increased soil compaction can lead
to higher soil bulk density, which can reduce water infiltration, reduce soil surface strength,
increase runoff and erosion potential, and reduce site productivity (Braunack 1986). The severity
of compaction will depend on several factors, including intensity and type of military use, climate,
soil properties, and soil moisture at time of impact.
Several methods exist for characterizing soil compaction. The most common field measurements
include soil bulk density and soil penetration resistance (Miller et al. 2001; Herrick and Jones
2002). Specific sampling guidelines for soil compaction assessment are provided in Jones and
Kunze (2004a).
4.3.4.1
Soil Bulk Density
Soil bulk density is defined as the mass per unit volume of soil (Grossman and Reinsch 2002). The
180
mass is usually considered the oven dry mass, which is obtained by drying a soil sample at 105oC
for 24 hr. The volume usually refers to the <2 mm particle fraction. Bulk density is typically
expressed as Mg/m3 or g/cm3.
Several field methods are available for collecting bulk density samples. The most common method
consists of inserting a small core cylinder of a known volume into the soil (Grossman and Reinsch
2002). Samples are usually weighed in the field to obtain wet mass for soil moisture
determination. Samples are dried in the laboratory at 105oC for 24 hr and reweighed. Samples are
then screened through a 2-mm sieve to separate rock fragments and course organics. The oven-dry
mass of the >2 mm fraction is weighed and then divided by a particle density (usually 2.65 g/cm3)
to obtain the mass and volume needed to subtract from the original mass and volume of the
sample. Further details on bulk density field measurement and laboratory analysis techniques are
found in Grossman and Reinsch (2002).
4.3.4.2
Soil Penetration Resistance
Soil penetration resistance can also be used to characterize soil compaction (Herrick and Jones
2002). Penetration resistance is measured using cone penetrometers. Penetrometers are used to
locate zones of compaction within a soil profile because of their easy, rapid, and economical
operation. Furthermore, since there is generally a strong correlation between penetration resistance
and bulk density, albeit one that is sensitive to changes in soil moisture, penetrometers can serve
as a rapid tool for estimating soil bulk density (Miller et al. 2001).
Several factors affect penetration resistance, including soil moisture, soil texture, and bulk density
(Vazquez et al. 1991; Miller et al. 2001; Halvorson et al. 2003). In general, penetration resistance
should increase with increasing bulk density, decreasing soil moisture, and increasing fine soil
fragment (Lowery and Morrison 2002). Penetration resistance may also change with soil depth, as
compacted layers may exist at several points throughout a soil profile due to changes in soil
moisture or texture (Halvorson et al. 2003). Data on bulk density, soil moisture, and soil texture
can be collected as corollaries to compaction. Methods for determining soil moisture are described
by Topp and Ferre (2002), and methods for determining soil texture are described by Gee and Or
(2002).
There are three general classes of penetrometers: static cone, dynamic cone, and drop cone. A
brief overview of each type follows.
4.3.4.2.1
Static Cone Penetrometers
Static cone penetrometers measure the force required to push a metal cone through the soil at a
constant velocity (Herrick and Jones 2002). The force is usually measured by a load cell or strain
gauge coupled with an analog dial or pressure transducer for readout (Herrick and Jones 2002).
The force is commonly expressed in kilopascals (kPa), an index of soil strength referred to as the
cone index (ASAE 1999; ASAE 2004). As the operator pushes down on the penetrometer, cone
index values are recorded at a specified depth increment. A static cone penetrometer with a 30o
cone has been recommended by the American Society of Agricultural Engineers (ASAE) as the
standard measuring device for characterizing penetration resistance (ASAE 2004).
Although the methods for static cone penetrometer operation have been standardized, there are
several disadvantages which may limit their use for monitoring (Herrick and Jones 2002). Static
penetrometers can be relatively expensive (≥$600), particularly for models with digital recording
181
capability. More importantly, since static penetrometers must be moved through the soil at a
constant velocity, different rates of insertion by the operator can yield variable results and affect
repeatability (Herrick and Jones 2002). Operator strength may also limit the use of static
penetrometers in dry soils. Advantages of static cone penetrometers include well-documented and
standardized methods and ease of use. Examples of studies that have used static cone
penetrometers to measure soil compaction in military training areas include Halverson et al.
(2001) and Halverson et al. (2003). Specifications and vendor information for selected static cone
penetrometers is provided in Jones and Kunze (2004a).
4.3.4.2.2
Dynamic Cone Penetrometers
Dynamic cone penetrometers (DCPs) apply a known amount of kinetic energy to the cone, which
causes the penetrometer to move through the soil (Herrick and Jones 2002). DCPs do not rely on
constant penetration velocity, as most use a slide hammer of fixed mass and drop height to apply
consistent energy with each blow. Either the number of blows required to penetrate a specified
depth or the depth of penetration per blow is recorded. The energy delivered is dependent on the
weight of the hammer, slide distance, and cone angle, all of which can be adjusted to local
conditions (e.g., soft vs. hard soils). Operation guidelines and calculations for DCPs are provided
in Herrick and Jones (2002).
DCPs tend to yield much more consistent results and have a greater range of repeatability than
static cone penetrometers because they are not subject to operator variability (Herrick and Jones
2002). Dynamic penetrometers also have fewer limitations in dry soils and tend to be less
expensive than static penetrometers. Examples of studies that have used static cone penetrometers
to measure soil compaction include Herrick and Jones (2002) and Kunze and Jones (2004a).
Specifications and vendor information for selected dynamic cone penetrometers is provided in
Jones and Kunze (2004a).
4.3.4.2.3
Drop Cone Penetrometers
Drop cone penetrometers represent a third type of penetrometer (Godwin 1991). The device
consists of a 30 o metal cone and lifting rod, a 1 m long PVC or acrylic guide tube, and an
aluminum millimeter ruler inlaid in the holding rod. The cone is machined with a collar to ensure
that it falls perpendicularly through the guide tube. To take a measurement, the base of the tube is
placed on the ground and the cone is lifted until it is flush with the top of the tube. The cone is
released and the depth of penetration is recorded. The device is inexpensive, easy to use, and
highly repeatable. The disadvantage is that only surface soil resistance is measured and nothing
can be inferred about the underlying soil profile. Examples of studies that have used drop cone
penetrometers to evaluate the effects of military training on soil compaction include Jones (2000).
182
4.3.5
Land Uses
Resource condition is highly influenced by land-use activities, many of which are inherently
destructive. Land-use information can be collected in the field or obtained from a variety of
sources including remotely sensed imagery and range operations and scheduling data. Information
may be qualitative, semi-quantitative, or quantitative in nature. The intensity or amount of usage
can be difficult to measure or evaluate. For this reason, land use data is often used in qualitative
ways. Some approaches attempt to integrate qualitative and quantitative assessment.
4.3.5.1
Military Training and Testing
Training and testing use can be assessed by evaluating a number of attributes. The primary types
of data are type of use, frequency of use, and intensity of use. In most cases, the type of use can be
identified through experience and knowledge about the range of possible disturbances/uses. Types
of common training disturbances include:
•
•
•
•
•
•
•
tracked or wheeled off-road maneuver damage – dispersed/extensive
tracked or wheeled off-road maneuver damage –concentrated/intensive
travel lanes or corridors where multiple vehicle passes took place (linear features)
turn-out or turn around areas adjacent to maintained roads and trails
mechanical excavation sites (vehicle fighting positions, anti-tank ditches)
bivouac sites and hand excavation disturbance
firing range disturbance
Categories of land use may vary by installation and by training or testing mission. Training
facilities such as vehicle driving courses, dig sites, engineering construction sites, and firing
ranges represent very specific types of uses. Land uses are sometimes inferred from evidence
present at a site. For example, communications wire and MRE wrappers may indicate small unit
maneuver or reconnaissance activities.
If an activity can be documented as having taken place in the recent past, then a relationship to
land condition may be established by comparing sites that received a particular use with sites that
were undisturbed or did not receive that use. It is sometimes difficult to assess the age of the
evidence for a particular land use. In arid regions especially, soil disturbance and vehicle ruts may
persist for decades or longer. Another consideration is the amount or intensity of the use. Can the
observer distinguish between two and four passes of a vehicle? Can disturbance by private offroad vehicles (ORVs) be distinguished from that caused by wheeled military vehicles?
Thus far the discussion has raised issues related to the difficulties of qualitative assessment of the
type of use and the drawbacks associated with estimating intensity of use. Another issue is the
determination of the spatial extent of the disturbance. For example, how much of the site is
impacted? The original RTLA methods addressed this issue by collecting presence-absence data
along a line transect to estimate the percent of the area disturbed by different activities. This
approach provides a semi-quantitative measure of disturbance at the sample location, but gives no
indication of the geographic extent of the disturbance, assuming it is definable. Other approaches
to assessing the areal extent of disturbance include surveying and mapping using GPS, or
delineating disturbances on remotely sensed images either manually (aerial photos) or through
automated procedures (satellite imagery). Different approaches should be used depending on data
needs and the level of detail needed to assess disturbance.
183
Military land use is commonly documented by Range Operations (e.g., Range Control) personnel
as part of the scheduling and training land allocation process. Most installations operate some
form of scheduling and use software such as the Range Facility Management Support System
(RFMSS). Programs such as RFMSS contain data on scheduling, use, unit type, training type, and
duration of use. This information is valuable in refining field assessment techniques by comparing
actual use data with observational or measured data. Inferences also may be made by examining
duration, type, and frequency of use and their associated impacts on resources over time. It is
hoped that RFMSS and similar data will be refined in order to provide data that is detailed enough
for monitoring applications and analysis.
4.3.5.2
Non-Military Land Uses and Management
Many of the issues raised in Section 4.3.5.1 are relevant to the discussion of non-military land uses
and their associated impacts. In the field it may be possible to assess if a land use took place, but
difficult to determine the time since the activity took place, it’s duration, or intensity. Examples of
non-military uses include activities by the public, natural resource management, and land repair
and maintenance activities. Public land uses include recreation, hunting, and ORV use. Land
management activities include prescribed burning, livestock grazing, forestry activities,
revegetation projects, and habitat alteration for wildlife. Examples of land maintenance activities
include herbicide application and mowing. Information about the types and extent of land
management activities should be available through installation land management staff.
4.3.6
Road Condition Assessment and Its Relation to Erosion and
Sedimentation
Unpaved roads provide key support to training and other land management activities on military
lands (White 1997). Surface runoff and erosion associated with roads can degrade the road surface
and lead to high maintenance costs and a poor-quality road network. Unpaved roads can also be a
significant source of sediment, degrading water quality and impairing aquatic habitat (Reid and
Dunne 1984; Gucinski et al. 2001).
Given the importance of unpaved roads in military training and their potential environmental
effects, RTLA programs may need to inventory and monitor road condition to support future land
management decisions. Inventorying and evaluating unpaved road condition can help RTLA
coordinators support a high-quality road network to military trainers, identify and prioritize sites
for rehabilitation, and ensure that the environmental effects of roads are minimized.
4.3.6.1
U.S. Army Unsurfaced Road Maintenance Management System
A standardized protocol for evaluating the condition of unpaved roads on military lands was
developed by the U.S. Army Corps of Engineers Cold Regions Research and Engineering
Laboratory (Eaton and Beaucham 1992). The methods were later incorporated into a U.S. Army
Technical Manual (US Army 1995). The procedure consists of identifying road segments and
using a rating system to evaluate a series of road surface distresses. Distresses include improper
cross-section, inadequate roadside drainage, corrugations, dust, potholes, ruts, and loose
aggregate. Distresses are rated low, moderate, or high severity on a sample unit of a road segment.
An Unsurfaced Road Condition Index (URCI) is calculated by measuring the density of each
distress type on the sample unit. The severity ratings and URCI can be used to determine a
maintenance and repair strategy for the road network. The procedure has been used in pilot studies
184
at Fort Leonard Wood, Missouri (Isaacson et al. 2001) and Eglin Air Force Base, Florida
(Albertson et al. 1995).
4.3.6.2
Unpaved Road Condition Assessment Protocol
The Center for Environmental Management of Military Lands (CEMML) developed a field
protocol to help RTLA coordinators facilitate data collection on unpaved roads (Kunze and Jones
2004b). The protocol supplemented the U.S. Army methods with additional attributes from the
Washington Watershed Analysis Manual (WFPB 1997). The procedure consists of: (1) identifying
and characterizing road segments, (2) assigning condition ratings to road segments, (3) identifying
site-specific road problems, (4) managing and analyzing data, and (5) periodically resurveying and
updating road condition data. The field methods of the protocol were tested at Fort Leonard Wood,
Missouri and Fort Jackson, South Carolina in August and September 2004 (Kunze and Jones,
2004b). A brief discussion of the procedure follows.
(1) Identifying and characterizing road segments
A road segment is identified as section of road with generally similar characteristics along its
length. Criteria that distinguish road segments include surfacing material, road size, traffic use,
topography, road condition, and construction history. Segment breaks are made due to changes in
surface material, traffic use, slope, road size or dimensions, or road condition. Surveys are
conducted by slowly driving the length of each segment and collecting data with a GPS data
dictionary.
Several attributes of the road segment are recorded: (1) reason for segment break at start/end, (2)
road width, (3) road position on hillslope, (4) road surface material, (5) road prism cross-section,
(6) traffic use intensity, (7) number of stream crossings, (8) hardened stream crossing locations,
(9) culvert locations, and (10) firebreak intersections. A photograph of the road segment is also
taken.
(2) Assigning condition ratings to road segments
Two interrelated condition ratings are assigned to each road segment: (1) a road surface condition
rating, and (2) a drainage structure rating that relates the drainage and erosion status of the road.
Several indicators are used for each condition rating, and each indicator is qualitatively assigned a
severity rating (low, moderate, high, or NA-not applicable). An overall rating for both the road
surface condition and drainage structure is assigned using the “preponderance of the evidence”
approach. The indicators for the road surface rating include: (1) corrugations (washboards), (2)
potholes, (3) ruts, and (4) loose aggregate/surface roughness. The indicators for the drainage
structure rating include: (1) improper cross-section, (2) overland flow, (3) rills, (4) gullies, (5) cut
and fill erosion, (6) ditches, (7) ditch relief turnouts, and (8) culverts.
(3) Identifying site-specific road problems
Site-specific road problems are severe distresses that may impair water quality or render the road
not drivable without immediate attention. Problem locations are described, photographed, and
mapped with the GPS in order to be addressed by maintenance activities or Best Management
Practices (BMPs). Examples of site-specific drainage or erosion problems include: (1) deep
potholes, (2) deep ruts, (3) rills, (4) gullies, (5) ditch erosion and plugging, (6) culvert inlet and
outlet erosion, (7) blocked culverts, (8) erosion at low water crossings or stream crossings.
(4) Managing and analyzing data
185
Data are stored and managed using a GIS geodatabase. Data analysis options include generating
maps to show the ratings for each road surface indicator and the overall road condition ratings for
each road segment. These maps could be used to identify areas where erosion may be occurring
and help prioritize maintenance activities and/or locations for BMPs.
(5) Periodically resurveying and updating road condition data
The final step in the protocol is to periodically resurvey the road network and update the database
and maps accordingly. Roads should be resurveyed at least every year, preferably during the same
time of year to ensure consistency in the data collection. Re-inspection data can be stored in a
separate table in the geodatabase, allowing for multiple entries of a single road segment. The
updated roads condition database and maps will be helpful in prioritizing maintenance activities,
identifying locations for BMPs, and tracking progress.
4.3.6.3
USDA Forest Service Roads Analysis Procedure
In 2001, the USDA Forest Service directed every National Forest System administrative unit to
conduct a forest-scale roads analysis to be completed by January 2003. An integrated ecological,
social, and economic-based approach was developed to analyze road networks on Forest Service
lands (USDA Forest Service 1999). The approach is conceptually similar to the federal watershed
assessment procedure. By completing roads analysis, land managers can generate maps and
narratives that show management opportunities for changing the current road system to better
address future needs, budgets, and environmental concerns (USDA Forest Service 1999).
The roads analysis procedure is comprised of six steps:
Step 1 — Setting up the analysis. The analysis is designed to produce an overview of the road
system. Interdisciplinary (ID) teams are formed and the proper analytic scales are identified. The
output from this step includes the assignment of ID team members, a list of information needs, and
a plan for the analysis.
Step 2 — Describing the situation. The existing road system is described in relation to current
management goals and objectives. Products from this step include a map of the existing road
system, descriptions of access needs, and information about physical, biological, social, cultural,
economic, and political conditions associated with the road system.
Step 3 — Identifying issues. Important road-related issues and the data needed to address these
concerns are identified. The output from this step includes a summary of key road-related issues, a
list of screening questions to evaluate them, a description of status of available data, and what
additional data are needed to conduct the analysis.
Step 4 — Assessing benefits, problems, and risks. The major uses and effects of the road system
are examined, including the environmental, social, and economic effects of the existing road
system. The output from this step is a synthesis of the benefits, problems, and risks of the current
road system and the risks and benefits of building roads into unroaded areas.
Step 5 — Describing opportunities and setting priorities. The ID team identifies management
opportunities, establishes priorities, and formulates technical recommendations that respond to the
issues and effects. The output from this step includes a map and descriptive ranking of
management options and technical recommendations.
186
Step 6 — Reporting. The ID team produces a report and maps that portray management
opportunities and supporting information needed for making decisions about the future
management of the road system. This information sets the context for developing proposed actions
to improve the road system and for future amendments and revisions of management plans.
Roads analyses have been published for several national forests, including the Deschutes and
Ochoco National Forests in Oregon (USDA Forest Service 2003a) and the Nantahala and Pisgah
National Forests in North Carolina (USDA Forest Service 2003b). The roads analysis procedure
may be useful for military land mangers who need to conduct an installation-wide analysis of the
road network.
4.3.7 Bivouac and High-Use Area Monitoring
4.3.7.1 Introduction
This approach was initially developed for the Fort Leonard Wood RTLA Program (Jones and
Robison 2004) and shares some components with bivouac monitoring programs at Fort A.P. Hill
(Jason Applegate, pers. comm.). It is designed for high-use bivouac and other heavily used areas
and is appropriate for application elsewhere in its entirety or portions thereof. The design and
methods were selected based on site-specific considerations and management needs, and are
intended as an example of applying monitoring concepts to military land management. The
approach focuses on three components:
1) Soil and site (i.e., hydrologic) stability
2) Vegetation structure and tree condition
3) Training environment and value
Areas of interest include heavy maneuver areas, combat service and service support (CS/CSS
sites), bivouac areas, firing points, engineering and equipment training areas, and other high-use
training sites. Repeated driving, foot traffic, hand and mechanical excavation, and other
disturbances result in considerable soil disturbance, compaction, and vegetation damage. A
combination of semi-quantitative and qualitative measures are used to evaluate the condition of
sites relative to management objectives and monitor their condition over time. In larger open areas
dominated by herbaceous vegetation, soil and site stability are the primary attributes of interest.
Attributes selected for monitoring are related to both tactical value and long-term training
sustainability, and are affected by excessive training usage.
4.3.7.2 Management and Monitoring Objectives
Management Goal: Maintain bivouac areas and other high use areas to sustain realistic training
environments and minimize soil erosion and off-site impacts to surface water quality.
Management Objective: Maintain upland watershed function, vertical structure/tree canopy and
forest health, and minimize soil erosion in delineated bivouac and high-use areas relative to
reference areas.
Monitoring Objective 1: Delineate perimeter(s) of high use for each bivouac site or other
area/facility of interest to determine extent of site and evaluate changes over time.
187
Monitoring Objective 2: Estimate soil erosion and soil compaction in identified high-use areas
with 85% certainty that the estimate is within 15% of the true value. USLE or RUSLE model data
requirements will be collected in the field and from other sources.
Monitoring Objective 3: Qualitatively estimate soil and hydrologic stability in high-use areas
using qualitative data such as evidence and severity of rills, gullies, water flow patterns, and soil
loss.
Monitoring Objective 4: Estimate vegetation cover (overstory, understory, and ground
level/herbaceous strata), tree density by size class, tree regeneration, and tree crown condition
indicators with 85% certainty that the estimate is within 15% of the true value. Tree damage
indicators are assessed in a more semi-quantitative or qualitative manner and have no associated
precision.
Monitoring Objective 5: Identify and mark snag and hazardous trees. Marked trees will be
removed by Range Maintenance or ITAM staff.
4.3.7.3 Sampling Design
The sampling design consists of a systematic sampling grid within the area of interest. The origin
of the grid is random, therefore the points, although systematically arranged, are located in
unbiased, random locations. Point locations are generated using GIS software and are numbered
sequentially. The procedure for generating systematic point grids in ArcGIS is presented in Jones
and Robison (2004).
The following areas are excluded from sampling within high-use training areas:
•
•
•
•
riparian areas
maintained roads and trails (e.g., graded, graveled)
impact area and small arms ranges
other off-limits, developed, or administrative areas
Plots are circular with a 5m radius and an area of 78.5m2. Trees evaluated for canopy condition
and damage are assessed within a 10m radius plot (314m2 =1/32nd ha or 1/13th ac). The
predetermined GPS sample point is the center of the plot. Unless noted, values are averaged across
the plot. Sampling consists of 20 circular plots per area. Sample locations are considered
temporary even though they may be relocated within 2-5m of the previous survey. The distance
between sample points will vary with the size of the area. The sample size of twenty points is
based partly on budgetary considerations and the need to sample a minimum number of points
within each area. This sample size may be adjusted for particular areas following initial sampling
if the number of samples is too high or low, based on calculated sample size requirements. If a
point surveyed in the field is located on a feature/area that is outside the population of interest
(maintained road, parking area, etc.), the classification will be noted on the data sheet and the plot
will not be surveyed. A 5m buffer is placed on the perimeter of each area delinated with the GPS
so that the sampled area does not extend beyond the area boundary.
When GPS coverage is unavailable, surveys can be done using compass and pacing, provided that
the perimeter of the area has been delineated and the acreage calculated. Once the area is known,
the distance between samples to provide 20 samples is determined. The map of the area is
examined and the boundaries reconnoitered in the field. Once the layout of the site is known, the
188
surveyor determines an area to begin the sampling process and plans a series of walking transects.
Pacing and compass or GPS are used to determine sampling locations. The placement of samples
in an unbiased manner is much more important than their systematic or exact placement.
4.3.7.4 Data Collection
4.3.7.4.1
Staffing and Equipment
Data can be collected by one or two people. The time required to survey a particular disturbed site
will vary primarily with the size of the site and the remoteness of the location/travel time.
Equipment List for High-Use Area Assessments
•
•
•
•
•
•
•
•
•
•
•
•
4.3.7.4.2
GPS with sample grid waypoint file uploaded, or pre-determined distance between
sample points based on the size of the area and 20 sample points minimum per area.
digital camera
compass (if not using GPS navigation)
clinometer
data forms, clipboard, pencils
BAF 20 prism
DBH tape for calibrating ocular estimates of tree diameter
plot center marking pin (e.g., pin flag) or staff
50m metric fiberglass tape
sampling protocol and supplemental Forest Health Monitoring information
soil penetrometer and data sheets
pre-printed field maps showing labeled sample locations and most recent aerial
photograph as a backdrop reference
Methods
Delineating Perimeter of High-use Area
A GPS is used to delineate the perimeter of each high-use area using surveys on foot. Criteria for
what qualifies as “high use” include recent hand or mechanical excavation, significant loss of
groundcover and herbaceous vegetation, evidence of vehicle travel and disturbance, soil
disturbance, surface soil compaction, vegetation loss, etc. Photos taken as part of the disturbance
and vegetation sampling will provide additional evidence showing the high-use area for future
reference. The GPS file is imported into a GIS and the area is calculated. The size of the area (i.e.
number of hectares) becomes the baseline value to compare with future monitoring, and is
necessary to determine the distance between sample locations for 20 samples per site. This step is
necessary in order to establish sampling locations. The resulting data can be extrapolated to all
areas within the polygon boundary, even though it is recognized that there may be patches of
intact vegetation or little disturbance in some areas. Each site/area is assigned a name for reference
(e.g., Bivouac 232).
Site Description and Training Evidence
For each area surveyed, the site name, date, and name of surveyor(s) is recorded on each data
sheet. At each point visited the following information is recorded:
189
•
•
•
•
•
•
•
Point number (from map)
Slope steepness (%)
Slope length (m)
Topographic position
Aspect
Military land use (list provided) occurring within the 5m plot radius
Training use severity – note the condition that most commonly occurs across the plot
using severity classes provided
For ocular cover estimates, a 10-class cover scale is used:
Cover Class Scale
Class Code
Percent Cover
Class
Midpoint
0
0
0.0
trace
0-1
0.5
1
1-5
3.0
2
5-15
10.0
3
15-25
20.0
4
25-40
32.5
5
40-60
50.0
6
60-75
67.5
7
75-95
85.0
8
95-100
97.5
For reference purposes, for a 10m diameter (5m radius) plot, 1% coverage = .79m2, 5% coverage
= 3.9m2, and 10% coverage = 7.9m2. For example, for bare ground, if all bare soil (with no litter or
rock cover) were consolidated into one polygon, what would the size of that patch be in square
meters? This cover class system provides more consistent estimates than methods employing a
higher-resolution system, but is also less sensitive to changes than finer or continuous cover
scales.
Soil/Site Stability Attributes
The departure from reference condition for 5 qualitative indicators of soil/site or hydrologic
stability is assessed for the area within the 5m radius. The observer assigns a departure rating of
none to slight, moderate, or extreme to each indicator by placing an “x” in the appropriate box on
the field data sheet. The overall score for a sample site is determined by the most commonly
occurring departure rating among the five indicators. The indicators described below and degree
of departure from reference conditions, with the exception of soil loss evidence, are described in
detail in Pellant at al. (2000).
•
•
•
•
Water Flow Patterns
Rills
Gullies
Bare Ground
190
•
Soil Loss Evidence
This qualitative assessment technique, in association with quantitative monitoring information,
does not identify the causes of resource problems and should not be used to monitor land or
determine trends, but the indicator ratings can provide early warning of problems and help identify
areas that are potentially at risk for degradation.
Soil Compaction
Soil compaction data can be collected at the same sample locations as those used for vegetation
and land use. See Section 4.3.4 Soil Compaction Assessment for details on tools to assess soil
compaction.
Groundcover and Canopy Cover/Structure
Ground Cover (assign cover class)
• Percent bare ground (may be less, more, or equal to maneuver disturbance)
• Percent litter (detached leaves, stems, woody material in contact with the
ground)
• Percent rock/gravel (fragments >2mm diam.)
Vegetation Cover (assign cover class)
• Total vegetation cover (up to 100%) – includes vertical projection of all
structural layers to the ground.
Vegetation Structure
• Percent canopy cover for each stratum:
ƒ
overstory (>8.5m)
ƒ
understory (3-8.5m)
ƒ
shrub layer (all species) (0.5-3m)
ƒ
low/herbaceous (<0.5m)
Tree Density, Condition, and Regeneration
• Tree density by DBH and species – prism sampling. BAF conversion factors
are used to calculate densities from point counts. Tree species >4” DBH using
4” DBH classes.
Forest Health Attributes:
Most attributes and methods are adapted or excerpted directly from the USDA Forest Service
Forest Health Monitoring (FHM) Methodology (USDA Forest Service 1997a; USDA Forest
Service 2002b). Several attributes have since been eliminated from the FHM program and others
were excluded for site-specific reasons.
o
o
o
Tree Regeneration
Tree Crown Condition
o
live crown ratio
o
crown density
o
crown dieback
Tree Damage Condition
o
open wounds
o
damaged roots
191
o
o
o
early decay
advanced decay
broken/dead branches
Tree species are those that regularly attain a tree form (at least 5” or 13 cm DBH) and a height of
at least 5m (16’) (USDA Forest Service 1997a, p. A-1).
Photo Point
One photograph is taken at each sample point from a position 5m south of the plot center and
facing due north toward the plot center. A focal length of approximately 35mm is used. Photos
should also be taken of representative situations, including soil loss, damaged or stressed trees,
low density trees, high density trees, high value concealment, etc.
4.3.7.5 Data Analysis and Reporting
Quantitative results should be presented with confidence intervals or the results of quantitative
statistical procedures. Results should also be presented spatially for sample points or training sites
using red-amber-green or poor-fair-good type classifications. Quantitative results could also be
presented spatially (e.g., show erosion status for all areas surveyed) using different colors or icons.
Both quantitative and qualitative results can be grouped into poor, fair, good rating systems.
Photos should be used liberally to illustrate conditions and show changes over time. The following
summaries should be generated for each high-use area sampled. Future monitoring data can be
compared to initial data using t-tests (comparing two periods) and/or Analysis of Variance
(ANOVA) (comparing more than two periods).
Area Size, Training Evidence, and Damage Severity
• Size of high-use area (ha/ac)
• % of plots with training evidence, by type
• % of plots in each damage severity category
Erosion and Soil/Site Stability
• Mean USLE or RUSLE estimate expressed as % of soil loss tolerance, with
90% confidence interval.
• Percentage of plots in each departure category for each of the five attributes
• Use preponderance of evidence to assign an overall rating for the site for each
attribute.
Vegetation Structure
• Mean tree density (# live trees/ac) by size (DBH) class by genus, with 90%
confidence interval.
• Mean spacing of trees (all species) >=4” (10cm) DBH, with 90% confidence
interval.
• Mean seedling and sapling density (#/acre) by genus, with 90% confidence
interval.
• Mean percent canopy cover for each stratum (overstory, understory, shrub,
low/herbaceous), with 90% confidence interval for each.
Forest Health
Critical thresholds for the percentage of trees classified in each crown condition and damage
category have been developed by the U.S. Forest Service (USFS), and are described in Steinman
192
(2000) and Applegate (2003). The percent of trees classified in each class are compared to the
USFS thresholds and the area is assigned an overall rating of good (green), fair (amber), or poor
(red) condition. Thresholds for condition ratings are presented in Table 4-11.
Table 4-11. Rating system for each site assessment using tree crown and damage condition data.
USFS Threshold
% of Trees Meeting Threshold
Good
Fair
Poor
Condition
Condition
Condition
Crown Condition
Live crown ratio (LCR)
Crown density (CD)
Crown dieback (CDB)
<30% = low LCR
<30% = low CD
>10% = high CDB
<10%
<10%
<10%
10-20%
10-20%
10-20%
>20%
>20%
>20%
Single damage/tree
Single damage/tree
Single damage/tree
Single damage/tree
Single damage/tree
% of trees with any
type damage
% of trees with
damage index >2
<10%
<10%
<10%
<10%
<10%
<30%
10-20%
10-20%
10-20%
10-20%
10-20%
30-40%
>20%
>20%
>20%
>20%
>20%
>40%
<10%
10-20%
>20%
Damage Condition
Open wounds
Root damage/exposure
Cankers/galls
Advanced decay
Broken/dead branches
Any damage
Damage Index Rating*
* Combined category damage index for open wounds, root damage, cankers, and advanced decay, where none-slight=0,
slight-moderate=1, and moderate-severe=2. The USFS damage rating index cannot be used because actual percent damage
is not recorded for each tree. A damage index of 2 approximates the damage score of 50% combined damage described by
Steinman (2000). Trees with a damage rating index greater than 2 have a high likelihood of premature mortality.
Training Value
Training value is assessed for the amount of tree canopy cover >8.5m (aerial concealment),
horizontal cover (0.5-3 m concealment), and spacing and size of trees (index of vehicle mobility).
Green-amber-red maps of condition for the following attributes can be generated to illustrate
conditions and/or changes in training value over time. The values can be illustrated for each point
surveyed to show patterns within the area (e.g., bivouac, maneuver area) or averaged and the mean
value applied to the entire area (and presented with a confidence interval).
•
Aerial concealment >8.5m:
green (good) = 75-100%
amber (fair)
= 40-75%
red (poor)
= <40%
•
Horizontal cover (0.5-3 m concealment): This value is related to concealment in the
context of obscuring dismounted troops and visual assessment of photographs.
green (good) = 75-100%
amber (fair)
= 40-75%
red (poor)
= <40%
•
Vehicle mobility (average spacing between trees >4” in diameter):
green (good) = >8m spacing
amber (fair)
= 4-8m spacing
193
red (poor)
4.3.8
= <4
Water Quality Monitoring
Water quality monitoring can be an important component of an overall monitoring strategy to
evaluate the effects of land use activities at military installations. The quality of water flowing past
a given point in a stream is related to upstream and upland conditions. Water quality monitoring
can therefore be a useful tool for evaluating land condition at the watershed scale. In particular,
water quality monitoring can be used on military installations to: (1) assess the effects of military
training on water quality, (2) comply with water quality standards and criteria set by state or
federal agencies, (3) assess the effectiveness of Best Management Practices (BMPs) or other
rehabilitation treatments, and (4) assess long-term changes in water quality due to changes in land
use or climate.
A variety of federal laws provide the legal context for water quality monitoring. The Federal
Water Pollution Control Act Amendments of 1972, the Clean Water Act of 1977, and the Water
Quality Act of 1987 are collectively referred to as the Clean Water Act. Specific sections of the
Clean Water Act deal with non-point source pollution and BMPs, dredge and fill activities, and
requirements for states to list water bodies not meeting water quality standards. Further
background information on water quality legislation and the Clean Water Act is provided by
Novotny and Olem (1994) and the EPA Clean Water Act website,
http://www.epa.gov/r5water/cwa.htm.
The purpose of this section is to discuss the key elements of a water quality monitoring program.
Other tools to assess water quality such as aquatic biomonitoring are discussed in Section 4.5.4
Aquatic Biomonitoring. A water quality monitoring program has many of the same elements and
design considerations as any other monitoring project. Key elements of a water quality monitoring
program include:
•
•
•
•
•
•
•
4.3.8.1
Monitoring program strategy
Monitoring objectives
Monitoring design
Water quality attributes
Quality assurance/quality control
Water quality data management and analysis
Reporting and program evaluation
Monitoring Program Strategy
The monitoring program strategy is a long-term implementation plan that provides the ‘big
picture’ overview of the water quality monitoring effort. The strategy is developed before starting
the monitoring and should describe how each of the remaining program elements will be
addressed. The strategy can be referenced throughout the course of a monitoring project to help
maintain consistency and provide documentation to others. The strategy should be comprehensive
in scope and identify the technical issues and resource needs that are currently impediments to an
adequate monitoring program (USEPA 2003a). Guidance for developing a monitoring program
strategy is provided by Potyondy (1980), MacDonald et al. (1991), USDA NRCS (1997a), and
USEPA (1997).
194
4.3.8.2 Monitoring Objectives
The most important step in a water quality monitoring project is to develop clear, explicit, and
realistic monitoring objectives. The monitoring objectives address the why of water quality
monitoring. Monitoring objectives should originate from management objectives (Section 2.3
Management and Monitoring Objectives) and/or from other identified water quality problems. The
monitoring objectives will largely define the remainder of the monitoring project, including the
cost, attributes, sampling locations, sampling frequency, and data analysis techniques (MacDonald
et al. 1991).
The following are examples of broad questions that may form the basis for water quality
monitoring objectives:
•
•
•
•
•
•
•
What is the overall quality of water on the installation?
Does water quality on the installation meet water quality standards?
What are the effects of military training on water quality?
To what extent is water quality changing over time?
What are the problem areas and areas needing protection?
What level of protection is needed?
How effective are Best Management Practices (BMPs) or other rehabilitation
treatments?
Further guidance for developing water quality monitoring objectives is provided by Potyondy
(1980), MacDonald et al. (1991), USDA NRCS (1997a), and USEPA (1997). General guidance
for developing monitoring objectives is provided in Section 2.3 Management and Monitoring
Objectives.
4.3.8.2.1 Water Quality Standards, Criteria, and Designated Uses
Water quality standards, criteria, and designated uses are important to consider when
implementing a monitoring program and formulating broad monitoring objectives. Water quality
standards are legal requirements that define the goals for a waterbody by designating its uses,
setting criteria to protect those uses, and establishing provisions to protect water quality from
pollutants (USEPA 1994). Standards are set by each state in conjunction with the EPA. A water
quality standard consists of four basic elements:
•
•
•
Designated uses of the waterbody. Examples of designated uses include recreation, water
supply, aquatic life, agriculture, and navigation.
Water quality criteria to protect designated uses. Water quality criteria consist of
numeric limits or narrative descriptions of water quality. The EPA Gold Book provides
summaries of each contaminant for which the EPA has developed criteria
recommendations (USEPA 1986).
Antidegradation policy to maintain and protect existing uses and high quality waters.
The policy has three tiers or levels of protection. The lowest tier requires that existing uses
be fully supported. The middle tier requires maintenance and protection of high-quality
waters. The highest tier applies to waters designated by states as Outstanding Resource
Waters. These waters are the highest quality waters in the state and no degradation of
water quality is allowed.
195
•
General policies addressing implementation issues (e.g., low flows, variances, mixing
zones).
Regional and state information on water quality standards can be obtained from the EPA Office of
Water website, http://www.epa.gov/waterscience/standards/. Additional background information is
provided in MacDonald et al. (1991), Novotny and Olem (1994), USEPA (1994), and USEPA
(2003b).
4.3.8.3 Monitoring Design
The monitoring design will address the where, how, and when of water quality monitoring. A
number of different monitoring designs will be available to address the monitoring objectives. For
example, the EPA’s Environmental Monitoring and Assessment Program (EMAP) and the U.S.
Geological Survey’s National Water Quality Assessment (NAWQA) Program use different
monitoring designs to assess water quality at the national-level (USEPA 2003a).
4.3.8.3.1 Types of Monitoring
Three types of monitoring were previously described in Section 2 Introduction to Resource
Monitoring: implementation, effectiveness, and validation. In addition to these, MacDonald et al.
(1991) identify four types of monitoring relevant to water quality monitoring: (1) trend, (2)
baseline, (3) project, and (4) compliance. Trend monitoring is used at regular, well-spaced
intervals to determine the long-term trend in a particular water quality attribute. Baseline
monitoring is used to characterize existing water quality conditions, and establish a database for
planning or future comparisons. Project monitoring is used to assess the effects of a particular
activity or project. This is often accomplished by comparing data taken above and below, or
before and after the project. Compliance monitoring is used to determine whether specific water
quality standards or criteria are being met. MacDonald et al. (1991) note that these seven types of
monitoring are not mutually exclusive; a given monitoring project may yield data that that support
several monitoring types. USDA NRCS (1997a) provides further information on different types of
water quality monitoring.
4.3.8.3.2 Site Selection
The selection of monitoring sites will first depend on the hydrologic feature of interest.
Monitoring locations can include surface water sites such as streams, rivers, lakes, and reservoirs;
groundwater quality monitoring will require the selection of well sites. Sites for monitoring
atmospheric deposition or precipitation chemistry will require additional considerations. Factors to
consider when selecting a sampling site include monitoring objectives, accessibility, sources of
contamination, flow regime, mixing, and other physical characteristics of the watershed or stream
channel. The location of streamflow gauging stations may dictate the selection of monitoring sites
on streams and rivers.
Further guidance on site selection for water quality monitoring is provided by Potyondy (1980),
USDI Geological Survey (1982), Stednick (1991), MacDonald et al. (1991), USDA NRCS
(1997a), and USDI Geological Survey (1998).
196
4.3.8.3.3 Sampling Frequency
Sampling frequency is a function of monitoring objectives and data variability and will reflect any
budgetary constraints. Sampling frequency can be determined by calculating the minimum sample
size for a given error and confidence term (Section 3.2 Sampling Design; USDA NRCS 1997a).
Sampling frequency in water quality monitoring is also influenced by the relationship between
streamflow and the concentration of the constituent(s) of interest (Stednick 1991; MacDonald et
al. 1991). Normal unbiased sampling is best when there is no relation between streamflow and
constituent concentration (Figure 4-7a). Relations may also be flow-driven, where concentration
increases with streamflow (Figure 4-7b). Suspended sediment, orthophosphate phosphorous, and
fecal coliform may have this relationship. The opposite relation is flow-dilution, when
concentrations decrease as flow increases (Figure 4-7c). Nitrate or potassium may have this
relation (Stednick 1991). Finally, the flow-dilution-driven relation is when concentrations
decrease as flow increases, and then begin to increase as the flow continues to increase (Figure
4-7d). The latter increase is due to flow routing mechanisms; alkalinity or conductivity may have
this relation (Stednick 1991). A water quality monitoring program must either sample over the
range of variability for a given constituent or determine the most critical period and then
consistently sample at this time (MacDonald et al. 1991). This may require an initial period of
intensive sampling to determine the most sensitive period for a particular constituent.
Concentration (mg/L)
Concentration (mg/L)
(a) No relationship
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
25
50
75
100
125
(b) Flow-driven
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
150
0
25
(c) Flow-dilution
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
25
50
75
100
Streamflow (cfs)
50
75
100
125
150
Streamflow (cfs)
Concentration (mg/L)
Concentration (mg/L)
Streamflow (cfs)
125
150
(d) Flow-dilution-driven
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
25
50
75
100
125
150
Streamflow (cfs)
Figure 4-7. Typical relationships between streamflow and constituent concentration (Stednick
1991).
Further guidance on sampling frequency for water quality monitoring is provided by Potyondy
(1980), USDI Geological Survey (1982), Stednick (1991), MacDonald et al. (1991), USDA NRCS
(1997a), USEPA (1997), Stednick and Gilbert (1998), and USDI Geological Survey (1998).
197
4.3.8.3.4 Sample Collection
Sampling procedures vary according to the hydrologic feature of interest. Precipitation samples
are usually collected in glass or plastic containers. Containers may be continually exposed to
collect an integrated wet and dry sample (bulk precipitation collectors) or may be preferentially
exposed to allow separation of the wet and dry components (wet/dry precipitation collectors)
(USDI Geological Survey 1982). Samples from streams and rivers can be collected using grab
samples, depth-integrated samplers, or automated pumping samplers. Samples from lakes and
reservoirs can be taken at any depth using Kemmerer-type samplers (USDI Geological Survey
1982). Groundwater samples can be collected from wells using bailers or various types of pumps.
Many attributes can also be recorded using in situ sensors and probes (see next sub-section).
Further guidance on procedures for sampling surface water, groundwater, and precipitation is
provided by USDI Geological Survey (1982), Stednick (1991), USDA NRCS (1997a), USEPA
(1997), Stednick and Gilbert (1998), and USDI Geological Survey (1998).
4.3.8.4
Water Quality Attributes
The selection of water quality attributes to sample should be based on specific monitoring
objectives. Other considerations include water quality standards and criteria, designated uses,
legislative or intuitional mandates, and the available budget. A literature review will help answer
questions about which parameters to sample in a monitoring program (Stednick 1991).
4.3.8.4.1 Streamflow Measurement
Streamflow is one of the most important attributes in water quality monitoring (Stednick 1991).
Streamflow is a measure of the volume of water passing a given point on a stream per unit of time,
typically expressed in cubic feet per second (cfs), cubic meters per second (cms), or liters per
second (L/sec). A streamflow measurement should be made each time a water quality sample is
collected. Streamflow data are needed to calculate constituent loads and provide context to other
water quality data attributes.
Streamflow can be measured by number of different methods. The primary method is the velocityarea method, which uses the equation (Buchanan and Somers 1969):
Q = AV
where:
Q = streamflow
A = cross-sectional area of stream
V = stream velocity
The channel cross-section is divided into a number of subsections depending on depth and
velocity and the degree of precision required. The depth of each subsection is determined with a
wading rod. Velocity measurements are collected with a current meter at each subsection. For
sections deeper than 0.76 m (2.5 ft), two velocity measurements are taken at 0.2 and 0.8 times the
depth; otherwise a single velocity measurement is taken at 0.6 times the depth. Buchanan and
Somers (1969) provide further guidelines on using the velocity-area method for streamflow
measurement. Other direct measurements such as volumetric and tracer techniques are described
198
in Herschy (1995). Indirect methods such as the slope-area method (Manning equation) for
estimating peak discharge are described in Dalrymple and Benson (1976).
Streamflow measurements are usually made at established gauging stations where continuous
stage (water level) data are collected. The selection of a station type will depend on site conditions
such as stream size, but will also depend on specific monitoring objectives, data accuracy needs,
and cost. The major station types include weirs, flumes, and natural channels. Weirs and flumes
are artificial structures used on smaller streams, while natural stream channels are used when flow
is too large for an artificial structure. Stage is recorded with floats, bubblers, or pressure
transducers (USDA NRCS 1997a). Data are typically logged at 15 min intervals, although a
shorter time interval is recommended for flashy streams. A series of simultaneous stage and
streamflow measurements are used to construct a rating curve. Once a rating curve has been
established for a gauging site, stage measurements can be used to construct a continuous
streamflow record. Most flumes and weirs have known rating curves based on their cross-sectional
area, while it may take up to several years to establish a rating curve for a natural stream channel.
Herschy (1995) provides a comprehensive overview of streamflow measurement using different
types of stations and data collection systems.
4.3.8.4.2 Field-Measured Attributes
Certain water quality attributes should be measured in the field whenever possible. These
constituents can undergo chemical change during transport due to sensitivity to environmental
conditions. For example, changes in water temperature can affect pH, conductivity, alkalinity, and
dissolved oxygen (Stednick and Gilbert 1998). The following are the most common water quality
attributes measured in the field:
•
•
•
•
•
•
Water temperature is important because it may indicate thermal pollution and influences
several physical, chemical, and biological processes. Liquid-in-glass thermometers are
commonly used to measure water temperature in the field. Automated recorders can also
be placed in the stream to continuously monitor water temperature.
pH is the negative logarithm of the hydrogen ion activity in the water. pH is important in
the toxicity and solubility of many constituents, such as ammonia (Stednick and Gilbert
1998). pH is measured in the field using electrometric meters. Probes can also be mounted
in the stream to continuously recorded pH.
Turbidity is the capacity of a water sample to scatter light, recorded in nephelometric
units (NTU). Turbidity can be used as a surrogate measure for suspended sediment.
Turbidity is measured in the field using battery-operated turbidimeters. Turbidimeter
probes can also be mounted in the stream to continuously recorded turbidity.
Conductivity is the ability of water to carry an electric current, reported in
micromhos/centimeter (umhos/cm) or milliSiemens/meter (mS/m). Conductivity is
dependent on water temperature and is useful for estimating the concentration of dissolved
solids in water. Conductivity is measured in the field using a specific conductance meter.
Probes can also be mounted in the stream to continuously recorded conductivity.
Dissolved oxygen is a measure of the amount of oxygen in the water, reported in mg/L.
Adequate dissolved oxygen is needed to maintain aquatic life. Several methods are
available for measuring dissolved oxygen, including meters, the Winkler analysis, and
Hach method.
Alkalinity is the capacity of water to neutralize an acid to a specified pH (typically 4.5).
Alkalinity is reported as mg/L CaCO3 or milliequivalent/liter HCO3-C. Alkalinity is often
199
used to assess water sensitivity to acidic inputs, such as acid rain or acid mine drainage
(Stednick and Gilbert 1998).
A number of references provide further information on specific methods used to measure these
and other attributes in the field: USDI Geological Survey (1982), Stednick (1991), MacDonald et
al. (1991), USDA NRCS (1997a), Stednick and Gilbert (1998), and USDI Geological Survey
(1998).
4.3.8.4.3 Laboratory-Measured Attributes
Some water quality attributes either cannot be measured in the field or are typically measured in
the laboratory where the conditions allow for better analytical work. Some of these attributes may
also be measured in the field. The following are the most common water quality attributes
measured in the laboratory (Stednick and Gilbert 1998):
•
•
•
•
•
•
•
•
•
•
•
Acid neutralizing capacity (ANC) is the capacity of water to neutralize an acid to a
specified pH endpoint. ANC differs from alkalinity since the pH equivalence point is
determined analytically rather than fixed (i.e., 4.5).
Alkalinity can also be measured in the laboratory (see above sub-section for description).
Bedload is the portion of sediment that rolls along the streambed. Bedload is an important
component of the overall sediment yield from a watershed. Field methods for collecting
bedload samples are described in Section 4.3.1.1.3.1 Instream Sediment Monitoring and
Edwards and Glysson (1988). Bunte and MacDonald (1999) provide a comprehensive
overview of sampling considerations for bedload monitoring.
Chloride ion, dissolved is significant due to the corrosivity it can cause and its relevance
to the Safe Water Drinking Act.
Chlorophyll a is used to estimate phytoplankton biomass, reported in ug/L.
Fecal coliform bacteria can indicate the presence of waterborne pathogens, reported as
the number of bacterial colonies per 100 mL. Fecal coliforms can also be measured in the
field.
Hardness is the capacity of water to precipitate or waste soap, reported in mg/L CaCO3.
Some metals such as copper and zinc may be more toxic when hardness is low.
Nitrogen (as N), ammonia dissolved is the reduced form on nitrogen in solution,
reported as mg/L NH4-N. Ammonia is a major limiting nutrient in most aquatic systems,
and increases may result in eutrophication. High concentrations are indicative of
agricultural pollution.
Nitrogen nitrate (as N), dissolved is the oxidized form of aqueous nitrogen, reported as
mg/L NO3 or mg/L NO3-N. Nitrogen is a major limiting nutrient in most aquatic systems,
and increases may result in eutrophication. High concentrations are indicative of
agricultural pollution. The drinking water standard is 10 mg/L.
Nitrogen (as N), Kjeldahl is the sum of nitrogen contained in the free ammonia and other
nitrogen compounds which are converted to ammonium sulfate, reported in mg/L.
Nitrogen is a major limiting nutrient in most aquatic systems, and increases may result in
eutrophication. High concentrations are indicative of agricultural pollution.
Phosphorous (P), orthophosphate is the quantity of orthophosphate (phosphate molecule
only) in water, reported in mg/L. Phosphorus is often a limiting nutrient in aquatic
systems. A minor increase in concentration can significantly affect water quality. Sources
include sediments, fertilizers, and soaps.
200
•
•
•
•
•
•
Phosphorous (P), total includes orthophosphate, condensed phosphates, and organically
bound phosphates, reported as mg/L PO43-. Phosphorus is often a limiting nutrient in
aquatic systems. High concentrations are indicative of agricultural pollution.
Solids, total dissolved (TDS) is the concentration of dissolved solids in water, reported in
mg/L. TDS is operationally defined as the material that passes through a filter with a pore
size of 0.45 micron. Conductivity measurements provide a rapid indication of TDS, and
many conductivity probes have TDS reporting capabilities built in to them.
Suspended sediment is the portion of sediment load suspended in the water column,
reported in mg/L. Field methods for collecting suspended sediment samples are described
in Section 4.3.1.1.3.1 Instream Sediment Monitoring and Edwards and Glysson (1988).
Bunte and MacDonald (1999) provide a comprehensive overview of sampling
considerations for suspended sediment monitoring.
Sulfate, dissolved is the oxidized form of aqueous sulfur in water, reported as mg/L SO4.
Sources include acid mine drainage and acid rain. High concentrations can affect human
health and contribute to poor water taste in the presence of sodium and magnesium.
Trace metals include elements such as Ag, Al, As, B, Be, Cd, Co, Cr, Cu, Fe, Mg, Mn,
Mo, Ni, Pb, Sb, Se, Th, Va, and Zn, reported in mg/L or ug/L. High concentrations can
affect human and aquatic life.
Turbidity can also be measured in the laboratory (see above sub-section for description).
The standard reference for water quality laboratory analysis is the Standard Methods for the
Examination of Water and Wastewater (APHA 1998). Other useful references include USDI
Geological Survey (1982), Stednick (1991), Stednick and Gilbert (1998), and USDI Geological
Survey (1998).
In many cases, a contract laboratory may be needed to analyze water quality samples. The choice
of a laboratory should be made after reviewing the QA/QC program, sample turnaround time, and
cost. Some laboratories are EPA or state-certified to perform water quality analytical services, but
certified results will be more expensive than results from a non-certified laboratory.
4.3.8.5 Quality Assurance and Quality Control
Quality assurance and quality control (QA/QC) procedures should be followed in all phases of
water quality monitoring (USDA NRCS 1997a). A QA/QC plan should be developed and
incorporated into the overall water quality monitoring plan. QA/QC plans should indicate
appropriate levels of accuracy and precision, as well as reporting limits for each water quality
attribute collected. Guidance for the development of QA/QC plans is provided by USEPA (1997)
and USEPA (2001). An example QA/QC plan is provided by Hallock and Ehinger (2003). Water
quality laboratories will typically follow a series of QA/QC procedures, and these should be
obtained before sending samples in for analysis. Typical QA/QC procedures for the field and
laboratory include:
•
•
Chain of custody documentation provides a paper trail used to ensure the proper handling
of samples and sampling equipment.
Calibration and maintenance is necessary on all field and laboratory equipment.
Procedures for calibrating field instruments are usually provided by the equipment
manufacturer. Additional guidelines are provided by USDI Geological Survey (1982)
Stednick and Gilbert (1998), and USDI Geological Survey (1998).
201
•
•
•
•
•
Sample preservation and transport. All water quality constituents have a designated
preservation and holding time criteria. The primary preservation methods are acidification,
refrigeration, filtration, and preventing light from reaching the sample (APHA 1998).
Further guidance on sample preservation and holding times for various constituents is
provided by USDI Geological Survey (1982), USDA NRCS (1997a), and APHA (1998).
Field blanks are samples consisting of deionized water which are processed as actual
samples. Field blank results are expected to be below the method reporting limit. High
results may indicate sample contamination.
Field replicates consist of repeating the entire sampling procedure about 20 minutes after
initial samples have been collected. The field replicates will indicate variability due to
short-term instream processes, sample collection and processing, and laboratory analysis.
Field splits consist of splitting samples from a single sampling event (usually the field
replicate sample). The split samples eliminate the instream variability and isolate the field
processing and laboratory variability.
Laboratory spikes consist of samples that have been spiked with a known quantity of
each analyte being measured (Stednick and Gilbert 1998). Spiked samples are used to
estimate within-batch accuracy and the calibration of analytical instruments.
Further guidance on QA/QC procedures for the field and laboratory are provided by the references
listed in this subsection.
4.3.8.6
Water Quality Data Management and Analysis
All water quality data should be error-checked and validated before uploading into a database
application such as MS Access. An adequate description of all data collection and data
management procedures should be kept so the data can be understood and analyzed, even several
years after collection. USDA NRCS (1997a) provides further guidance on data management in
water quality monitoring.
A data analysis approach should be identified in the monitoring strategy before data are collected.
The approach should be appropriate to address the monitoring objectives and provide answers to
management issues and concerns. USEPA (1997), USDA NRCS (1997a), and USDA NRCS
(2002) provide a comprehensive overview of different statistical and graphical techniques that can
be used in water quality data analysis. These include traditional statistical approaches, as well as
methods for single watersheds, above and below watersheds, multiple watersheds, paired
watersheds, and trend stations. Section 3 Introduction to Sampling and Section 7 Data Analysis
and Interpretation provide a general background on data analysis techniques in natural resources
monitoring.
4.3.8.6.1 Statistical Software
Several statistical software packages were developed specifically for water quality data analysis.
These packages include WQStat-plus, DETECT, SDS, and ESTREND. USDA NRCS (2002)
provides a review of each of these software packages. Other general statistical packages are
described in Section 7.1.1 Software for Statistical Analysis.
4.3.8.6.2 Water Quality Databases
A number of national water quality databases exist and may be useful for obtaining additional data
202
for analysis and comparison. EPA’s STORET (STOrage and RETrieval)
(http://www.epa.gov/STORET/) is a repository for water quality data collected by state and other
federal agencies, universities, private citizens, and small volunteer groups. Users also have the
ability to upload data to the database. The USGS NAWQA Program
(http://water.usgs.gov/nawqa/index.html) has collected and analyzed water quality data from more
than 50 major river basins and aquifers across the U.S since 1991. A variety of data are available
for download. The EPA’s EMAP (http://www.epa.gov/nheerl/arm/) has conducted aquatic
resource monitoring throughout the U.S. A variety of data are available for download.
4.3.8.6.3 Water Quality Models
Several water quality models are available to simulate the transport of pollutants and their effects
on aquatic ecosystems. These models include BASINS, AQUATOX, CORMIX, WASP6, and
QUAL2K. Each model has its own unique purpose and data needs. The EPA’s water quality
modeling website, http://www.epa.gov/waterscience/wqm/ contains additional information and
links for each of these models. The EPA’s Watershed and Water Quality Modeling Technical
Support Center provides further information on water quality models
(http://www.epa.gov/athens/wwqtsc/index.html).
4.3.8.7
Reporting and Program Evaluation
A final report should be prepared as soon as possible after the monitoring project is completed. A
brief progress report or summary can be used to track progress on longer-term monitoring efforts.
The report should address each element of the monitoring strategy, particularly how well the
monitoring design and results address the monitoring objectives. Section 7.1.3 Guidelines for
Reporting Monitoring Results provides further guidance on reporting in natural resource
monitoring.
A water quality monitoring program should be periodically reviewed to determine how well the
program serves its objectives and management needs. This should involve an evaluation of each of
the program elements. It may be easier to consider the first season or set of data collection
activities as a pilot project. This will allow more flexibility to adapt the monitoring design to the
conditions and variability found in the field (MacDonald et al. 1991).
4.3.9
Fuels and Fire Effects Monitoring
4.3.9.1
Background
Naturally occurring or human-caused wildland fires are a common disturbance on many military
installations. Military training increases the risk of fire because munitions can act as ignition
sources. Consequently, the number of potential ignitions on military lands is often far greater than
on other public lands. Past fire suppression has also increased fuel loads and there is potential for
larger and more severe fires in many vegetation types, particularly in the western U.S. (DeBano et
al. 1998). Large, severe fires on military installations can present a serious risk to people,
infrastructure, and training lands. Fire effects on natural resources can include changes in
vegetation structure and composition, increased runoff and erosion rates, and altered wildlife
habitat. Smaller, less severe fires may be beneficial in that they help reduce fuel loads, prepare
203
seedbeds, thin overstock stands, increase forage production, and improve wildlife habitat (DeBano
et al. 1998). The ecological effects of fire are dependent on management objectives, vegetation
factors, and the relationship between the current and historic fire regimes (e.g., fire extent,
frequency, severity).
The purpose of this section is to discuss some common monitoring techniques that RTLA
coordinators can use to monitor fuels and fire effects. Other aspects of wildland fire management
such as fire behavior monitoring, fire suppression, and fuels reduction strategies are not discussed.
Several other resources provide this information (e.g., DeBano et al. 1998; USDI NPS 2003;
Peterson et al. 2004; USDI Geological Survey 2004). Furthermore, these aspects of wildland fire
management are typically addressed by other directorates and personnel at military installations.
RTLA coordinators should work closely with these other programs when implementing fuels and
fire effects monitoring projects.
4.3.9.2 Management and Monitoring Objectives
As with other resource monitoring, fuels and fire effects monitoring should be guided by specific
management and monitoring objectives. The monitoring objectives address the why of monitoring,
and should originate from management objectives (see Section 2.3 Management and Monitoring
Objectives) and/or from other identified wildland fire issues. The monitoring objectives will
largely define the remainder of the monitoring project, including the cost, attributes, sampling
locations, sampling frequency, and data analysis techniques (MacDonald et al. 1991).
The following are examples of broad questions that may form the basis for fuels and fire effects
monitoring:
•
•
•
•
•
•
•
What are the representative fuel types and fuel loading in important training areas?
What are the successional changes in burned areas over time?
How long does it take for vegetation cover to return to pre-fire conditions?
Are weedy species invading burned areas?
What are the effects of wildfire on runoff and erosion rates at the hillslope and
watershed scales?
Does prescribed burning have an effect on erosion rates at the hillslope scale?
How effective are postfire seeding treatments at increasing vegetation cover over
time?
Further guidance for developing fuels and fire effects monitoring objectives is provided by
Anderson et al. (2001) and USDI NPS (2003). These references also provide guidance on other
monitoring design considerations for fuels and fire effects monitoring. General guidance for
developing monitoring objectives is provided in Section 2.3 Management and Monitoring
Objectives.
4.3.9.3 Fire Regime and Fire History
Knowledge of the historical and current fire regime on the installation will provide background
information to fuels and fire effects monitoring efforts. The fire regime is determined by a variety
of factors, including fire frequency, fire size, fire interval, fire season, fire intensity, and fire
severity (Table 4-12; DeBano et al. 1998; Romme et al. 2003). Each of these components is
further controlled by a variety of other factors such as weather, climate, and fuels.
204
Table 4-12. Components of a fire regime (modified from Romme et al. 2003).
Component
Description
Fire frequency
Number of fires occurring within a specified area during a specified time period (i.e., number of
fires at Yakima Training Center, WA per year).
Fire size or extent
The size (hectares) of an individual fire, or the statistical distribution of individual fire sizes, or
the total area burned by all fires within a specified time period (i.e., total hectares within Fort
Leonard Wood that burned in 2004).
Fire interval
The number of years between successive fires, either within a specified landscape or at any
single point within the landscape.
Fire season
The time of year at which fires occur (i.e., spring and fall fires, when most plants are semidormant and relatively less vulnerable to fire injury, or summer fires when most plants are
metabolically active and relatively more vulnerable to fire injury).
Fire intensity
Amount of heat energy released during a fire. Fire intensity is rarely measured directly, but
sometimes is inferred indirectly from fire severity.
Fire severity
Fire effects to soil and vegetation (see Section 4.3.9.5.1)
Determining the fire regime and fire history of an area does not require monitoring, but extensive
literature searches and gathering of current and historical records may be needed. Fire regimes for
several vegetation types are relatively well-known, and may be determined from a literature
review and/or consulting with a local fire ecologist. Records on the frequency, size, interval, and
season of fires on a particular installation may be kept by various programs such as Range Control
or the federal fire department. In some cases, fire return intervals can be determined by analyzing
tree rings in increment cores or cross-sections, or by analyzing sediment deposition layers.
4.3.9.4 Fuels Monitoring
Along with heat and oxygen, fuels must be present in order for a fire to ignite. Fuels available for
burning consist of all vegetative materials that can potentially be ignited by a heat source. Fuels
monitoring is used to characterize the amount and type of fuels available for burning. The results
can be used to create fuels maps and assess fire behavior and risk. Fuels monitoring is typically
used before a fire takes place, but can also be used after a fire to evaluate the characteristics of
fuels consumed. Fuels are described according to their properties, complexes, and classification
(DeBano et al. 1998).
4.3.9.4.1 Fuel Properties
Fuel properties affect the manner in which a fire burns. Physical properties include the quantity
(fuel loading), size and shape, compactness, and arrangement of fuels. Fuel loading is the total dry
weight of fuel per unit surface area, and is a measure of the potential energy that might be released
by a fire (DeBano et al. 1998). Fuel loading in natural ecosystems range from 0.5 to over 400
Mg/ha (DeBano et al. 1998). Sites with higher fuel loadings, which burn at higher intensity,
generally have more severe impacts than those with lower fuel loadings. Chemical properties of
fuels affect their heat content and the types of pollutants emerging from a fire.
205
4.3.9.4.2 Fuel Complexes
Fuel complexes are associations of fuel components described in terms of vegetative cover or
habitat types (DeBano et al. 1998). A forest fuel complex might consist of a combination of
overstory trees, understory trees, herbaceous plants, litter, dead and downed fuels, or some
combination of these fuels. A rangeland fuel complex might only consist of grasses or a
combination of grasses and shrubs.
4.3.9.4.3 Fuels Classification
Fuels can be classified a number of different ways. One classification is based on the vertical
position of fuels (DeBano et al. 1998; Anderson et al. 2001):
•
•
•
•
Ground fuels are all combustible materials below the surface litter layer. These fuels
may be partially decomposed, such as duff, dead moss and lichen layers, and deep
organic layers (peat), or may be living plant material, such as tree and shrub roots.
Fire spread in a ground fire is usually slow because of compactness of ground fuels,
with burning by smoldering combustion.
Surface fuels are those on the surface of the ground, consisting of leaf and needle
litter, dead branch material, downed logs, bark, and tree cones. Many fires are ignited
in, and carried by, surface fuels.
Aerial fuels include the strata above the surface fuels and include all parts of tree and
tall shrub crowns. The aerial fuel layer consists of needles, leaves, twigs, branches,
stems, and bark, and living and dead plants that occur in the crowns such as vines,
moss, and lichens.
Ladder fuels bridge the gap between surface and aerial fuels. Fuels such as tall
conifers can carry a fire from the surface fuel layer into tree crowns.
Fuels can also be classified according to the dominant fuel type (e.g., forest, shrub, herbaceous,
litter, duff) or dominant plant species (e.g., pine, oak, chaparral). Fuels classification can also be
based on the fuel state, which is defined by the moisture condition of the fuel (DeBano et al.
1998). Moisture largely determines the amount of fuel available for burning at any given time.
Fuel classifications based on fuel state include categories used in the National Fire Danger Rating
System (NFDRS). This includes live fuels (grouped by category as woody or herbaceous fuels)
and dead fuels (grouped by size class as 1, 10, 100, or 1000 hour timelag classes). Additional
information on fuels is provided by Anderson et al. (2001) and USDI Geological Survey (2004).
4.3.9.4.4 Fuel Load Sampling Methods
The standard reference for inventorying and sampling fuels is the Handbook for Inventorying
Surface Fuels and Biomass in the Interior West (Brown et al. 1982). Despite the title, the
procedure is applicable almost anywhere. The procedure is used to determine the biomass of
several vegetation categories, including duff, litter, herbaceous vegetation (grasses and forbs),
shrubs, standing trees (< 3m in height), and downed woody material. A brief discussion of the
sampling procedure for each of these categories follows.
•
Duff. Duff is the O2 horizon or humus layer of the forest floor. The duff layer lies
below the litter layer and above the mineral soil. A randomly positioned line transect
is used to measure duff. The depth of duff is measured to the nearest 0.1 inch at a
206
•
•
•
•
•
defined interval along the transect. Duff measurements are taken at the same location
as downed woody material measurements.
Litter. Litter includes freshly fallen leaves, needles, bark flakes, cone scales, fruits,
dead matted grasses, and a variety of other vegetative parts. Litter cover is estimated
using four 30 x 60 cm plots. Percent litter cover is recorded on the subplot (standard
plot) with the greatest quantity of litter. Cover is estimated on the other three subplots
as a percent of that on the standard plot. Cover is estimated using a class system such
as Daubenmire cover classes. A litter sample is also collected from half of the
standard plot. The sample is oven-dried at 95° C for 24 hr and weighed.
Herbaceous vegetation. Herbaceous vegetation is measured on the same plots as
litter. Herbaceous cover is estimated on the standard plot and the other subplots. A
sample is then clipped from the standard subplot. The sample is oven-dried at 95° C
for 24 hr and weighed.
Shrubs. Shrub data are collected on two ¼ milacre (11 ft2) plots. Within each subplot,
percent cover of live and dead shrubs is estimated using the established cover classes.
The average shrub height is recorded to the nearest inch in each subplot. The number
of stems by species and basal diameter class are also counted. Seven basal diameter
classes are used; classes start at 0 to 0.5 cm and rise in increments of 0.5 cm.
Standing trees (< 3m in height). The number of trees per acre is measured by species
and height on a 1/300 acre plot. The biomass of foliage and branchwood by size class
can then be calculated from weight and height relationships. Diameter at breast height
(DBH) can also be measured, and tree age can be determined using cores. Larger trees
can also be included if desired.
Downed woody material. Downed woody material consists of all dead twigs,
branches, stems, and boles of shrubs and trees that have fallen and lie on or above the
ground. Downed woody material is sampled using a planar intersect. This technique
involves counting intersections of woody pieces with vertical sampling planes. First,
the number of 0-0.6 cm and 0.6-2.5 cm particles that intersect the sampling plane are
counted. Second, the number of 2.5-7.6 cm particles that intersect the sampling plane
are counted. Lastly, the diameter of all particles 7.6 cm and larger are estimated or
measured. Sound and rotten pieces are recorded separately.
The protocol also includes measuring slope, elevation, aspect, cover type, and habitat type. Once
the data have been collected, there are a series of calculations that are used to estimate biomass
and fuel loadings. Brown et al. (1982) provide further detail on the sampling procedure, equipment
needed, calculations, and examples. Other resources describe similar procedures for fuel load
sampling: Anderson et al. (2001), USDI NPS (2003), and USDI Geological Survey (2004).
4.3.9.5 Fire Effects Monitoring
A key step in assessing fire effects is to identify the resource(s) of interest. The resource(s) of
interest should be clearly stated in the monitoring objectives. Typical monitoring situations may
include assessing the effects of fire on vegetation attributes (i.e., cover, density, frequency, and
composition), soils, runoff, erosion, and wildlife habitat. A number of resources provide
background information on the effects of fire on different attributes in different vegetation types.
A literature review will aid in the design and implementation of any monitoring project to evaluate
the effects of fire. DeBano et al. (1998) provide a good summary of fire effects in a variety of
vegetation types in the U.S., with detailed information on effects to vegetation, soils, and water.
Other good resources include the Fire Effects Guide published by the National Interagency Fire
207
Center (Anderson et al. 2001), the U.S. Fish and Wildlife Service Fuels and Fire Effects
Monitoring Guide http://www.fws.gov/fire/downloads/monitor.pdf, and the USDA Forest
Service publications Effects of Fire on Flora (Brown and Smith 2000), Effects of Fire on Fauna
(Smith 2000), and Effects of Fire on Air (Sandberg et al. 2002). A source of information on fire
effects to individual plant and animal species is provided by the USDA Forest Service Fire Effects
(FEIS) database (http://www.fs.fed.us/database/feis/index.html). The FEIS provides updated
literature reviews of nearly 900 plant species, about 100 animal species, and 16 Kuchler plant
communities found in North America.
4.3.9.5.1 Burn Severity
Burn severity reflects the magnitude of disturbance to soils and vegetation (Wells et al. 1979;
USDA Forest Service 1995). Burn severity is a function of several variables including fire
duration, fuel loading, topography, and weather (Robichaud et al. 2000). Stratifying sites by burn
severity may be useful for monitoring the effects of fire on a resource of interest.
Burn severity is classified as low, moderate, or high based on the postfire appearance of litter and
soil (Ryan and Noste 1983; DeBano et al. 1998) (Table 4-13). In a low severity fire, most of the
organic matter is not consumed, but the litter surface is at least charred. The duff layer is still
intact and there are no visible effects of the fire at the soil surface (Table 4-13). In a moderate
severity fire, most of the organic matter is consumed; ash and scorching are found on the soil
surface (Table 4-13). High severity fires heat the mineral soil and consume all litter, duff, and
woody debris, leaving only ash at the soil surface (Table 4-13).
Burn severity is usually assessed for larger fires by flying over the burn area and using a GPS to
map severity classes. Satellite imagery is also becoming an important tool for mapping burn
severity over large areas. These methods use overstory tree scorching and not litter and soil
conditions to assign burn severity classifications (Robichaud et al. 2003). However, in some cases
the condition of the canopy may be similar to the condition of ground cover and soils. It is
recommended that monitoring at burned sites use ground-truthing to compare litter and soil
conditions to canopy conditions in cases where burn severity maps have been created.
Table 4-13. Burn severity classification based on the postfire appearance of litter and soil (modified
from Ryan and Noste (1983) and Robichaud et al. (2003)).
Burn severity
Soil and litter parameter
Low
Moderate
High
Litter
Duff
Small woody debris
Large woody debris
Ash color
Mineral soil
Scorched, charred, consumed
Intact, surface char
Partly consumed, charred
Charred
Black
Not changed
Consumed
Deep char, consumed
Consumed
Charred
Light colored
Not changed, scorched
Consumed
Consumed
Consumed
Consumed, deep char
Reddish orange or white
Altered structure, porosity, etc.
4.3.9.5.2 Vegetation
The magnitude of fire effects to vegetation is a function of several variables, including fire
frequency, fire duration, fire temperature, fuel loading, and the season of burning (DeBano et al.
1998; Brown and Smith 2000; Anderson et al. 2001). The ability of plants to survive a fire
depends mostly on their tolerance to heat and their resistance to fire (DeBano et al. 1998).
208
Knowledge of the fire regime of the site will help indicate how the plant community will respond
to burning. Fire can have varying levels of mortality to the crown, stems, and roots of plants.
Fire effects to vegetation communities may be expressed in terms of changes to cover, density,
frequency, weight, species composition, number, height, vigor, growth stages, age classes, and
phenology (Anderson et al. 2001). Section 4.1 Vegetation Attributes discusses methods for
sampling these attributes. Plant mortality and injury to trees also need consideration because they
are directly or indirectly related to fire effects. Injury to trees can be assessed in terms of char
height, scorch height, and percent crown scorch (USDI NPS 2003):
•
•
•
Char height is the maximum height of charred bark on each overstory tree. The
maximum height is measured even if the char is patchy.
Scorch height is the maximum height at which leaf mortality occurs due to radiant or
convective heat generated by a fire. Below this height, all needles are brown and dead;
above it, they are live and green.
Percent crown scorch is the percent of browning needles or leaves in the crown of a tree,
caused by heat from a fire.
Additional guidance on forest and tree measurements is provided by USDI Geological Survey
(2004) and Section 4.25 Forest and Tree Measurements. The USDA Forest Service Forest Health
Monitoring (FHM) methods may also be helpful in postfire vegetation monitoring (see Section
4.5.1 Forest Health Monitoring and Section 4.3.7 Bivouac and High-Use Area Monitoring).
USDI NPS (2003) suggests a number of attributes to monitor in prefire and postfire monitoring.
The attributes to monitor depend on the vegetation type; USDI NPS (2003) lists variables for
grassland plots, brush/shrub plots, and forest plots (Table 4-14).
A successful monitoring program should establish control plots outside the burned area so that
other factors such as climate or insect infestations can be separated from effects of the fire itself
(Anderson et al. 2001). For prescribed burns, control plots must be in similar vegetative
communities, with similar physical characteristics (e.g., slope, aspect, etc.), as the area to be
burned in order to make valid conclusions regarding fire effects. USDI NPS (2003) recommends
using photo points to help document changes and recovery in vegetation structure and
composition over time (see Section 4.2.6 Photo Monitoring for guidance). Additional guidance for
prefire and postfire vegetation monitoring is found in Anderson et al. (2001), USDI NPS (2003),
USDI Fish and Wildlife Service (2004), and USDI Geological Survey (2004).
209
Table 4-14. Attributes suggested for prefire and postfire monitoring in grasslands,
brush/shrublands, and forests (modified from USDI NPS 2003).
Vegetation Type
Attribute
Strata
Grassland
Herbaceous cover
Burn severity
Photographs
Brush/Shrubland
Forest
Herbaceous cover
Shrub density
Burn severity
Photographs
Tree density
Overstory
Pole
Seedling
DBH1/DRC2
Overstory
Pole
Live/dead count
Overstory
Pole
Fuel load
Herbaceous cover
Herbaceous density
Shrub cover
Shrub density
Burn severity
Photographs
Percent crown scorch
1
2
Overstory
Pole
Scorch height
Overstory
Pole
Char height
Overstory
Pole
Diameter at breast height
Diameter at root crown
4.3.9.5.3 Soil Water Repellency
A hydrophobic or water repellent layer can develop at or near the soil surface following burning
(DeBano 1981). The heat produced by the combustion of the litter layer vaporizes hydrophobic
compounds, which then move downward into the soil until they condense on cooler underlying
soil particles (DeBano 1981; DeBano 2000). A postfire water repellent layer can limit infiltration,
increasing runoff and erosion rates (DeBano et al. 1998; Robichaud et al. 2000).
The strength of postfire soil water repellency is controlled by several factors, including burn
severity, vegetation, soil texture, soil moisture, and time since burning (DeBano 1981; Huffman et
al. 2001). Soil water repellency is generally stronger and more extensive after high severity fires
because more organic compounds are vaporized. Vegetation type influences the amount and types
of hydrophobic compounds in plant materials. Coarse-textured soils have stronger water
repellency than finer textured soils due a lower specific surface. Most soils have a soil moisture
threshold at which they cease to be water repellent. Soil water repellency can return to pre-burn
conditions fairly rapidly (1-2 years) (Huffman et al. 2001), and usually persists no longer than six
years (DeBano 1981).
210
Soil water repellency can be assessed using several methods. The most common methods include
water drop penetration time (WDPT) and critical surface tension (CST) (DeBano 1981; Letey et
al. 2000). The WDPT is the time needed for a drop of de-ionized water to be absorbed into the
soil. The CST is determined by placing drops of varying concentrations of de-ionized water and
pure ethanol on the soil. If the drops are not absorbed within a given period of time (usually 5
seconds), drops with successively greater concentrations of ethanol are applied. The CST of the
soil is recorded as the surface tension of the solution that penetrates the soil (Letey et al. 2000).
Surface tension decreases with increasing ethanol concentrations, and lower CST values indicate
stronger soil water repellency (Letey et al. 2000). CST is the recommended method for assessing
soil water repellency because it is easier to measure and yields more consistent results than the
WDPT (Scott 2000; Huffman et al. 2001).
Many studies have shown that postfire soil water repellency can be highly variable over space,
even when site characteristics and burn severity classifications are similar (Scott 2000; Huffman et
al. 2001). Furthermore, some soils are naturally water repellent when dry. These factors require
consideration when developing sampling plans to assess postfire soil water repellency.
4.3.9.5.4 Postfire Runoff and Erosion
The disturbance to soil and reduction in ground cover following burning can trigger dramatic
increases in runoff and erosion (Tiedemann et al. 1979; DeBano et al. 1998). The magnitude and
duration of the increase will depend on fire severity, topography, soils, geology, vegetation type,
the amount and character of precipitation, and the proportion of the watershed that burned
(Robichaud et al. 2000). Postfire effects at the watershed scale include increased peak runoff rates
and sediment yields. These increases have the potential to affect infrastructure, water quality, and
aquatic habitat in areas downstream of fires.
Postfire runoff and erosion can be monitored using several methods. The selection of a method
will depend on several variables including monitoring objectives, site conditions, data accuracy
and precision needs, cost, time requirements, and the scale of the analysis. Methods for measuring
runoff and collecting water quality samples are discussed in Section 4.3.8 Water Quality
Monitoring. Methods for monitoring erosion at the plot, hillslope, and watershed scales are
discussed in Section 4.3.1 Soil Erosion. There are very few erosion prediction models currently
available that deal explicitly with postfire erosion. The Rocky Mountain Research Station of the
USDA Forest Service is in the process of developing several models to predict erosion after
wildfires (Elliot 2004). The Disturbed WEPP interface in particular can be used to predict postfire
erosion at the hillslope scale. An example of modeling postfire erosion using a modified version of
the Revised Universal Soil Loss Equation (RUSLE) and GIS is discussed in MacDonald et al.
(2000) and Miller et al. (2003).
4.3.9.5.5 Postfire Rehabilitation Treatments
Postfire rehabilitation treatments are used to help minimize runoff and erosion and reduce the
invasion of undesirable species following fires (Robichaud et al. 2000). Common hillslope
treatments include seeding, mulching, and contour log felling. On federal lands, Burned Area
Emergency Rehabilitation (BAER) teams assess fire effects and implement rehabilitation
treatments following large fires (USDA Forest Service 1995).
211
RTLA coordinators should work with LRAM staff to determine whether rehabilitation treatments
are needed. The effectiveness of postfire rehabilitation treatments was recently reviewed by
Robichaud et al. (2000). Many hillslope, channel, and road treatments have not been extensively
studied (Robichaud et al. 2000), although recent monitoring on large fires in the western U.S. has
provided more information on the effectiveness of several treatments (Robichaud et al. 2003).
RTLA coordinators may conduct effectiveness monitoring after treatments have been applied. The
results can be used to design better rehabilitation treatments based on the specific ecological
characteristics of the installation. Effectiveness monitoring techniques depend on the resource of
interest and the scale of the analysis. Methods to assess treatment effectiveness on reducing runoff
and erosion are discussed in Section 4.3.8 Water Quality Monitoring and Section 4.3.1 Soil
Erosion. An example of a monitoring project to evaluate the effectiveness of seeding treatments
on vegetation recovery is provided by Jones and Kunze (2004b).
4.3.9.5.6 Fire Effects Models
A variety of models are available to assess fire effects. Data needs and complexity vary by model,
and many are available for downloading at no cost. One model, the First Order Fire Effects Model
(FOFEM) predicts the effects of surface fire on tree mortality, fuel consumption, mineral soil
exposure, smoke production, and soil heating (Sutherland 2004). It uses four geographical regions:
Pacific West, Interior West, Northeast, and Southeast. Vegetation types provide an additional level
of resolution within each region. Other fire effects models are listed at: http://www.fire.org/, the
Fire Research and Management System (FRAMES) http://www.frames.gov/tools/, and
http://www.fs.fed.us/fire/planning/nist/distribu.htm#Distribution.
4.4
Overview of Original LCTA Design and Methods 9
A general overview of the development and objectives of RTLA is presented in Section 1.1
Introduction to the RTLA Program. A detailed description of the sample survey design and field
methodology is presented in the USACERL Technical Report entitled U.S. Army Land ConditionTrend Analysis (LCTA) Plot Inventory Field Methods (Tazik et al. 1992). The USACERL manual
also contains protocols for the collection of small mammal, reptile, amphibian, and songbird data,
as well as guidelines for vascular plant collections. Sampling animal populations is not addressed
in this document.
4.4.1
Sampling Design and Plot Allocation
Permanent field plots are used in LCTA to inventory and monitor the condition of natural
resources. By using permanent plots, variability from year to year is reduced compared to using
non-permanent plots. Plots are divided into two types. Core plots are located randomly and are
used to characterize the installation. Special-use plots may be random stratified, and are located in
a more subjective manner to address specific management issues related to resource trends and/or
military impacts.
The sampling design is a stratified random sample designed to ensure objectivity, randomness, and
representation (Warren et al. 1990). The procedure incorporates satellite imagery (e.g., SPOT,
LANDSAT) and digital soil survey information using a GIS. The Geographic Resources Analysis
Support System (GRASS) was commonly used for original LCTA plot allocations. Imagery dates
9
The term “LCTA” is used in this section to reflect the time period described.
212
are recommended to coincide with peak phytomass for each installation. An unsupervised
classification producing up to 20 landcover categories is performed on the image based on
reflectance values in the green, red, and near infrared wavelength bands. Within a GIS, the
landcover classification is superimposed on the soil survey (e.g., soil series) to produce unique
landcover/soil combinations occurring in discreet areas referred to as polygons. Polygons less than
two hectares in size are eliminated because they are difficult to locate in the field (Tazik et al.
1992).
Core plots are subsequently allocated randomly within each landcover/soil combination.
The number of plots allocated to each unique land cover/soil type combination is proportional to
the amount of land area represented by each category. For example, a category occupying 5
percent of the land area would receive 5 percent of the plots. The plot allocation is intended to
ensure that each unique category is represented so that the installation as a whole can be
characterized (Tazik et al. 1992). This stratified random approach was not designed to sample and
make inferences about each polygon on the allocation map. Field crews are provided with plastic
overlays registered to USGS 7.5 minute (1:24,000) quadrangle maps. The plastic overlays are
color coded with the eligible polygons and plot locations. Field crews establish the plots as close
as possible to the locations marked on the overlay. Where plot locations are inaccessible, the crew
leader may substitute a comparable location having the same landcover/soil combination, slope
steepness, and aspect (Tazik et al. 1992).
Recommended plot density was approximately one plot per 200 hectares, up to maximum of 200
core plots. The maximum number of plots is based on cost and logistical factors.
For specific concerns or situations requiring more intensive, site-specific samples, plots referred to
as “special use” plots are placed more subjectively. Special use plots are not included in data
analysis for installation-level summaries. Issues prompting the use of special-use plots include
evaluating land rehabilitation success, fire impacts, training impacts, and monitoring of specific
habitats. Special-use plots also are placed in areas excluded from training activities in order to
provide control plots for comparison with plots where training occurs.
4.4.2
Data Collection Methods
A summary of the original LCTA field methods for disturbance and vegetation is presented below.
4.4.2.1
Vascular Plant Inventory
One of the initial steps often taken in original LCTA implementation is to conduct a
comprehensive vascular plant (floristic) survey of an installation. The survey is conducted
typically over a period of several years and during various growing seasons in order to document
and collect the highest possible number of plant species. The survey provides a species list that is
based on voucher specimens for use in environmental documentation, and as an educational
resource for training natural resources staff and LCTA field crews. Plant surveys also are
beneficial in documenting the presence and locations of endangered, threatened, or rare species on
the installation.
4.4.2.2
Plot Establishment
Core plots are located in the field using Mylar overlays on United States Geological Survey
(USGS) 7.5 minute topographic maps. After determining the location, a random azimuth is chosen
213
for the direction of the transect, which should not cross into a different soil type/land cover
category. The beginning point is permanently marked with a metal pipe driven into the ground.
The azimuth of the plot is selected randomly, and excludes azimuths which would result in the
transect leaving the landcover/soil polygon. A total of five metal rods are driven into the ground at
25m intervals, and a 100m fiberglass tape is attached along the transect. Plots are relocated in
subsequent years with the aid of photographs, maps, and a metal detector.
4.4.2.3
Plot Inventory
The vegetation inventory is designed to determine the dominance of different types of vegetation
based on ground and canopy cover, as well as the density of succulent, shrubby, and woody
plants. Surface disturbance, ground cover, and aerial plant cover are measured at one meter
intervals along the tape using a modified point intercept method (Bonham 1989), also referred to
as the point-quadrat method (PQM) (Mueller-Dombois and Ellenberg 1974). Ground cover
categories are: basal cover by plant species, prostrate cover by species, litter by life form (grass,
forb, shrub), duff, rock, gravel and bare ground. Canopy cover is estimated by recording
vegetation that contacts the measuring rod from ground level up to 8.5m. Canopy cover is
recorded at 1 decimeter intervals up to 2m and then at 5 decimeter intervals to 8.5m. The
uppermost species above 8.5m also is recorded. Canopy cover categories are: foliar cover by
species, litter by life form, and dead wood. All woody plants within 3m of each side of the transect
(6m wide) are recorded by location, height, and species. The belt width and minimum height
requirements can be modified in extremely dense woody vegetation. A composite soil sample is
collected for textural, organic matter, and pH analyses. Additional qualitative information gathered
includes estimated soil depth, aspect, slope length and gradient, military and non-military uses,
land maintenance activities, and evidence of wind and/or water erosion. Standardized categories
and definitions for qualitative information is presented in Tazik et al. (1992). Photographs are
taken and specific location maps are drawn for each plot.
4.4.2.4
Plot Monitoring
To detect trends in land use, amount of disturbance, and vegetation cover and abundance, plots are
monitored periodically after the initial inventory. Short-term monitoring consists of recording land
use, maintenance, and erosion evidence in addition to performing an abbreviated belt and line
transect. Canopy cover is recorded as present or absent and as annual or perennial cover (or both)
above each point. The belt transect is monitored by tallying woody species by 1m height classes
up to 4m or in a height class of greater than 4m. Long-term monitoring at some installations
consists of initial inventory line procedures and short-term belt procedures. Short-term monitoring
is designed to document changes in ground cover and canopy characteristics. Long-term
monitoring is intended to document changes in species composition that cannot be determined
from short-term monitoring data.
4.4.3
Findings of the Independent Review Panel
In 1989, a panel of natural resources experts, composed of professionals from academic
institutions and government agencies, was hired to evaluate the LCTA approach and
methodologies in light of program goals and objectives articulated at the time. Comments
submitted by panel members in response to specific topics were synthesized in a report (Cook
1989). The Report of LCTA Review is presented in Section 4.8 Appendix Report of 1989 LCTA
Review. Although some concerns were voiced regarding specific aspects of the program, the panel
found the LCTA techniques to be valid for Army installations that are being managed as training
214
centers. It was concluded that the information generated by LCTA had worthwhile applications for
land managers and trainers, environmental compliance documentation, and land acquisition
evaluation. Moreover, the panel postulated that in terms of costs incurred in training area repairs
and noncompliance, the price of not implementing a resource monitoring program such as LCTA
would far exceed program implementation costs.
4.4.4
Strengths and Weaknesses of the Original LCTA Approach
LCTA was begun as a top-down initiative emphasizing uniform methodologies regardless of
differences in ecoregion characteristics or military training activities (i.e., disturbance types and
patterns). There are a number of benefits to having a standardized approach, including the ability
to prepare regional, MACOM, or national-level assessments using data that has been collected
uniformly. As stated in the Appendix Report of 1989 LCTA Review, the approach is considered
valid for meeting stated objectives. However, as discussed in Section 1.1, there have been
shortcomings in the application of standardized methods across a wide variety of environments,
and data quality control, summarization, and application has been lacking, partly because no
reporting requirements were instituted or required. As a result, the emphasis has historically been
on implementation and data collection, not needs analysis or evaluation at the installation level.
Some of the technical strengths and weaknesses of the original LCTA methodology are presented
below. For additional information regarding specific methods, see Section 4.2
4.4.4.1
Strengths
Precision and repeatability: The point intercept method has been documented as one of the most
repeatable methods for measuring vegetation for a wide variety of environments. As a result, year
to year variability and observer bias is minimized. Density data for woody species is generally
highly repeatable, especially where densities are low to moderate, primarily because the surveyor
is counting stems or individuals rather than estimating a parameter.
Permanent plots: Permanent plots, although requiring more time to relocate in the field, increase
sampling efficiency by minimizing spatial variability over time. The use of permanent plots allows
for the use of paired t-tests and repeated-measures ANOVA in data analysis.
Standardized protocols: The uniform application of methods helps to ensure that data are
comparable over time and among installations. Data that are used to calculate indicators of
condition (e.g., erosion status) is collected in a manner that is consistent.
Software for data management and summary: Data management and analysis software tools have
been developed to assist LCTA managers with data analysis and report preparation. Data such as
the handheld computed software minimizes the need for data transcription and facilitates loading
data into a database. Quality control procedures can be streamlined by using the available software
tools such as the ACCESS LCTA Program Manager (CEMML).
Promotes development of other essential data: Plot allocation needs and data collection promote
the development of other data that aids with land management and planning activities. This
includes satellite imagery and soil surveys, endangered species locations, and the mapping of
training types and intensities.
215
Photo documentation: LCTA plot photographs provide a consistent visual record of resource
condition, and are valuable in supporting trends or changes illustrated by data collected on the
plot.
4.4.4.2
Weaknesses
Inadequate sample size and spatial replication: The original LCTA sampling design does not
provide enough samples to detect desired changes in some attributes of interest. Larger sample
sizes are generally necessary, especially where within-type variability is relatively high due to
land-use disturbances.
Less common species are not well documented: The point intercept method is intended to provide
consistent estimates of abundance for the dominant plant species in a community. Less common
and rare species are not detected consistently or at all. Therefore, data analysis applications
emphasizing species composition or diversity are not well-suited to the line data. This is not true
for the belt transect data, however, which should adequately document species composition.
Methods that sample a larger area (not single points) tend to provide more complete composition
information, and are more appropriate where documenting species diversity is desired.
Clonal species : Clonal and sprouting woody species are difficult to count because it is often
difficult to determine what constitutes an individual. Rules of thumb should be developed for
some species to ensure repeatability in data collection. For example, stems meeting a minimum
diameter can be counted.
Best suited to grasslands/shrublands: Although point intercept methods are suited to a wide
variety of environments, they were developed in, and are best suited for grassland and shrubland
environments.
Slope length estimation: Slope length estimations made in the field for calculating erosion
estimates are often inconsistent. This is not truly a weakness of the method, but may be a result of
inconsistent or poor training of field staff. Although the guidelines for estimating slope lengths are
widely accepted, the application of the definition varies considerably among RTLA coordinators
and scientists alike. One way to minimize the effects of such “errors” is to apply the definition
uniformly at an installation over time.
Subjective estimates of plot attributes: Qualitative estimates of land uses, military and non-military
disturbance, and fire may vary greatly among observers. Moreover, the information collected is
generally presence-absence data that reveals little about the intensity, timing, or age of a particular
disturbance. Ground disturbance categories such as “vehicle pass” are only rough indicators of the
level of disturbance that occurred at a particular point.
Plot allocation needs change over time: Because plot allocation strata do not include training
areas, it is difficult to make inferences about specific training areas from the data collected.
Training areas may be an important strata to incorporate into plot allocation designs if condition
summaries by training area are desired. Additionally, maps such as vegetation maps that were
unavailable during initial implementation have subsequently been developed. Vegetation maps
and ecological classification maps may be valuable strata for plot re-allocation. Original
allocations can be modified through the addition or subtraction of random plots in (re)defined
strata.
216
Lack of reporting guidelines and requirements: Few or no requirements were established
regarding reporting of LCTA data or the contents of LCTA reports, and data were often left in
paper form and unavailable for summarization or examination. The significant lags in data entry,
quality control, reporting, and recommendations gave the false impression that the data and LCTA
as a whole had little or no value. Moreover, because little was done with LCTA databases on
many installations, evaluation of LCTA sampling design and methods was delayed.
4.5
4.5.1
Integrative Approaches
Forest Health Monitoring
Forest Health Monitoring (FHM) is a cooperative, multi-agency program using a multi-layered
approach consisting of several levels of monitoring at different spatial scales (Burkman 1992,
USDA Forest Service 1997a, USDA Forest Service 1997b, http://fhm.fs.fed.us/). The purpose of
FHM is to make assessments, monitor, and report on the long-term status, changes, and trends in
the health of forest ecosystems. Nationwide implementation is planned (USDA Forest Service
1997a). Data are intended to assist with national and regional assessments of forest condition, with
emphasis on National Forest Lands within states and ecological regions. The goal is to integrate
plot and survey data to provide a more complete picture of forest conditions. The national forest
health monitoring program is not intended for detecting small to medium changes at stand to
watershed scales. The program consists of four components:
(1) Detection Monitoring applies to all forested lands. It consists of a plot component and survey
component. The plot component surveys are performed on a system of permanent ground plots
assessed annually. Plots are based on the EPA EMAP grid (approx. 4600 forest plots nationally approx. 27 km apart). Each plot consists of 4 circular subplots (1/60 ha) placed 36 m apart. Each
subplot includes a 1/750 ha microplot (Figure 4-8). Indicators used to measure condition or change
are: lichen communities, ozone bioindicator plants, tree damage, tree mortality, vegetation
structure, plant diversity, tree crown condition, tree growth, and tree regeneration. Feature of the
survey component are symptoms of stress including defoliation, foliage discoloration, tree
dieback, main stem and branch breakage, and tree mortality. The survey component consists of
ground and aerial surveys of insect pests, diseases, and other stressor effects present in the vicinity
of the plot. The purpose of detection monitoring is to estimate baseline conditions and detect
changes and trends over time. Data from FHM plots and surveys are analyzed with other data to
determine if conditions and changes are within the normal range of variation, are improving, or are
cause for concern.
(2) Evaluation Monitoring involves more intensive and specific assessments of changes in forest
conditions identified by detection monitoring. It examines the extent, severity, and probable
causes of undesirable changes or improvements in forest health beyond the determinations made
in detection monitoring. Reports attempt to identify cause and effect relationships, associations
between forest health and forest stress indicators, and document the consequences of management
activities.
(3) Intensive Site Ecosystem Monitoring (ISEM) examines a standard set of indicators to
determine key components of ecosystem processes. Monitoring occurs at a small number of sites
(nationally) that represent important forest types. The H.J. Andrews Experimental Forest
217
(Oregon), the Fraser Experimental Forest (Colorado), and the Coweeta Hydrological Laboratory
(North Carolina) are ISEM sites.
(4) Research on Monitoring Techniques (ROMT) is research specifically designed to improve
the three other monitoring activities. ROMT attempts to identify new indicators, improve current
indicators, evaluate sampling designs for repeated sampling, and improve reporting protocols.
Figure 4-8. Forest health monitoring plot design. Non-destructive sampling occurs on the four
subplots; the annular plots are used for more invasive sampling (from USDA Forest Service
1996a).
To date, detection monitoring has been the focus of FHM activities. Sampling density is very low
(FMH plot density = 1 plot per 63,942 ha), providing little statistical power to extrapolate beyond
the specific plot areas. However, summaries are relevant to detecting status and trends at state,
regional, and national scales. Detection monitoring is most useful for high level information
reporting where little detail is necessary. Higher plot densities could provide some information at
the management unit level, and supplemental plots and measurements have been adapted to local
needs. Strengths of the FHM approach include an emphasis on standard protocols and reporting
procedures across states, and long-term continuity in methodology and data integrity. FMH also
uses numerous indicators to assess condition, thereby avoiding possible pitfalls associated with
using only one or several indicators.
Version 2.0 of the FIA field methods guide was released in January 2004, and describes the
current set of nationally consistent core field procedures used by all FIA units for projects which
begin after the date of release. This version of the field guide covers all core and core optional
measurements collected on Phase 2 plots (the standard base FIA sample grid, 1 sample location
per roughly 6,000 acres). Protocols for FIA Phase 3 measurements (also known as Forest Health
Detection Monitoring plots) were also updated in 2004, and can be downloaded from the USFS
FIA library at http://www.fia.fs.fed.us/library/field-guides-methods-proc/. These measurements
are collected on a 1/16th subset of the standard base FIA grid plots.
Critical thresholds for the percentage of trees classified in each crown condition and damage
category have been developed by the U.S. Forest Service (USFS), and are described in Steinman
218
(2002) and Applegate (2003). The percent of trees classified in each class are compared to the
USFS thresholds and the area is assigned an overall rating.
4.5.2
Rangeland Health
Rangeland health is defined as “the degree to which the integrity of the soil, vegetation, water, air,
as well as the ecological processes of the rangeland ecosystem, are balanced and sustained” (SRM
Task Group on Unity in Concepts and Terminology 1995). The concept and framework for
rangeland health arose from a perceived need to better assess and manage public rangelands. The
National Research Council Committee on Rangeland Classification was formed in 1989 to
examine the methods used by federal agencies to classify, inventory, and monitor rangelands. In
its report, released in January 1994 (Committee on Rangeland Classification 1994), the committee
recommended that the U.S. Department of Interior and the U.S. Department of Agriculture jointly:
(1) define and adopt a minimum standard for what constitutes acceptable conditions on
rangelands, (2) develop consistent criteria and methods of data interpretation to evaluate whether
rangeland management is meeting this standard, and (3) implement a coordinated and statistically
valid national inventory to periodically evaluate the health of federal and nonfederal rangelands.
Rangeland health was defined as the degree to which the integrity of the soil and ecological
processes of rangeland ecosystems are sustained (Committee on Rangeland Classification 1994).
The committee concluded that a “lack of a consistently defined standard of acceptable conditions
of rangeland ecosystems is the most significant limitation to current efforts to assess rangelands”,
and recommended that “the minimum standard for rangeland management should be to prevent
human-induced loss of rangeland health.”
The committee also concluded that the current confusion over the status of U.S. rangelands is
exacerbated because different individuals, groups, and agencies use different definitions, methods,
and terminology to evaluate and report the condition of rangelands. A solution to this problem
would require that government agencies, range users, and the public use the same criteria and
interpret those criteria in the same way to evaluate rangeland health (Busby and Cox 1994).
4.5.2.1
Indicators, Criteria, and Thresholds
Early thinking on the evaluation of rangeland health classifies sites as: (1) healthy if an evaluation
of the soil and ecological processes indicates that the capacity to satisfy values and produce
commodities is being sustained, (2) at risk if the assessment indicates an increased vulnerability to
degradation, and (3) unhealthy if the assessment indicates that degradation has resulted in an
irreversible loss of capacity to provide values and commodities (Committee on Rangeland
Classification 1994). Categorization of rangelands within these categories required defining the
boundary distinguishing healthy from at-risk rangelands and the boundary distinguishing at-risk
from unhealthy rangelands. Thresholds of rangeland health are boundaries between ecological
states that, once crossed, are not easily reversed. The initial criteria chosen for assessing rangeland
health include the stability of soils and watersheds, the integrity of nutrient cycles and energy
flows, and the functioning of ecological processes that enable rangelands to recover from damage.
Because direct measurement of site integrity and processes can be difficult and/or timeconsuming, biological and physical attributes are used as indicators of site integrity and ecological
function.
Further examination, development, and practical testing of these concepts resulted in a revised
framework described in Pellant (2000) and further refined by Pyke et al. (2002). The assessment
219
framework was changed to evaluate the degree of departure of each indicator from reference
conditions particular to that ecological site. The three attributes related to the seventeen indicators
of rangeland health are: 1) soil and site stability, 2) hydrologic function, and 3) biotic integrity
(Table 4-15). Degree of departure descriptions for selected soil stability and hydrologic indicators
are presented in Table 4-16.
Table 4-15. Standard indicators included in the rangeland health protocol and attributes to which
each indicator applies (Pyke et al. 2002 updated with Pellant et al. 2005 information).
220
Table 4-16. Degree of departure descriptions for soil and hydrologic stability indicators (from Pellant et al. 2000).
Indicator
Degree of Departure from Undisturbed/Reference Area
Extreme
Rill formation is severe
and well defined
throughout most of the
area.
Moderate to Extreme
Rill formation is
moderately active and
well defined throughout
most of the area.
Moderate
Active rill formation is
slight at infrequent
intervals, mostly in exposed
areas.
Slight to Moderate
No recent formation of rills;
old rills have blunted or
muted features.
None to Slight
Current or past formation of
rills is as expected for the site.
2. Water
Flow
Patterns
Extensive and numerous;
unstable with active
erosion: usually
connected
More numerous than
expected; deposition and
cut areas common;
occasionally connected
Nearly matches what is
expected for the site;
erosion is minor with some
instability and deposition.
Matches what is expected for
the site; some evidence of
minor erosion. Flow patterns
are stable and short.
Matches what is expected for
the site; minimal evidence of
past or current soil deposition
or erosion.
3.
Pedestals/
Terracettes
Abundant active
pedestaling and
numerous terracettes.
Many rocks and plants
are pedestalled, exposed
plant roots common.
Moderate active
pedestaling; terracettes
common. Some rocks and
plants are pedestalled
with occasional exposed
roots.
Slight active pedestaling;
post pedestals are in flow
paths and interspaces and/or
on exposed slopes.
Occasional terracettes
present.
Current or past evidence of
pedestalled plants or rocks as
expected for the site.
4. Bare
Ground
Much higher than
expected for the site.
Bare areas are large and
generally connected.
Moderately to much
higher than expected for
the site. Bare areas are
large and occasionally
connected.
Moderately higher than
expected for the site. Bare
areas are of moderate size
and sporadically connected.
Active pedestaling or
terracette formation is rare;
some evidence of past
pedestal formation,
especially in water flow
patterns and/or on exposed
slopes.
Slightly to moderately higher
than expected for the site.
Bare areas are small and
rarely connected.
5. Gullies
Common with indications
of active erosion and
downcutting; vegetation
is infrequent on slopes
and/or bed. Nickpoints
and headcuts are
numerous and active.
Moderate to common
with indications of active
erosion; vegetation is
intermittent on slopes
and/or bed. Headcuts are
active; downcutting is not
apparent.
Moderate in number with
indications of active
erosion; vegetation is
intermittent on slopes
and/or bed. Occasional
headcuts may be present.
Uncommon with vegetation
stabilizing the bed and
slopes; no signs of active
headcuts, nickpoints, or bed
erosion.
Drainages are represented as
natural stable channels; no
signs of erosion with
vegetation common.
1. Rills
Amount and size of bare areas
nearly to totally match that
expected for the site.
221
Table 4-16. Degree of departure descriptions for soil and hydrologic stability indicators (from Pellant et al. 2000).
Indicator
6. Litter
Movement
7. Soil
Surface
Resistance
to Erosion
Degree of Departure from Undisturbed/Reference Area
Extreme
Extreme; concentrated
around obstructions.
Most classes of litter
have been displaced.
Extremely reduced
throughout the site.
Biological stabilization
agents including organic
matter and biological
crusts virtually absent.
Moderate to Extreme
Moderate to extreme;
loosely concentrated near
obstructions. Moderate to
small size classes of litter
have been displaced.
Significantly reduced in
most plant canopy
interspaces and
moderately reduced
beneath plant canopies.
Stabilizing agents present
only in isolated patches.
Moderate
Moderate movement of
smaller size classes in
scattered concentrations
around obstructions and in
depressions.
Significantly reduced in at
least half of the plant
canopy interspaces, or
moderately reduced
throughout the site.
Slight to Moderate
Slightly to moderately more
than expected for the site
with only small size classes
of litter being displaced.
None to Slight
Matches that expected for the
site with a fairly uniform
distribution of litter.
Some reduction in soil
surface stability in plant
interspaces or slight
reduction throughout the site.
Stabilizing agents reduced
below expected.
Matches that expected for the
site. Surface soil is stabilized
by organic matter
decomposition products
and/or a biological crust.
222
4.5.2.2
Evaluation Procedure
Rangeland health evaluation involves six steps (Pellant 2000):
1. Identify the evaluation area and verify soils and ecological site for the area.
2. Develop expected indicator ranges for the ecological site. Visually familiarize yourself with
the 17 indicators at an Ecological Reference Area and rate the reference area against the
Ecological Reference Worksheet.
3. Review or modify descriptors of indicators.
4. Characterize the vegetation found at the evaluation area.
5. Rate the 17 indicators.
6. Determine the functional status of the three rangeland health attributes.
Each site is evaluated relative to its potential as an ecological site. An ecological site (formerly called
a rangesite) is “a kind of land with specific physical characteristics which differs from other kinds of
land in its ability to produce distinctive kinds and amounts of vegetation and in its response to
management” (SRM Glossary Update Task Group, 1998). The ecological site description or reference
area information may be derived from published information, personal experience, or actual reference
locations. In some cases, good site descriptions or reference site information may not exist and may
make the assessment process difficult or imprecise. At state and county levels, the NRCS is in the
process of developing ecological site descriptions and reference worksheets for ecological sites. It is
recommended that local managers revise the reference descriptions as necessary to reflect local
considerations. In the field, attribute ratings are based on the estimated departure from ecological site
description or ecological reference area(s).
Departure categories, from good to poor, are: none to slight, slight to moderate, moderate, moderate
to extreme, and extreme. In other words, an extreme departure from reference conditions is highly
undesirable. The rangeland health departure rating for each of the three attributes is determined using
the “preponderance of evidence” approach (i.e., selecting the most commonly occurring departure
rating among the indicators). In some cases, a subjective averaging procedure may be used, since all
indicators do not carry the same ecological significance.
This qualitative assessment technique, in association with quantitative monitoring information, can be
used to provide early warnings of resource problems in upland vegetation. These indicators do not
identify the causes of resource problems and should not be used to monitor land or determine trends,
but do provide early warning of problems and help identify areas that are potentially at risk for
degradation. Program developers Examples of quantitative measures that could be used to support
rangeland health assessments are presented in Table 4-17. Developers of the rangeland health
approach cite a lack of repeated attribute ratings at a single location and the fact that there are good
quantitative techniques for measuring indicators as reasons that this method should not be used for
monitoring (Pyke et al. 2002). Issues associated with observer knowledge and bias, problems of scale
(e.g., assessing sites vs. watersheds), and defining and visualizing a “reference” site (John Mitchell 10
pers. comm., David Pyke 11 pers. comm., John Buckhouse 12 pers. comm.) have also been raised.
Overall, the abiotic assessment portion appears less prone to observer bias and error than the biotic
portion.
10
Dr. John Mitchell, Range Scientist, USDA Forest Service Rocky Mountain Research Station, Lecture presented at Colorado
State University, Fort Collins, CO. November 1997.
11
Dr. David Pyke, Research Scientist, National Biological Service. Spring 1995.
12
Dr. John Buckhouse, Professor, Rangeland Resource Science, Oregon State University. April 1995.
223
Table 4-17. Examples of quantitative measurements and indicators related to the 17 qualitative
rangeland health indicators (Pyke et al. 2002).
224
Version 4 of Interpreting Indicators of Rangeland Health, Technical Reference 1734-6 (Pellant et al.
2005) is an updated version of the Version 3 Manual (Pellant et al. 2000), following the
recommendations of Pyke et al. (2002) and additional testing and workshop evaluation. The
indicators are unchanged in the 2005 Version. Although there are a number of minor changes, the
most significant change is the replacement of the Ecological Reference Area Worksheet with the
Reference Sheet. The Reference Sheet improves the assessment process on each ecological site by
integrating all available information to generate a single range of reference conditions for each
indicator.
4.5.2.3
State Standards and Guidelines
The following quotations from the Federal Register, Vol. 60, No. 35, pages 9955-9956, February
1995, describe the purpose of rangeland health standards and grazing guidelines and their
implementation:
“The fundamentals of rangeland health, guiding principles for standards and the fallback standards
address ecological components that are affected by all uses of public rangelands, not just livestock
grazing … The guiding principles for standards and guidelines require that State or regional standards
and guidelines address the basic components of healthy rangelands… The Department will use a
variety of data including monitoring records, assessments, and knowledge of the locale to assist in
making the ‘significant progress’ determination. It is anticipated that in many cases it will take
numerous grazing seasons to determine direction and magnitude of trend.”
The regulations on rangeland health specify fundamental principles that provide direction to States,
districts, land managers, and users in the management and use of rangeland ecosystems. Authorized
officers are directed by the Fundamentals of Rangeland Health, stated in 43 CFR 4180, to ensure that
the following conditions of rangeland health exist:
(a) Watersheds are in, or are making significant progress toward, properly functioning
physical condition, including their upland, riparian-wetland, and aquatic components;
soil and plant conditions support infiltration, soil moisture storage, and the release of
water that are in balance with climate and landform and maintain or improve water
quality, water quantity, and timing and duration of flow.
(b) Ecological processes, including the hydrologic cycle, nutrient cycle, and energy flow,
are maintained, or there is significant progress toward their attainment, in order to
support healthy biotic populations and communities.
(c) Water quality complies with State water quality standards and achieves, or is making
significant progress toward achieving, established BLM management objectives such
as meeting wildlife needs.
(d) Habitats are, or are making significant progress toward being, restored or maintained
for Federal threatened and endangered species, Federal Proposed, Category 1 and 2
Federal candidate, and other special status species.
In response to the requirements and intent of the Federal Regulations cited above, a number of states
or multi-state regions have developed standards for rangeland health, often associated with guidelines
for grazing management or administration. Standards are specific objectives for the desired condition
of the biological and physical components and characteristics of rangelands. The standards are
intended to be measurable and attainable. Guidelines are management approaches, methods, and
practices that are intended to achieve a standard. Guidelines typically identify and describe methods
of influencing or controlling specific land uses, are developed and consistent with the desired
225
condition and site capability, and may be adjusted over time. Standards and guidelines have been
developed and approved for a number of states and regions, including Washington and Oregon
(Bureau of Land Management 1997a), Arizona (BLM 1997b), California and Northwestern Nevada
(BLM 1997c), and Montana, North Dakota, and South Dakota (BLM 1998).
The standards and guidelines include criteria for meeting each standard, and lists of potential
indicators that can be used to assess criteria. However, standard methods (qualitative or quantitative)
and evaluation procedures are lacking.
4.5.3
Watershed Assessment
4.5.3.1
Background
Watershed assessment is an integrative process for collecting and analyzing information about
hydrological and ecological conditions at the watershed scale (REO 1995). Watershed assessments
are used to help establish causal relationships between physical and biological processes in order to
assess the effects of past, current and future land use activities. The assessments are used to support
ecosystem management and to guide monitoring and restoration of watersheds. The process includes
steps for characterizing the watershed, identifying a set of key management issues, evaluating various
resources within the watershed, and making management recommendations. The scale for watershed
assessments varies from approximately 20 to 200 square miles (12,800 to 128,000 acres) (REO 1995).
4.5.3.1.1
Cumulative Watershed Effects
The concept of Cumulative Watershed Effects (CWEs) is closely tied to watershed assessment.
Watershed assessment procedures were initially developed to evaluate cumulative effects from
forestry practices (Reid 1998). Cumulative effects result from the combined effect of multiple
activities over space or time (Reid 1993; MacDonald 2000). The National Environmental Policy Act
(NEPA) requires consideration of cumulative effects in addition to specific impacts when planning
projects on federal lands (Council on Environmental Quality 1997). CWEs are a special type of
cumulative effect that involve watershed processes such as the generation or transport of water and
sediment (Reid 1993). Watershed assessment can combine a background analysis with a procedure to
use the results to plan land-use activities that are intended to avoid cumulative impacts (Reid 1998).
4.5.3.2
Watershed Assessment Procedures
There is no universal procedure for conducting watershed assessment (Montgomery et al. 1995).
Every watershed will have a unique set of characteristics and management concerns, and no single
method will be appropriate for every location (Reid 1994). A variety of state and federal government
agencies, including Canada, have published manuals for conducting watershed assessment (Table
4-18). Most assessment manuals have been published by agencies in the western U.S. for lands
managed for timber production. More recently, procedures have been developed for other regions
such as the northeast and southeast U.S. Although there are slight differences in the scope and
methods of each assessment approach, the overall process is similar. An interdisciplinary team
approach is used regardless of the procedure followed. RTLA coordinators may choose to follow one
manual or combine methods and modules from several manuals. The EPA’s guide for tribes (USEPA
2000) is a good example of a method that incorporates ideas from several other manuals. For the
purposes of this discussion, the procedure used on Forest Service and BLM lands in the Pacific
Northwest (REO, 1995; REO, 1996) will be described. This procedure has been used to conduct a
226
multitude of watershed assessments and several examples are available online at agency websites.
Table 4-18. Watershed assessment manuals published by different government agencies.
Agency/Organization
Title
Reference
Ecosystem Analysis at the Watershed Scale, Section IVers. 2.2, Section II-Vers. 2.3
REO 1995; REO 1996
East-Wide Watershed Assessment Protocol
USDA Forest Service 2000
Hydrologic Condition Analysis
McCammon et al. 1998
Watershed Analysis and Management (WAM) Guide for
Tribes
USEPA 2000
Standard Methodology for Conducting Watershed
Analysis, Vers. 4.0
WFPB 1997
Oregon Watershed Assessment Manual
WPN 1999
Cumulative Watershed Effects Process for Idaho
IDL 2000
State of California
North Coast Watershed Assessment Program Methods
Manual
Bleier et al. 2003
State of Vermont
Vermont Stream Geomorphic Assessment: Phase I
Handbook: Watershed Assessment
VANR 2004
Community Watershed Assessment Handbook
CBP 2003
Watershed Assessment Procedure Guidebook, Vers.
2.1
BCMF 2001
Federal
USDA FS Region 6 and USDI BLM
USDA FS Regions 8 and 9
USDA FS and USDI BLM
USEPA Region 10
State
State of Washington
State of Oregon
State of Idaho
Regional
Chesapeake Bay Program
Canada
British Columbia Ministry of Forests
4.5.3.2.1
Watershed Characterization
The first step in any watershed assessment is to characterize the watershed. The purpose of watershed
characterization is to identify the dominant physical, biological, and human processes or features of
the watershed (REO 1995). Watershed characterization provides the ‘big picture’ overview of the
watershed and sets the stage for the remainder of the assessment. Watershed characterization is one of
the information and data gathering phases of the assessment, and can take considerable amounts of
time. GIS capabilities can greatly speed up the characterization process, provided that spatial data are
available for the watershed of interest. DeBarry (2004) provides a good review of the use of GIS in
watershed assessment and lists GIS layers and data sources needed for conducting watershed
assessment. Schubauer-Berigan et al. (2000) also provide a list of the types and sources of data
needed for watershed assessments.
REO (1995) suggests addressing seven core topics in the watershed characterization process (Table
4-19). These topics are also addressed in subsequent steps of the watershed assessment. Each core
topic has one or more questions to help guide characterization (Table 4-19). A variety of data sources
are needed to help address the topics, and in some cases little or no information will exist. Additional
core topics can be included such as climate and soils.
227
Table 4-19. Core topics and questions used in watershed characterization (modified from REO 1995).
Core topic
Questions
Erosion
processes
What are the dominant erosion processes within the watershed? Where have
they occurred or are they likely to occur?
Hydrology
What are the dominant hydrologic characteristics (e.g., total discharge, peak
flows, base flows) in the watershed? How is the hydrology influenced by climate
and other physical basin characteristics?
Vegetation
What is the array and landscape pattern of plant communities and seral stages
in the watershed (riparian and upland)? What disturbances influence these
patterns (e.g., fire, wind, climate)?
Stream
channel
What are the basic morphological characteristics of streams and the general
sediment transport and deposition processes in the watershed?
Water quality
What are the beneficial uses of water in the watershed? What is the state of
water quality? Which water quality parameters are of interest?
Species and
habitats
What is the relative abundance and distribution of species of concern in the
watershed (e.g. threatened and endangered species)? What is the distribution
and character of their habitats (both terrestrial and aquatic)?
Human uses
What are the major human uses in the watershed? Where do they occur?
4.5.3.2.2
Identify Key Issues and Formulate Questions
The next step in a watershed assessment is to identify a set of key issues of concern in the watershed.
Key issues are needed to provide direction and focus to the analysis. Sources for key issues include
information from existing management plans or assessments, discussions with other government
agencies, and state and federal water quality standards (REO 1995). Key issues may be prioritized to
ensure the assessment can be completed in a reasonable amount of time. Examples of key issues
include flooding, water quality degradation, streambank erosion, sediment deposition, diminished
streamflows, and habitat and aquatic life degradation (DeBarry 2004).
Once key issues have been identified, a set of guiding questions are formulated to address the issues
and frame the analysis. Questions are expected to be answered by the analysis and there may be more
than one question for a given issue. Examples of questions include:
•
•
•
4.5.3.2.3
What are the effects of roads on watershed hydrology and water quality?
What are the dominant sediment sources in the watershed and where are they located?
How have changes in habitat conditions influenced certain fish stocks?
Current Conditions
The next step in a watershed assessment is to document current conditions relevant to the key issues
and questions developed in step 2. This step is similar to the watershed characterization step, although
more detail is needed. In some cases, field data collection or field verification may be necessary. REO
(1995) suggests using the same set of core topics (Table 4-19) to help guide description of current
conditions in the watershed.
228
4.5.3.2.4
Reference Conditions
One of the more challenging steps in a watershed assessment is to describe the reference conditions of
the watershed and explain how conditions have changed over time due to human influence or natural
disturbances. Reference conditions can help establish goals and objectives to be used in management
plans. Identifying reference conditions may be hampered by the lack of historic data. The seven core
topics (Table 4-19) can also be used to provide a framework for discussing reference conditions.
4.5.3.2.5
Synthesis and Interpretation
The most important step in a watershed assessment is to synthesize and interpret information about
the current and reference conditions of the watershed and to integrate these findings with the key
issues and questions. This is probably the most challenging step in the assessment procedure because
of the difficulty in synthesizing across disciplines. Several watershed assessments on federal lands
have focused exclusively on identifying and assembling information (watershed characterization) and
spent little time on synthesis and analysis (Reid et al. 1996). Summarizing key findings in relation to
the guiding issues and questions provides a basis for management recommendations in the next step.
4.5.3.2.6
Recommendations
The final step in a watershed assessment is to make management recommendations based on the
results of the assessment. Recommendations can be used to modify land use activities or identify and
prioritize restoration activities in the watershed. Monitoring and research opportunities are identified.
Important data gaps and information needed for making future land management decisions are also
identified.
4.5.3.2.7
Watershed Assessment Modules
Many watershed assessment manuals have specific modules that are used to examine the physical and
biological processes in a watershed
Table 4-20). The modules can be used as part of the characterization process or may be used to help
answer key questions. The topics for modules are similar among many manuals, although the data
needs, level of detail, and time requirements can vary considerably. Some modules require extensive
data collection and analysis, but many can be conducted in the office using GIS.
Table 4-20 shows modules by assessment manual for hydrology, erosion/sediment sources, stream
channel, riparian vegetation, water quality, fish habitat, and land use. It may also be necessary to use
other accepted assessment tools and models to complete a watershed assessment. The EPA Watershed
Tools Directory provides a comprehensive list of resources that may be helpful for conducting
watershed assessment (http://www.epa.gov/OWOW/watershed/tools/).
229
Table 4-20. Watershed assessment modules by assessment manual.
Modules
Assessment Manual
Hydrology
Erosion/
Sediment
Sources
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Federal
USDA FS Region 6 and USDI BLM
USDA FS Regions 8 and 9
USDA FS and USDI BLM
USEPA Region 10
State
State of Washington
State of Oregon
State of Idaho
State of California
State of Vermont
Regional
Chesapeake Bay Program
Canada
British Columbia Ministry of Forests
x
x
Stream
Channel
Riparian
Vegetation
Water
Quality
Fish
Habitat
Land
Use
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
4.5.3.3 Watershed Assessment Examples
Several watershed assessment examples are available online at agency websites. Table 4-21 lists
website addresses for watershed assessments completed using four of the manuals discussed in this
section. Schubauer-Berigan et al. (2000) also review ten different watershed assessments and describe
the data types, sources of data, data reliability, as well as other information used in each of the ten
assessments.
Table 4-21. Examples of watershed assessments by agency/organization.
Agency
Web link
Federal
USDA FS Region 6
Deschutes NF
(10 assessments)
http://www.fs.fed.us/r6/centraloregon/projects/planning/major-plans/index.shtml
State
State of California
(Several
assessments)
State of Vermont
(2 assessments)
4.5.4
http://cwam.ucdavis.edu/
http://www.anr.state.vt.us/dec/waterq/rivers/htm/rv_geoassess.htm
Aquatic Biomonitoring
Aquatic biomonitoring is used to evaluate the biological health and ecological integrity of a
waterbody. The presence of different types of benthic macroinvertebrates, fish, and periphyton can
reflect past as well as current water quality conditions. Biomonitoring data can be used to develop
narrative or numeric biocriteria to help protect water resources at risk from chemical, physical, or
biological impacts (Gibson et al. 1996; Barbour et al. 1999; USEPA 2002). These narrative and/or
230
numeric biocriteria may be formally adopted into water quality standards. Aquatic biomonitoring can
be used as a surrogate for more traditional forms of water quality monitoring, or it can be used as a
component of an overall water resources monitoring program. Aquatic biomonitoring has been
incorporated into the RTLA programs at Camp McCain and Camp Shelby, Mississippi (Howell 2001)
and Indiantown Gap, Pennsylvania. USEPA (2002) provides a comprehensive summary of aquatic
biomonitoring programs used by different states and tribes.
4.5.4.1 Aquatic Bioindicators
The EPA’s Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers (Barbour et al.
1999) is the standard reference for aquatic biomonitoring in riverine systems. The Rapid
Bioassessment Protocols (RBPs) provide field methods for sampling the common bioindicators used
in aquatic biomonitoring such as benthic macroinvertebrates, fish, and periphyton. Aquatic
biomonitoring strategies for lakes and reservoirs are discussed in USEPA (1998).
4.5.4.1.1 Benthic Macroinvertebrates
Benthic macroinvertebrates are the most popular bioindicator used in aquatic biomonitoring.
Macroinvertebrates are good indicators of habitat health and water quality because they: (1) live in
the water for all or most of their life, (2) stay in areas suitable for their survival, (3) are easy to
collect, (4) differ in the their tolerance to the amount and type of pollution, (5) are relatively easy to
identify, (6) often live for more than one year, and (7) have limited mobility (USEPA 1997; Klemm
et al. 1998; Barbour et al. 1999).
The EPA RBP identifies two field methods for sampling macroinvertebrates: the Single Habitat
Approach: Meter Kick Net, and the Multihabitat Approach: D-Frame Dip Net (Barbour et al. 1999).
Barbour et al. (1999) provide guidance on laboratory processing and list taxonomic references for
identifying macroinvertebrates.
A variety of benthic metrics can be calculated using data collected from macroinvertebrate surveys.
Metrics provide a measure of habitat or stream quality; in most cases an increase in disturbance will
reduce the value of the index used. Indices are used in part to develop biocriteria for water quality
standards. Some metrics have been developed for specific locations or regions while others are
appropriate over wide geographic areas. Barbour et al. (1999) present five general categories of
metrics:
•
•
•
Taxa richness measures represent the number of distinct taxa, or diversity within a sample.
Taxa richness usually consists of species-level identifications but can also be evaluated as
higher taxonomic groups (i.e., genera, families, orders). Increasing diversity correlates with
increasing health of the assemblage and suggests that niche space, habitat, and food source
are adequate to support survival and propagation of many species.
Composition measures are characterized by several classes of information such as identity,
key taxa, and relative abundance. Identity is the knowledge of individual taxa and associated
ecological patterns and environmental requirements. Key taxa provide information that is
important to the condition of the targeted assemblage. Measures of composition (or relative
abundance) provide information on the make-up of the assemblage and the relative
contribution of the populations to the total fauna.
Tolerance/intolerance measures are intended to be representative of relative sensitivity to
perturbation and may include numbers of pollution tolerant and intolerant taxa or percent
composition. The tolerance/intolerance measures can be independent of taxonomy or can be
specifically tailored to taxa that are associated with pollution tolerances. 24, 1
231
•
•
Feeding measures encompass functional feeding groups and provide information on the
balance of feeding strategies in the benthic assemblage. Examples involve the feeding
orientation of scrapers, shredders, gatherers, filterers, and predators. Trophic dynamics are
also considered and include the relative abundance of herbivores, carnivores, omnivores, and
detritivores.
Habit measures denote the mode of existence among benthic macroinvertebrates. Habit
categories include movement and positioning mechanisms such as skaters, planktonic, divers,
swimmers, clingers, sprawlers, climbers, burrowers.
Further details on specific metrics under these categories are provided in USEPA (1997) and Barbour
et al. (1999).
4.5.4.1.2 Fish
Fish are useful for aquatic biomonitoring because (Barbour et al. 1999): (1) fish are good indicators
of long-term effects and broad habitat conditions because they are relatively long-lived and mobile,
(2) fish assemblages generally include a range of species that represent a variety of trophic levels, (3)
fish are at the top of the aquatic food web and are consumed by humans, making them important for
assessing contamination, (4) fish are relatively easy to collect and identify, (5) environmental
requirements of most fish are comparatively well known, (6) aquatic life uses in water quality
standards are typically characterized in terms of fish, and (7) fish account for nearly half of the
endangered vertebrate species and subspecies in the United States.
The EPA RBP identifies electrofishing as the standard method for sampling fish populations (Barbour
et al. 1999). Fish surveys should yield a representative sample of the species present at all habitats
within a sampling reach that is representative of the stream. Barbour et al. (1999) recommend that a
habitat assessment be performed and physical/chemical parameters be measured concurrently with
fish sampling. Field procedures for these assessments are provided in Barbour et al. (1999). Barbour
et al. (1999) provide guidance on laboratory processing and list taxonomic references for identifying
fish species.
The Index of Biotic Integrity (IBI) is the common fish assemblage assessment approach (Karr 1981).
The IBI incorporates the zoogeographic, ecosystem, community, and population aspects of the fish
assemblage into a single ecologically-based index. The IBI is an aggregation of 12 biological metrics
that are based on the fish assemblage’s taxonomic and trophic composition and the abundance and
condition of fish. A summary of each of these metrics is provided in Barbour et al. (1999).
4.5.4.1.3 Periphyton
Periphyton is useful for aquatic biomonitoring, despite not having been incorporated in many
monitoring programs. Periphyton are a community of organisms (soft algae, algal or filamentous
mats, and diatoms) that adhere to and form a surface coating on stones, plants, and other submerged
objects in water (USEPA 1997). Periphyton is advantageous in aquatic biomonitoring because
(Barbour et al. 1999): (1) algae generally have rapid reproduction rates and very short life cycles,
making them valuable indicators of short-term impacts, (2) as primary producers, algae are most
directly affected by physical and chemical factors, (3) sampling is easy, inexpensive, requires few
people, and creates minimal impact to resident biota, (4) standard methods exist for evaluation of
functional and non-taxonomic structural characteristics of algal communities, and (5) algal
assemblages are sensitive to some pollutants which may not visibly affect other aquatic assemblages,
or may only affect other organisms at higher concentrations.
232
The EPA RBP identifies two field protocols for periphyton (Stevenson and Bahls 1999). The first
protocol is a standard approach in which species composition and/or biomass of a sampled
assemblage is assessed in the laboratory. The second protocol is a field-based survey of periphyton
biomass and coarse-level taxonomic composition (e.g., diatoms, filamentous greens, blue-green
algae) and requires little taxonomic expertise. The first protocol has the advantage of providing more
accuracy in assessing biotic integrity and in diagnosing causes of impairment, but it is labor-intensive.
Stevenson and Bahls (1999) provide guidance on laboratory processing and list taxonomic references
for identifying periphyton.
Periphyton indices of biotic integrity have been developed and tested in several regions. Stevenson
and Bahls (1999) summarize nine metrics that can be used to assess biotic integrity, and six metrics
that can be used to diagnose causes of impaired biotic integrity.
4.6
Monitoring Ecological Integrity on Public Lands
A number of approaches to evaluating resource condition have been developed and implemented to
varying degrees by agencies which manage or regulate public lands as well as those that provide
outreach and technical services to private landowners. Most programs were developed and
implemented as top-down approaches to satisfy regional or national-level reporting requirements.
Currently, most land management and support agencies use a variety of approaches, including
standardized top-down approaches, to assess resource condition and determine the direction of
resource trends.
Rangeland resource assessments have the longest history of data collection for the purpose of
measuring indices of integrity or condition. Approaches for assessing the condition of grasslands and
semi-arid rangelands were developed primarily by the BLM and the NRCS (Dyksterhuis 1949 and
1958, Pendleton 1989), and provide a lengthy and continuous source of resource condition. Most
methods for assessing rangeland condition, such as the system of range condition classification, were
based on the Clementsian concepts of plant community succession and climax. Reviews of the
historical development of rangeland monitoring are presented in West (1985b) and Joyce (1993). In
light of changing management goals, recent criticisms of the traditional range condition concept have
stressed the need to assess condition from a more objective, holistic perspective which integrates
attributes related to biodiversity, soil erosion, and other ecosystem characteristics (Committee on
Rangeland Classification 1994, Joyce 1993, Lauenroth 1985, West 1993) and the use of standard
terminology and benchmarks (Range Inventory Standards Committee 1983, Society for Range
Management 1990). A comparison of rangeland reference terms used by the U.S. Forest Service,
Bureau of Land Management, and Natural Resources Conservation Service is presented in Table
4-22.
233
Table 4-22. Comparison of terms and definitions used by three public land management agencies
(Committee on Rangeland Classification 1994).
234
Table 4-22. Continued.
235
Other approaches to grassland and rangeland monitoring are the Parker Procedure (Moir 1989) and
the short-lived Soil-Vegetation Inventory Method (Bureau of Land Management 1979). Standardized
methodologies applied in a range of environments (e.g., grasslands, shrublands, woodlands, etc.)
include the Land Condition-Trend Analysis approach (Tazik et al. 1992), EMAP (USEPA 1992),
Forest Health Monitoring (Burkman and Hertel 1992) (see Section 4.5.1), and fire monitoring
protocols (USDI NPS 2004). A comparison of standardized approaches for assessing resource
condition is presented in Table 4-23.
236
Table 4-23. Comparison of monitoring systems (modified from West et al. 1994). Question marks
indicate that some uncertainty is associated with the information contained in the cell.
Range
Condition
(NRCS*)
Purpose
Scope
Implementation
conserve natural
resources while
sustaining military
training
military installations
entire nation and states:
(mainly ARMY)
individual ranches, private and state-owned
one point in time
nationwide and
data collected as
agricultural lands.
survey on allotments
overseas: data
funding is available Surveys about every 5collection every 1-3
10 years
yrs
top-down at first, but
top-down initially,
bottom-up
MACOM and
bottom-up
top-down
modifications have installation initiatives
emerged over time
more recently
1950s-present
Conceptual
underpinnings
succession to
climax
Trend
LCTA
(DoD)
assess soil and water
allocate forage for all
conserve soil while conservation needs and
animals on
grazing livestock
set priorities for action
rangelands
and budgeting
Implementation
status
Indicators
Soil –Vegetation
National Resource
Inventory Method
Inventory (USDA)
(BLM)
plants, ranked by
response to
livestock grazing,
range condition
estimated
1967-present
1979-1984
succession to climax for
succession to climax
vegetation
soil surface characters
and plant community
composition
range condition and
ecological status
apparent trend
based on
determined from
apparent trend based
comparison of
successive 10-year data on plant species and
condition data
bases of condition
surface soil factors
collected over time
Benchmarks
relict areas and
intuition
comparison with earlier
surveys
begins with first
inventory
Ecological
Integrity
condition:
excellent, good,
fair, poor
condition: excellent,
good, fair, poor
condition: excellent,
good, fair, poor
Parker 3-Step
Procedure (USFS)
sustain lands for
multiple uses
livestock grazing
allotments
bottom-up
1989-present
1956-present
(procedure varies by
region)
soil stability, none
specified for
vegetation
succession to desired
plant community and
protection of soil
soil erosion and
vegetation change
plants ranked by
response to grazing,
soil surface condition
real trend based on
baseline data
apparent trend (?)
initial baseline from
one or more years of
relict areas and
data, soil erosion
intuition (?)
status
(erosion/tolerance)
only specified for soil
high seral, mid seral,
erosion as % of soil
low seral
loss tolerance
237
Table 4-23 Continued.
Purpose
EMAP
(EPA)
Unity in
Concepts and
Terminology
(SRM)
Rangeland Health
(NAS)
Forest Health
Monitoring
(USFS)
Fire Monitoring –
Western Region
(NPS)
develop a national
report card of
environmental
"health"
inventory land for
sustainability
assess rangeland
ecosystems
assess long-term
status and trends in
forests
evaluate the effects
of prescribed fire
programs
systematic national
individual ranches
national survey with
survey: each point
and national survey
no time frame
visited once each 4-5
collected periodically
defined
years
Scope
national survey done National parks in the
western U.S.
at 3 scales. 1/3 of
Surveys performed
plots surveyed
before and after fires;
annually. State,
four levels of
regional, and
monitoring intensity.
national reports
Implementation
top-down
top-down
intermediate
top-down. Plots
surveyed within 2
years of burn
top-down
Implementation
status
late 1980s-present
recommended in
1990
recommended in
1994 – ongoing
implementation
1990-present
1990-present
Conceptual
Underpinnings
--
soil stability
soil stability
Indicators
Trend
Benchmarks
Ecological
Integrity
*NRCS
BLM
DOD
USFS
EPA
SRM
NAS
NPS
ecological responses succession following
to stressors
fire
plant, soil, and other soil stability, nutrient
indicators being
cycling, recovery
tested (?)
indicators
forest pathogens,
tree damage and
mortality, crown
condition, biological
diversity, others
tree density and size,
fuel loads, fire
characteristics,
herbaceous and
brush cover and
density, ratio of nonnatives/natives
real trend, based on
baseline data
real trend based on
baseline data (?)
real trend based on
baseline data (?)
real trend based on
baseline data
real trend based on
pre-burn data
initial baseline
site potential (?)
"reference areas" not
standardized
initial baseline
pre-burn baseline,
control plots
nominal or
subnominal
sustainable,
marginal,
nonsustainable
healthy, at risk,
unhealthy
no fixed descriptors
no fixed descriptors
many being tested
Natural Resources Conservation Service
Bureau of Land Management
Department of Defense
United States Forest Service
Environmental Protection Agency
Society for Range Management
National Academy of Sciences
National Park Service
238
4.7
References
Ainsworth, N. 1999. Monitoring and evaluation of environmental weeds. Plant Protection Quarterly
14(3): 117-119.
Albertson, P.E., A.J. Bush III, S.L. Webster, J.P. Titre, D.M. Patrick, and J.W. Brown. 1995. Road
Management Plan and Workshop, Eglin Air Force Base, Florida. Miscellaneous Paper GL-95-13.
U.S. Army Corps of Engineers Waterways Experiment Station, Vicksburg, MS. 37 pp.
Anderson, L.D., R.G. Clark, J. Findley, R.C. Hanes, L. Mahaffey, M. Miller, K. Stinson, and G.T.
Zimmerman. 2001. Fire Effects Guide. NFES 2394.
http://www.fws.gov/fire/downloads/monitor.pdf National Interagency Fire Center, Boise, ID. 313
pp.
APHA (American Public Health Association). 1998. Standard Methods for the Examination of Water
and Wastewater. 20th Edition. American Public Health Association, Washington, D.C.
Applegate, J. 2003. Forest Health Monitoring Technical Reference Guide – Fort A.P. Hill. ITAM
Program, Fort A.P. Hill, Virginia.
APRS Implementation Team. 2000. Alien Plants Ranking System Version 5.1. Jamestown, ND:
Northern Prairie Wildlife Research Center Online.
http://www.npwrc.usgs.gov/resource/2000/aprs/aprs.htm
ASAE (American Society of Agricultural Engineers). 1999. Procedures for Using and Reporting Data
with the Soil Cone Penetrometer. ASAE Standard EP542. American Society of Agricultural
Engineers: St. Joseph, MI.
ASAE (American Society of Agricultural Engineers). 2004. Soil Cone Penetrometer. ASAE Standard
S313.3. American Society of Agricultural Engineers: St. Joseph, MI.
ASCE (American Society of Civil Engineers). 1975. Reservoir Deposits. Pages 349-382 in
Sedimentation Engineering. American Society of Civil Engineers, New York.
Avery, T.E. and H.E. Burkhart. 1995. Forest Measurements, 3rd Ed. McGraw-Hill Publishing
Company. New York.
Bailey, A.W. and C.E. Poulton. 1968. Plant communities and environmental relationships in a portion
of the Tillamook burn, northwestern Oregon. Ecology 49: 1-13.
Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid Bioassessment Protocols for
Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates and Fish. Second
Edition. EPA 841-B-99-002. U.S. Environmental Protection Agency, Washington, D.C. 339 pp.
BCMF (B.C. Ministry of Forests). 2001. Watershed Assessment Procedure Guidebook, Version 2.1.
Second Edition. British Columbia Ministry of Forests, Victoria, B.C. 40 pp.
Beers, T. and C. Miller. 1973. Manual of Forest Mensuration. T & C Enterprises, West Lafayette, IN.
239
Bleier, C., S. Downie, S. Cannata, R. Henly, R. Walker, C. Keithley, M. Scruggs, K. Custis, J.
Clements, and R. Klamt. 2003. North Coast Watershed Assessment Program Methods Manual.
California Resources Agency and California Environmental Protection Agency, Sacramento, CA. 191
pp.
Bonham, C.D. 1989. Measurements for Terrestrial Vegetation. John Wiley, New York.
Borman, M.M. 1995. Photo Monitoring. Presentation given at the Ecological Monitoring Workshop;
Assessment of Aridland Ecosystems, Warm Springs, Oregon, April 24-28, 1995. Department of
Rangeland Resources, Oregon State University.
Braunack, M.V. 1986. The residual effects of tracked vehicles on soil surface properties. Journal of
Terramechanics 23: 37-50.
Braun-Blanquet, J. 1965. Plant Sociology: the Study of Plant Communities. Translated, revised, and
edited by C.D. Fuller and H.S. Conard. Hafner, London. 439 pp.
Brayshaw, T.C. 1996. Plant Collecting for the Amateur. Royal British Columbia Museum, Victoria,
BC. 44 pp.
Brewer, L. and D. Berrier. 1984. Photographic Techniques for Monitoring Resource Change at
Backcountry Sites. General Technical Report NE-86. USDA Forest Service, Northeastern Forest
Experiment Station. 13 pp.
Brown, J.K, R.D. Oberheu, and C.M. Johnston. 1982. Handbook for Inventorying Surface Fuels and
Biomass in the Interior West. General Technical Report INT-129. USDA Forest Service,
Intermountain Forest and Range Experiment Station, Ogden, UT. 48 pp.
Brown, J.K. and J.K. Smith (eds). 2000. Wildland Fire in Ecosystems: Effects of Fire on Flora.
General Technical Report RMRS-GTR-42-Vol. 2. USDA Forest Service, Rocky Mountain Research
Station, Ogden, UT. 257 pp.
Buchanan, T.J. and W.P. Somers. 1969. Discharge Measurements at Gaging Stations. Techniques of
Water-Resources Investigations of the United States Geological Survey, Chapter A8, Book 3,
Applications of Hydraulics. USDI Geological Survey, Washington, D.C. 65 pp.
Bunte, K. and L.H. MacDonald. 1999. Scale Considerations and the Detectability of Sedimentary
Cumulative Watershed Effects. Technical Bulletin 776. National Council for Air and Stream
Improvement, Inc. (NCASI), Research Triangle Park, NC. 328 pp.
Bureau of Land Management (BLM). 1979. Physical Resource Studies: Soil-Vegetation Inventory
Method, Bureau of Land Management Manual 4412. Denver Service Center, Denver, CO.
Bureau of Land Management. 1996. Sampling Vegetation Attributes: Interagency Technical
Reference. BLM National Applied Resource Sciences Center, BLM/RS/ST-96/002+1730. Supersedes
BLM Technical Reference 4400-4, Trend Studies, dated May 1995. 163 pp.
240
Bureau of Land Management. 1997a. Standards for Rangeland Health and Guidelines for Livestock
Grazing Management for Public Lands Administered by the BLM in the States of Oregon and
Washington. Draft document dated August 12, 1997.
Bureau of Land Management. 1997b. Arizona Standards for Rangeland Health and Guidelines for
Grazing Administration. USDI Bureau of Land Management, Arizona. 18 pp.
Bureau of Land Management. 1997c. Rangeland Health: Standards and Guidelines for California and
Northwestern Nevada: Draft EIS. USDI Bureau of Land Management.
Bureau of Land Management. 1998. Standards for Rangeland Health and Guidelines for Livestock
Grazing Management for Montana, North Dakota, and South Dakota. USDI Bureau of Land
Management.
Burkman, W.G. and G.D. Hertel. 1992. Forest health monitoring: National program to detect,
evaluate, and understand change. Journal of Forestry 90: 27.
Busby, F.E. and C.A Cox. 1994. Rangeland health: New methods to classify, inventory, and monitor
rangelands. Renewable Resources Journal, Spring 1994: 13-19.
Canfield, R. 1941. Application of the line interception method in sampling range vegetation. Journal
of Forestry 39: 388-394.
Center for Ecological Management of Military Lands (CEMML). 1996. Protocol for Floristics
Inventories. Floristics Laboratory, Department of Forest Sciences, Colorado State University, Fort
Collins, Colorado.
CBP (Chesapeake Bay Program). 2003. Community Watershed Assessment Handbook. Chesapeake
Bay Program, Annapolis, MD. 96 pp.
Clinton, W. 1999. Executive Order 13112. 3 February 1999. Invasive Species. Federal Register 64:
6183-6186.
CNAP (Colorado Natural Areas Program). 2000. Creating An Integrated Weed Management Plan; A
Handbook for Owners and Managers of Lands with Natural Values. Colorado State Parks, Colorado
Department of Natural Resources, Division of Plant Industry, and Colorado Department of
Agriculture. Colorado State Parks, Denver, Colorado.
Colby, B.R. 1956. Relationship of Sediment Discharge to Streamflow. U.S. Geological Survey OpenFile Report. USDI Geological Survey, Washington, DC. 170 pp.
Collins, B., P.S. White, and D.W. Imm. 2001. Introduction to ecology and management of rare plants
of the Southeast. Natural Areas Journal 21(1): 4-11.
Committee on Rangeland Classification. 1994. Rangeland Health: New Methods for Classifying,
Inventorying, and Monitoring Rangelands. Board on Agriculture, National Research Council,
National Academy Press, Washington, D.C.
Cook, C.W. 1989. Report of LCTA Review. Unpublished report submitted to USACERL, December
1989. U.S. Army Construction Engineering Research Laboratory, Champaign, IL.
241
Cottam, G. and J. T. Curtis. 1956. The use of distance measures in phytosociological sampling.
Ecology 37: 451-460.
Council on Environmental Quality (CEQ). 1997. Considering Cumulative Effects Under the National
Environmental Policy Act. Council on Environmental Quality, Executive Office of the President,
Washington, D.C. 64 pp. plus appendices.
Crocker, R.L. and N.S. Tiver. 1948. Survey methods in grassland ecology. J. Br. Grassland Soc. 3: 126.
Curtis, J.T. and R.P. McIntosh. 1950. The interrelations of certain analytic and synthetic
phytosociological characters. Ecology 31: 434-455.
Dalrymple, T. and M.A. Benson. 1976. Measurement of Peak Discharge by the Slope-Area Method.
Techniques of Water-Resources Investigations of the United States Geological Survey, Chapter A2,
Book 3, Applications of Hydraulics. USDI Geological Survey, Washington, D.C. 12 pp.
Davis, G.E. 1989. Design of a long-term ecological monitoring program for Channel Islands National
Park, California. Natural Areas Journal 9(2): 80-89.
Daubenmire, R.F. 1959. Canopy coverage method of vegetation analysis. Northwest Scientist 33: 4364.
DeBano, L.F. 1981. Water Repellent Soils: A State-of-the-Art. General Technical Report PSW-46.
USDA Forest Service, Pacific Southwest Research Station, Berkeley, CA. 21 pp.
DeBano, L. F., D.G. Neary, and P.F. Ffolliott. 1998. Fire’s Effects on Ecosystems. John Wiley and
Sons, New York. 333 pp.
DeBano, L.F. 2000. The role of fire and soil heating on water repellency in wildland environments: a
review. Journal of Hydrology 231/232: 195-206.
DeBarry, P.A. 2004. Watersheds: Processes, Assessment, and Management. John Wiley and Sons,
Hoboken, NJ. 699 pp.
de Becker, S. and D. Mahler. 1986. Photographing quadrats to measure percent vegetation cover.
Natural Areas Journal 6(1): 67-69.
Diersing, V.E., Shaw, R.B., and D.J. Tazik. 1992. U.S. Army Land Condition-Trend Analysis
(LCTA) program. Environmental Management 16: 405-414.
Dingus, R. 1982. The Photographic Artifacts of Timothy O’Sullivan. University of New Mexico
Press, Albuquerque, N.M.
Doe, W.W., D.S. Jones, and S.D. Warren. 1999 (Coordinating Draft). Soil Erosion Modeling Guide
for Military Land Managers: Analysis of Erosion Models for Natural and Cultural Resources
Applications. Prepared for the Tri-Services Natural and Cultural Resources Field Working Group by
the Center for Ecological Management of Military Lands, Colorado State University, Fort Collins,
CO.
242
Dutton, A.A. and D.T. Bunting. 1981. Arizona Then and Now: A Comprehensive Rephotographic
Project. Ag. Press, Phoenix, AZ.
Dyksterhuis, E.J. 1949. Condition and management of rangeland based on quantitative ecology.
Journal of Range Management 2: 104-115.
Dyksterhuis, E.J. 1958. Range conservation based on sites and condition classes. Journal of Soil and
Water Conservation 13: 151-155.
Eaton, R.A. and R.E. Beaucham. 1992. Unsurfaced Road Maintenance Management. Special Report
92-26. U.S. Army Corps of Engineers Cold Regions Research and Engineering Laboratory (CRREL),
Hanover, NH.
Edwards, T.K. and G.D. Glysson. 1988. Field Methods for Measurement of Fluvial Sediment. U.S.
Geological Survey Open-File Report 86-531. USDI Geological Survey, Reston, VA. 118 pp.
Elliot, W.J. 2004. FS WEPP Interfaces. http://forest.moscowfsl.wsu.edu/fswepp/. USDA Forest
Service, Rocky Mountain Research Station, Moscow, ID. Accessed 13 December 2004.
Elzinga, C.L., D.W. Salzer, and J.W. Willoughby. 1998. Measuring and Monitoring Plant
Populations. USDI Bureau of Land Management, in partnership with Alderspring Ecological
Consulting, The Nature Conservancy of Oregon, and the Bureau of Land Management California
State Office. BLM Technical Reference 1730-1. BLM National Applied Resource Sciences Center,
Denver, CO.
Erich, H. 1997. Guide to Monitoring Exotic and Invasive Plants. Ecological Monitoring and
Assessment Network, Environment Canada. http://www.emanrese.ca/eman/ecotools/protocols/terrestrial/exotics/intro.html. Accessed 19 October 2004.
ESRI. 2004. What's New in ArcGIS 9. http://www.esri.com/software/arcgis/about/whats-new.html.
ESRI, Redlands, CA. Accessed 27 December 2004.
Farrell, G.S. and W.M. Lonsdale. 1997. Measuring the Impact of Biological Control Agents on
Weeds. In Handbook on Weed Biocontrol, M.H. Julien and G. White, (eds). ACIAR, Canberra.
Forcella, F. and S. Harvey. 1988. Patterns of weed migration in northwestern USA. Weed Science 36:
194-201.
Floyd, D.A., and J.E. Anderson. 1987. A comparison of three methods for estimating plant cover.
Journal of Ecology 75: 221-228.
Foster, M.S., C. Harrold, and D.D. Hardin. 1991. Point versus photo quadrat estimates of the cover of
sessile marine organisms. Journal of Experimental Marine Biology and Ecology 146: 193-203.
Ganey, J.L. and W.M. Block. 1994. A comparison of two techniques for measuring canopy closure.
Western Journal of Applied Forestry 9(1): 21-23.
243
Garton, E.O. 1984. Cost-Efficient Baseline Inventories of Research Natural Areas. Pages 40-45 in
J.L. Johnson, J.F. Franklin, and R.G. Krebill, coordinators, Research Natural Areas: Baseline
Monitoring and Management. Proceedings of a Symposium, March 21, 1984, Missoula, Montana.
General Technical Report INT 173. USDA Forest Service, Intermountain Forest and Range
Experiment Station, Ogden, Utah.
Gee, G.W. and D. Or. 2002. Particle-size analysis. Pages 255-293 in J.H. Dane and G.C. Clarke,
(eds.), Methods of Soil Analysis: Part 4-Physical Methods. Soil Science Society of America, Inc.,
Madison, WI.
Gibson, G.R., Jr., M.T. Barbour, J.B. Stribling, J. Gerritsen, and J.R. Karr. 1996. Biological Criteria:
Technical Guidance for Streams and Small Rivers, Revised Edition. EPA 822-B-96-001. U.S.
Environmental Protection Agency, Washington, D.C. 162 pp.
Godwin, R.J., N.L. Warner, and D.L. Smith. 1991. The development of a dynamic drop-cone device
for the assessment of soil strength and the effects of machinery traffic. Journal of Agricultural
Engineering Research 48: 123-131.
Goodall, D.W. 1952. Some considerations in the use of point quadrats for the analysis of vegetation.
Australian Journal. Sci. Res., Series B 5: 1-41.
Grantham, W.P. 2000. Tracked Vehicle Impacts to Vegetation Structure and Wind Erodibility of
Soils. MS Thesis, Colorado State University, Fort Collins, CO.
Greig-Smith, P. 1983. Quantitative Plant Ecology. Third Edition. Blackwell Scientific, London. 359
pp.
Griffith, B. and B.A. Youtie. 1988. Two devices for estimating foliage density and deer hiding cover.
Wildlife Society Bulletin 16: 206-210.
Grossman, R.B. and T.G. Reinsch. 2002. Bulk density and linear extensibility. Pages 201-228 in J.H.
Dane and G.C. Clarke, (eds.), Methods of Soil Analysis: Part 4-Physical Methods. Soil Science
Society of America, Inc., Madison, WI.
Gruell, G.E. 1980. Fire’s Influence on Wildlife on the Bridger-Teton National Forest, Wyoming. Vol.
I – Photographic Record and Analysis. USDA Forest Service Research Paper INT-235.
Gruell, G.E. 1983. Fire and Vegetation Trends in the Northern Rockies: Interpretations from 18711982 Photographs. General Technical Report INT-206. USDA Forest Service.
Gucinski, H., M.J. Furniss, R.R. Ziemer, and M.H. Brookes. 2001. Forest roads: A Synthesis of
Scientific Information. General Technical Report PNWGTR-509. U.S. Department of Agriculture,
Forest Service, Pacific Northwest Research Station, Portland, OR. 103 pp.
Guthery, F.S., T.B. Doerr, and M.A. Taylor. 1981. Use of a profile board in sand shinnery oak
communities. Journal of Range Management 34:157-158.
244
Haber, E. 1997. Guide to Monitoring Exotic and Invasive Plants. Environment Canada, Ecological
Monitoring and Assessment Network.
http://www.eman-rese.ca/eman/ecotools/protocols/terrestrial/exotics/intro.html
Hagen, L.J. 1991. A wind erosion prediction system to meet user needs. Journal of Soil and Water
Conservation 46(2):106-111.
Hagen, L.J. et al. 1995. Wind Erosion Prediction System: Technical Description. In Proceedings,
WEPP/WEPS Symposium August 9-11, 1995. Soil and Water Conservation Society, Ankeny, IA.
Hagen, L.J. 1997. Wind Erosion Prediction System Application to DOD Lands. In Evaluation of
Technologies for Addressing Factors Related to Soil Erosion on DoD Lands. USACERL Technical
Report 97/134, U.S. Army Corps of Engineers Construction Engineering Research Laboratories,
September 1997, 100 pp.
Hall, F.C. 2001. Photo Point Monitoring Handbook: Part A-Field Procedures; Part B-Concepts and
Analysis. General Technical Report PNW-GTR-526. USDA Forest Service, Pacific Northwest
Research Station, Portland, OR. 134 pp.
Hallock, D. and W. Ehinger. 2003. Quality Assurance Monitoring Plan: Stream Ambient Water
Quality Monitoring. Publication No. 03-03-200. Washington Department of Ecology, Olympia, WA.
28 pp.
Halvorson, C.H. 1984. Long-Term Monitoring of Small Vertebrates: a Review With Suggestions.
Pages 11-25 in J.L. Johnson, J.F. Franklin, and R.G. Krebill, coordinators, Research Natural Areas:
Baseline Monitoring and Management. Proceedings of a Symposium, March 21, 1984, Missoula,
Montana. General Technical Report INT 173. Forest Service Intermountain Forest and Range
Experiment Station, Ogden, Utah. 84 pp.
Halvorson, J.J., D.K. McCool, L.G. King, and L.W. Gatto. 2001. Soil compaction and over-winter
changes to tracked-vehicle ruts, Yakima Training Center, Washington. Journal of Terramechanics 38:
133-151.
Halvorson, J.J., L.W. Gatto, and D.K. McCool. 2003. Overwinter changes to near-surface bulk
density, penetration resistance and infiltration rates in compacted soil. Journal of Terramechanics 40:
1-24.
Hanley, T.A. 1978. A comparison of the line interception and quadrat estimation methods of
determining shrub canopy coverage. Journal of Range Management 31(1): 60-62.
Heady, H.F., R.P. Gibbens, and R.W. Powell. 1959. A comparison of the charting, line intercept, and
line-point methods of sampling shrub types of vegetation. Journal of Range Management 12: 180188.
Herrick, J.E. and T.L. Jones. 2002. A dynamic cone penetrometer for measuring soil penetration
resistance. Soil Science Society of America Journal 66: 1320-1324.
Herschy, R.W. 1995. Streamflow Measurement. Second Edition. E & FN Spon, London. 524 pp.
245
Hiebert, R.D. and J. Stubbendieck. 1993. Handbook for Ranking Exotics Plants for Management and
Control. Natural Resources Report NRMWRO/NRR-93/08. USDI National Park Service, Denver,
CO.
Hinds, W.T. 1984. Towards monitoring of long-term trends in terrestrial ecosystems. Environmental
Conservation 11(1): 11-18.
Hironaka, M. 1985. Frequency approaches to monitor rangeland vegetation. Pages 84-86 in
Symposium on Use of Frequency and for Rangeland Monitoring, W.C. Kreuger, Chairman.
Proceedings, 38th Annual Meeting of the Society for Range Management, Feb. 1985. Salt Lake City,
UT. Society for Range Management.
Hobbs, R. J. and S. E. Humphries. 1995. An Integrated Approach to the Ecology and Management of
Plant Invasions. Conservation Biology 9:761-770.
Holecheck, J.L., R.D. Pieper, and C.H. Herbel. 1989. Range Management – Principles and Practices.
Prentice Hall, Englewood Cliffs, New Jersey.
Howell, F.G. 2001. Aquatic biomonitoring: An integrated part of Mississippi Military Department’s
environmental program. The Bridge (ITAM Newletter) 14: 16-18.
Huffman, E.L. L.H. MacDonald, and J.D. Stednick. 2001. Strength and persistence of fire-induced
soil hydrophobicity under ponderosa and lodgepole pine, Colorado Front Range. Hydrological
Processes 15: 2877-2892.
Humphrey, R.R. 1987. Ninety Years and 535 Miles: Vegetation Changes Along the Mexican Border.
University of New Mexico Press, Albuquerque, N.M.
IDL (Idaho Department of Lands). 2000. Forest Practices: Cumulative Watershed Effects Process for
Idaho. Idaho Department of Lands, Boise, ID.
Isaacson, J.B., A.E. Hurst, D.L. Miller, and P.E. Albertson. 2001. Unsurfaced road investigation and
management plan, Fort Leonard Wood, Missouri. Pages 177-190 in J. Ehlen and R.S. Harmon, (eds.),
The Environmental Legacy of Military Operations. Geological Society of America Reviews in
Engineering Geology, Volume XIV, Boulder, CO.
Jensen, M.E., W. Hann, R.E. Keane, J. Caratti, P.S. Bourgeron. 1994. ECODATA – A Multiresource
Database and Analysis System for Ecosystem Description and Analysis. In M.E. Jensen and P.S.
Bourgeron, (eds.), Eastside Forest Ecosystem Health Assessment, Volume II: Ecosystem
Management: Principles and Applications. USDA Forest Service General Technical Report GTRPNW-318. USDA Forest Service, Portland, OR.
Johnson, K.L. 1987. Rangeland Through Time. University of Wyoming Agric. Exp. Sta. Misc. Publ.
50.
Johnston, A. 1956. Comparison of the line interception, vertical point quadrat, and loop methods as
used in measuring basal area of grassland vegetation. Canadian Journal of Plant Science 37: 34-42.
Jones, D.S., D. Kowalski, and R.B. Shaw. 1996. Calculating Revised Universal Soil Loss Equation
(RUSLE) Estimates on Department of Defense Lands: A Review of RUSLE Factors and U.S. Army
246
LCTA Data Gaps. CEMML TPS 96-8. Center for Ecological Management of Military Lands,
Colorado State University, Fort Collins, CO. 9 pp.
Jones, D.S. 2000. Impacts of the M1A1 Abrams Tank on Vegetation and Soil Characteristics of a
Grassland Ecosystem at Fort Lewis, Washington. CEMML TPS 00-1. Center for Environmental
Management of Military Lands, Colorado State University, Fort Collins, CO.
Jones, D.S. and M.D. Kunze. 2004a. Guide to Sampling Soil Compaction using Hand-Held Soil
Penetrometers. CEMML TPS 04-1. Center for Environmental Management of Military Lands,
Colorado State University, Fort Collins, CO. 8 pp.
Jones, D.S. and M.D. Kunze. 2004b. Evaluation of Seeding Treatments at Yakima Training Center,
Washington: 2002-2004 Field Data. CEMML TPS 04-17. Center for Environmental Management of
Military Lands, Colorado State University, Fort Collins, CO. 31 pp.
Jones, D.S. and L. Robison. 2004. Range and Training Land Assessment (RTLA) Protocols for Fort
Leonard Wood, Missouri. August 2004. Unpublished CEMML report. Center for Environmental
Management of Military Lands, Colorado State University, Fort Collins, CO.
Jones, R.E. 1968. A board to measure cover used by prairie grouse. Journal of Wildlife Management
32: 28-31.
Jones, K.B. 1986. The Inventory and Monitoring Process. Pages 1-9 in A.Y. Cooperrider, R.J. Boyd,
and H.R. Stewart, (eds)., Inventory and Monitoring of Wildlife Habitat. USDI Bureau of Land
Management Service Center, Denver, CO. 858 pp.
Joyce, L.A. 1993. The life cycle of the range condition concept. Journal of Range Management 46:
132-138.
Karr, J.R. 1981. Assessment of biotic integrity using fish communities. Fisheries 66: 21-27.
Klemm, D.J., J.M. Lazorchak, and P.A. Lewis. 1998. Benthic Macroinvertebrates. Pages 147-160 in
Lazorchak, J.M., D.J. Klemm, and D.V. Peck (eds)., Environmental Monitoring and Assessment
Program - Surface Waters: Field Operations and Methods for Measuring the Ecological Condition of
Wadeable Streams. EPA/620/R-94/004F. U.S. Environmental Protection Agency, Washington, D.C.
Krajina, V.J. 1933. Die Pflanzengesellschaften des Mlyncia-Tales in den Vysoke Tatry (Hohe Tatra).
Mit besonderer Beruckscichtigung der Okologischen Verhaltnisse. Botan. Centralbl., Beih., Abt. II
50: 774-957; 51: 1-224.
Krebs, C.J. 1989. Ecological Methodology. Harper Collins Publishers, New York.
Kunze, M.D. and D.S. Jones. 2004a. Soil Penetration Resistance and Bulk Density Sampling at
Grafenwoehr Training Area, Germany: Results from April 2004 Pilot Project. Unpublished CEMML
report. Center for Environmental Management of Military Lands, Colorado State University, Fort
Collins, CO. 16 pp.
247
Kunze, M.D. and D.S. Jones. 2004b. Unpaved Roads Condition Assessment Protocol. CEMML TPS
04-16. Center for Environmental Management of Military Lands, Colorado State University, Fort
Collins, CO. 41 pp.
Lal, R. 1994 (ed). Soil Erosion Research Methods, 2nd Edition. Soil and Water Conservation Society,
St. Lucie Press, Delray Beach, FL, 340 pp.
Lancaster, J. 2000. Guidelines for Rare Plant Surveys. Alberta Native Plant Council, Edmonton,
Alberta.
Lattanzi, A.R., L.D. Meyer, and M.F. Baumgardner. 1974. Influences of Mulch Rate and Slope
Steepness on Interrill Erosion. Soil Science Society of America Proceedings 38: 946-950.
Lauenroth, W.K. 1985. New Directions for Rangeland Condition Analysis. Pages 101-106 in Selected
Papers Presented at the 38th Annual Meeting of the Society for Range Management, Feb. 1985. Salt
Lake City, UT. Society for Range Management, Denver, CO.
Laycock, W.A. 1965. Adaptation of distance measurements for range sampling. Journal of Range
Management 18: 205-211.
Leonard, G.H. and R.P. Clark. 1993. Point quadrat versus video transects estimates of the cover of
benthic red algae. Marine Ecology Press Series 101: 203-208.
Lemmon, P.E. 1956. A spherical densiometer for estimating forest overstory density. Forest Science
2: 314-320.
Lesica, P. 1987. A technique for monitoring nonrhizomatous, perennial plant species in permanent
belt transects. Natural Areas Journal 7(2): 65-68.
Letey, J., M.L.K. Carrillo, and X.P. Pang. 2000. Approaches to characterize the degree of water
repellency. Journal of Hydrology 231-232: 61-65.
Levy-Sachs, R. and M. Robinson. 2004. Using Digital Photographs in the Courtroom—
Considerations for Admissibility.
http://www.securitymanagement.com/library/feature_August2004.pdf. Accessed 14 December 2004.
Lewis, J. 1996. Turbidity-controlled suspended sediment sampling for runoff-event load estimation.
Water Resources Research 32: 2299-2310.
Lowery, B. and J.E. Morrison, Jr. 2002. Soil penetrometers and penetrability. Pages 363-388 in J.H.
Dane and G.C. Clarke, (eds)., Methods of Soil Analysis: Part 4-Physical Methods. Soil Science
Society of America, Inc., Madison, WI.
Lund, H.G. 1983. Change: - Now You See it – Now You Don’t! Pages 211-213 in J.F. Bell and T.
Atterbury, (eds)., Renewable Resource Inventories for Monitoring Changes and Trends. Proceedings
of an International Conference, August 15-19, 1983, Corvallis, Oregon. College of Forestry, Oregon
State University. 737 pp.
248
MacDonald, L.H., A.W. Smart, and R.C. Wissmar. 1991. Monitoring Guidelines to Evaluate Effects
of Forestry Activities on Streams in the Pacific Northwest and Alaska. EPA 910/9-91-001.
Environmental Protection Agency, Region 10, Seattle, WA. 166 pp.
MacDonald, L.H., R. Sampson, D. Brady, L. Juarros, and D. Martin. 2000. Predicting post-fire
erosion on a landscape scale: A case study from Colorado. Journal of Sustainable Forestry 11: 57-87.
MacDonald, L.H. 2000. Evaluating and managing cumulative effects: Process and constraints.
Environmental Management 26: 299-315.
Magill, A.W. 1989. Monitoring environmental change with color slides. General Technical Report
PSW-117. USDA Forest Service, Pacific Southwest Forest and Range Experiment Station, Berkeley,
CA. 55 pp.
Marston. R.A. 1986. Maneuver-Caused Wind Erosion Impacts, South Central New Mexico. In W.G.
Nickling (ed.), 1986, Proceedings of the 17th Annual Binghamton Geomorphology Symposium.
McCammon, B., J. Rector, and K. Gebhardt. 1998. A Framework for Analyzing the Hydrologic
Condition of Watersheds. BLM/RS/ST-98/004+7210. USDI Bureau of Land Management, Denver,
CO. 37 pp.
McGinnies, W.J., H.L. Shantz, and W.G. McGinnies. 1991. Changes in Vegetation and Land Use in
Eastern Colorado: a Photographic Study, 1904 to 1986. USDA, Agricultural Research Service, ARS85.
Menges, E.S. and D.R. Gordon. 1996. Three levels of monitoring intensity for rare plant species.
Natural Areas Journal 16(3) 227-237.
Meyer, L.D. 1994. Rainfall simulations for soil erosion research. Pages 83-103 in R. Lal, (ed)., Soil
Erosion Research Methods. Second Edition. Soil and Water Conservation Society, Ankeny, IA.
Miller, J.D., J.W. Nyhan, and S.R. Yool. 2003. Modeling potential erosion due to the Cerro Grande
Fire with a GIS-based implementation of the Revised Universal Soil Loss Equation. International
Journal of Wildland Fire 12: 85-100.
Miller, R.E., J. Hazard, and S. Howes. 2001. Precision, accuracy, and efficiency of four tools for
measuring soil bulk density or strength. General Technical Report PNW-RP-532. USDA Forest
Service Pacific Northwest Research Station, Portland, OR. 16 pp.
Mitchell, J.E., W.W. Brady, and C.D. Bonham. 1994. Robustness of the Point-Line Method for
Monitoring Basal Cover. Research Note RM-528. USDA Forest Service Rocky Mountain Forest and
Range Experiment Station.
Mitchell, Wilma A., and Hughes, H. Glenn. 1995. Visual Obstruction: Section 6.2.6, U.S. Army
Corps of Engineers Wildlife Resources Management Manual. Technical Report EL-95-23, U.S. Army
Engineer Waterways Experiment Station, Vicksburg, MS.
Moir, W.H. 1989. History of Development of Site and Condition Criteria for Range Condition Within
the U.S. Forest Service. Pages 49-76 in W.K. Lauenroth and W.A. Laycock (eds.), Secondary
Succession and the Evaluation of Rangeland Condition. Westview Press, Boulder, CO.
249
Montgomery, D.R., G.E. Grant, and K. Sullivan. 1995. Watershed analysis as a framework for
implementing ecosystem management. Water Resources Bulletin 31: 369-386.
Morse, L.E., J.M. Randall, N. Benton, R. Hiebert, and S. Lu. 2004. An Invasive Species Assessment
Protocol: Evaluating Non-Native Plants for Their Impact on Biodiversity. Version 1. NatureServe,
Arlington, Virginia. http://www.natureserve.org/library/invasiveSpeciesAssessmentProtocol.pdf
Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and Methods of Vegetation Ecology. John
Wiley, New York. 547 pp.
Mutchler, C.K., C.E. Murphree, and K.C. McGregor. 1994. Laboratory and field plots for erosion
research. Pages 11-37 in R. Lal, (ed)., Soil Erosion Research Methods. Second Edition. Soil and
Water Conservation Society, Ankeny, IA.
NAMWA (North American Weed Management Association). 2002. North American Invasive Plant
Mapping Standards. May 7, 2002. NAWMA Mapping Standards Committee.
http://www.nawma.org/documents/Mapping%20Standards/Mapping%20Standards%20Index.html
National Park Service, Invasive Species Monitoring Resources. 2002. NPS Invasive Plant Inventory
and Monitoring Guidelines.
http://science.nature.nps.gov/im/monitor/invasives/index.cfm.
Nelson, J.R. 1984. Rare Plant Field Survey Guidelines. In: J.P. Smith and R. York, Ed. Inventory of
Rare and Endangered Vascular Plants of California. 3rd. Ed California Native Plant Society,
Berkeley, CA. 174 pp.
Nelson, J.R. 1987. Rare Plant Surveys: Techniques for Impact Assessment. In: T.S. Elias (ed.),
Conservation and Management of Rare and Endangered Plants, pp. 159-166. Sacramento, California.
Novotny, V. and H. Olem. 1994. Water Quality: Prevention, Identification, and Management of
Diffuse Pollution. John Wiley and Sons, Inc., New York. 1054 pp.
Nudds, T.D. 1977. Quantifying the vegetation structure of wildlife cover. Wildlife Society Bulletin 5:
113-117.
Owen, W. and R. Rosentreter. 1992. Monitoring rare perennial plants: techniques for demographic
studies. Natural Areas Journal 12(1): 32-38.
Owensby, C.E. 1973. Technical notes: Modified step-point system for botanical composition and
basal cover estimates. Journal of Range Management 26(4): 302-303.
Palmer, M.E. 1987. A critical look at rare plant monitoring in the United States. Biological
Conservation 39: 113-127.
Pellant, M., D. Shaver, D.A. Pyke, and J.E. Herrick. 2000. Interpreting Indicators of Rangeland
Health, Version 3. TR-1734-6. USDI Bureau of Land Management, Denver, CO.
Pellant, M., D. Shaver, D.A. Pyke, and J.E. Herrick. 2005. Interpreting Indicators of Rangeland
Health, Version 4. TR-1734-6. USDI Bureau of Land Management, Denver, CO.
250
Pendleton, D.T. 1989. Range condition as used in the Soil Conservation Service, Pages 17-34 in
W.K. Lauenroth and W.A. Laycock, (eds)., Secondary Succession and the Evaluation of Rangeland
Condition. Westview Press, Boulder, CO.
Peterson, D.L. M.C. Johnson, J.K. Agee, T.B. Jain, D. McKenzie, and E.D. Reinhardt. 2004. Fuel
Planning: Science Synthesis and Integration—Forest Structure and Fire Hazard. General Technical
Report PNW-GTR-xxx (in press). USDA Forest Service, Pacific Northwest Research Station,
Portland, OR.
Pieper, R.D. 1988. Rangeland Vegetation Productivity and Biomass. Pages 449-467 in P.T. Tueller,
(ed)., Vegetation Science Applications for Rangeland Analysis and Management. Handbook of
Vegetation Science, Volume 14. Kluwer Academic Publishers, Dordrecht.
Porterfield, G. 1972. Computation of Fluvial-Sediment Discharge. Techniques of Water-Resources
Investigations of the United States Geological Survey, Chapter C3, Book 3, Applications of
Hydraulics. USDI Geological Survey, Washington, D.C. 66 pp.
Potyondy, J.P. 1980. Technical Guide for Preparing Water Quality Monitoring Plans. USDA Forest
Service, Intermountain Region, Ogden, UT. 74 pp.
Proudfoot, M.J. 1942. Sampling with transverse traverse lines. American Statistical Association
Journal 37: 265-270.
Pyke, D.A., J.E. Herrick, P. Shaver, and M. Pellant. 2002. Rangeland health attributes and indicators
for qualitative assessment. J. Range Management. 55(6):584-597.
Quinn, J.F. and C. van Riper III. 1990. Design Considerations for National Park Inventory Databases.
Pages 5-13 in C. van Riper III, T.J. Stohlgren, S.D. Veirs, Jr., and S.C. Hillyer, (eds)., Examples of
Resource Inventory and Monitoring in National Parks of California. Trans. and Proc. Ser. No.8. US
Department of the Interior, Washington, D.C.
Range Inventory Standards Committee. 1983. Guidelines and Terminology for Range Inventories and
Monitoring. Society for Range Management. Denver, CO.
Reich, R. M., C. D. Bonham, and K. K. Remington. 1993. Technical notes: double sampling
revisited. Journal of Range Management 46: 88-90.
Reid, L.M. and T. Dunne. 1984. Sediment production from road surfaces. Water Resources Research
20(11): 1753-1761.
Reid, L.M. 1993. Research and Cumulative Watershed Effects. General Technical Report PSW-GTR141. USDA Forest Service, Pacific Southwest Research Station, Albany, CA. 118 pp.
Reid, L.M. 1994. Watershed analysis…whatever that is. Watershed Management Council Newsletter
6(2): 1,16.
Reid, L.M., R.R. Ziemer, and T.E. Lisle. 1996. What a long strange trip it’s been – or – who took the
synthesis out of analysis? Watershed Management Council Newsletter 6(4): 6-7.
251
Reid, L.M. 1998. Cumulative watershed effects and watershed analysis. Pages 476-501 in R.J.
Naiman and R.E. Bilby (eds)., River Ecology and Management: Lessons from the Pacific Coastal
Ecoregion. Springer-Verlag, New York.
Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., and D.C. Yoder, coordinators. 1997.
Predicting Soil Erosion by Water - A Guide to Conservation Planning With the Revised Universal
Soil Loss Equation (RUSLE). USDA-ARS Agriculture Handbook No. 703. 404 pp.
Renfro, G.W. 1975. Use of erosion equations and sediment-delivery ratios for predicting sediment
yield. Pages 33-45 in Present and Prospective Technology for Predicting Sediment Yields and
Sources. Proceedings of the Sediment-Yield Workshop, USDA Sedimentation Laboratory, Oxford,
Mississippi, November 28-30, 1972. ARS-S-40. USDA Sedimentation Laboratory Oxford, MS.
REO (Regional Ecosystem Office). 1995. Ecosystem Analysis at the Watershed Scale: Federal Guide
for Watershed Analysis. Version 2.2. Regional Ecosystem Office, Portland, OR. 26 pp.
REO (Regional Ecosystem Office). 1996. Ecosystem Analysis at the Watershed Scale: Federal Guide
for Watershed Analysis: Section II, Analysis and Methods Techniques. Version 2.3. Regional
Ecosystem Office, Portland, OR. 87 pp.
Riggins, R.E. and L.J.Schmitt. 1984. Development of Prediction Techniques for Soil Loss and
Sediment Transport at Army Training Areas. USACERL Interim Report N-181. U.S. Army
Construction Engineering Research Laboratories, Champaign, IL. 36 pp.
Robel, R.J., J.N. Briggs, A.D. Dayton, and L.C. Hulbert. 1970. Relationships between visual
obstruction measures and weight of grassland vegetation. Journal of Range Management 23: 295-297.
Roberts, E., D. Cooksey and R. Sheley. 1999. Montana Noxious Weed Survey and Mapping System
Weed Mapping Handbook. Bozeman, MT. Montana State University.
http://www.montana.edu/places/mtweeds/pubs.html
Robichaud, P.R., J.L. Beyers, and D.G. Neary. 2000. Evaluating the Effectiveness of Postfire
Rehabilitation Treatments. General Technical Report RMRS-GTR-63. USDA Forest Service, Rocky
Mountain Research Station, Fort Collins, CO. 85 pp.
Robichaud, P.R. and R.E. Brown. 2002. Silt Fences: An Economical Technique for Measuring
Hillslope Soil Erosion. General Technical Report RMRS-GTR-94. USDA Forest Service Rocky
Mountain Research Station, Fort Collins, CO. 24 pp.
Robichaud, P.R., L.H. MacDonald, J. Freeouf, D. Neary, D. Martin, and L. Ashmun. 2003. Postfire
Rehabilitation of the Hayman Fire. Pages 293-313 in R.T. Graham (ed)., Hayman Fire Case Study.
General Technical Report RMRS-GTR-114. USDA Forest Service, Rocky Mountain Research
Station, Fort Collins, CO.
Robson, D.B. 1998. Guidelines for Rare Plant Surveys. Native Plant Society of Saskatchewan.
http://www.npss.sk.ca/inforesource/rareplant.html.
Rosenzweig, M.L. and J. Winakur. 1969. Population ecology of desert rodent communities: habitats
and environmental complexity. Ecology 50: 558-572.
252
Ryan, K.C. and N.V. Noste. 1983. Evaluating Prescribe Fires. Pages 230-238 in J.E. Lotan, B.M.
Kilgore, W.C. Fischer, and R.W. Mutch, (technical coordinators), Symposium and Workshop of
Wilderness Fire. General Technical Report INT-182. USDA Forest Service, Intermountain Research
Station, Ogden, UT.
Sandberg, D.V., R.D. Ottmar, J.L. Peterson, and J. Core. 2002. Wildland Fire in Ecosystems: Effects
of Fire on Air. General Technical Report RMRS-GTR-42-Vol. 5. USDA Forest Service, Rocky
Mountain Research Station, Ogden, UT. 79 pp.
Schubauer-Berigan, J.P., R.J.F. Bruins, V.B. Serveiss, J. Little, and D.L. Eskew. 2000. Gathering
Information for Watershed Ecological Assessments: A Review of Ten Watershed Assessments.
NCEA-C-0847. U.S. Environmental Protection Agency, Cincinnati, OH. 66 pp. plus appendices.
Schwegman, J. 1986. Two types of plots for monitoring herbaceous plants over time. Natural Areas
Journal 6: 64-66.
Scott DF. 2000. Soil wettability in forested catchments in South Africa; as measured by different
methods and as affected by vegetation cover and soil characteristics. Journal of Hydrology 231-232:
87–104.
Shimwell, D. W. 1971. The Description and Classification of Vegetation. University of Washington
Press, Seattle, WA. 322 pp.
Silsbee, D.G. and D.L. Peterson. 1991. Designing and Implementing Comprehensive Long-Term
Inventory and Monitoring Programs for National Park System Lands. Natural Resources Report
NPS/NRUW/NRR-91/04. USDI National Park Service. Denver, CO.
Skidmore, E.L. 1994. Wind Erosion. In R. Lal, ed., Soil Erosion Research Methods. Second Edition.
Soil and Water Conservation Society, Ankeny, IA.
Smith, J.K. (ed). 2000. Wildland Fire in Ecosystems: Effects of Fire on Fauna. General Technical
Report RMRS-GTR-42-Vol. 1. USDA Forest Service, Rocky Mountain Research Station, Ogden,
UT. 83 pp.
Smith, R.L. 1980. Ecology and Field Biology. Harper & Row, Publishers, New York. 835 pp.
Society for Range Management (SRM). 1990. Report of Unity in Concepts and Terminology
Committee. Society for Range Management, Denver, CO.
Society for Range Management (SRM) Glossary Update Task Group. 1998. Glossary of Terms Used
in Range Management, 4th Edition. Society for Range Management, Denver, Colorado.
SRM Task Group on Unity in Concepts and Terminology. 1995. New concepts for assessment of
rangeland condition. Journal of Range Management 48:271-282.
Stednick, J.D. 1991. Wildland Water Quality Sampling and Analysis. Academic Press, San Diego,
CA. 217 pp.
253
Stednick, J.D. and D.M. Gilbert. 1998. Water Quality Inventory Protocol: Riverine Environments.
Technical Report NPS/NRWRD/NRTR-98/177. National Park Service, Water Resources Division
and Servicewide Inventory and Monitoring Program, Fort Collins, CO. 104 pp.
Stein, B. and S. Flack. 1996. America’s Least Wanted: Alien Species Invasions of U.S. Ecosystems.
The Nature Conservancy, Arlington, WA. 31 pp.
Steinman, J. 2000. Tracking the Health of Individual Trees Over Time on Forest Health Monitoring
Plots, in M. Hansen and T. Burke, (eds)., Integrated Tools for Natural Resources Inventories in the
21st Century: An International Conference on the Inventory and Monitoring of Forested Ecosystems
held in Boise, ID, August 16-19, 1998. General Technical Report NRCS-212.
Stevenson, R.J and L.L. Bahls. 1999. Periphyton Protocols. Pages 6-1 to 6-23 in Rapid Bioassessment
Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates and Fish.
Second Edition. EPA 841-B-99-002. U.S. Environmental Protection Agency, Washington, D.C.
Stohlgren, T.J., D. Barnett, and S. Simonson. 2002. Beyond NAMWA Standards. Unpublished
manuscript available at http://www.nawma.org/
Sutherland, S. 2004. First Order Fire Effects Model (FOFEM). Research Note RMRS-RN-23-2WWW. USDA Forest Service, Rocky Mountain Research Station, Missoula, MT. 2 pp.
Sutter, R. 1997. Monitoring Changes in Exotic Vegetation. Presentation at the Conference on Exotic
Pests of Eastern Forests, April 8-10, 1997, Nashville, TN. Edited by Kerry O. Britton, USDA Forest
Service & TN Exotic Pest Plant Council
Tanner, G.W., J.M. Inglis, and L.H. Blankenship. 1978. Acute impact of herbicide strip treatment on
mixed-brush white-tailed deer habitat on the northern Rio Grande Plain. Journal of Range
Management 31: 386-391.
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (LCTA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
Thomas, L., M. DeBacker, and J.R. Boetsch. 2002. Considerations for Developing Invasive Exotic
Plant Monitoring. Prairie Cluster Prototype Monitoring Program, June 2002.
http://science.nature.nps.gov/im/monitor/Meetings/FtCollins_02/PRCL_InvasivePlantMonitoring.pdf
Tiedemann, A.R., C.E. Conrad, J.H. Dieterich, J.W. Hornbeck, W.F. Megahan, L.A. Viereck, and
D.D. Wade. 1979. Effects of Fire on Water: A State-of-the-Knowledge Review. General Technical
Report WO-10. USDA Forest Service. 28 pp.
Todd, J.E. 1982. Recording Changes: A Field Guide to Establishing and Maintaining Permanent
Camera Points. R6-10-095-1982. USDA Forest Service, Pacific Northwest Region, Portland, OR.
Topp, G.C., and P.A. Ferre. 2002. Methods for measurement of soil water content:
Thermogravimetric using convective oven-dying. Pages 422-424 in J.H. Dane and G.C. Clarke,
(eds)., Methods of Soil Analysis: Part 4-Physical Methods. Soil Science Society of America, Inc.,
Madison, WI.
254
Travis, J. and R. Sutter. 1986. Experimental designs and statistical methods for demographic studies
of rare plants. Natural Areas Journal 6(3): 3-12.
University of California Cooperative Extension. 1994. “How To” Monitor Rangeland Resources.
Division of Agriculture and Natural Resources, Intermountain Workgroup Publication 2.
US Army. 1995. Unsurfaced Road Maintenance Management. Technical Manual 5-626. Department
of the Army Headquarters, Washington, DC.
USACE (U.S. Army Corps of Engineers). 2003a. Guidance for Non-Native Invasive Plant Species on
Army Lands: Western United States. Public Works Technical Bulletin 200-1-18. Washington, DC
USACE (U.S. Army Corps of Engineers). 2003b. Guidance for Non-Native Invasive Plant Species on
Army Lands: Eastern United States. Public Works Technical Bulletin 200-1-19. Washington, DC
USACERL (U.S. Army Corps of Engineers Research Laboratory). 1997. Evaluation of Technologies
for Addressing Factors Related to Soil Erosion on DoD Lands. USACERL Technical Report 97/134,
U.S. Army Corps of Engineers Construction Engineering Research Laboratories, September 1997,
100 pp.
USAEC (U.S. Army Environmental Center). 1996. Land Condition Trend Analysis II: Report for
Workshop held January 23-25, 1996 in Linthicum, Maryland. Prepared by the Center for
Environmental Management of Militayr Lands (CEMML), OClorado State University, Fort Collins,
CO for the U.S. Army Environmental Center
USDA Forest Service. 1995. Burned Area Emergency Rehabilitation Handbook. Forest Service
Handbook No. 2509.13-95-6. USDA Forest Service, Washington, D.C. 91 pp.
USDA Forest Service. 1996a. Forest Health Monitoring Information Booklet. USDA Forest Service,
National Forest Health Monitoring Program, Research Triangle Park, NC.
USDA Forest Service. 1996b. Rangeland Analysis and Management Training Guide. USDA Forest
Service Rocky Mountain Region, Denver, CO.
USDA Forest Service. 1997a. Forest Health Monitoring 1997 Field Methods Guide. USDA Forest
Service, National Forest Health Monitoring Program, Research Triangle Park, NC.
USDA Forest Service. 1997b. Forest Health Monitoring Program Implementation Plan for Fifty
States. April 1997. USDA Forest Service, Wash. D.C.
USDA Forest Service. 1999. Roads Analysis: Informing Decisions about Managing the National
Forest Transportation System. Miscellaneous Report FS-643. USDA Forest Service, Washington,
D.C. 222 pp.
USDA Forest Service. 2000. East-Wide Watershed Assessment Protocol for Forest Plan Amendment,
Revision, and Implementation. USDA Forest Service, Southern Region, Atlanta, GA. 17 pp.
USDA Forest Service. 2002a. Field Guide – Invasive Plant Inventory, Monitoring and Mapping
Protocol. USDA Forest Service.
255
USDA Forest Service. 2002b. Forest Health Monitoring Phase 3 Manual. USDA Forest Service,
National Forest Health Monitoring Program, Research Triangle Park, NC.
USDA Forest Service. 2003a. Roads Analysis Report, Forest-Wide Assessment: Ochoco National
Forest, Deschutes National Forest, and Crooked River National Grassland.
http://www.fs.fed.us/r6/centraloregon/projects/planning/roadsanalysis/index.shtml. Accessed 29
November 2004.
USDA Forest Service. 2003b. Roads Analysis Process Report: Pisgah and Nantahala National
Forests. Pisgah and Nantahala National Forests, Ashville, NC. 88 pp. plus appendices.
USDA NRCS (Natural Resources Conservation Service). 1997a. National Handbook of Water
Quality Monitoring. 450-VI-NHWQM. Part 600, National Water Quality Handbook. USDA Natural
Resources Conservation Service, National Water and Climate Center, Portland, OR.
USDA NRCS (Natural Resources Conservation Service). 1997b. National Range and Pasture
Handbook. USDA Natural Resources Conservation Service, Washington D.C.
USDA NRCS (Natural Resources Conservation Service). 2002. Analysis of Water Quality
Monitoring Data (Draft). 450-VI-NWQM. Part 615, National Water Quality Handbook. USDA
Natural Resources Conservation Service, National Water and Climate Center, Portland, OR.
USDI Fish and Wildlife Service. 2004. Fuel and Fire Effects Monitoring Guide.
http://www.fws.gov/fire/downloads/monitor.pdf. Accessed 12 January 2005.
USDI Geological Survey. 1982. Chapter 5 – Chemical and Physical Quality of Water and Sediment.
Pages 5-1 – 5-194 in National Handbook of Recommended Methods for Water-Data Acquisition.
USDI Geological Survey, Reston, VA.
USDI Geological Survey. 1998. National Field Manual for the Collection of Water Quality Data,
Section A. Techniques of Water-Resources Investigations, Book 9, Handbooks for Water-Resources
Investigations. USDI Geological Survey, Washington, D.C.
USDI Geological Survey. 2004. FIREMON (Fire Effects Monitoring and Inventory Protocol).
http://fire.org/index.php?option=com_content&task=category&sectionid=5&id=18&Itemid=42.
Accessed 12 January 2006.
USDI NPS (National Park Service). 2003. Fire Monitoring Handbook. Fire Management Program
Center, National Interagency Fire Center, Boise, ID. 274 pp.
USEPA (U.S. Environmental Protection Agency). 1986. Quality Criteria for Water (Gold Book).
EPA 440/5-86-001. U.S. Environmental Protection Agency, Washington, D.C.
USEPA (U.S. Environmental Protection Agency). 1987. Guidelines for developing quality assurance
project plans, Appendix B. Environmental Research Laboratory, Corvallis, Oregon.
USEPA (U.S. Environmental Protection Agency). 1992. EMAP Monitoring. EPA 600/M-91-051.
U.S. Environmental Protection Agency, Office of Research and Development, Washington, D.C.
USEPA (U.S. Environmental Protection Agency). 1994. Water Quality Standards Handbook: Second
256
Edition. EPA 823-B-94-005. U.S. Environmental Protection Agency, Washington, D.C.
USEPA (U.S. Environmental Protection Agency). 1997. Monitoring Guidance for Determining the
Effectiveness of Nonpoint Source Controls. EPA/841-B-96-004. U.S. Environmental Protection
Agency, Washington, D.C.
USEPA (U.S. Environmental Protection Agency). 1998. Lake and Reservoir Bioassessment and
Biocriteria: Technical Guidance Document. EPA-841-B-98-007. U.S. Environmental Protection
Agency, Washington, D.C. 88 pp.
USEPA (U.S. Environmental Protection Agency). 2000. Watershed Analysis and Management
(WAM) Guide for Tribes. U.S. Environmental Protection Agency, Seattle, WA.
USEPA (U.S. Environmental Protection Agency). 2001. EPA Requirements for Quality Assurance
Project Plans. EPA QA/R-5. U.S. Environmental Protection Agency, Washington, D.C. 24 pp. plus
appendices.
USEPA (U.S. Environmental Protection Agency). 2002. Summary of Biological Assessment
Programs and Biocriteria Development for States, Tribes, Territories, and Interstate Commissions:
Streams and Wadeable Rivers. EPA-822-R-02-048. U.S. Environmental Protection Agency,
Washington, D.C.
USEPA (U.S. Environmental Protection Agency). 2003a. Elements of a State Water Monitoring and
Assessment Program. EPA 841-B-03-003. U.S. Environmental Protection Agency, Washington, D.C.
14 pp.
USEPA (U.S. Environmental Protection Agency). 2003b. Strategy for Water Quality Standards and
Criteria: Setting Priorities to Strengthen the Foundation for Protecting and Restoring the Nation’s
Waters. EPA-823-R-03-010. U.S. Environmental Protection Agency, Washington, D.C. 37 pp.
Vales, D.J. and F.L. Bunnell. 1988. Comparison of methods for estimating forest overstory cover. I.
Observer effects. Canadian Journal of Forest Research 18: 606-609.
Van Horn, M. and K. Van Horn. 1996. Quantitative photomonitoring for restoration projects.
Restoration and Management Notes 14(1): 30-34.
VANR (Vermont Agency of Natural Resources). 2004. Vermont Stream Geomorphic Assessment:
Phase I Handbook: Watershed Assessment. Vermont Agency of Natural Resources, Waterbury, VT.
82 pp. plus appendices.
Vazquez, L., D.L. Myhre, E.A. Hanlon, and R.N. Gallaher. 1991. Soil penetrometer resistance and
bulk density relationships after long-term no tillage. Communications in Soil Science and Plant
Analysis 22: 2101-2117.
Vora, R.S. 1988. A comparison of the spherical densiometer and ocular methods of estimating canopy
cover. Great Basin Naturalist 48(2): 224-227.
Walling, D.E. 1994. Measuring sediment yield from river basins. Pages 39-80 in R. Lal, (ed)., Soil
Erosion Research Methods. Second Edition. Soil and Water Conservation Society, Ankeny, IA.
257
Warren, S.D., V.E. Diersing, P.A. Thompson, and W.D. Goran. 1989. An erosion-based land
classification system for military installations. Environmental Management 13(2): 251-257.
Warren, S.D., M.O. Johnson, W.D. Goran, and V.E. Diersing. 1990. An automated, objective
procedure for selecting representative field sample sites. Photogrammetric Engineering and Remote
Sensing. No. 56: 333-335.
Wells, C.G., R.E. Campbell, L.F. DeBano, C.E. Lewis, R.L. Fredriksen, E.C. Franklin, R.C. Froelich,
and P.H. Dunn.1979. Effects of Fire on Soil: A State-of-the-Knowledge Review. General Technical
Report WO-7. USDA Forest Service. 27 pp.
Wenger, K.F. (ed). 1984. Forestry Handbook. Second Edition. Wiley, NY. 1360 pp.
West, N.E. 1985a. Shortcommings of Plant Frequency-Based Methods for Range Condition and
Trend. Pages 87-90 in Selected Papers Presented at the 38th Annual Meeting of the Society for Range
Management, Feb. 1985, Salt Lake City, UT. Society for Range Management, Denver, CO.
West, N.E. 1985b. Origin and Early Development of the Range Condition and Trends Concepts.
Pages 75-78 in Selected Papers Presented at the 38th Annual Meeting of the Society for Range
Management, Feb. 1985. Salt Lake City, UT. Society for Range Management, Denver, CO.
West, N.E. 1993. Biodiversity of rangelands. Journal of Range Management 46: 2-13.
West, N.E., K. McDaniel, E. LaMar Smith, P.T. Tueller, and S. Leonard. 1994. Monitoring and
Interpreting Ecological Integrity on Arid and Semi-Arid Lands of the Western United States. New
Mexico Range Improvement Task Force. New Mexico State University, Las Cruces New Mexico.
Westbrook, C. and K. Ramos. 2005. Under Siege: Invasive Species on Military Bases. National
Wildlife Federation
WFPB (Washington Forest Practices Board). 1997. Board Manual: Standard Methodology for
Conducting Watershed Analysis. Version 4.0. Washington Forest Practices Board, Olympia, WA.
White, P.S. and S.P. Bratton. 1984. Monitoring Vegetation and Rare Plant Populations in U.S.
National Parks and Preserves. In Synge, H. Ed. The Biological Aspects of Rare Plant Conservation
(1981). John Wiley and Sons.
White, S. 1997. Maintenance and Control of Erosion and Sediment along Secondary Roads and
Tertiary Trails. USACERL Special Report 97/108. U.S. Army Corps of Engineers Construction
Engineering Research Laboratories, Champaign, IL. 236 pp.
Whitman, W.C. and E.I. Siggeirsson. 1954. Comparison of line interception and point contact
methods in the analysis of mixed grass range vegetation. Ecology 35(4): 431-436.
Wilson, L. M. and C. B. Randall. 2003. Biology and Biological Control of Knapweed. USDA-Forest
Service FHTET-2001-07. 2nd Edition. http://www.invasive.org/weeds/knapweed/
Winward, A.H. and G.C. Martinez. 1983. Nested frequency - An Approach to Monitoring Trend in
Rangeland and Understory Timber Vegetation. In Proceedings of International Conference on
258
Renewable Resource Inventories for Monitoring Changes and Trends (15-19 August, 1983), J.F. Bell
and T. Atterbury, (eds). Oregon State University, Corvallis. 737 pp.
Wischmeier, W.H. and D.D. Smith. 1978. Predicting Rainfall Erosion Losses - A Guide to
Conservation Planning. Agriculture Handbook No. 537. USDA, Washington, D.C.
WNHP (Washington Natural Heritage Program). 2004. Washington Natural Heritage Program, Dept.
of Natural Resources, PO Box 47014, Olympia, WA 98504-7014
http://www.dnr.wa.gov/nhp/refdesk/plants.html.
Woodruff, N.P. and F.H. Siddoway. 1965. A Wind Erosion Equation. Soil Science Society of
America Proceedings 29: 602-608.
WPN (Watershed Professionals Network). 1999. Oregon Watershed Assessment Manual. Prepared
for the Governor’s Watershed Enhancement Board, Salem, OR.
259
4.8
Appendix Report of 1989 LCTA Review
This report, dated December 14, 1989, was prepared by a panel of natural resource experts (listed at
end of document) for the U.S. Army Construction Engineering Research Laboratory (USACERL).
C.W. Cook, is the author.
REPORT OF LCTA REVIEW
by
U.S. Army Land Inventory Advisory Committee
OVERVIEW
LCTA merits high priority treatment with appropriate reward for those who contribute most
meaningfully to it. Its success depends on stability of leadership from top to bottom, especially at the
installation level.
It is generally agreed by all members of the advisory committee, that the U.S. Army is in critical need
of a comprehensive multi-resource inventory and analysis system capable of maintaining or restoring
land and resource conditions on training and testing installations. Training that simulates realistic
battlefield conditions is essential for the preparation and maintenance of effective military forces. For
training to be realistic, installations must provide a variety of conditions ranging from arid grasslands
to humid, tree-covered landscapes. Perhaps more importantly, these conditions must be maintained
over the long run if training objectives are to be accomplished year after year. A training installation
that loses its tree cover quickly becomes useless for concealment training.
In addition to the need to maintain that long-term integrity of land and resource conditions in order to
support the training and testing missions, the U.S. Army, like other federal agencies, is required to
observe federal and state laws relating to protection of the environment, and to cooperate with local
officials and with the public to ensure that environmental problems, especially those relating to
pollution, cumulative effects, and off-site degradation are quickly and responsible addressed. Of
particular importance in this regard is the National Environmental Policy Act (NEPA) of 1969 (P.L
91190, 83 Stat. 852). Among other things, NEPA requires extensive documentation, including
quantitative data, relating to the environmental effects of operations such as military training
exercises. NEPA also requires that the effects of such operations be monitored in order to ensure that
the cumulative impacts of the operations can be mitigated and that, if necessary, work can be
undertaken to restore land and resource conditions.
Both to satisfy the requirements of NEPA and to ensure that training and testing installations continue
to provide realistic conditions over the long run, it is essential (a) to develop and implement a
comprehensive multi-resource inventory system capable of measuring land and resource conditions
260
and of determining changes in those conditions over time; and (b) to develop analytical procedures
that will help estimate, in advance, the environmental consequences of military testing and training
operations.
In the past several years, the perception of the general public of the protection and uses of natural
resources on military installations has not been favorable. The reasons are many, but two among them
are important. The first is that very little information gets to the public about the Army's programs to
preserve the natural resources. The public generally believes that training areas are bombed and
shelled out and the Army has no conscience or programs to protect them. The conservation
community at large receives very little information on the Army's programs to protect scarce life, to
protect fragile lands, and to renovate disturbed lands. Perhaps this perception is a natural consequence
of military demeanor; secrecy and silence are customary attributes of military activities. Secondly,
uses of natural resources at Army installations in the past have been perceived as a perk for general
command officers and politicians. Army installations have been perceived as almost private hunting
clubs by the general public. As might be expected, some resentment has been felt about this, and
natural resources which might have been managed have sometimes been left to their own devices.
The LCTA program will indeed improve the image of resource management on Army training and
testing installations.
Answers to Specific Requests
1. Are the LCTA methods technically sound?
For the most part, the answer is yes. It is suggested that the following items be considered: 1) identify
to some taxonomic level the soils and vegetation that the plots are located on; 2) how far a steel rod is
hammered into the ground is not a very good indicator of soil depth; 3) only two very limited
observations of soils top 15 cm at two locations is not appropriate for soil classification; 4) the use of
additional plots (special use plots, control plots) or other sampling techniques where appropriate to
augment the basic sampling scheme for some high priority items such as endangered or threatened
species, or sensitive areas that may be missed or under sampled in the Core sampling procedure
should be used; 5) as now constituted, it appears that each installation will be sampled with the same
number of samples without regard to ecological types and sizes of the areas. While 200 samples may
be adequate for some areas with rather homogeneous vegetation and physical features, some
installations with diverse habitats will not be adequately sampled. This is to say that some
consideration should be given to sampling effort by considering ecological variation within types and
sizes of the areas to be sampled; 6) a total of 200 sample plots is generally adequate for most
extensive inventories of the timber resource. However, since multiple resources are being inventoried,
the intensity may need to be strengthened. Keep in mind that at least two plots per stratum are needed
to compute sampling errors and that all strata should be sampled when using the current LCTA
design. Consider doing some testing on variation within stratum and then using optimum allocation to
keep costs down to a minimum; 7) keep subjectivity to a minimum in locating plots in the field and in
selecting an azimuth to run the transect. Remember, if stratified sampling is being used, the plot is to
represent the stratum and not necessarily the polygon in which it occurs. Try to locate the plot at the
same location that the computer selects. Consider using global positioning systems to assist in plot
location if locating the plot on the ground is a problem. As to the transect direction, either randomly
select an azimuth from 0-359 degrees or pre-select an azimuth in advance of the sample by which all
transects will be established. This keeps personal bias to a minimum; 8) linkage of ground data with
imagery/soil classes will come about. However, the Army can anticipate difficulty in using the
finished map products and GIS data bases without re-shaping the user's thinking on traditional land
261
cover/use classes; 9) for trend analysis, recognized different items of concern will change at different
rates. Consider using periodic large-scale aerial photography (including 35 mm) or video coverage of
sample sites to denote changes in land cover and use; and 10) resource managers on military lands are
in a unique position to conduct treatment control experimentation to elucidate impacts of training on
their lands. It is strongly recommend that they pursue this route. Perfect controls are admittedly
unrealistic, but some control should be achieved.
2. Do the LCTA methods address the needs expressed by military land managers and
trainers/testers?
To a remarkable degree it is believed that they do. Inevitably there are differences of opinion between
land managers and trainers/testers about how the land should be managed and therefore what kinds of
inventory information are necessary. It is essential for land managers to recognize that LCTA is
designed as a tool to facilitate dominant use while conforming to legal requirements for multiple use
of federal lands, such as NEPA. That is, the training and testing missions must be considered the
primary purpose for lands on these installations. The objective of land management therefore must be
to provide lands suitable for training and testing over the long run; i.e., those lands cannot be
permitted to degrade to the extent that realism in training and testing suffers.
LCTA will provide information that should be useful in constructing management practices/plans, but
how realistic those practices/plans are will probably be the test on how successful LCTA is. Hastily
constructing management practices/plans is a concern. It is suggested that a concerted effort be made
to identify, develop and test management/ training practices that can be used to mitigate and reduce
the environmental impact of the Army's training mission. This effort should not be attempted without
the user (trainer/ operators) involved, so that management practices can be designed in concert with
the user to enhance where possible the Army's mission. It is not clear as to what the final management
objective is: 1) to manage for native plant communities at some successional stage for multiple use
purposes; 2) to manage for the enhancement of endangered and threatened species including the
introduction of species not currently found on military controlled lands and for other multiple use
purposes; or 3) to manage to maintain the soil resource and some assemblage of native plants,
allowing for introduced plant species, and the use of intensive agronomic and engineering
management practices to mitigate the impacts of training and enhance some selected other uses as
well. Is the army required by law or by some other reason to manage for multiple uses? If not, why
accept multiple use management in all of its conventional phases? Detailed management
practices/plans should be developed that will reduce/mitigate impacts to soil erosion, vegetation,
endangered/threatened species, and water quality, which should include intensive agricultural and
engineering practices.
Inventories and tolerance limits for land resource uses should be understood by both trainers and
resources managers. Furthermore, guidelines for training to protect natural resources must be
reasonable and acceptable to the command officers.
The training program should be an integrated product of both training officers and the natural
resource managers. Communication between the command and resource managers must be cordial
and reciprocal in a give and take resolution to conflicts. Planning must be oriented toward training as
the primary objective, but within due tolerance of the natural resource tolerance, primarily the land
itself.
262
3. Does the Army appear to be adequately staffed to conduct the LCTA program? If not, what
are the minimum requirements?
This is a difficult question outside reviewers to answer without extensive first-hand involvement.
However, based on papers presented at the meetings which suggested that land and resource
conditions arc degrading significantly at many installations, and on discussions with both
management and training personnel at the meeting in Colorado Springs, it is concluded that the Army
is not adequately staffed to properly conduct the program. More importantly, this implies that the
Army is not adequately staffed to meet the requirements of NEPA and to ensure that the land and
resources of training and testing installations can be maintained as needed over the long run to
support the training and testing missions.
Staffing and budget appear to be the greatest problems for successfully accomplishing the LCTA
mission. The Advisory Group was impressed by the caliber of personnel that presented papers at the
meeting. They all seemed dedicated, knowledgeable and enthusiastic, but in every presentation the
comment was made that “we do not have the personnel or budget to fulfill our mission.” The LCTA
program is destined to failure unless this need can be addressed. Not only do staffs and budgets need
to be bolstered, but incentive and reward systems need to be put in place to encourage the competent
staff to remain on the job. It seems efficient to have a central staff developing computer software and
data analysis techniques for deployment throughout the system. There is a need to increase this staff,
as well as the field staff at each training installation. A core crew of resource managers at each
training installation is desirable. This crew would be responsible for implementation of LCTA. Field
crews for seasonal tasks can be readily obtained from universities. Use of such temporary crews
would save money over a large permanent staff. It is not suggested that the Army contract out all their
LCTA work to universities because it would be unwise to lose control over the data gathering and
analysis.
The Advisory Committee encourages hiring natural resource specialists who are well trained in the
various disciplines of natural resource management. There appeared to be an under-representation of
soil scientists and range conservationists. The training and experience should match the resource they
are asked to manage. There also appeared to be a frequent turnover of natural resource specialists in
the program. The program needs to be developed to the point that it will attract and retain some of the
better talent. In the case of LCTA, it appears that a real need exists for upgrading numbers and
expanding capabilities of staff. Rather than employing generic natural resources persons, specific
expertise should be sought to do specific jobs; i.e., foresters for forests, wildlife scientists for wildlife;
soil scientists for soils, etc.
Budgets also need to be increased and set up in a manner such that they cannot be diverted for other
uses. If this cannot be accomplished then the success of LCTA is in jeopardy. LCTA has no chance of
succeeding without command endorsement and without financial arrangements that funds cannot be
siphoned off as monies drift downward to those who must do the work.
The second part of this question is even harder for an outside reviewer to address in a credible way.
Again, based largely on discussions with both management and training personnel, it is believed that
a reasonable goal for ongoing programs would be to approximately double the number of people
presently assigned full-time to land management activities on training and testing installations.
Furthermore, it is suggested that the responsibilities of these personnel be devoted strictly to land
management; it seems unreasonable to expect land managers to also assume responsibility for
263
environmental matters such as disposal of chemical wastes or replacement of asbestos, etc.
As mentioned before, in the short run, some of the additional personnel requirements of LCTA,
particularly those relating to the actual inventory measurements, could be met by contracting with
universities to provide students in natural resource programs who could do the inventory work during
the summers. Some of these students might later be hired on as permanent employees and their
experience in doing the inventory work would likely prove invaluable. Graduate students and faculty
members might also be utilized to conduct research related to the LCTA program that would provide
information about the robustness of the inventory methodology and its applicability to a wide variety
of ecosystems.
4.What would be the greatest cost to the Army - to implement LCTA or not implement LCTA?
Almost certainly the greater cost would result from a failure to implement LCTA or some program
very much like it. This cost would arise from two principle sources: (a) loss of land resources on
which to conduct realistic training; and (b) impairment of the military training and testing missions as
a result of environmentally related lawsuits, citations, and appeals associated with NEPA
requirements.
In light of current environmental laws, resources available to environmental groups to enforce those
laws, and the apparent commitment by upper management in the Army to sound environmental
management, it would appear disastrous if LCTA were not implemented. It is critical that the Army
develops a land stewardship plan of which LCTA has to be the basis. The Army cannot afford to
place our military readiness in jeopardy because of non-compliance to environmental laws.
5. Can data collected by LCTA procedures be used to fulfill environmental compliance
documentation (EAs, EISs, endangered species, etc.) and to develop land management plans?
This is the primary reason that similar procedures are being developed by the USDA Forest Service
for implementation on the national forests. Management of the national forests has been crippled
during the past four or five years because of the excessive number of NEPA-related appeals
associated with the development of forest management plans. Much of this is due to a failure to base
the plans on data from multi-resource inventories. LCTA has the potential to help the Army avoid
similar problems on its training and testing installations.
The methods employed by LCTA and the types of data collected are precisely what are needed for all
land management agencies. By establishing a long-term database, better information will be available
to fill such needs. LCTA should form the basis for such activities. How well the step is made from
LCTA to realistic, useful, and defendable management practices/plans is critical to this entire process.
At least the procedures will provide a firm base that can be built upon if necessary to meet future
local needs.
6. Should the LCTA methods be used early on in the land acquisition process identify suitable
lands for purchase?
It is in the best interest of the tax payer and national security that new military lands be matched as
closely as possible with current and future training needs by whatever means is appropriate. The use
of LCTA methods for assessing the suitability of lands for potential acquisition by the Army seems to
be one of the promising uses for LCTA. Because of the intensive use made of lands on the training
264
and testing installations, it is critical for the Army to select parcels for acquisition that represent the
most resilient possible ecosystems for training and testing purposes. The inventory procedure of
LCTA is the most important part for use in land acquisition. Probably the most critical facets of this
inventory are terrain, soils and threatened or endangered species. These latter two components have
been identified as weaknesses in the current LCTA system. A thorough suitability analysis of lands
needs to be conducted prior to their acquisition for training grounds, but it can be questioned whether
all LCTA methods and approaches would be best suited for this. If it is only a couple hundred of
acres, then some other means may be quicker and more economical.
7. Other comments and recommendations
It is enthusiastically believed by the Advisory Committee that the Army is off to an excellent start in
land management by the initiation of LCTA on its training and testing installations.
The following comments are not listed in any particular order of importance, but it is hoped that each
will receive consideration by those associated with the direction and development of LCTA.
(a) The climate diagrams currently being used in LCTA utilize temperature and precipitation. This
choice has undoubtedly been made because both are easily measured and data are likely to have been
kept over a long period of time. However, the use of temperature can be misleading because it is at
best an indirect measure of soil moisture. It seems that potential evapotranspiration (PET) might be
considered for this purpose as well.
(b) The idea of using climatic data to schedule major exercises as described in "A climatic basis for
planning military training operations and land maintenance activities" could be greatly enhanced by
the adoption of water balance modeling technology such as the ERHYM-11 or WEPP models. An
additional advantage is the compatibility of the modeling technology with GIS.
(c) USA-CERL should stay with USDA’s wind erosion prediction project and water erosion
prediction project for new technology that could be used in the near future to augment and replace
USLE in LCTA. Likewise, don’t shy away from channel structural management practices to control
on-site gully erosion and off site sediment impacts.
(d) Mapping of soils on core plots and any specialized plots to the series level should be a priority
consideration, with the surface mineral horizon being sampled in each series. Soil measurements that
would be useful for establishing trends include bulk density, infiltration rate and penetrometer
measurements. The present policy would be strengthened by testing a composite surface sample
collected randomly from 3 to 5 location within each series within a 5 m radius of the 25 and 75 m
points on the line transect if the soil series distribution in not determines. Special plots for evaluation
of erosion losses, especially from marginal sites, should be considered.
(e) Soil scientists with soil mapping and land use interpretation experience should be employed for
the interpretative and data collection aspects of LCTA. Present sampling provides for no trend
analysis of soil properties except as revealed through vegetation. Typically, two or more soil
taxonomic units (soil series) arc included in soil mapping units for low intensity surveys. Many others
may occur as inclusions within delineations of the soil mapping unit. Vegetation trends must be
associated with taxonomic units, such as soil series, if they are to have maximum extrapolative value,
thus the need to map the occurrence of soil series in each plot. Explore cooperative agreements with
the Soil Conservation Service for facilitating the LCTA effort, particularly for characterizing the soils
265
occurring on core plots.
(f) Integrated land management by definition requires an interdisciplinary approach. Because of the
unique situation associated with the management of Army installations, training and testing command
personnel represent one discipline that must become actively involved in the land management
planning process. Therefore, it is believed that command personnel must be made part of a land
management team and consulted regularly as part of tile planning process if LCTA is to be successful.
(g) Better communication should be established between resource managers of potential training
sites and persons testing new equipment. There seemed to be a general lack of understanding about
impacts of new innovations such as the “HUM-V.” Evaluation of impacts on training lands cannot be
made a priori without input regarding the nature of the insult. Better communication would help
remedy this problem.
(h) Most of the work that has been accomplished thus far in the LCTA program has been done inhouse. Although this is admirable, and certainly impressive because of the enormous amount of
progress that has been made in a short period of time, it is urged that CERL make use of outside
expertise whenever possible to improve the process and speed up implementation. Much related work
is underway by federal agencies such as the Forest Service, the Bureau of Land Management, the
Agricultural Research Service, the Soil Conservation Service, the National Fish and Wildlife Service,
and others. In addition, many universities have strong research and teaching programs relating to the
type of work being undertaken in the LCTA program, and most of them would be more than happy to
share their experience in order to contribute to this program. As examples of current problem areas
associated with the LCTA program where outside expertise could be of assistance, the following two
that were cited during the Colorado Springs meeting are listed:
Soil compaction -- determination of the relative degree of soil compaction associated with tracked
versus wheeled vehicles used in Army maneuvers is needed to schedule maneuvers and to plan
restoration work. This is closely related to similar problems that have been studied thoroughly by
forest engineers in many different parts of the country as part of the evaluation process associated
with timber harvesting operations.
Tree cover problems -- several military installations are facing rapid loss of tree cover, which presents
major problems for providing realistic training in concealment. Much research has been done at
forestry schools in the United States and abroad on reforestation of semi-arid landscapes with fastgrowing exotic species such as acacias and eucalyptus. The use of such species might conceivably be
able to alleviate some of the deforestation problems that plague some military installations.
266
PANEL OF NATURAL RESOURCE EXPERTS
contributing to LCTA Review Document
Soils Management
Dr. Murray Milford
Professor of Agronomy
Department of Soils and Crop Sciences
Texas A&M University
Dr. Will Blackburn
Project Leader
Northwest Watershed Research Center
USDA-ARS
Grassland Management
Dr. C. Wayne Cook
Professor Emeritus
Department of Range Science
Colorado State University
Dr. Ardell Bjugstad
Range Scientist Liaison
Rocky Mountain Forest and Range Experiment Station
South Dakota School of Mines
Forestry Sciences Laboratory
USDA Forest Service
Forest Management
Dr. Dennis Dykstra
Professor of Forestry
Northern Arizona University
School of Forestry
Mr. H. Gyde Lund
Inventory Forester
USDA/Forest Service
Timber Management Staff
Washington, D.C.
Wildlife Management
Dr. Bill Alldredge
Department of Fisheries and Wildlife Biology
Colorado State University
Dr. James G. Teer
Director, Welder Wildlife Foundation
Sinton, Texas
267
5 Data Management
In order to provide accurate and insightful information to land managers, range managers, and
trainers, an installation must collect, store, retrieve, and analyze data such as topographic features,
soil characteristics, climatic variables, vegetation, and wildlife information. Accurate information can
only be obtained from reliable and valid data. Reliable and valid data is only achieved through proper
data management processes, which are covered in this chapter.
5.1
Data Administration
Data has become an integral part of organizations. In many cases an organization depends heavily on
the reliability and validity of its data. To protect their data, organizations have created organizational
elements charged with all aspects of data management. Two key functions that fall under these
information resource management elements are data administration (DA) and database administration
(DBA). Listed below are some of the responsibilities of each.
DA Roles and Responsibilities
Establish and implement policies, procedures and standards for handling data in a consistent
manner
Provide education and training to personnel to promote data management skills
Strategic planning for development of data resource management policies consistent with the
organization's goals
Develop policies, standards and procedures to ensure the integrity of the data resources
Technical and administration control of the data to improve documentation and coordination
Structure analysis including data modeling and database design
Develop migration strategies for the migration of data to promote usability and sharing of data
DBA Roles and Responsibilities
Deal with technical operation of physical maintenance of the data resources and database
management systems
Carry out policies, standards, and procedures for the management of the organization's data
resources
Provide support to users, programmers and analyst
Improve the quality, accuracy and integrity of the data
Help to establish and implement data management policies
As seen from the partial lists above data administration can quickly become a serious and large
function of an organization.
5.2
RTLA Data Administration
Although the RTLA component is not a large organization it still relies heavily on the reliability and
validity of its data and still has a need for data administration functions. However, the structure of the
RTLA information resource management umbrella differs. The role of the data administrators in
268
many cases have been performed external to the installation. This would include the U.S. Army
Construction Engineering Research Laboratories (USACERL) and support provided by the Army
Environmental Center (AEC) through the Center for Environmental Management of Military Lands
(CEMML). The RTLA or ITAM coordinators, who are also responsible for data administration at the
installation, usually perform database administration roles.
Often, only one or two people at an installation are responsible for the information resource
management element of RTLA. And because RTLA requires accurate information this role becomes
very important in the RTLA structure. Unfortunately data administration of RTLA is the most
overlooked role in RTLA. Attempts have been made at higher Army levels to implement an external
data administration element in the form of a central database repository. A major objective of the
central repository was providing support for data management. Unfortunately installations were not
mandated to participate in this repository and gained little or no benefit from it. The remainder of this
chapter will present information and techniques for managing RTLA data and ensuring its integrity.
5.3
5.3.1
RTLA Data Management
A Priori
One key process that reduces the need of extensive data management, after data collection, is proper
training of the field crew. Ensuring the field crew understands the data they are collecting, collection
methodologies, and valid data requirements ensures good data going into the database. Some a priori
data management tasks are listed below.
Vegetation Identification
Make sure the field crew is familiar with all common species found on the installation. Use voucher
specimens if available for examples. Explain the coding used for vegetation identification and provide
a list of valid codes.
Data Collection Methodologies
Thoroughly cover the methodologies used on the RTLA plots. If using the standard RTLA
methodology acquaint the field crew with USACERL technical report U.S Army Land Condition Trend Analysis (LCTA) Plot Inventory Field Methods (Tazik et al. 1992). If additional, or different,
methodologies are used create an installation RTLA inventory methods manual outlining all
requirements.
Acceptable Codes
In addition to vegetation codes much of the RTLA data utilizes codes. Examples include military
disturbance, basal cover categories, and vegetation condition. The RTLA handheld data logger
restricts the entry of data to only acceptable codes in most cases. Be prepared for those instances
when paper forms are required due to data logger failure or when paper forms are used to collect
installation specific needs. Also develop a protocol for coding unknown species.
Metadata Collection
Instruct the field crew to record any information that will help explain discrepancies in the data later.
For example, if an unknown plant species is found and collected for later identification, the field crew
should note the code used when entering the data and any site information that may be helpful to
identification.
269
5.3.2
During The Collection Process
The best time to find errors in the data is while the field crew is still on the installation. What at first
inspection appears to be an error could be some special circumstance that the field crew failed to
document. The RTLA coordinator should develop procedures for entering the data into the database
weekly and checking the data for errors. Some of the more common errors to check include unknown
vegetation codes, missing data, improperly recorded data, and missing plots.
If data is collected with the handheld data loggers the process of loading data is easy. If data is
recorded on field sheets check them each week for any errors. Once the data has been transcribed and
loaded into the database a second check is needed to find any transcription errors. In the next section
data management tasks are discussed that will help facilitate the weekly data checks.
5.3.3
Database Design
A properly designed database can help to ensure data integrity in a database. Following are a few
concepts related to the design of a database.
Referential integrity constraints are utilized to ensure data integrity in a database. These constraints
are defined between a parent table and a child table by defining primary and foreign keys. A parent is
defined with a data element, or set of elements, as a primary key. The primary key is a unique value,
or set of values, that constrain the entry of data into the dependent child table. For example, PlotMast
is a parent table to PlotSurv with PlotID as the primary key in PlotMast and the foreign key in
PlotSurv. If a particular value for PlotID does not exist in PlotMast data for that plot can not be added
to the PlotSurv table.
It is worth mentioning at this time that other rules exist that can be used to ensure data integrity.
These rules include column constraints, check constraints, unique rule, and others. The most
important column constraint is the not null rule. This rule forces a value to be entered for a data
element before that observation of data is applied to the database. The not null rule is specified in the
database schema, or structure. Check constraints test the rows of a table against a logical expression.
Not all database servers utilize this rule. The unique and primary key rules are important to relational
database theory and are somewhat related. The unique rule ensures no duplicate values will exist for a
data element. Unless specifically defined a data element with the unique rule applied can have only
one null value. The primary rule is used in referential integrity constraints as discussed above. Only
one primary key can exist for a table, but many data elements can make up the primary key. Each
column, or group of columns, that define the primary key must have the unique and not null rules
applied.
Further information about database design and creating databases is available in the following
documents on the Center for Environmental Management of Military Lands (CEMML) web site:
Database Design Primer: A Beginners Guide to Creating a Database
http://www.cemml.colostate.edu/files/TPS_05_10_DatabaseDesignPrimer.pdf
Converting SQLBase Databases to Microsoft Access
http://www.cemml.colostate.edu/files/ConvertSQLBasetoAccess.pdf
Structured Query Language (SQL) Tutorial
http://www.cemml.colostate.edu/files/LCTASQL.pdf
270
Bruce, T.A. 1992. Designing Quality Databases with IDEF1X Information Models.
Dorset House Publishing.
Relational Systems Corporation. 1998. How to Design a Relational Data Model,
Extended Relational Analysis, Workshop Version 6.2. Relational Systems Corporation,
Birmingham, MI.
5.4
Database Interface Programs for RTLA
5.4.1
LCTA/RTLA Program Manager
5.4.1.1
Background
In December 1991, several installations collecting RTLA (then LCTA) data received a Beta Version
of the LCTA Program Manager (Anderson et al. 1995a; Anderson et al. 1995b). The Beta release
prototype was a DOS character-based program having all of the statistical capabilities of the current
release, Version 1.1. This Beta release provided the developers of LCTA Version 1.1 with vital user
responses and requirements. The responses acquired from the Beta version helped to establish
software design specifications for Version 1.1.
By using a graphical user interface (GUI), the infrequent user can utilize the system to the fullest
because of its ease of use and reduced training requirements. Providing standardized analysis ensures
proper error checking. In addition, a report of errors found in the data during processing provides
increased data integrity. By minimizing complex operations of data management and analysis, the
land manager has more time to interpret the results.
The LCTA Program Manager was designed to be used with data collected using the original LCTA
protocols as outlined in U.S. Army Land Condition-Trend Analysis (LCTA) Plot Inventory Field
Methods (Tazik et. al. 1992).
The LCTA Program Manager interfaces with the SQLBase database. Since the development of the
LCTA Program Manager, many new database tools have become available. Some of these new tools
are packaged with complete office software solutions and are now very popular. Microsoft Access,
for example, has risen to the forefront of popularity in the database arena. Many installations
expressed interest in a Microsoft Access database and user interface because of its ease of use and
availability. To respond to this desire, the Center for Environmental Management of Military Lands
(CEMML) developed Access LCTA, an LCTA database and user interface in Microsoft Access (see
Section 5.4.2).
Microsoft Access is recommended for RTLA over SQLBase for the following reasons:
•
•
•
•
Microsoft Access is more user friendly
Microsoft Access is more readily available
Microsoft Access may be supported on the installation
With Microsoft Access it is easier to share data with other Microsoft Access users and other
271
Microsoft Office programs
Given these reasons it is recommended that any existing RTLA databases still stored in the SQLBase
database management system be converted to Microsoft Access. If the RTLA database is consistent
with the original RTLA protocols the Microsoft Access program Access RTLA may be used in place
of the RTLA Program Manager.
5.4.1.2
Program Files and Documentation
Program files and documentation for the RTLA Program Manager are available for download on the
CEMML web site at the following location:
http://www.cemml.colostate.edu/itamhelpdesk.htm
5.4.2
Access RTLA
5.4.2.1
Introduction
Access RTLA was developed by the Center for Environmental Management of Military Lands
(CEMML) in response to installation requests for a Microsoft Access based RTLA user interface
(Sprouse and Anderson 1995). Access RTLA takes full advantage of the Windows environment
producing a very user-friendly program. Access RTLA, and its' corresponding RTLA database, was
designed to be used with data collected using the original RTLA protocols as outlined in U.S. Army
Land Condition-Trend Analysis (RTLA) Plot Inventory Field Methods (Tazik et. al. 1992) and .
5.4.2.2
Program Files and Documentation
Access RTLA Version 1.5 requires Microsoft Access 97 while Access RTLA Version 2.x requires
Microsoft Access 2000/XP. Program files and documentation for the Access RTLA Program are
available for download on the CEMML web site at the following location:
http://www.cemml.colostate.edu/accessRTLA/accessRTLA.htm
5.5
References
Anderson, A.B., W. Sprouse, P. Guertin, and D. Kowalski. 1995a. RTLA Users Interface Program
Users Manual: Version 1.0. USACERL ADP Report 95/24. Champaign, IL.
Anderson, A.B., W. Sprouse, D. Kowalski, and R. Brozka. 1995b. Land Condition Trend Analysis
(RTLA) Data Collection Software Users Manual: Version 1.0. USACERL ADP Report 95/13.
Champaign, IL.
Diersing, V.E., R.B. Shaw, and D.J. Tazik. U.S. Army Land Condition-Trend Analysis (RTLA)
Program. Environmental Management 16: 405-414.
Sprouse, W. and A.B. Anderson. 1995. Land Condition Trend Analysis (RTLA) Program Data
Dictionary: Version 1.0. USACERL ADP Report EN-95/03. Champaign, IL.
272
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (RTLA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
273
6 Electronic Data Collection Tools
6.1
Introduction
Field computers are a valuable tool for recording data for resource inventory and monitoring. A wide
variety of technologies are available to collect natural resources data electronically,including
computers that are specifically designed to withstand the rigors of outdoor use (e.g., moisture,
temperature, dust, shock-resistance), non-ruggedized computers and personal digital assistants
(PDAs), and GPS with data collection capability. The reliability of these devices has been proven
over many years of use. Numerous benefits and applications are associated with field computers, but
their use is not always optimal or cost effective. Some potential pros and cons of using field
computers are listed below.
Potential Advantages of Using Field Computers
• Reduced office data entry time requirement.
• Quality control can be incorporated into the data collection interface. For example, pull
down lists or reference lists can limit the values entered in a field, and the user can be
prompted if all fields are not filled.
• Integration of spatial data with monitoring data.
• Elimination of transcription errors and effort.
• Field data collection with ruggedized computers under wet conditions is easier than with
paper (although write-in-the-rain paper can be used effectively).
• Hardware and software development is very dynamic.
Potential Disadvantages of Using Field Computers
• Typically increased field time requirement depending on the intensity of data collection
and the ease-of use of the field computer. Paper data collection is sometimes the best
choice for some projects and when field time is limited. Field data entry errors can still be
made but minimized using data dictionaries and drop-down menus.
• Possible data loss due to battery failure, physical damage, or system/software crashes.
Field computers in use today are much more reliable than their predecessors, but they are
not immune to failure. However, data collected on paper can also be lost.
• Additional field training time required.
• Additional time, expertise, and computer system required to download and manage data
at the end of the day.
• Up-front costs can be high for equipment, software, and development of custom
interfaces. These costs may more than compensated for over the course of several years
by lower data transcription costs.
• Modifications to sampling protocol may be cumbersome or expensive if the interface is
not flexible and easily changed. This is an important consideration, since software should
not dictate methodology.
• Hardware and software development is very dynamic. For example, DOS-based custom
software will often not operate on the new operating systems being used on today’s field
computers.
274
6.2
6.2.1
Considerations when Choosing Field Computers
What Field Computer Meets Your Needs?
When choosing a computer for use in the field, it is important to consider your particular needs and
understand how different choices will meet those needs. Quite often, a single field computer will not
meet all of your needs. Computers that are ergonomically designed for rapidly logging data may lack
the performance or display features needed for more advanced computing. Computers that perform
the functions of a full performance PC, such as a laptop or tablet, are cumbersome and not
ergonomically designed for optimal data logging. With both platforms a Global Positioning System
(GPS) can usually be integrated that performs as well as a stand alone GPS (Figure 6-1). The
manufacturers and products discussed in this section were chosen because they have led the industry
and set the standard for natural resource inventory. The inclusion of a product in this discussion does
not constitute an endorsement by the authors.
6.2.2
Major Types of Field Computers
The major types of field computers are defined by their physical characteristics and operating
systems. They include data loggers and DOS-based systems, PDA systems, mobile PC systems, and
GPS. Three general operating systems (OS) are widely used on field computers. Microsoft DOS or
DOS emulating OS have been in use since the inception of field computers. Microsoft Windows
personal computer (PC) versions have been in use on rugged laptop and tablet computers for several
years. PDAs use either the Palm or Windows OS. All of these computer types have ruggedized
versions. There are numerous software packages designed for creating custom data collection and
management interfaces using Palm and Windows OS computers.
6.2.2.1
Data Loggers
Ruggedized data loggers have historically been the workhorse of field data collection. They predate
the advent of palmtop, tablet, and PDA style machines; function well in practically all weather
conditions; and enable direct transfer of data to the office computer. The development of rugged
computers for natural resource inventory dates back to the 1980s. Over the next decade the rugged
field computers became popular for collecting natural resource data, most notably in the forest
products industry. Early data loggers used MS DOS-like operating systems with command line
interfaces to run simple spreadsheet and database programs, as well as custom-built task-specific
software. The operating systems, processor power, display, and data storage capacity of this class of
computer have developed rapidly. Features previously found in PDAs and palmtop computers are are
now incorporated into ruggedized models for use in the field. Larger keyboards with more keys on
some models have improved the usability for data collection. This new generation of field computers
has much of the utility of our office based PCs, and go far beyond the capability of early data loggers.
DOS and DOS emulating operating systems have been the standard on data loggers since their
inception. DOS and DOS emulating OS use a keyboard based interface rather than a tactile type of
interface, such as the mouse or touch screen. These systems are still in wide use today and probably
will be for many years to come. As a platform for keyboard-based data entry in the field these
systems perform very well. The keypad is laid out in a configuration optimal for data entry.
275
Figure 6-1. Examples of field computers and Global Positioning Systems (GPS).
276
It is relatively easy to create custom software for DOS operating systems, and commercial
spreadsheet and database software is widely available. Power system longevity is measured in days of
use, often up to a week depending on usage, temperature, display type, and battery capacity. The
LCD screen typically measures about 2” x 3”, and has a black and white display. Memory is
expandable and some systems feature an external memory card slot. Most DOS-based loggers are also
compatible as a GPS controller/data logger.
The introduction of PDAs, palm-top, and tablet computers in the mid 1990s presented new
possibilities for portable computing. Touch screen interfaces made it possible to create a powerful
palm-sized computer that would function independent of a keyboard. Resource managers began using
the inexpensive PDAs and compact computer for a variety of tasks, but the lack of a keyboard
interface limited its usefulness as a data logger.
6.2.2.2
PDA Systems
Personal digital assistants (PDA) use PALM, Pocket PC, or Windows CE Net operating systems. All
three OS are comparable in functionality. They rely on a touch screen interface and use pull down
menus and icons to initiate functions and programs. These OS vary in system requirements, but
generally require more system random access memory (RAM) than the DOS OS. In addition to the
palm sized PDA, this type of OS is now being used on computers with a keypad optimal for data
logging. Many of the DOS based data logger hardware platforms are now being offered with the PDA
OS. Standard layout keyboards are available as an add-on accessory for most PDA platforms,
although they can still be cumbersome to use in the field. Most systems are arranged so you can hold
the unit in both hands and operate the keypad with both thumbs.
Software is widely available for the PDA OS. Because early PDAs used the Palm OS, most software
was originally written for that operating system. As Windows-based PDAs gained popularity, more
software was developed for them. Currently, there is abundant software available for both platforms.
Some software packages are available in one or the other but not both. Many of the modern software
programs such as spreadsheet packages are available in an abbreviated form for the PDA. Most
software is highly compatible with the common desktop OS. The PDA platform can be utilized for
several wireless applications, including cell phone, wireless internet, Bluetooth, infra red, and two
way radio communications. Plug-in modules are also available for use as a digital camera or GPS.
Power system longevity is generally limited to less than a day with most PDA platforms. The LCD
screen is usually around 2.5” x 3.5”, and has a color display. Memory is very expandable and most
systems feature an external memory card slot that will accept micro-drives with gigabytes of memory
capacity. Some PDA systems are highly ruggedized for use in most weather and temperature
extremes. More economical PDA systems can often be fitted with ruggedized casings to protect
against moderate weather conditions. Most are also compatible as a GPS controller/data logger, and
can utilize state-of-the-art GPS survey software.
6.2.2.3
Mobile PC Systems
The laptop and tablet personal computers are powerful computing platforms that allow the resource
manager to take the functionality of a desktop computer into the field. The large display and desktop
PC operating system and software make it very convenient to use as a field computer. Some laptop
systems are fully ruggedized and feature a removable screen panel that functions as a touch screen
tablet PC via wireless link to the CPU in the keyboard panel. Shoulder harnesses are available to hold
the tablet in a comfortable position for two handed operation.
277
The mobile PC platform can be utilized for several wireless applications, including cell phone,
wireless internet, Bluetooth, infra red, and two way radio communications. Special mobile operating
systems are available that increase power system longevity up to 6 hours on a single battery. Most are
also compatible as a GPS controller/data logger, and can utilize state of the art GPS survey software.
6.2.2.4
Global Positioning Systems (GPS)
GPS originally designed to document locations on the earth, is now capable of supporting a wide
range of other data collection needs. GPS is necessary to map the locations of features and has helpful
navigation features. Some GPS companies have developed sotware modules that allow the user to
collect additional data that is associated with location data. These interfaces are commonly referred to
as data dictionaries. The data dictionary creation extension allows the user to create a form on the
office PC that can be downloaded and utilized on the PDA platform’s GPS/GIS software for detailed
feature data collection in the field. Data dictionaries utilize pull down menus, text entry fields, and
fields that are automatically updated with date, time, GPS positions, and user defined defaults. They
utilize the touch screen interface and allow the user to rapidly select pre-defined variables from pull
down menus as well as enter data using a touch screen or keyboard. A with most data collection
software, the efficiency of data collection diminishes as the need to type text increases. Data
dictionaries are less useful when the variables in a pull down menu feature are numerous, and take
longer to locate in the menu than to enter the text with a key pad. Data dictionaries have been created
for rare species surveys, invasive weed surveys, plant community mapping, road condition
assessments, and other applications.
6.2.3
Cost
When considering the affordability of a field computer it is important to evaluate its features relative
to your needs. Many of the available computers are comparable in terms of computing performance
regardless of their field ruggedness. A common PDA or laptop will perform all of the operations of a
ruggedized system for a fraction of the cost. These hardware platforms are usually the first to
integrate technological advancements. These systems are not designed for the rigors of field work, but
can be made semi-rugged with the addition of cases that protect the unit from most environmental
hazards. Some PDAs and palmtops have been successfully used in the field for several seasons using
resealable plastic bags for protection. If it is critical that the equipment be able to function in all field
situations, or it is difficult to replace damaged equipment (even if it costs a fraction of a ruggedized
version) then ruggedized equipment is appropriate. The number of companies manufacturing
computers with the ruggedness and functionality required by resource managers is limited. Several of
the most popular rugged field computers, data loggers, and GPS units in use are listed in Table 6-1.
This is by no means a comprehensive list or an endorsement of particular brands. Also note that
specifications and costs change over time. For a thorough description of each product, visit the
websites listed in Section 6.4.
278
Table 6-1. Comparison matrix of ruggedized field computers and data loggers. Information is current as of January 2005.
Make &
Model
Type
Cost
Rugged Op. System Range
Speed
Bat./Life
Panasonic
Toughbook P1
Handheld *Mil. Std.
Computer
810 F
Lithium Ion
64 MB SDRAM, 32 1,400 mAHr. 8
MB Flash EPROM
to 12 hour
Panasonic
Toughbook
P18
Field
Windows XP.
Lithium Ion
Laptop / *Mil. Std.
Intel
$3,500- Pentium M. 256-758 SDRAM. 40 6,600 mAHr. 6
Tablet PC
810 F
Centrino
4,000
900 Mhz
Gig. HD.
hours
Windows
Pocket PC
$1,5002,000
Ext.
Wt. Mem. Display
3.5"
320x240
Yes
NA
4.5
Lb.
Yes
10.4"
Color LCD
Full
1024X768 with light Yes Keyboard
Yes
3.8"
320X240
Color LCD
with light Yes
52 key
with
numeric
Color LCD
with light Yes
57 key
with
numeric
Color LCD
with light Yes
No
23
Oz.
Handheld *Mil. Std. Windows CE $2,000- Strong Arm 64 MB SDRAM, 32
Data Logger 810 F Net 4.0 on CF 2,500
206 Mhz
MB Flash EPROM
NIMH 3800
mAHr. 30
hours
-13-140
deg. F.
40
Oz.
No
3.8”
320X240
NIMH 3800
mAHr.
-22 - 140
deg. F.
17
Oz.
Yes
3.8”
240X320
16 bit AT 1 MB Flash EPROM. NiCad 2,100 -40 - 129
bus
Up to 8 MB Ram Disk mAHr. 9 hours deg. F.
29
Oz.
XScale 64 MB SDRAM, 64 to
$1,000- 200 to 400
128 MB Flash
1,500
Mhz
Memory
Corvallis
Microtech.
PC5L
Handheld *Mil. Std.
Data Logger 810 F
MS Dos 5.0
$1,5002,000
Corvallis
Microtech.
PC5L
Handheld *Mil. Std.
Data Logger 810 F
$1,0001,500
CMT ROS
5 Mhz
384 Kb. Standard. Up
to 8 MB Ram Disk.
NiCad 2,100
mAHr. 90
hours
-40 - 129
deg. F.
29
Oz.
Color LCD
with light Yes
38 key.
No
numeric
NA
-4-140
deg. F
Windows
Pocket PC
Display
Type *T.S. Key Pad
16
Oz.
Lithium Ion
2,800 mAHr.
?hours
Tripod Data
*Mil. Std.
Systems Recon Field PDA
810 F
Intel
Centrino
933 Mhz
Temp.
Range
128 MB DRAM, 64
MB CF for OS
Itronix
Handheld *Mil. Std. Windows XP. $2,000GoBook Q-200 Data Logger 810 F
Tablet PC
2,500
Tripod Data
Systems
Ranger
XScale
PXA263
400 Mhz
Memory
Yes
3.5X2” B&W LCD
8X21 lines with light.
No
3.5X2" B&W LCD
8X21 lines with light.
Windows Ce. $1,500XScale
Lithium Ion
Symbol Tech. Handheld
Net or Pocket 2,000
PXA255 Up to 128 Mb RAM, 2,200 mAHr. -4 - 122
25
MC9000-K
Data Logger *IP64 Std.
PC
400 MHz
64 Mb ROM
Removable
deg. F.
Oz.
No
*Mil. Std. 810 F. Dust and moisture resistant casing. Sealed port and connector covers. Scratch resistant case and screen. Shock resistant
*IP64 Std. Resistant to wind blown dust and rain.
*T.S. = Touch Screen.
3.8”
240X320
B&W or
Color with
light
No
55 key
with
numeric.
No
55 key
with
numeric.
Yes
28,43, 53
key with
numeric
279
Table 6-1. Comparison matrix of ruggedized global positioning systems. Information is current as of January 2005.
Make /
Model
Ruggedness
Cost External
Range Antenna
Accuracy
Range
Memory
Weight
Bat. /
Life
Temp.
Range
Display /
Pixels
2MB Ram
Disk. 512 Kb
EPROM
2 lb.
NiCad, 5
hours
-10 to 129 2X3", 8X21 B&W LCD. Back
deg. F.
lines
light.
No
14 Key
Display Type
Touch
Screen Keyboard
Corvallis
Microtech.
March-II-E
$3,000*Mil. Std. 810 F 3,500
Optional
2.5 cm to 5
meters
Corvallis
Microtech.
Alto-G12
*Mil. Std. 810 F
$4,5005,000
Optional
2.5 cm to 1
meters
4 to 8MB
3 lb.
NiCad, 8
hours
-17 to 129
deg. F.
3.5X2"
128X200
B&W LCD. Back
light.
No
55 Key
Centimeter
64 MB RAM,
512 MB
internal flash
25 lb.
2 VHS
bats
-4 to 129
deg. F.
320X240
Color LCD. Back
light.
Yes
57 key with
numeric.
1.6 lb
Lithium
Ion, 8
hours
14 to 122
deg. F.
240X320
Color LCD. Back
light.
Yes
No
Trimble Nav.
$7,500Pro XRS/TSCe *Mil. Std. 810 F 8,000
Trimble Nav.
Geo XT & Geo
XM
Yes
64 MB RAM,
Sub-meter with
512 MB
DGPS
internal flash
Ip 54*
$3,5004,000
Optional
Trimble Nav.
BoB
*Mil. Std. 810 F
$1,5002,000
NA
Real time
DGPS
NA
UK
8 hours
UK
NA
NA
NA
NA
Trimble Nav.
Hurricane
Antenna
*Mil. Std. 810 F
$5001,000
NA
Sub-meter with
DGPS
NA
0.86 lb
NA
-40 to 158
deg. F.
NA
NA
NA
NA
Trimble Nav.
Patch Antenna
*Mil. Std. 810 F
$0-50
NA
Sub-meter with
DGPS
NA
2 Oz.
NA
-40 to 185
deg. F.
NA
NA
NA
NA
Trimble Nav.
Pathfinder
Power
*Mil. Std. 810 F
$40004,500
NA
Sub-meter with
DGPS
NA
1.38 lb.
External. -22 to 140
8 hours
deg. F.
NA
NA
NA
NA
Garmin. Etrex
Legend
IPX7*
$150200
No
3 meter
8 MB
5.6 Oz.
18 Hours
?
1.25X2
B&W LCD. Back
light.
No
No
Magellan.
Meridian
IPX7*
$300350
No
3 meter
16 MB
8 Oz
13 hours
?
1.25X2
Color LCD. Back
light.
No
No
*Mil. Std. 810 F. Dust and moisture resistant casing. Sealed port and connector covers. Scratch resistant case and screen. Shock resistant
*IPX7 ruggedness standard. Immersible in 1 meter of water for 30 seconds
*Ip 54 ruggedness standard. Resistant to wind driven rain and dust.
280
6.3
6.3.1
GPS/GIS Software from Field to Office
Software Development and Systems Integration
With the new generation of field computers in place, software companies are developing software
packages that fully utilize field computer potential and facilitate the transfer of data from field
GIS/GPS systems to the office GIS. A variety of powerful software is now available for data
collection and management for the field and office. Cooperation between hardware and software
developers continues to integrate field computer systems with office systems. The integration of the
PDA technology and operating systems into rugged field computers is helping to bridge the software
gap between field and office computer functionality that existed with the DOS OS.
6.3.2
Software Solutions for Increased Productivity
Several companies have developed software that is purchased separately or as a bundled package. The
software share similar features, although each has unique features that make it useful for particular
applications. These software systems are fully integrated with each other, and import/export of data is
generally seamless. The following information is current as of January 2005:
•
•
•
•
•
•
•
•
•
•
6.3.3
ESRI: ArcPad for Windows CE, Pocket PC. GPS/GIS for PDA platforms. $495.00
ESRI: ArcGIS 9.0. for Windows. Office GIS/Mobile PC GPS/GIS $1,500.00
Corvallis Microtechnology (CMT): Field CE GPS/GIS for Windows Ce. $985.00
CMT: Bluetooth wireless GPS module for PDA platform. $150.00
CMT: PC-GPS 3.8 for Windows. Office GIS. $395.00
CMT: PC-Mapper 5.6 for Windows. Mobile GPS/GIS for tablet PC $1,995.00
Trimble Navigation: GPS Pathfinder Office for Windows. Office PC GPS/GIS. $945.00.
$2245.00 when bundled with Terra Sync.
Trimble Navigation: Terra Sync field GPS/GIS/Data Dictionary for Windows Ce/Pocket
PC. $1500.00 when purchased separately from GPS Pathfinder Office.
Trimble Navigation: GPS Analyst Extension for ArcGIS. $2500.00
Trimble Navigation: GPScorrect Extension for ArcPad. $445.00
Overview of Software Features
The software components listed represent the state of the art in GPS/GIS technology for resource
management inventory. For a thorough description of each, please visit their website in the references
for this section. The single most innovative feature of the new GPS/GIS field software for the PDA
platform is the ability to display geo-referenced background images and feature data layers in a “real
time” GPS environment. It is possible to display digital topographic maps, digital orthographic photo
maps, satellite imagery, and other background images that are geo-referenced. Feature data layers
may be created or edited on the background layer. Feature layers are created using the GPS to record
positions and contain fields for identifying the feature type. A digital camera can be interfaced with
the GPS to attach an image to the specific feature. Audio notes may also be attached to features. The
graphical navigation feature is useful for relocating GPS positions or navigating to a place on the
background map. The graphical route logger is useful for getting back the way you came in. Special
software or extensions are also available to use the office-based GIS on a GPS enabled laptop or
tablet PC for an instantaneous interface with your data layers.
281
6.3.4
RTLA Monitoring Software
Installations using the original RTLA protocols as outlined in U.S. Army Land Condition-Trend
Analysis (RTLA) Plot Inventory Field Methods (Tazik et. al. 1992) may use the RTLA (now RTLA)
field data logger programs. These programs are DOS-based and are most commonly used on the
Corvallis MicroTechnology (CMT) MC or PC series of data loggers. The software was designed to
display and interface with the assorted forms used in the RTLA plot inventory. Files generated by the
program are uploaded to existing RLTA databases via the AccessRTLA software program. The data
logger program automated many of the repetitive actions inherent to recording data on paper forms,
and replaced tedious data transcription with instant downloading into the database. The program is
still widely used today for collection of RTLA plot data.
Further information about these programs is available from Anderson et al. (1995a), Anderson et al.
(1995b), and in documents available from the Center for Environmental Management of Military
Lands (CEMML) website:
Field Data Logger Programs
http://www.cemml.colostate.edu/files/RTLAFieldDataLogger.pdf
Requesting Software
http://www.cemml.colostate.edu/itamhelpdesk.htm#Programs
6.4
Website Resources for Electronic Data Collection
http://catalog2.panasonic.com/webapp/wcs/stores/servlet/ModelList?storeId=11201&catalogId=13051&catGro
upId=12871
http://www.tdsway.com/products/ranger/related_articles http://www.tdsway.com/products/recon
http://www.cmtinc.com/nav/frprod.html
http://trl.trimble.com/docushare/dsweb/Get/Document-128929/
http://www.trimble.com/bob.html
http://www.trimble.com/mgis_hurricane.html
http://www.garmin.com/products/etrexLegendc/
http://www.magellangps.com/en/products/meridian.asp?bhcp=1
http://www.cmtinc.com/nav/frprod.html
http://www.tdsway.com/
http://www.esri.com
http://www.symbol.com/products/mobile_computers/kb_mc9000k.html
6.5
References
Anderson, A.B., W. Sprouse, D. Kowalski, and R. Brozka. 1995a. Land Condition Trend Analysis
(RTLA) Data Collection Software Users Manual: Version 1.0. USACERL ADP Report 95/13.
Champaign, IL.
Anderson, A.B., W. Sprouse, P. Guertin, and D. Kowalski. 1995b. RTLA Users Interface Program
Users Manual: Version 1.0. USACERL ADP Report 95/24. Champaign, IL.
Tazik, D.J., S.D. Warren, V.E. Diersing, R.B. Shaw, R.J. Brozka, C.F. Bagley, and W.R. Whitworth.
1992. U.S. Army Land Condition-Trend Analysis (RTLA) Plot Inventory Field Methods. USACERL
Technical Report N-92/03. Champaign, IL.
282
7 Data Analysis and Interpretation
This instructional document is intended as a generic guide to help ITAM personnel and others address
issues related to data analysis and interpretation in the context of the Integrated Training Area
Management (ITAM) Program and DoD land management. For this reason, it does not specifically
address Land Condition-Trend Analysis (RTLA) program goals and objectives, which may change
over time. Examples of data analyses are taken from a variety of sources ranging from traditional to
innovative and simple to complex in nature. Methods presented here are equally appropriate for
examining training-related and conservation-related issues or problems, and examples draw from both
types.
7.1
Introduction and General Guidance
Data analysis and interpretation should be related directly to management and monitoring objectives
as outlined in implementation plans and monitoring protocols. A discussion and examples of
management and monitoring objectives are presented in Chapter 2. The selection of a statistical
procedure must consider a number of variables, including independence of samples (temporary versus
permanent samples), normality of data, equality of variances, and type of data. Analysis also depends
on the amount of data available (e.g., initial year, two years, etc.).
Just as management and monitoring goals and objectives determine the selection of data collection
methods and sampling designs, they can also be used to formulate specific questions that direct data
analysis approaches and procedures. For example, are you interested in comparing mean values with
a threshold value, detecting changes over time, or examining cause-and-effect or correlative
relationships? Some monitoring objectives specify the type of statistical comparison (if any) and the
level of confidence or other test requirements. In some cases a number of different analyses can be
applied to explore relationships and differences, both temporal and spatial. Monitoring may document
changes and/or relationships that were unforeseen at the outset of implementation, thus necessitating
a re-evaluation of approaches and methodologies. Where inventories of areas or populations are
made, the counts, acreages, etc. can be directly compared to one another over time, assuming the
same inventory procedure is used.
Quantitative approaches to data analysis consist of descriptive and inferential statistics (statistical
tests). Descriptive statistics organize and summarize information in a clear and effective way (e.g.,
means and measures of variability). Inferential statistics analyze population differences, examine
relationships between two or more measured factors, examine the effect of one factor on one or more
other factors, and examine whether a management action is having the desired effect. Statistics allow
the user to make inferences about the population from a sample because it provides a measure of
precision or variation of the sample data. Sample estimates without measures of variation have
limited use because it is not possible to know the proximity of the sample mean to the “true” value.
Data analysis and interpretation should be documented and performed in as straightforward a manner
as possible, allowing for replication of procedures and comparisons of future analyses with results
from previous years. Presentation (i.e., reporting of results) should be done at a level that is
appropriate to the audience or reader. Examples of different audiences include the military training
283
community (e.g., Training Directorate, Range Control), natural resources staff, the public, or
scientific/professional forums. The framework of a monitoring or RTLA report is discussed in
Section 7.1.3 Guidelines for Reporting Monitoring Results.
When analyzing and presenting results for quantitative data (e.g., continuous or frequency), a
measure of sampling precision or the results of a statistical test should be provided. Without a
measure of precision (standard error, confidence interval) or the results of a statistical test such as a
the Student t-Test, it is not possible to determine if there is a difference between averages or changes
over time. Means and a measure of precision should be presented in addition to graphs illustrating
results. If analysis does not result in statistical differences being found, especially where observation
and experience indicate that changes have occurred, then the sampling approach may be inadequate to
document those changes. In this situation, it is advisable to evaluate monitoring data to document
whether or not differences of a specified magnitude can be detected.
"Statistically significant" is not the same as "ecologically important". Before interpreting a test
statistic (e.g., P or t statistic) or confidence interval, the user must think about the size of the
difference having ecological or management implications. How large a difference is ecologically
important? How small a difference is not important, especially within the context of natural
variability over time and space? Other resources and expert knowledge may be required to address
these questions. The level of statistical significance, and desired change detection where applicable,
should be determined based on needs and practical considerations. The scientific literature often
promotes the use of statistical significance levels or minimum detectable changes that are impractical
(and often excessive) for natural resources applications. A confidence level of 80% to 90%
(significance level of 0.10 to 0.20) should be considered for many applications.
7.2
7.2.1
Analyzing Monitoring Data
Overview of Statistical Applications
The choice and application of analysis tools is largely determined by the level of monitoring that is
being used and the type of data that is collected, which in turn is related to the monitoring objectives
that have been specified. Descriptions and applicability of different levels of monitoring are presented
in Section 2.2 Levels of Monitoring. Quantitative approaches to data analysis consist of descriptive
and inferential statistics. Descriptive statistics consists of methods for organizing and summarizing
information in a clear and effective way (e.g., means and measures of variability). Inferential statistics
consists of methods of drawing conclusions about a population or relationships based on information
obtained from a sample of the population. Inferential statistics can be used to analyze population
differences, make associations between two or more factors that have been measured, examine the
effect on one factor from changes to other factors, and examine whether a management action is
having the desired effect. Statistics allow the user to make inferences about the population from a
sample because it provides a measure of precision or variation with regard to the sample data. Sample
estimates without measures of variation have limited use because it is not possible to know the
proximity of the sample mean to the “true” value.
Monitoring objectives generally focus on parameter estimation and change detection over time. The
primary inferential procedures for addressing these issues are confidence intervals and statistical tests.
Confidence intervals can be used for both point estimates (e.g., estimates for a single point in time)
284
and estimating changes over time. Statistical tests are a way to determine the probability that a result
occurs by chance alone.
7.2.2
Types of Data
The primary types of data that will be considered in this section are abundance data and frequency
data. Although the approaches to interpreting these types of data are similar (parameter estimation
and testing for differences), the methods of statistical analyses are different. Abundance data includes
density and cover information. This data is considered interval or continuous data, where quantities
are counted or estimated on a continuous scale (i.e., height, density, cover, length). Data on the
number of individuals or items falling into various categories is considered frequency data (e.g., in
how many of 50 frames surveyed did species A occur?). Frequency is based on presence or absence,
and is not a true measure of abundance. For example, if a particular species or condition is present,
there is no way to compare different levels of abundance (or degree) among different quadrats.
Frequency data can be analyzed according to the normal distribution or the binomial distribution,
depending on the sampling design and distribution of the data.
7.3
Confidence Intervals
Sampling error arises from estimating a population characteristic by looking at a subset of the
population rather than the entire population. It refers to the difference between the estimate derived
from a sample survey and the 'true' population value. There are no sampling errors in a census
because the calculations are based on the entire population. Sample variance is commonly used to
quantify sampling error, and is related to the sampling and estimation methods used in the survey as
well as other factors. The relationship between sample variance, precision, and statistical power is
described in Section 3.1.6. Confidence intervals and standard errors are two representations of
sampling error.
A confidence interval for the population mean, based on sample data, provides information about the
accuracy of the sample mean. The confidence level of a confidence interval for a population mean
signifies the confidence of the estimate. That is to say, it expresses the confidence we have that the
estimated value actually lies within the confidence interval. The width of the confidence interval
indicates the precision of the estimate; wide confidence intervals indicate poor precision (or high
variability), while narrow confidence intervals indicate good precision. For a fixed sample size, the
greater the required level of confidence, the greater the width of the confidence interval. Commonlyused levels of confidence are 80%, 90%, 95%, and 99%. For natural resources management purposes,
confidence levels of more than 95% are generally impractical, expensive, and unnecessary. The
confidence level chosen should be reflective of the amount of risk you are willing to accept in making
a false conclusion based on a confidence interval (i.e., the confidence interval does not in fact contain
the true population mean).
The standard error also characterizes the error associated with a sample taken from a population. It
measures the average difference between the statistic and population parameter from the array of all
possible random samples from a given population (Huck 2000; PSU 2004). While the standard
deviation is a measure of the amount of variability in a data set, standard error measures the amount
of variability a statistic has from sample to sample (Devore and Peck 1986). The standard error helps
show the range of possible values of a parameter, and represent only an educated guess of the
parameters true value (Huck 2000). An approximate confidence interval may be generated using the
sample estimate + 2 standard errors, but like the standard error, it is limited to situations when the
285
sample size is sufficiently large. It is generally true for 95% of all random samples from a population
and thus represents an approximate 95% confidence for a parameter (PSU 2004).
While the width of both the standard error and confidence interval provide important information on
how precise the estimate of the parameter is (Devore and Peck 1986), the confidence interval is easier
to interpret because the probability that the interval of possible values will include the true value of
the parameter is specified/known (Huck 2000). Also, confidence interval strength or reliability is not
dependent on sample size, whereas standard error size may vary with the size of the sample,
especially for smaller sample sizes. For larger sample sizes, the variability in the standard error
estimates becomes less significant, as does the advantage of using confidence intervals versus
standard errors (Huck 2000).
7.3.1
Assumptions
Confidence intervals are a form of parametric statistics (data are assumed to have an approximate
normal distribution) that rely on several assumptions in order to interpret results with the appropriate
level of confidence. If the assumptions are violated, the validity of the confidence intervals may be
suspect. Several visual approaches for testing assumptions for parametric statistics are presented in
Figure 7-1.
The basic assumptions and guidelines for using confidence intervals are:
a) The data (samples) have a normal distribution. In statistics, the Central Limit Theorem (CLT)
states that the sampling distribution of means will be approximately normally distributed for
large samples even if the population is not normally distributed. Thus confidence intervals
can often be used despite non-normal parent distributions, as long as the departure from
normality is not too severe and the sample size is large enough. In community and
population-level monitoring, the CLT is usually applicable where n ≥ 10 (The Nature
Conservancy 1997).
b) Samples are independent. It is important that the samples are not related or correlated.
Random sampling helps to ensure that samples are independent. This assumption is violated
by samples from permanent plots, where the value of a measurement will often be related to
the value of subsequent measurements.
c) Samples are random and unbiased. Restricting data collection to representative, typical, or
“key” areas does not constitute a random sample (Green 1979).
d) Variances are equal. This assumption applies only to comparisons of two or more samples.
Samples are typically assumed to have similar variances. Large differences in sample sizes
can contribute to unequal variances.
When one or more of these assumptions about the population is seriously violated, then
nonparametric statistics are used.
286
A.
0
20
40
60
80
PERCOV93
100
120
0
20
40
60
80
PERCOV93
100
120
20
40
60
80
PERCOV93
100
120
B.
Expected Value for Normal Distribution
C.
3
2
1
0
-1
-2
-3
0
Figure 7-1. Graphical approaches to examining data distribution. A. Box plot, B. Dot histogram (dit)
plot, C. Normal probability plot.
287
7.3.2
Calculating Confidence Intervals
The confidence interval for the estimated population mean is calculated using the following equation:
X ± tα ,v s X
where:
t = the critical t value for a confidence level of 1-α and n-1 degrees of freedom
v = number of degrees of freedom = n-1
S x = standard deviation of the estimated mean or standard error of the mean (SE or
SEM))
where: S x =
standard deviation
n
A two-tailed t table is used (see Appendices). In words, we can say that we are 1-α confidence that
the confidence interval contains the true mean. When referring to a confidence interval, the quantity
1-α (e.g., 1-0.1 = 0.90 or 90%) is referred to as the confidence level. Note that as the standard
deviation of the mean becomes smaller, the confidence interval also becomes smaller. Also, as sample
size n increases, standard deviation of the mean typically gets smaller. As the confidence level
increases (i.e., as α gets smaller), the confidence interval becomes larger. A large α produces a
narrower confidence interval.
7.3.3
Comparing a Point Estimate to a Threshold Value
Often, it is desirable to know if a resource has achieved a particular status or condition, sometimes
referred to as a management threshold. By specifying management thresholds expressed as numerical
goals, land managers have a benchmark against which progress or lack of progress can be measured.
For example, a management objective may specify a minimum population size for a particular species
of concern. Constructing a confidence interval for a point estimate is the most straightforward
application of confidence intervals. If the threshold value is included in the confidence interval (i.e.,
the confidence interval overlaps with the threshold value), there is no statistical difference at the
specified confidence level (e.g., 90%).
7.3.3.1
Example 1: Cover/Abundance (normal distribution)
Canopy cover of perennial grasses was measured on grassland plots on a parcel of land in Georgia
(Table 7-1). Sampling was conducted in 1991 and 1993 on 25 vegetation transects. The management
objective was to maintain at least 70% perennial plant cover. The monitoring objective was to
determine whether perennial plant cover was at least 70%. The following steps are necessary: (1)
calculate the mean ( x ), standard deviation (s), and standard error of the mean (SE) for each sample;
(2) calculate the confidence interval as x ± (tα , v ×SE ) . Sample data is presented in Table 7-1.
288
Table 7-1. Sample data for calculating confidence intervals.
Sample
ID
Percent Perennial Cover
1991
1993
10
16
21
22
24
44
60
61
62
66
67
73
90
101
103
104
106
121
124
125
136
158
161
187
188
mean
standard deviation
standard error
57
72
84
70
37
46
80
2
2
43
30
32
79
63
66
45
47
7
69
48
79
53
33
76
30
63
99
47
58
67
60
100
0
0
69
34
64
82
89
40
47
63
20
97
73
85
45
26
80
71
50
59.16
24.39
28.04
4.88
5.61
A table of values of the Student’s t distribution is presented in Section 7.15 (Appendix – Statistical
Reference Tables).
The 90% confidence interval for 1991 mean perennial vegetation cover is:
50 ± (1.71)(4.88)
= 50 ± 8.34
= 41.66 to 58.34
The 90% confidence interval for 1993 mean perennial vegetation cover is:
59.16 ± (1.71)(5.61)
= 59.16 ± 9.59
= 49.57 to 68.75
Figure 7-2 illustrates that for both 1991 and 1993, perennial plant cover was less than 70% on
grassland plots, with a 90% level of confidence. Based on these results, the management objective has
not been achieved.
289
80
perennial cover (% )
70
60
50
40
30
20
10
0
1991
1993
y e ar
Figure 7-2. Percent perennial grass cover - means and 90% confidence intervals.
7.3.3.2
Example 2: Frequency or Proportional Data (binomial distribution)
The following example uses data from a Great Basin installation to examine the frequency of
Centaurea diffusa (diffuse knapweed) within a particular watershed over a five year period. Data was
collected on 100m –long transects, placing the frequency frame (in this case 60 cm X 60 cm) at 50
locations on either side of the transect for a total of 100 frames per sample. All frames were
aggregated within a watershed where diffuse knapweed was considered a land management concern.
Data is presented in Table 7-2.
Table 7-2. Frequency and confidence limits for diffuse knapweed over a five-year period.
year
1991
1992
1993
1994
1996
# frames with diffuse # plots surveyed total # of frequency
knapweed present
(100 frames/ plot) quadrats surveyed
830
798
1378
879
1488
22
30
38
35
34
2200
3000
3800
3500
3400
proportion of frames
95% lower
with diffuse
limit
knapweed
0.38
0.27
0.36
0.25
0.44
0.34
0.24
0.32
0.21
0.40
95% upper
limit
0.42
0.30
0.40
0.28
0.48
The results are presented graphically in Figure 7-3. If the monitoring objective is to detect whether
knapweed frequency exceeds 0.4 (or 40% of samples), then the threshold is exceeded in 1991, almost
exceeded in 1993, and exceeded again in 1996.
290
proportion of frames with
diffuse knapweed
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1991
1992
1993
1994
1996
year
Figure 7-3. Frequency and confidence limits (95% level) for diffuse knapweed over a five-year period.
In this case, a 95% confidence level was chosen. Binomial confidence limits were taken from
published tables (Sokal and Rohlf 1981) (see Section 7.15 Appendix Statistical Reference Tables).
7.3.4
Comparing Two Independent Samples
This approach illustrates how confidence intervals are used to evaluate changes over time or
differences between samples at the same point in time. For example, is the sample in year 1 different
from the sample in year 2? There are two methods to address this type of question. The first uses
confidence intervals for each point estimate (i.e., each time period). If the confidence intervals
overlap greatly, the samples are not different, especially if the confidence interval of one sample
includes the mean value of the other. If the confidence intervals do not overlap at all or are widely
spaced, the samples are probably different. This method is an extension of the approach for pointestimate confidence intervals, discussed in Section 7.3.3.
A second, more effective method, is to estimate the amount of change by developing a confidence
interval for the difference between the two means. If the confidence interval for the mean difference
does not contain zero, then the samples are different at the specified level of confidence.
7.3.4.1
Example 1: Confidence Intervals for Means (independent samples)
This example uses data from a Great Basin installation to determine if there are differences in shrub
density on shrubland plots between areas that receive training and areas that are unavailable for
training (Figure 7-4). The confidence intervals for the two samples do not overlap and there is some
distance between them. In this case we would conclude that the samples are different at a 95% level
of confidence; shrub density is higher on plots with no military use.
291
30000
shrub de nsity (ste m s/ha )
25000
20000
15000
10000
5000
0
training
no training
Figure 7-4. 95% confidence intervals for shrub density on land that is used for military training and
land where training is excluded (1991 Idaho data).
7.3.4.2
Example 2: Confidence Interval for the Mean Difference (independent samples)
Using the same data collected at the Idaho site (22 plots in training areas, 22 plots in control areas) in
the above example, we can use a more exact approach to examine the difference between two
independent samples. Figure 7-5 illustrates that the 95% confidence interval for the mean difference
does not contain zero. We can therefore conclude that the means are different. This result agrees with
and provides a less subjective interpretation than the findings based on the comparison of confidence
intervals for the sample means (Figure 7-4).
20000
18000
m e a n diffe re nce
(ste m s/ha )
16000
14000
12000
10000
8000
6000
4000
2000
0
diff train/no train
Figure 7-5. Mean and 95% confidence interval for difference in shrub density in trained and untrained
areas, 1991.
7.3.5
Comparing Two Non-independent Samples (permanent plots)
When permanent plots are remeasured, then measurements are not independent from one another.
Instead of comparing confidence intervals for point estimates, a confidence interval is constructed
around the mean of the differences between each pair of plots using the standard error of the mean
292
difference. This approach is appropriate for examining changes over time. If the confidence interval
for the mean difference does not contain zero, then the resampled plots are significantly different at
the specified confidence level.
7.3.5.1
Example: Calculate a CI Around the Mean Paired Difference
This example compares data collected from permanent plots (paired data) to determine if there has
been a change over time. Data was collected in Idaho on land used for training and adjacent land
where no training occurs. Shrub densities were counted on permanent plots in 1991 and 1997.
Confidence intervals for the mean difference in shrub densities between 1991 and 1997 were
calculated for both groups of plots (Figure 7-6). The results indicate that densities did change
significantly (neither confidence interval overlaps with zero). Densities on plots where training
occurred increased significantly from 1991 to 1997. Plots located where no training occurs had a
significant decrease in shrub density at the 90% confidence level.
1000
Diffe re nce 1991-1997
shrub de nsity (ste m s/plot)
800
600
400
200
0
-200
-400
-600
-800
-1000
training
no training
Figure 7-6. Mean and 90% confidence interval for change in shrub density on permanent plots.
The data can be organized in a simple table based on descriptive statistics that can be calculated by a
spreadsheet, a statistical package, or by hand (Table 7-3).
Table 7-3. Density of shrubs (live individuals/plot).
1991
training
no training
n
22
22
Mean
467.5
1338.9
St. Dev.
332.0
1243.3
1997
Mean
1166.4
891.9
St. Dev.
918.3
798.9
Paired Difference
Mean
701.4
-447.0
St. Dev.
955.9
854.4
St. Error
111.1
182.2
% Diff.
+150
-33
Percent difference is calculated as the change between 1997 and 1991 relative to 1991 (relative
change), and is calculated as:
relative difference (%) =
(1997 mean - 1991 mean)
× 100
1991 mean
293
7.4
Statistical Tests for Monitoring Data
Statistical tests provide technically defensible and statistically sound means to examine conditions
relative to threshold or ‘desired’ values, evaluate the magnitude and significance of changes in
resource conditions over time, examine of cause-and-effect relationships, and evaluate the adequacy
of sampling designs using inventory and monitoring data.
7.4.1
Caveats for Statistical Tests
Statistical software makes calculations easy. However, care must be exercised to adhere to the
assumptions associated with statistical tests. Fowler (1990) suggests ways to avoid some common
statistical errors:
1. Explain the experimental design and how the statistical analysis was done.
2. Avoid doing lots of separate statistical tests (e.g., do not do a large series of t-tests when an
ANOVA is appropriate).
3. Be aware of the assumptions associated with the tests used.
4. Don't pool data without justification.
5. Use multiple comparison tests correctly (i.e., if nonsignificance is found in an ANOVA, do not
break-up the data to identify significant differences between individual pairs).
By definition, it is possible to carry out a parametric test if there are at least two samples. With two
samples the degree of freedom is 1; however, the quality of the information is questionable (i.e., the
sample mean and estimate of variability may not be very representative). The larger the sample size
the greater the chance the data represent the population.
294
Table 7-4. Statistical terms and their definitions.
7.4.2
Terms
Definition
parameter
A measure of a population, such as the mean, standard deviation,
proportion, or correlation.
statistic
A descriptor of a sample, such as mean, standard deviation, proportion, or
correlation.
hypothesis
Part of a test for significance is that the hypothesized value of the sample
(statistic) is or is not equal to the population (parameter) value.
Type I Error
The null hypothesis is rejected when true. By setting a low level of
significance, the chance of a Type I Error is reduced, but the probability of
Type II Error increases.
Type II Error
A null hypothesis is accepted when it is false. In this case, the two means
really are not equal.
one-tailed test
When a parameter in a hypothesis is stated to be greater than or less than a
given value, the test is said to be one-tailed. A one-tailed test considers the
results in one direction, such as is μ1-μ2 > 0 or biomass is greater on plots
with less than 30% tracking. The probability at a given level of significance
is half that of a two-tail test, therefore, a one-tail test is more rigorous
(powerful) and less susceptible to a Type II error.
two-tailed test
When a parameter in a hypothesis is equal or not equal, then the test is said
to be two-tailed. A two-tailed test is preferred if either deviation would be
cause for action. In this case, both tails of the sampling distribution are of
concern, such as μ1-μ2 = 0. The probability at a given level of significance
is twice that of a one-tailed test and, therefore, less rigorous.
variable
Any measured characteristic or attribute, such as percent bare ground, litter,
or plants/plot.
independent variable
A measured characteristic or attribute thought to be the controlling variable
in the relation.
dependent variable
A manipulated characteristic or attribute determined by another variable.
variance
Variance is the measure of variability in a population. The value of a
variance around a mean ranges from zero (when all measurements in the
population have the same as the mean) to plus infinity (Woolf 1968).
Statistical Significance and Confidence Levels
Statistical significance level and confidence are often used interchangeably. Biological significance is
not equivalent to statistical significance. While there is a scientific need for using the terms
“significant” or “P<0.05”, neither may accurately describe a biological significant situation (Yoccoz
1991). While two sample means may differ statistically, the result may be the consequence of a small
or unrepresentative sample, nonrandom data, dependency between samples, or unequal variances. For
this reason, the importance of incorporating biological meaning or significance into program
objectives should not be understated. However, a proper level of biological significance is often
difficult to determine. Determining what constitutes a biologically significant change requires reviews
295
of available scientific information and professional judgment. Ideally, significance levels are set prior
to looking at the data to avoid bias.
Typically, statistical test results state whether the probability level is greater than or less than the test
level (α). In other words, if α = 0.05, then the test value is displayed as P<0.05 or P≥0.05. One goal
of statistical tests is to minimize the chance of committing a Type I error (i.e., rejecting the null
hypothesis when it is true - a false change error). By setting α = 0.20, there is a greater chance of
committing a Type I error. By setting a lower α such as 0.01, there is less chance of committing a
Type I error, but a greater chance of committing a Type II error (i.e., accepting the null hypothesis
when false; a missed change error).
If change detection is an objective, then particular attention should be paid to setting Type I and Type
II error rates. Without a priori information, it may be difficult at the beginning of a monitoring
program to set realistic power and Type I error rates simultaneously. One approach is to set Type I
and Type II error rates at the same level. The minimum detectable change would be set by the
affordable sample size and the observed variance. If the affordable sample size and the minimum
detectable change size are unacceptable, then the method or design is inadequate and must be
reconsidered (Hinds 1984). Type I and II error rates can also be adjusted (within limits) in order to
reach a balance with affordability and minimum detectable change. Sometimes minimum detectable
change size and Type II error are ignored altogether while the sensitivity of the analysis to falsechange errors is examined exclusively (Hinds 1984).
There is nothing immutable about the values of 0.01, 0.05, and 0.10, which correspond to confidence
levels of 99, 95, and 90 percent, respectively (Yoccoz 1991). A biologically significant difference
may be accurate at a lower confidence level (e.g., α = 0.10 or α = 0.20). An α = 0.20 may describe a
biologically significant difference between a control and training area attribute better than a P value
less than an α = 0.05. Traditionally, the two types of errors are not treated equally; Type I errors are
often considered more severe. For example, a 5% chance of a Type I error and a 20% chance of Type
II error may be accepted in relation to a given amount of change (Snedecor and Cochrane 1980). In
each case, the consequences of making the two errors should be considered. Hinds (1984) suggests
that traditional rates for both types of errors of 1 and 5% were suitable for experimental work and
domestic (i.e., controlled) conditions where the costs for making errors were quantifiable, and that
realistic and adequate error rates for monitoring projects may be significantly higher (10 to 15%) and
still produce credible results.
When reporting statistical results include means, a measure of variability (e.g., standard error,
standard deviation, confidence interval), the estimated difference set prior to the test (often zero), and
the confidence level. Also realize that with further sampling, the probability level and required
sample sizes may change. Understanding biological systems requires multiple years of data collection
to assess both spatial and temporal variability.
7.4.3
Hypothesis Testing
Hypothesis testing is necessary to correctly interpret statistical results. Some uses of statistics such as
confidence intervals do not involve hypothesis testing. Prior to beginning an "experiment," a
researcher states the anticipated result or statistical hypothesis. Typically, the statement is that a
parameter (population) represented by one sample group of data will or will not be equal to a second
group of sample data. The statement is written about the population (parameter). The initial
hypothesis, or null hypothesis, is stated and the alternative hypothesis(es) follows. A question such as
296
-- Are military impacts similar between Training Area X and Training Area Y?, would translate into:
The mean value of the response variable (e.g., vegetation cover, bare ground) in areas subjected to
training impacts in X equals the mean value of the response variable in areas subjected to training
impacts in Y, or Ho μ1=μ2, where Ho stands for the null hypothesis, μ1 is the mean of the population
represented by the first sample, and μ2 represents the mean of the population represented by the
second sample. An alternative hypothesis might be: The mean for the training impacts in X is
different from the mean training impact in Y, or H1 μ ≠μ .
1
2
The basic steps in performing a hypothesis test are:
ƒ
ƒ
ƒ
ƒ
ƒ
7.5
State the null and alternative hypotheses.
Decide on the significance level, α.
Determine the decision rule.
Apply the decision rule to the sample data and make the decision.
State the conclusion in words.
Choosing a Statistical Procedure
Different monitoring objectives and types of data necessitate that the user choose an analysis
approach from a number of possible approaches. Monitoring objectives often focus on parameter
estimation and detecting change over time. The selection of a statistical procedure must consider a
number of variables, including independence of samples, distribution of data, equality of variances,
and type of data. Decision keys for the selection of a statistical procedure for non-independent
samples and independent samples are presented in Figure 7-7 and Figure 7-8, respectively.
7.5.1
Normality Assumptions
Examining the normality of sample data involves comparing the distribution of samples to that of a
normal distribution. A normal distribution is a specific mathematical function with a bell-like shape,
which can be expressed by the mean and the standard deviation. The distribution curve may vary in
the height and width; however, the mean, median, and mode are all at the same point. Many
biological variables follow a normal distribution. Survivorship curves, rates, and size variables, tend
to follow a Poisson or other distribution functions, as do other continuous variables related to time
and space. Because the common statistical tests are based on a normal sampling distribution, some
investigators test their data for "normality." For these data types, a "goodness-of-fit" test is
performed. Goodness-of-fit measures the degree of conformity between the sample data to the
hypothesized distribution (D'Agostino and Stephens 1986).
297
Permanent or Paired Plots (Non-independent Samples)
Statistical
assumptions
met?
parametric
Paired t-test
non-parametric
Wilcoxon signed rank test
Sign test
frequency
(binomial)
McNemars test (paired)
yes
2 Samples
no
Statistical
assumptions
met?
parametric
Repeated
measures
ANOVA
F test
significant
non-parametric
Friedman test
F test
significant
yes
>2 Samples
no
frequency
(binomial)
Cochrane's test (Q test)
Figure 7-7. Decision key to statistical analysis of monitoring data from permanent or paired plots.
298
Bonferroni method
Duncan's multiple range test
Fisher's LSD
Scheffe's test
Student-Newman-Keuls (SNK)
Tukey's method
Temporary Plots (Independent Samples)
Statistical
assumptions
met?
parametric
2 sample t-test
non-parametric
Mann Whitney U
frequency
(binomial)
Chi-Square test of Independence
yes
2 Samples
no
Statistical
assumptions
met?
parametric
ANOVA
F test
significant
non-parametric
Kruskal-Wallis
F test
significant
yes
>2 Samples
no
frequency
(binomial)
Bonferroni method
Duncan's multiple range test
Fisher's LSD
Scheffe's test
Student-Newman-Keuls (SNK)
Tukey's method
Chi-Square test of Independence
Figure 7-8. Decision key to statistical analysis of monitoring data from temporary plots.
299
Data can either be graphed or tested to determine if the data approximates a normal distribution
pattern. Parametric tests such as an analysis of variance (ANOVA) and t-tests assume the data are
normally distributed (i.e., a bell shape distribution, or if a cumulative distribution is plotted on normal
probability paper, linear). Nonparametric tests do not require a normal data distribution. However,
using nonparametric statistics (i.e., rankings) for analyzing continuous data can be problematic,
leading to erroneous results. Graphing a data set is a quick method to evaluate the pattern of
distribution (Figure 7-9). Statistical tests are easily performed and included in a number of statistical
software packages. Three commonly used tests are the Kolmogorov-Smirnov (KS) test, Pearsons ChiSquare, and Log-likelihood ratio.
When data do not conform to a particular probability distribution, there are two courses of action. The
first is the use of a nonparametric test, such as Kruskal-Wallis or Friedman. A second possibility, is to
transform the variable to meet the assumption of normality (Sokal and Rohlf 1981). By transforming
the data to another scale, a standard analysis can be used. An appropriate transformation may be a
logarithmic scale for data that are multiplicative on a linear scale. The use of square root
transformation works well for areas, reciprocals for pH and dilution series, and arcsine transformation
for percentages and proportions. Scale of measurement is arbitrary and transformation of the variable
helps satisfy the assumptions of parametric tests (Sokal and Rohlf 1981).
7.5.2
Frequency/Binomial Tests
The chi-square test of independence and McNemar’s test are nonparametric tests for detecting
differences between proportions. Discussion and examples of these tests are provided in Section
7.5.4, Non-parametric Tests.
7.5.3
Parametric Tests
Parametric tests involve the calculation of the t statistic (i.e., the sample average divided by the
estimated standard error, or the number of estimated sample standard deviations the test statistic ( x )
is from its hypothesized value) or the F statistic (i.e., Treatments Mean Square divided by Error Mean
Square). In both cases, the parameter of interest is the population mean (μ). Because the value of μ is
reflected in the sampling distribution of x (sampling mean), and because x follows a normal
distribution with a sufficient sample size, the t and F tests are good tests for identifying differences
between and among sample means.
7.5.3.1
The T-Test
A t-test is a measure of a random sample mean and an unbiased estimate of a population. The
sampling distribution of the data set should be normal, or a close approximation. t-tests, and other
parametric tests are not as robust with small sample sizes (e.g., less than 12) as they are with larger
samples (e.g., 100 or more). The larger the sample size the closer the sample distribution approaches
to a normal distribution. Large samples are robust (i.e., there is a greater chance the P value is
accurate), powerful (i.e., correctly rejecting a false null hypothesis, or 1- β), and can discriminate
between a normal and a non-normal distribution (GraphPad Software, Inc. 1998). Small samples
often do not have enough information and, in some cases, statistical testing may be inappropriate.
300
Expected Value for Normal Distribution
A. Cumulative distribution; 1994
B. Cumulative distribution; 1997
3
3
2
2
1
1
0
0
-1
-1
-2
-2
'normal' distribution
-3
-3
0 10 20 30 40 50 60 70 80 90
0 10 20 30 40 50 60 70 80 90 100
C. Density distribution; 1994
D. Density distribution; 1997
80
60
70
50
60
x
Count
positive skew
40
x
50
40
30
30
20
20
10
10
0
0
0 10 20 30 40 50 60 70 80 90 100
F. Log transformation, 1997 data
E. Log transformation, 1994 data
70
60
60
50
x
50
Count
0 10 20 30 40 50 60 70 80 90
x
40
40
30
30
20
20
10
10
0
0
1
2
3
LITTER
4
5
0
1
2
3
LITTER
4
5
Figure 7-9. Comparison of differences in cumulative, density, and log-transformed density
distribution of litter ground cover; 1994 and 1997. A. Linearity of cumulative distribution as an
estimation of normality of data in 1994. B. Lack of linearity of cumulative distribution of data in 1997,
suggesting non-normality. C. Density distribution of 1994 data approximates a Gaussian, or bellshape distribution. D. Positive skew of the density distribution of the data in 1997, suggesting nonnormality. The box at the top of graphs C and D are the confidence intervals. The means are shown
( x ). E and F. Log transformations.
The generalized formula for a t-test is (Rice Virtual Lab 1998):
t=
× − μh
Statistic − Hypothezised value
OR t df = n −1 =
Estimated standard error of the statistic
sx 2
n
where
x = the mean of the random sample
μh = the statistical hypothesis of the population mean
sx2 = the estimated population variance
n = sample size
301
A typical null hypothesis associated with a t-test is stated as the mean equal to zero (Ho x = 0), or any
other value. The value should represent a target or threshold that has real-world significance. For
example, if we want warm season grass cover to be at least 30%, then we would state the null
hypothesis as Ho x ≥ 30%. In this case a one-tailed test would be employed. The alternative
hypothesis may be the mean does not equal a value (H1 x ≠ 0) or that the mean of the population is
greater than a given value (Ho x > 0). If the alternative hypothesis is H1 x ≠ 0, then a two-tailed test
of significance is required. If the alternative hypothesis is H1 x > 0, then a one-tailed test of
significance is appropriate. The difference is whether or not both or only one side of the normal curve
is considered by the test. When a hypothesis states greater than or less than, only one-tail of the curve
is considered. When a hypothesis states a condition either equal to or unequal to some value then both
tails of the curve must be considered (two-tailed test).
7.5.3.1.1
One Sample T-Test Example
Consider plant litter estimated on ten plots and from each plot determine the difference XI from μ
when α = 0.05 (Table 7-5). The previous year, litter ground cover was estimated at 25%. We want to
know if litter cover this year is significantly different from 25% The null hypothesis is Ho μ = 25
(i.e., the average amount of litter ( x ) represented by the sample is equal to or greater than the
threshold value specified for the population). The alternative is H1 , μ ≠ 25.
Table 7-5. Percent litter cover for 10 samples.
Statistical Parameter
Sum of the percent litter on 10 plots = ∑XI
Sample mean =
∑
Xi
n
Xi
Xi2
43
58
62
24
29
33
34
85
26
42
1,849
3,364
3,844
576
841
1,089
1,156
7,225
676
1,764
436
43.6
where n = the sample size
Sum of each sampled squared = ∑Xi2
Sample variance = s 2 =
=
302
22,384
∑(X
i
− x)
n −1
(436) 2
10
10 − 1
22,384 −
∑ Xi −
2
2
=
(∑ X i ) 2
n −1
n
=
22,384 − 19009.6
9
= 374.9
Standard deviation (s) = s 2 = 374.9 = 19.36
Estimated standard error = s x =
Calculated t value
t=
s2
=
n
374.9
= 6.12
10
x − μ 43.6 − 25
=
= 3.04
sx
6.12
The estimated sample mean ( x ) is 43.6, the hypothesized population mean is 25, and the standard
error ( s x ) is 6.12, and the calculated t is +3.04. The critical t value for 10-1 = 9 degrees of freedom at
P (0.05) is 2.26 and for P (0.01) is 2.82 (see Table 7-37, Critical Values of the Two-tailed Student’s
T-Distribution). Because the calculated t is greater than the critical t value, we reject the null
hypothesis. The probability value for the calculated t is therefore smaller than 0.01. If only one tail of
the curve is considered, and the alternative hypothesis is H1 μ ≥ 25, the probability at 0.05 would be
half the stated t-value or 1.13 and 1.41 for P (0.01).
To set the confidence limits at 95% for the population mean from which the sample was drawn, the t
value at the 0.05 level for n - 1 degrees of freedom is 2.26.
L1 = x - t0.05 s x = 43.6 - 2.26(6.12) = 29.77
L2 = x + t0.05 s x = 43.6 + 2.26(6.12) = 57.43
The probability is 95% that the true population mean is between 29.77 and 57.43.
Confidence limits are useful measures of the reliability of a sample statistic, but are not commonly
stated in scientific publications. Generally, the statistic plus and minus (+/-) its standard error are
cited along with the sample size upon which the standard error is based (Sokal and Rohlf 1981).
However, in monitoring, confidence intervals are very useful for comparing a mean to a threshold or
target value. If the target value fall outside the confidence interval, then you can be 1-α % confident
that the mean is greater than, less than, or no different from the threshold value, whatever the case
may be.
Statistical books have additional examples. The primary source used for this discussion was
Principles of Biometry (Woolf 1968).
7.5.3.1.2
Comparison Test Involving Two Sample Means
A common test is to compare sample means from random samples. If μ1 = μ2, i.e., the samples means
are the same, then any differences are due to sampling variation. The point at which the samples
describe different populations is based on the level of significance set prior to the test (e.g., α = 0.05).
303
A group comparison test requires the samples to be independent, normally distributed, and to have
equal population variances. The null and alternative hypotheses are: Ho μ1 = μ2 and H1 μ1 ≠ μ2. The
degrees of freedom are (n -1) + (n - 1). The alternative hypothesis calls for a two-tailed t-test. For the
following question, the significance level is α = 0.05.
1
2
Given plant litter cover data was estimated on an installation in June and September of the same year,
do the means represent the same population (Ho μ1 = μ2)? or the alternative hypothesis -- do the
means represent different populations (H1 μ1 ≠ μ2), α = 0.05 (Table 7-6)?
304
Table 7-6. Plant litter cover data collected across an installation in June and September 1998.
June
X12
3364
3721
2916
2916
2704
1936
7569
5041
4225
5041
6724
3844
2809
1600
4356
400
2704
2704
3364
1156
4225
1936
2304
4761
2304
5476
5476
1225
2809
5929
X1
58
61
54
54
52
44
87
71
65
71
82
62
53
40
66
20
52
52
58
34
65
44
48
69
48
74
74
35
53
77
∑xi
∑xi2
xi
ni
1723
September
X22
729
484
16
289
361
361
1024
1089
1024
1764
1369
484
625
1089
1089
256
256
1600
289
196
900
625
625
900
X2
27
22
4
17
19
19
32
33
32
42
37
22
25
33
33
16
16
40
17
14
30
25
25
30
610
105,539
57.4
30
17444
25.4
24
The formula for the estimated pooled variance is:
s p2 =
∑X
1
2
−
(
∑X ) +∑X
n
1
1
2
2
2
−
(
∑X
2)
2
n2
(n1 − 1) + (n2 − 1)
372,100 ⎤
2,968,729 ⎤ ⎡
⎡
+ ⎢17,444 −
⎢105,539 −
⎥
24 ⎥⎦
30
⎦ ⎣
=⎣
(30 − 1) + ( 24 − 1)
305
=
6,581.4 + 1,939.8
52
= 163.87
The calculated t value is --
=
x1 − x2
⎛
sP2⎜⎜
1⎞ ⎛ 1 ⎞
⎟⎟ + ⎜⎜ ⎟⎟
⎝ n1 ⎠ ⎝ n2 ⎠
=
57.4 − 25.4
163.87*(29+ 23)
= 9.13
Since α = 0.05, the corresponding t statistic is between 2.00 and 2.02 for 52 degrees of freedom.
Because t is greater than 2.00 and 2.02, the null hypothesis is rejected, or P<0.05 for 52 df; that is, the
difference in litter cover between June and September is statistically significant.
The confidence intervals based on the pooled variance for the individual population means at 95%
are:
μ = x ± t 0.05
μ1 : L1 = 57.43 − 2.01
L 2 = 57.43 + 2.01
ni
163.87
= 52.7
30
163.87
= 62.1
30
μ 2 : L1 = 25.42 − 2.01
L2 = 25.42 + 2.01
s p2
163.87
= 20.2
24
163.87
= 30.7
24
The confidence intervals can be determined using individual variances for each mean rather than the
pooled variance; however, the pooled variance is a better estimator of the population's variance.
7.5.3.1.3
Paired T-Test
When samples are not independent and when there is a positive correlation between the two sample
means, a paired design is appropriate. Also, a paired test requires equal sample sizes. There is no
assumption that the variances are equal; however, the differences between the samples should have a
consistent variance (i.e., the variance of the differences does not increase as the differences
themselves increase).
306
Given percent litter ground cover was determined on plots both in June and September, do the means
represent the same population (Ho μ1 = μ2) or, the alternative hypothesis, do the means represent
different populations (H1 μ1 ≠ μ2) at α = 0.05 (Table 7-7).
Table 7-7. Litter ground cover data collected at 30 permanent plots in June and September.
Plot
June
September
Difference
d
d2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
58
61
54
54
52
44
87
71
65
71
82
62
53
40
66
20
52
52
58
34
65
44
48
69
48
74
74
35
53
77
57.43
xi
∑Xd
∑Xd2
nd
27
22
4
17
19
19
32
33
32
42
37
22
25
33
33
16
16
40
17
14
30
25
25
30
24
40
11
11
8
39
31
39
50
37
33
25
55
38
33
29
45
40
28
7
33
4
36
12
41
20
35
19
23
39
24
34
63
24
45
38
24.76
961
1521
2500
1369
1089
625
3025
1444
1089
841
2025
1600
784
49
1089
16
1296
144
1681
400
1225
361
529
1521
576
1156
3969
576
2025
1444
32.66
980
36930
30
The variance of the sample differences is --
s p2 =
=
∑
Xd2 −
∑( X
d)
2
nd
nd − 1
(980) 2
30
30 − 1
36,930 −
307
=
36,930 − 32,013.3
29
=169.5
The estimated standard error of the mean difference is -s xd =
sd 2
= 2.38
n
The t value is -t=
xd
= 13.7
s xd
The t value is greater than the critical t value for α = 0.05; therefore P<0.05 for 29 degrees of
freedom. The null hypothesis is rejected (The means represent the same population, Ho μ1 = μ2) and
the alternative hypothesis is accepted (The means represent different populations, H1 μ1 ≠ μ2).
The 95% confidence interval is 27.8 to 37.5.
7.5.3.2
Analysis of Variance (ANOVA)
When more than two samples are compared, an Analysis of Variance (ANOVA) is an appropriate
parametric test. The ANOVA table accounts for the variation of selected factors of numerous samples
simultaneously. ANOVAs are often used to compare the effectiveness of different management
activities or other applied treatments, for example as in experimental designs. The experimental
design chosen will determine how the ANOVA table is constructed. Differences in the primary
factors of interest can be examined by removing variability in the data that can be attributed to other
factors (e.g., natural variability). ANOVAs can also be applied when comparing conditions or
measured variables in three or more management units or geographical areas, even if differences in
treatments, stressors, or land uses are unknown.
When permanent samples are measured over time, analysis using repeated measures ANOVA is more
appropriate than a simple random design ANOVA (ACITS 1997; Wendorf 1997). As with other
ANOVAs, the repeated measures ANOVA tests the equality of the means, but it models the
correlation between the measurements, thus avoiding violation of the assumption of independence. In
contrast to simple random design ANOVA, where subjects are nested within each group, repeated
measures subjects are crossed with each group since all subjects participate under all levels (Wendorf
1997). By doing so, repeated measures controls for individual differences between subjects within
each group, differences that otherwise would be combined with the treatment effects of the
experiment using the simple random design. A repeated measures design can reduce error variance in
cases where between sample variability is high. This reduction in error variance in the repeated
measures design results in a direct increase in statistical power (ACITS 1997; Wendorf 1997).
According to ACITS (1997), repeated measures ANOVA should be used if any repeated measure
factor is present in a design. A repeated measures design is comparable to a simple multivariate
design, in that both sample subjects on several occasions or trials. However, in a repeated measures
308
design each trial represents the measurement of the same characteristic under a different condition,
and in a multivariate design each trial represents the measurement of a different characteristic (ACITS
1997). Repeated measures would be used for example, when comparing vegetation condition at a
specific site over a period of years.
7.5.3.2.1
ANOVA Example
We want to test for differences among groups (or treatments) using ANOVA and a post-hoc multiple
comparison procedure.
Data were collected at 5 different training sites (locales) within the same soil type (Table 7-8). Cone
penetrometer readings were taken to indicate the degree of surface soil compaction. Shallow
penetration depths indicate increased compaction. A one way analysis of variance was performed to
test the hypothesis that the sample means for the five sites are not different from one another.
Table 7-8. Soil surface compaction data.
SITE
SAMPLE
DEPTH
1
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6.9
7.3
6.5
7.2
6.7
7.9
6.4
6.6
5.6
8.1
7.7
5.6
7.6
6.4
8.5
5.5
7.0
7.8
7.4
7.2
5.3
5.5
5.9
6.3
5.8
2
3
SITE
4
5
SAMPLE
DEPTH
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
6.6
5.6
5.6
5.8
5.9
6.7
7.0
8.9
7.1
5.5
6.6
5.9
7.0
5.8
5.3
5.3
6.2
6.9
6.7
5.7
4.6
5.3
6.9
5.6
6.5
The F-ration in the ANOVA table is used to test that hypothesis that the slope is zero. The F is large
(between group error is much larger than within group error) when the independent variable helps to
explain the variation in the dependent variable. Here there is a significant linear relationship between
compaction and site. Thus we reject the hypothesis that the slope of the regression line is zero. The
ANOVA results indicate that at the 0.05 level of significance, the means for the five sites are not all
the same (P value = 0.0034<0.05) (Table 7-9).This type of test is similar to doing a number of t-tests,
but is more powerful because the variances are pooled. Once the ANOVA is completed, a multiple
pairwise comparison can be performed to determine which site (or sites) is different from the others.
In this case, the Bonferroni procedure was applied at a 95% confidence level (see Section 7.5 for a
discussion of procedures). The matrix of pairwise comparison probabilities reveals several significant
309
differences among sites (Table 7-10). The Bonferroni test results indicate significant differences
between sites 1 and 3, sites 2 and 3, and sites 2 and 5.
Table 7-9. One-way ANOVA results using compaction data.
Source
Sum-ofSquares
df
MeanSquare
F-ratio
P
SITE
(between groups)
Error
(within groups)
12.3412
4
3.0853
4.6011
0.0034
30.1750
45
0.6706
Table 7-10. Matrix of pairwise comparison probabilities from a Bonferroni test. Significant pairwise
differences at α=0.05 are highlighted (P<0.05). Means with the same letter in the summary column
are not significantly different.
SITE
1
2
3
4
5
7.5.3.3
1
1.0000
1.0000
0.0468
1.0000
.1275
2
1.0000
0.0148
1.0000
0.0435
3
1.0000
0.4642
1.0000
4
1.0000
1.0000
5
1.0000
MEAN
6.92
7.07
5.83
6.58
5.97
Summary
ac
a
bd
ab
cd
Correlations and Regression
Correlation and regression analyses test the relationship and the degree of relationship of two
variables to each other. In correlation analysis, variables are independent of each other. In regression
analysis, the two variables may consist of one independent (i.e., a fixed variable by the investigator)
and one dependent (Model I), or each variable may be independent of the other (i.e., without
investigator control) (Model II). The difference between a Model II regression and correlation
analysis is the lack of units in correlation analysis. With regression analysis, the X and Y variables
are compared based on their units. Both tests will indicate whether a relationship exists between two
variables, but using different ways (Woolf 1968).
Regression and correlation analyses are two different tests. While many of the calculations are
similar, each addresses a very different question. In regression analysis, the intent is to examine the
possible causation of changes in Y by changes in X for purposes of prediction and to explain
variation. Causation of change is unknown in correlation analysis, rather the question is how much do
the variables vary together (covariance) (Sokal and Rohlf 1981)? There may be a relationship
between the variables examined by correlation analysis, but the mathematical model for that
relationship is not of direct concern.
Biological variables generally have a relationship to other variables. For example, litter cover is noted
on a series of plots with varying levels of tracked vehicle use. The differences in litter cover may or
may not be aligned with training activities. If differences in litter cover are related to training actives,
to what degree? In this case, litter cover would be the dependent variable and the amount of training
the independent variable. The type of analysis could be either correlation or regression analysis. If on
the other hand, if we pose the question: Is there a relationship between the amount of litter cover and
the amount of standing biomass on training lands?, then correlation or Model II regression analysis
310
would be appropriate. A Model II regression would consider each variable as independent (i.e.,
without investigator control).
A scatter diagram is a pictorial method of describing the relationship of two variables (Figure 7-10).
If there is no correlation between the two variables, the line that best fits the scatter of points is
horizontal. With a positive correlation, the slope of the line increases as each variable increases, and
with a negative correlation, as one variable increases, the second decreases. In some cases, a curved
line is the best fit. Only linear relationships will be discussed here.
No Correlation
Positive Correlation
Negative Correlation
Figure 7-10. The correlation between two variables can be none, positive (as one variable increases
so does the other), or negative (as one variable increases the other decreases).
Fitting a line to a scatter of data can lead to a bias interpretation of the data. An objective method used
in regression analysis is least squares; that is, fitting a line through points that is the minimum value
of the summation of the squared deviations (Woolf 1968).
7.5.3.3.1
Regression Example
The following example compares output for a regression analysis computed with MS Excel (Tool,
Data analysis, Regression) with a step-by-step presentation of the calculation of the same information.
On a military installation, areas are subjected to varying amounts of tracked vehicle impacts. We pose
the following question: Is there a relationship between the amount of litter and the amount of tracked
vehicle use (Table 7-11)?
The question can be addressed by using regression (Model I -- one variable is dependent on a second
variable) or by correlation analysis. The null hypothesis is Ho β=0, and the alternative hypothesis is
H1 β≠0. There are some assumptions involved: 1) the amount of litter within each level of tracked
vehicle intensity follows a normal distribution, and 2) the variances of the populations represented by
the various intensities are equal.
311
Table 7-11. Training intensity recorded as percent tracked vehicle disturbance and percent plant
litter cover on 15 plots.
PlotID
Samples
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
n=15
Training Intensity
Litter Cover
Xi2
Xi
23
28
22
29
35
42
43
48
39
54
52
61
72
68
83
3Xi=699
529
784
484
841
1225
1764
1849
2304
1521
2916
2704
3721
5184
4624
6889
80
78
65
67
53
58
52
49
46
35
38
32
23
34
29
3Yi=739
0=46.6
6400
6084
4225
4489
2809
3364
2704
2401
2116
1225
1444
1024
529
1156
841
∑Yi2=40811
∑X
∑
∑
∑
y2 =
1840
2184
1430
1943
1855
2436
2236
2352
1794
1890
1976
1952
1656
2312
2407
3XiYi=30263
y-=49.3
2
i =37339
x2 =
Xi Yi
Yi2
Yi
X i2 −
∑
Yi 2 −
(
∑X )
2
i
= 37,339 −
n
(
∑Y )
i
n
2
= 40,811 −
(699) 2
= 4,765.6
15
(739) 2
= 4,402.9
15
(
X )( Y )
(699)(739
∑ xy = ∑ X Y − ∑ n ∑ = 30,2063 − 15 = −4174.4
i
i
i i
The regression coefficient:
b=
∑ xy = − 4,174.4 = −0.88
∑ x 4,765.6
2
It is estimated that for a 1 unit increase in tracked vehicle use there is an 0.88 percentage point
decrease in litter.
If the value b is known, then a from the equation for a slope ( a = y − bx ) can be determined (Table
7-12).
a = y − bx = 49.3-(-0.88)(46.6) = 90.09
312
Table 7-12. The difference between the observed and the estimated litter ground cover.
Training
Intensity
Litter Cover –
Observed
Litter Cover -Estimated
Y − Yˆ
Y − Yˆ 2
samples
X
Y
Yˆ
d
d2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
23
28
22
29
35
42
43
48
39
54
52
61
72
68
83
80
78
65
67
53
58
52
49
46
35
38
32
23
34
29
69.9
65.6
70.8
64.7
59.4
53.3
52.4
48.0
55.9
42.8
44.5
36.7
27.0
30.5
17.4
10.1
12.4
-5.8
2.3
-6.4
4.7
-0.4
1.0
-9.9
-7.8
-6.5
-4.7
-4.0
3.5
11.6
101.22
154.77
33.81
5.37
41.31
22.13
0.18
0.92
98.48
60.60
42.73
21.65
16.14
12.10
134.97
PlotID
Sum =739
Sum =739 Sum = 0.0
Sum =
746.39
Where Yˆ = a + bX = 90.09 + (-0.88)X.
For PlotID 1 Y − Yˆ = 90.09 + (-0.88)(23) = 10.1
The sum of (Y − Yˆ ) 2 is also referred to as the error sum of squares or d2. The variance is:
sxy2= ∑ (Y − Yˆ ) 2 /n-2 = 746.4/13 = 57.41
From these calculations an analysis of variance table for a regression analysis can be filled.
313
If a statistical package is used, only the results are displayed. MS Excel includes the ANOVA table
and how the variance is partitioned is shown in the output. In most cases, a t test is all that is
necessary to determine which hypothesis is appropriate.
∑y2 = Total sum of squares (Total)
= 4402
∑d2 = Error sum of squares (Residual)
= 746.4
∑ Yˆ 2 = Regression sum of squares
(Regression) is the difference between the ∑y2 and ∑d2
= ∑y2 - ∑d2 = 3656.5
The degrees of freedom (df) are 1 for the Regression, 13 for the Residual (n-2), and 14 for the total
(n-1). The Mean Squares are calculated for the Regression and the Error (Residual) by dividing the
sum of squares by the degrees of freedom.
The F-value is the Regression Mean Square divided by the Error Mean Square, or
F = 3656.5/57.4 = 63.69
1 0.05] = 4.7 (look up in an F table), the null hypothesis would be rejected. A negative
Since F[13
relationship does exist between training intensity (tracked vehicles) and litter cover.
314
The t-value can be calculated a number of ways. One way is to divide the regression or b by the
standard error, or
t=
b
s xy
=
2
∑x
2
−0.876
57.4
4,765.6
= −7.98
The probability of t at 0.05 is 2.16; therefore P<0.05 for 13 degrees of freedom.
7.5.3.3.2
Correlation Example
While MS Excel includes the correlation coefficient and the coefficient of determination as part of the
regression analysis output, these two values are the products of correlation and not regression
analysis. The values used to calculate the correlation coefficient, however, follow the steps for
regression. The correlation coefficient (Multiple R as identified in MS Excel) is -r=
∑ xy
(∑ x )(∑ y
2
2
=
)
− 4,174.4
(4,765.6) * (4,402.9)
= −0.911
The coefficient of determination (R2) is -r2 = 0.8305
In other words, approximately 83.05% of the variance in Y (litter) can be attributed to X (training
intensity). If 83.05% of the variation in Y is due to the linear regression of Y on X then 16.95%
remains unexplained. A table of critical values for correlation coefficients is presented in Table 7-38
(Section 7.15 - Statistical Reference Tables).
7.5.4
Non-Parametric Tests
Parametric tests are based on assumptions, such as a random sample and equal variances. While an
investigator may presume the data meet these assumptions, data are rarely examined prior to the
execution of the desired test. These and other assumptions associated with parametric tests are very
stringent. Often assumptions are presumed to be met due to sample size alone. There are instances
when it is apparent the assumptions will not be met. These include a small number of samples and a
non-normal data distribution. In these cases, nonparametric tests may be appropriate.
Also known as distribution-free methods, non-parametric tests are not concerned with specific
parameters, such as the mean in an analysis of variance (ANOVA), but with the distribution of the
variates (Sokal and Rohlf 1981). Nonparametric analysis of variance is easy to compute and permits
freedom from the distribution assumptions of an ANOVA (i.e., the data may or may not follow a
Gaussian, bell shape distribution). These tests are less powerful than parametric tests when the data
are normally distributed. In such cases, P values tend to be higher, and there is a greater possibility of
making a Type II error (Woolf 1968). As with parametric tests, the larger the sample size, the greater
the power of a nonparametric test.
315
Some guidelines for deciding when to apply a nonparametric test (GraphPad Software, Inc. 1998):
1) Fewer than 12 samples.
2) Some values are excessively high or low.
3) The sample is clearly not normally distributed. Consider transforming the data to convert from
non-normal to a normal distribution and then using a parametric test.
4) A test for normality fails. Be aware that testing for normality requires a dozen or more
categories to be effective.
Keep in mind that nonparametric tests are counterparts to parametric tests (Table 7-13).
Table 7-13. Nonparametric test equivalents of parametric tests.
PARAMETRIC TESTS
NONPARAMETRIC EQUIVALENTS
Student's t-Test
Kolmogorov-Smirnov, One-Sample
Group Comparison
Multi-Response Permutation Procedure
Kruskal-Wallis H
Mann-WhitneyU
Kolmogorov-Smirnov, Two-Sample
Wilcoxon Signed Rank, Not Paired
Paired t
Wilcoxon Signed Rank, Paired
Two-way ANOVA; Randomized-Block
Friedman X r2
Correlation
Spearman Rank-Correlation Coefficient
Kendall Rank-Correlation
Another consideration for choosing between a parametric and a nonparametric test is the type of scale
used. Data based on descriptive scales (e.g., nominal scales -- short, tall; ordinal scales -- short,
medium, tall, very tall; and some interval scales that may not meet the assumptions of normal
distribution or homogeneous variances) should be tested using nonparametric tests. Data described by
an interval scale and interval scales with a true zero point (a ratio scale), should be examined with
parametric tests if the test assumptions are met (Woolf 1968).
The chi-square test and McNemar’s test are presented below. Both can be calculated using statistical
software programs that provide P values to directly determine statistical significance.
7.5.4.1
Chi-Square Test of Independence
The chi-squared test of independence is used to determine if there is an association or statistical
dependence between two characteristics of a population. It is appropriate for determining a change in
frequency (proportion) when using temporary sampling units or comparing differences in two or
more samples at a given point in time. This test can be used where quadrats (frequency sampling) or
points (point intercept sampling) are the sampling units and data is not consolidated to the “plot”
level. Actual frequency counts, not percentages, are the unit of measurement used by the chi-square
test.
316
In the example presented below (Table 7-14), the level of tracked vehicle use on different slopes was
categorized as either high or low. The data can be used to address the question: is there a relationship
between intensity and slope steepness? Based on that question, null and alternative hypotheses were
developed:
Null hypothesis (Ho): amount of tracked vehicle use is independent of slope steepness Alternative
hypothesis (Ha): amount of tracked vehicle use is dependent on slope steepness.
The test compares observed frequencies with the frequencies that would be expected if the null
hypothesis of independence were true. The test statistic employed to make the comparison is the chisquare statistic (X2):
X
2
∑ (O − E )
=
2
E
where O = observed frequency, and E = expected frequency
The observed frequency is the value recorded during data collection. The expected frequency is
calculated with the following equation:
Expected frequency =
row total * column total
sample size
Contingency tables are used to organize and analyze frequency (or binomial) data (Table 7-14).
Contingency tables are based on the concept that rows and columns are independent. The simplest
contingency table is made up of 2 columns and 2 rows (2x2).
Table 7-14. Qualitative data for tracked vehicle use on slopes.
Slopes < 25%
High Tracked Vehicle
Use
Low Tracked Vehicle
Use
Totals
Slopes > 25%
Totals
2
25
4
21
25
27
23
50
23
The data in the table suggest that there is a tendency for a tracked vehicle use to be higher on
shallower slopes than on steeper slopes. To test the hypothesis of the independence of the rows from
the columns, it is necessary to determine the expected frequency (E) for the four cells. Results are
presented in Table 7-15. For the first cell, the expected value is:
Ε=
25 x 27
= 13.5
50
317
Table 7-15. The observed and the expected frequency of tracked vehicle use in relation to slope
steepness. Expected values are in parentheses.
Slopes < 25%
Slopes > 25%
Totals
High Tracked Vehicle Use
23 (13.5)
2 (11.5)
25
Low Tracked Vehicle Use
4 (13.5)
21 (11.5)
25
27
23
50
Totals
The Chi-Square test is used to test the independence of the variables:
χ2 =
(23 − 13.5) 2 (2 − 11.5) 2 (4 − 13.5) 2 (21 − 11.5) 2
+
+
+
= 29.06
13.5
11.5
13.5
11.5
This chi-square value of 29.06 is compared to the chi-square value obtained from a table of critical
chi-square values (Section 7.15 Appendix Statistical Reference Tables). The degrees of freedom for
determining the critical chi-square value =[(# of rows-1)(# of columns-1)] d.f. In this case, the
degrees of freedom are equal to (2-1)*(2-1) = 1. The critical chi-square value from the table using
P<0.10 is 2.706. The calculated chi-square value of 29.06 is greater than the critical value, so we
reject the null hypothesis of no difference in amount of vehicle use. The rows (Tracked Vehicle Use)
and the columns (Slope Steepness) are not independent of each other. Based on the sample data, we
can conclude at a 90% level of confidence that tracked vehicle use is significantly higher on slopes
<25% compared to slopes <25%. The p-value can be interpolated using the chi-square table of critical
values.
Contingency tables and chi-square tests can be prepared for multiple samples and/or years. The
procedures are the same as those used for the 2x2 contingency table. Interpretation of the Chi-square
results follow those for interpreting ANOVA results: a rejection of the null hypothesis only indicates
that at least one of the proportions is significantly different. The results do not indicate which sample
proportion is different.
7.5.4.2
McNemar’s Test
McNemar’s test is applied to frequency data collected on permanent plots, where some independence
is assumed between samples. As with chi-square applications, the data consists of frequency data
where a quadrat or point is considered the sampling unit. The structure of the contingency table for
the McNemar’s test is identical to the setup for the chi-square test. McNemar’s test cannot be used to
compare more than two years of data (no more than 2x2 table). Table 7-16 contains sample frequency
data for Species X, which was sampled using 116 permanently-located frequency frames in both 1997
and 1999. Results of chi-square and McNemar’s test are presented in Table 7-17. For the equation for
calculating the McNemar statistic, see Zar (1996).
318
Table 7-16. Contingency table or cross-tabulation table for McNemar’s test.
1997
73
43
116
present
absent
totals
1999
64
52
116
totals
137
95
232
Table 7-17. Chi-square and McNemar’s test statistics and probabilities.
Test statistic
Value
df
Prob
Pearson Chi-square
McNemar Chi-square
1.444
3.528
1.000
1.000
0.230
0.060
The calculated P value is less than the threshold P value of 0.10. Therefore, we reject the null
hypothesis that the proportions are the same for 1997 and 1999; the frequency of species X is
significantly lower in 1999. If the same data were produced using temporary quadrats and the chisquare statistic were calculated, a significant difference would not have been found (accept Ha, P
value = 0.23>0.10). A threshold p value of 0.05 would have resulted in our not rejecting the null
hypothesis.
7.5.4.3
Wilcoxon Signed Rank Test
This test is used when there is a presumed underlying continuity in the data. The null hypothesis is:
There is no difference in the frequency distribution of plant litter between the spring and fall data
collection periods (Table 7-18). α= 0.05. There are 10 samples when the pair with a difference of
zero is dropped. If more than 20% of the observations are dropped, this procedure is not an
appropriate test.
Table 7-18. Plant litter cover estimated during spring and fall collections.
Plot Number
1
2
3
4
5
6
7
8
9
10
11
Spring
Fall
Difference
Rank
54
52
61
82
87
71
71
65
19
54
27
17
19
22
22
32
33
37
42
44
54
58
37
33
39
60
55
38
34
23
-25
--31
6
4
8
10
9
7
5
1
-2
--3
T = |5|
The difference per plot is determined and all samples are ranked by the absolute value of the
difference. Once ranked, the number of negative and positive samples is determined (in this example
8 of the samples are positive and 2 are negative). The values for the less frequent sign (negative) are
summed (-2 + -3 = |5|). This is T.
319
Using a Wilcoxon Signed Rank Test table, T α(n=10) = 8 (two-sided test); therefore, P<0.05 and the null
hypothesis is rejected. In other words, for 10 pairs a rank sum < 8 is required to reject the null
hypothesis at 5%. Because 5 is less than 8, the null hypothesis is rejected; There is no difference in
the frequency distribution of litter between the spring and fall data collection periods. Based on the
Wilcoxon Signed Rank Test for 10 pairs and an α = 0.05, we can conclude that there is a difference in
the frequency distribution of litter between the two seasons.
7.5.5
Multivariate Analyses
Multivariate Analysis encompasses a variety of statistical techniques that allow a user to examine
multiple variables using a single technique. For example, whereas traditional univariate comparison
techniques like t-tests and the chi-square test can be very powerful, one can only interpret the results
based on the analysis of one manipulation variable. Multivariate techniques allow for the examination
of many variables at once.
There are many different types of multivariate techniques that can be applied to vegetation analysis. It
has traditionally been used by researchers and managers to identify plant communities, define
successional trends, or pick out unusual plant assemblages as well as other uses. Several techniques
have been developed specifically with vegetation analysis in mind. These include Detrended
Correspondence Analysis (DCA) (Hill and Gauch 1980), Canonical Correspondence Analysis
(CANOCO) (ter Braak 1987), and TWINSPAN (Hill 1979). Other techniques that were developed for
applications other than natural resource management are Principal Component Analysis (PCA) (SAS
Institute 1996), Cluster Analysis(CA) (SAS Institute 1996), and various other discriminate analysis
techniques. PCA and CA will be examined in this section because they are effective tools which are
supported by affordable statistical software packages.
Multivariate techniques can be very powerful, but their results must be interpreted with care. Some
techniques are sensitive to particular data types and require that data be distributed normally. Others
cannot be used with non-linear (e.g. classification) variables. Sometimes the techniques only identify
trends, without statistical assurances regarding the confidence of the results. When using multivariate
techniques, it is important to understand their respective intended uses, strengths, and limitations.
For discussions of ordination and multivariate techniques, see Ludwig and Reynolds (1988), MuellerDombois and Ellenberg (1974), and Jongman et al. (1987). Information, references, and internet
links are provided by “The Ordination Web Page” at http://www.okstate.edu/artsci/botany/ordinate/.
7.5.5.1
Principal Component Analysis (PCA)
PCA is used to examine relationships among several quantitative variables. The technique is
particularly good at detecting linear relationships between plots of varying species composition,
density, and cover (SAS Institute 1996). For Example, plots of principal components are an excellent
way to conduct preliminary analysis of a vegetation classification scheme, in preparation for
developing a vegetation map for an installation.
Principal components are computed as linear combinations of the variables used in the analysis, with
the coefficients equal to the eigenvectors of the correlation or covariance matrix. The eigenvectors are
customarily taken with unit length. The principal components are sorted by descending order of the
eigenvalues, which are equal to the variances of the components.
320
When applied correctly, PCA is powerful for preliminary analysis of vegetation datasets, especially
for analysis of plant community data. It is particularly effective as a means for clustering survey plots
of similar composition, density, or cover. Univariate techniques such as analysis of variance
(ANOVA) can subsequently be used to compare the principal components of the ordination.
7.5.5.1.1
PCA Example
The goal of this analysis is to examine whether the plant community definitions we’ve applied to the
belt transect plots are reasonable and appropriate for these plots, by examining the similarities and
dissimilarities of the plots.
The results of a Principal Components Analysis of woody species cover for 208 transects from a
southeastern U.S. installation is presented in Figure 7-11. Raw data format is presented in Table 7-19.
The plots were classified according to vegetation map categories to facilitate visualization of results.
Principle Component Analysis, Example Dataset
1.5
2nd Principle Component
1
0.5
Deciduous Forest
Deciduous Woodlands
Evergreen Forest
0
Evergreen Woodland
Grassland
-0.5
Mixed Forest
Mixed Woodland
Pine Plantation
-1
Wet Deciduous Forest
-1.5
-2
-8
-7.8
-7.6
-7.4
-7.2
-7
-6.8
-6.6
1st Principle Component
Figure 7-11. Principal Component Analysis of woody species densities from RTLA plots on a
Southeastern U.S. Military Base. The X and Y axis variables are unitless.
321
Table 7-19. Subset of the data used in the principal component analysis and the cluster analysis
example. Values shown are cover % of plant species (shown in columns) for each plot.
Plot
1
2
3
4
5
6
7
8
9
10
11
ACNE2 ACRU ACSAF AEPA ALSE2 AMAR3
0
0
4
0
0
0
0
0
0
0
0
0
0
44
0
0
18
0
0
12
0
0
1
0
0
0
0
0
0
0
0
1
3
0
0
1
0
0
0
0
0
0
0
0
4
0
0
1
0
0
0
0
0
0
0
2
0
0
0
0
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
…
The PCA yields n-1 principal components for n variables in an analysis. The first two principal
components provide the most information, in this case accounting for over 96% of the variance in the
model. For this example we used canopy cover of all plant species with at least 10% cover. For
analyses of this type, uncommon taxa do not contribute substantially to vegetation classification
analysis since their effects are masked by the most abundant taxa. This is not to say that uncommon
taxa should not be evaluated using PCA. To do so, one should structure the dataset to include the taxa
of interest, and eliminate from the dataset the dominant taxa that may mask the effects of the
uncommon taxa on the ordination.
A visual examination of Figure 7-11 indicates that points for deciduous forest and wet deciduous
forest occupy the same space in the ordination. Mixed Forest types are clustered toward the left, and
grasslands and pine plantations tend toward the top. There is, however, substantial visual overlap
between the groupings, which suggests that shrub and understory species may be contributing
significantly to the ordination model.
Analysis of variance of the principal components based on the preliminary vegetation classification
indicates that most of the groups have a statistical basis for the groupings assigned by the field crews.
The analysis also points out, however, that there is substantial variation within the Pine Plantation
classification, suggesting that other factors other than the reforestation management regime may be a
factor in defining the dominant vegetation in the plots.
The ordination of the pine plantation data suggests that a single classification may not be appropriate
for those plots. Running a PCA ordination on just the Pine Plantation data yields the ordination
shown in Figure 7-12. Note that the locations for the points on this diagram are quite different from
those in Figure 7-11. It is important to note also that results of an ordination will vary greatly
depending upon the contents of the dataset. The resulting ordination will place plot points in very
different points in ordinal space depending on whether the data contains plots from a single
vegetation classification type or several different classification types.
322
Principle Component Analysis, Example Dataset
Pine Plantation Plots
0.16
0.14
2nd Principle Component
Outliers
0.12
0.1
0.08
0.06
0.04
0.02
0
0
0.05
0.1
0.15
0.2
0.25
0.3
1st Principle Component
Figure 7-12. Principal Component Analysis of plant cover from Pine Plantation monitoring plots on a
Southeastern U.S. Military Base. The X and Y axis variables are unitless.
An analysis of variance of the principal components indicates that the three points in the upper right
corner of the diagram are outliers. This suggests that these three plots may have vegetation
characteristics that distinguish them in some way from the other plots. At this point it seems
appropriate to examine the dominant vegetation in these plots and determine what the source of the
variation may be.
7.5.5.2
Cluster Analysis
Cluster analysis hierarchically clusters observations or samples based on the coordinates of the
observations. That is to say, it uses separating algorithms to analyze the differences and similarities of
a group of points in a two-dimensional space, and then separates the points in a hypothetical set of
clusters based on their locations in the x-y plane and their relative distances from one another.
In order to utilize cluster analysis, one must first use some type of ordination technique to produce
coordinates in ordinal (x, y) space that represent the various observations in your dataset, and then use
cluster analysis to define the various clusters that may be present. The PCA example used in Section
7.5.5.1.1 is just one type of ordination technique that can be used prior in cluster analysis. For this
example we will use the SAS CANDISC procedure, which is a form of canonical discriminate
analysis (SAS Institute 1996)
7.5.5.2.1
Cluster Analysis Example
The first step after conducting the ordination is to determine the hypothetical number of clusters that
the ordination has produced. Field crews identified nine different forest types, so it makes sense to
ask the cluster analysis algorithm to attempt to divide the data into nine clusters. The first example of
this, however, produces chaos on the ordination diagram. Figure 7-13 shows the results of the
analysis, without the clusters assigned. One can see that it is difficult to imagine nine different
clusters of points in the analysis, and Analysis of variance of the coordinates of the clusters indicates
that there is no statistical basis for producing nine different clusters. Careful inspection of the
323
ordination diagram indicates, however, that there appear to be three relatively obvious clusters of data
points.
Cluster Analysis
Results of CANDISC Procedure
6
4
2
0
-2
-4
-6
-6
-4
-2
0
2
4
6
8
Coordinate 1
Figure 7-13. Ordination diagram of the example dataset, showing x-y positions of the 208 RTLA plots
based on the first two coordinates produced by the ordination.
Specifying that only 3 clusters be produced in the cluster analysis produces the ordination diagram
shown in Figure 7-14. Analysis of variance of the x and y coordinates establishes that there is a
statistical basis at the .05 level for separating the RTLA plots into three clusters.
Table 7-20 shows how the relative proportions of the vegetation types are separated by the three
clusters. A subjective analysis of the forest types indicates that the analysis tends to cluster coniferous
forest/ woodland types toward the bottom of the diagram and deciduous types toward the top. There
are numerous other factors that appear to be involved, however, as there are exceptions to this trend.
Whereas pine plantation plots are placed in clusters 2-3-1 (in order of importance), evergreen forest
plots are placed in clusters 1-2-3 and evergreen woodland plots are placed in clusters 2-1-3.
Discussions with the field crews and the installation forester indicate that seeded pine plantations are
frequently located according to their proximity to roads within the installation. In many cases pine
plantations were started in areas that were previously cleared of deciduous vegetation or previously
were grasslands, and the dominant understory vegetation persists throughout the stands. The
vegetation classification is based primarily upon the dominant vegetation within the plots. The dataset
used for this analysis includes the several dozen smaller shrub, herb, and graminoid species in
addition to the tree or large shrub overstory species. Hence, one may conclude that the vegetation
classification system used should be expanded to include dominant understory vegetation as well as
overstory species. It might be prudent to examine other edaphic factors at the plots, such as soil type,
slope, aspect, and relative age of the forest in defining the dominant vegetation classification.
324
Cluster Analysis
Results of CANDISC Procedure
3 clusters evaluated
cluster1
Deciduous
cluster2
6
4
cluster3
2
0
-2
-4
Coniferous
-6
-6
-4
-2
0
2
4
6
8
Coordinate 1
Figure 7-14. Results of the cluster analysis with three hypothetical clusters specified.
Table 7-20 . Results of the Cluster Analysis. The relative percentage of the number of plots by
vegetation type are shown.
Forest Types*
Cluster
DF
DW
EF
EW
GR
MF
PP
WDF
1
2
3
12.21
2.33
11.63
1.16
3.49
3.49
0.58
3.49
1.16
1.74
9.3
2.33
4.07
7.56
8.14
0
9.88
0.58
2.91
0.58
0
4.07
2.33
5.23
*DF = deciduous woodland DW = deciduous woodland EF = evergreen forest EW = evergreen
woodland GR = grassland MF = mixed forest PP = pine plantation WDF = wet deciduous forest
7.5.5.3
Summary
This section described various analysis tools and demonstrated applications to monitoring data.
Natural resource managers are rarely limited in the type of analyses they can use. Whereas the
statistics mantra “the experimental design determines the type of statistical analysis” is very often
true, it does not restrict the analyst to conducting other types of analyses that he or she determines to
be prudent.
Whereas the above examples were based upon cover, a measured variable, other variables such as
stem density or biomass estimates could be used. Ordination can also be done on soil classification
data, soil loss estimates, or combinations of these types of data. Ordination is useful for combining a
large number of variables and using them to uncover trends or patterns that are otherwise not evident.
325
Ordination techniques described are very powerful, but the results can be misleading if misapplied. It
is imperative that the resource manager become familiar with the techniques and apply them as
carefully as possible to the resource management questions that he or she is investigating.
7.5.5.4
Multivariate Analysis Selected Bibliography
Burton, A.J., C.W. Ramm, K.S. Pregitzer, and D.D. Reed. 1991. Use of multivariate methods in forest
research site selection. Canadian Journal of Forest Research 21: 1573-1580.
Devore J. and R. Peck. 1986. Statistics, the exploration and analysis of data. 1983. West Publishing
Company, St. Paul, MN. 699 pp.
Digby, P.G.N. and R.A. Kempton. 1987. Population and Community Biology Series: Multivariate
Analysis of Ecological Communities. Chapman and Hall, London.
Foran, B.D., G. Bastin and K.A. Shaw. 1986. Range assessment and monitoring in arid lands: the use
of classification and ordination in range survey. Journal of Environmental Management 22: 67-84.
Gauch, H.G., Jr. 1982. Multivariate Analysis in Community Structure. Cambridge University Press,
Cambridge.
Hill, M.O. and H.G. Gauch Jr. 1980. Detrended correspondence analysis: an improved ordination
technique. Vegetatio 42: 47-58.
Hill, M.O. 1979. TWINSPAN. A FORTRAN program for arranging multivariate data in an ordered
two-way table by classification of the individuals and attributes. Ecology and Systematics, Cornell
University, Ithaca.
Huck, S.W. 2000. Reading Statistics and Research. 3rd edition. Adison Wesley Longman, Inc., New
York, NY. 688 pp.
Kent, M. and P. Coker. 1992. Vegetation Description and Analysis: A Practical Approach. Belhaven
Press, London.
Krebs, C.J. 1989. Similarity Coefficients and Cluster Analysis. In Krebs, C.J. (ed.), Ecological
Methodology. Harper Collins, New York.
Ludwig, J.A. and J.F. Reynolds. 1988. Statistical Ecology: A Primer on Methods and Computing.
Wiley, New York.
Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and Methods of Vegetation Ecology. Wiley,
New York.
Palmer, M. The Ordination Web Page. http://www.okstate.edu/artsci/botany/ordinate/
Penn State University (PSU). 2004. http://www.stat.psu.edu/~rho/stat200/chap12-p1.pdf.
Jongman, R.H.G., C.J.F. ter Braak, and O.F.R. van Tongeren, (eds). 1995. Data Analysis in
Community and Landscape Ecology. Pudoc, Wageningen, The Netherlands.
326
Pielou, E.C. 1984. The Interpretation of Ecological Data: A Primer on Classification and Ordination.
Wiley, New York.
SAS Institute 1996. SAS/STAT Guide.
Stroup, W.W. and J. Stubbendieck. Multivariate statistical methods to determine changes in botanical
composition. Journal of Range Management 36(2): 208-212.
ter Braak, C.J.F. 1987. CANOCO- a FORTRAN program for canonical community ordination by
[partial] [detrended] [canonical] correspondence analysis, principal component analysis and
redundancy analysis. TNO, Wateningen.
ter Braak, C.J.F. and I.C. Prentice. 1988. A theory of gradient analysis. Advanced Ecological
Research 18: 271-313.
7.6
Interpreting Results
The interpretation of monitoring results analyzed using confidence intervals is discussed in Section
7.3. Based on results, the interpretation can be made that a threshold is crossed/met, when dealing
with threshold objectives, or that the threshold or target was not met/crossed. Interpretation of
confidence intervals or limits is straightforward and the results are easy to communicate with a
variety of audiences. However, just because a threshold falls within the confidence interval for a
sample mean, there is still some possibility that the sample mean is either below or above the true
population mean, since the true population mean may fall anywhere within the confidence interval, at
the specified level of confidence.
Interpreting the results of statistical tests is superficially straightforward. However, interpretation goes
beyond simply stating the decision rule associated with the null hypothesis, especially when the
monitoring objective involves change detection. A decision key to interpreting quantitative
monitoring results is presented in Figure 7-15.
327
statistically significant result?
yes
no
change has
ecological
significance?
yes
RED flag - take
appropriate action
if change is
undesirable
perform post-hoc
power analysis
(one or both)
no
GREEN or AMBER
flag - no immediate
action required but
note trend
calculate power
high power
GREEN flag - have
confidence in results - there
probably was no change
low power
calculate minimum
detectable change
(MDC)
MDC too
large
MDC
acceptable
GREEN flag - have
confidence in results - there
probably was no change
AMBER OR RED flag - a change
may have been missed.
Precaution actions may be
appropriate. Modify monitoring
approach to increase power
Figure 7-15. Interpreting the results from a statistical test examining change over time (adapted from
The Nature Conservancy 1997).
Interpretation of results will depend on what constitutes an ecologically significant change in the
resource of interest and the statistical power or minimum detectable effect size (MDC) associated
with the sample data. In most cases, given the cost of monitoring, the MDC specified in the
monitoring objective should probably not be smaller than one that is biologically significant. Natural
variability plays a role in determining the minimum detectable effect size. For example, if number of
oak seedlings varies on average by 25% from year to year, then specifying a relative minimum
detectable change of 10% may result in a costly sampling program that detects a significant change
(10%) that is not ecologically significant. In the case of the oak seedlings, specifying a relative MDC
of 50% might be more realistically achieved at significantly reduced cost.
Statistical power or MDC size can be calculated using post-hoc power analysis. If the statistical test
of differences between time periods shows no statistical difference, it may be that a change has in fact
occurred but that the design used had low statistical power, which translates into the ability to only
detect a relatively large change. Therefore, before you conclude that the null hypothesis is true (no
change took place), determine the power of the test. Perhaps the sample sizes were too small, or the
background variation too large, to determine any but the largest differences between treatments or
samples. A nonsignificant but obvious trend suggests that the null hypothesis should not be accepted,
but it cannot be rejected either (Fowler 1990). Examples and discussion of post-hoc power analysis
and minimum detectable change calculation are provided in section 3.1.6 (Hypothesis Testing and
Power Analysis).
328
7.7
7.7.1
Climate Data Summarization
Sources of Climatic Data
Climatic data is available from a number of sources. They include published records available
through libraries, national weather service data, private sources which provide a wide range of
regional or global climatic data, state or county agencies or services, local airports and airfields, or
collection by the installation itself using meteorological equipment or stations. Extensive
meteorological data, including current conditions and long-term summaries, are available from a
number of sources at no cost on the World Wide Web. The principal variables of interest include
precipitation, temperature, wind direction and speed, evaporation rates, and relative humidity. In
some geographic regions, seasonal and yearly variations in climate greatly influence the response and
growth of vegetation. In these cases, and where sampling designs permit, tools such as analysis of
covariance may be helpful in accounting for variability due to climate vs. variability due to other
factors.
7.7.2
Probability of Weekly Precipitation and Climate Diagrams
Historical data can be used to predict the likelihood of climatic and soil moisture conditions during
the course of the year. If training exercises, especially those involving mechanized vehicles, are
scheduled during periods where the likelihood of wet soils is high, then vegetation loss, soil
compaction, and erosion losses are likely to be relatively high compared to periods of drier soils. In
general, damage to soil and vegetation associated with training activities is minimized during periods
of dry and frozen soils compared to moist or wet soils. Several graphic tools have been developed to
help understand seasonal patterns of precipitation and soil moisture: 1) probability of weekly
precipitation graphs, and 2) climatic diagrams developed by Walter (1985) and modified for
applications in military land management (Tazik et al. 1990). Probability of weekly precipitation
graphs and climatic diagrams are discussed and examples are presented in Tazik et al. (1990).
7.7.2.1
Probability of Weekly Precipitation
The probability of weekly precipitation is defined as the likelihood of receiving more than a given
amount of total precipitation during a specified 1-week period. Probabilities are typically based on
long-term records (25-30 years of data). A moving average can be used to smooth weekly values by
using the mean of the previous, current, and following weekly values for the current value. From a
military training standpoint, probability of weekly precipitation uses include planning of water
crossings on intermittent streams, testing equipment under wet or dry conditions, and minimizing the
need for rescheduling range activities, weapons instruction, and equipment use. In terms of land
management, the probability graphs are useful to identify optimum periods for seeding, tree and shrub
planting, and acquiring cloud-free satellite imagery (Tazik et al. 1990). A calendar of 1-week periods
and exemplary weekly data is presented in Table 7-21. Weekly probability data is presented in
graphic form in Figure 7-16. Data may also be presented using line graphs, with each line
representing a specified minimum precipitation threshold. For the location represented in Figure 7-16,
the probability of precipitation is generally high. Precipitation probabilities are lowest in October
(weeks 40-43). Precipitation generally peaks during March, July-August, and December. Seasonal
patterns of precipitation are highly influenced by geographic location and local or regional
physiography, and are therefore more pronounced in some locations relative to others.
329
Table 7-21. Example of format of weekly climatic data used to construct long-term averages and
calculate probabilities of weekly precipitation.
YEAR WEEK DATES
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
1965
330
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
JAN 1-7
JAN 8-14
JAN 15-21
JAN 22-28
JAN 29-FEB 4
FEB 5-11
FEB 12-18
FEB 19-25
FEB 26-MAR 4
MAR 5-11
MAR 12-18
MAR 19-25
MAR 26-APR 1
APR 2-8
APR 9-15
APR 16-22
APR 23-29
APR 30-MAY 6
MAY 7-13
MAY 14-20
MAY 21-27
MAY 28-JUN 3
JUN 4-10
JUN 11-17
JUN 18-24
JUN 25-JUL 1
JUL 2-8
JUL 9-15
JUL 16-22
JUL 23-29
JUL 30-AUG 5
AUG 6-12
AUG 13-19
AUG 20-26
AUG 27-SEP 2
SEP 3-9
SEP 10-16
SEP 17-23
SEP 24-30
OCT 1-7
OCT 8-14
OCT 15-21
OCT 22-28
OCT 29-NOV 4
NOV 5-11
NOV 12-18
NOV 19-25
NOV 26-DEC 2
DEC 3-9
DEC 10-16
DEC 17-23
DEC 24-21
PRECIP
(mm)
AVG MAX
(deg. C)
5.1
10.2
27.9
11.4
37.8
24.4
78.5
8.9
30.2
54.1
23.9
29.0
3.0
2.8
1.8
2.0
0.5
0.0
0.0
19.1
25.9
46.7
25.1
17.8
5.6
2.3
74.9
5.8
8.6
62.2
52.1
63.5
5.6
4.1
28.2
6.1
0.0
2.5
48.5
31.8
1.3
0.3
0.0
0.0
17.8
0.0
23.9
0.0
0.0
42.7
26.2
0.0
19
12
15
17
11
22
13
15
15
12
20
22
22
27
26
27
26
30
32
32
32
31
29
28
30
32
31
30
32
31
31
30
32
33
30
30
32
30
25
24
27
22
20
24
22
20
21
14
18
15
16
19
AVG MIN AVG MEAN
(deg. C)
(deg. C)
5
1
-1
1
-2
11
2
0
4
2
7
9
11
15
12
13
12
11
15
17
19
18
21
20
18
21
21
22
21
22
20
21
22
21
20
19
21
19
19
14
13
14
3
8
10
7
9
0
-2
9
1
3
12
7
7
9
4
17
7
7
10
7
13
15
17
21
19
20
19
21
23
24
26
24
24
24
24
27
26
26
26
27
25
25
27
27
25
24
26
24
22
19
20
18
12
16
16
13
14
7
8
12
8
11
100
Probability (% )
No Precipitation
80
60
40
20
0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
100
Probability (% )
Precipitation >13mm
80
60
40
20
0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
100
Probability (% )
Precipitation >25mm
80
60
40
20
0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
100
Probability (% )
Precipitation >38mm
80
60
40
20
0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
100
Precipitation >51mm
Probability (% )
80
60
40
20
0
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
W eeks of the Year
Figure 7-16. Probability of weekly precipitation of 0 mm (no precipitation), greater than 13 mm,
greater than 25 mm, greater than 38 mm, and greater than 51 mm. Data is from a southeastern Army
installation.
331
7.7.2.2
Climate Diagrams
Because soil moisture is greatly influenced by both the amount of precipitation received and
temperature, which greatly influences evaporation and transpiration, integrating the two variables
provides a useful management tool. The purpose of the modified Walter climate diagram is to : 1)
synthesize temperature and precipitation in order to represent soil moisture conditions, 2) illustrate
the approximate length of the growing season and period of frozen soils, and 3) characterize average
monthly precipitation and temperature. Use of the diagram can enhance land management efforts by
identifying periods where the risk of damage to vegetation and soils is elevated.
An interpretive guide to the climate diagram is presented in Figure 7-17. The diagram is constructed
by plotting both average monthly temperature (°C) and average monthly total precipitation
(millimeters) on an year-long time scale in months. Temperature and precipitation are scaled at 1 °C =
2 mm of precipitation. Where the temperature curve is above the precipitation curve, conditions are
increasingly arid. Where the temperature curve is below the precipitation curve, conditions are more
humid. Soil moisture should reflect these predominant climatic conditions. In cold climates, several
lines are added to the diagram to enhance utility to the training and land management community
(Figure 7-18). One line is placed across the diagram at 10 °C. The period where the temperature
exceeds 10 °C is generally considered the growing season, assuming soil moisture is available. The
second line, representing the point of freezing for soils, is drawn at 0 °C. If soil moisture is present,
soils will freeze at or below this temperature. During periods of frozen soils, off-road maneuvers
(including both mechanized and motorized vehicles) have minimal impacts to soil structure.
Additional examples of climatic diagrams for locations in the southeastern and northwestern U.S. are
presented in Figure 7-19 and Figure 7-20. Where a significant proportion of winter precipitation is
received as snow, soils may be wetter than indicated by the diagram during periods of snowmelt.
Graphic and or statistical examination of climatic data is also useful in understanding interannual
variability, especially in arid climates, where yearly fluctuations in temperature and precipitation or
patterns thereof can have significant ecological responses. Figure 7-21 illustrates year to year
variability in temperature and rainfall patterns relative to the long-term averages. It is evident that
while temperatures for all years approximated the long-term mean, patterns of precipitation varied
widely in some years. Spatial variability of climatic variables is also considerable where localized
events or physiographic effects result in large variations in climate within the same installation. For
example, Figure 7-22 contains precipitation data collected from seven weather stations within a
watershed on a Great Basin installation. Even though total annual precipitation may be the same for
different locations, the temporal distribution can differ significantly.
332
Figure 7-17. Interpretation guide to modified Walter climate diagrams. Reprinted from Tazik et al.
(1990).
Temperature (Celsius)
80
Precipitation (mm)
35
70
30
60
25
50
20
40
15
30
10
20
5
10
0
0
-5
-10
Jan
Precipitation (mm)
Temperature (Celsius)
40
-10
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
-20
Dec
Month
Figure 7-18. Climatic diagram for an installation in the Great Basin, based on long term (30 year)
temperature and precipitation records.
333
Temperature (Celsius)
P recipitation (mm )
160
70
140
60
120
50
100
40
80
30
60
20
40
10
20
0
P re cipita tion (m m )
Te m pe ra ture (Ce lsius)
80
0
Jan
Feb
Mar
A pr
May
Jun
Jul
A ug
Sep
Oct
Nov
Dec
Month
Figure 7-19. Climatic diagram for an installation in the southeastern United States, based on long
term (30 year) temperature and precipitation records.
Temperature (Celsius)
180
80
160
70
140
60
120
50
100
40
80
30
60
20
40
10
20
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Precipitation (mm)
Temperature (OC)
Precipitation (mm)
90
0
Dec
Month
Figure 7-20. Climatic diagram for an installation in the northwestern United States, based on longterm temperature and precipitation records.
334
30
A.
Temperature (Deg. C)
25
91
92
20
93
94
15
95
10
30 Yr Mean
5
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
400
B.
350
Precipitation (mm)
300
91
250
92
200
93
94
150
95
100
30 Yr Mean
50
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Figure 7-21. Monthly mean temperature (A) and precipitation (B) for a five year period compared to
long-term averages (southeastern U.S.).
1993-1994
1994-1995
1995-1996
90
80
Precipitation (mm)
70
60
50
40
30
20
10
0
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Month
Figure 7-22. Means and standard errors for precipitation by month for seven weather stations within
a large watershed (Great Basin data).
335
7.8
Extrapolating Results
Statistical extrapolation is the process of estimating or inferring beyond the known range based on a
sample of known values. Making inferences about an administrative unit, and management unit, or an
ecological unit requires that samples are collected and summarized according to certain principles.
Sampling design and plot allocation determine the targets and limitations associated with sampling,
and should guide the process of estimating parameters for the populations or communities
represented. Sampling design and the concept of target populations is introduced in Chapter 3.
Samples must be aggregated in order to make inferences about the population of interest. A minimum
of 2 samples is required to generate a measure of variability for a sample mean. Considerably larger
sample sizes are required to provide acceptable levels of precision.
The method of sample extrapolation is sometimes provided by the method of plot allocation. Simple
random samples are in no way constrained – every possible location has an equal chance of receiving
a sample. If any areas are excluded from the allocation, statistically speaking they are not represented
by the sample. Simple random samples can be aggregated (i.e., grouped) using any variable.
Allocation in proportion to area ensures that sampling intensities are equal within the specified
geographic boundaries of the study area. Grouping variables that reduce heterogeneity are the most
advantageous. Subjectively located plots, including macroplots, only represent the plot or area
sampled; results cannot be extrapolated statistically beyond plot boundaries.
7.8.1
Grouping or Pooling Data
Post-sampling stratification involves stratifying or grouping a sample after the data is collected
(Snedecor and Cochrane 1980). Grouping the data by strata typically results in groups that are more
homogeneous than the sample as a whole. Strata often used include vegetation types, soil types, and
land-use types.
Stratified random sampling typically draws upon existing knowledge, information, or pilot data to
delineate strata and assign a given sample size to each allocation category. In some cases the sample
sizes are equal among strata, but most often the sample sizes are unequal. Because all samples are
allocated randomly in a stratified random allocation, plot locations are by definition unbiased. For this
reason, it is not uncommon to re-aggregate sample data into groups that are different from those used
in the original allocation. In so doing, data analysis and examination is more flexible and may reveal
patterns or differences that are not readily apparent using the initial stratification scheme. However,
care must be taken to ensure that samples are both random and unbiased. For example, if a high
proportion of samples is allocated in a small portion of the area of interest, then that area is
represented by a larger proportion of samples than other areas or strata within the larger area of
interest. In such a case, aggregating samples using a simple average mathematical function may result
in bias because a large proportion of the samples represent a smaller area that may or may not be
representative of the larger area of interest.
Some strata are well-suited to some types of analyses and poorly suited to others. For example,
examining data using drainage or watershed boundaries is logical from the standpoint of hydrological
response, soil erosion, and water quality. However, watershed boundaries may not be logical divisors
of vegetation communities, soil types, or ecological communities (unless the divides consist of high
336
mountain divides that present significant obstacles to species migration, as in the case of weed
distribution and movement).
For example, data summaries are often requested using military training areas or other administrative
boundaries used by the training community, and to some extent the natural resource management
community. However, these boundaries are largely artificial in nature, often bearing no relationship to
natural features of the landscape and sometimes represented by roads and arbitrary straight lines. For
example, a large portion of most installation boundaries and impact areas are delineated using straight
lines that often correspond to grid coordinates, county boundaries, or other artificial divisions. Using
administrative boundaries as a grouping variable will likely result in high variability for the attributes
examined.
The list of possible grouping variables for summary and analysis is similar to that used for stratified
random plot allocation:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
military training areas or other administrative boundaries
types of training activity/land use
potential training type (maneuver, bivouac, dismounted only)
vegetation type/plant community classification type
ecological land classification
land maintenance activity
erosion evidence
evidence of burning
soil physical & chemical properties
soil series
erosion potential or current erosion status based on estimates
aspect
slope steepness
presence of plants of concern
rangesite (Western NRCS classification found in soil surveys)
range condition (objective or subjective, qualitative or quantitative)
watershed or subwatershed
habitat or community type (from local, state, regional, heritage or other classification)
landcover/soil type following original RTLA allocation methods (Tazik et al. 1992):
integrates historic and current disturbance with potential vegetation
historic land use
elevation
precipitation zone
multivariate analysis (vegetation composition, site characteristics, soil chars, etc.) – posthoc aggregation/grouping.
Care should be taken when using the sample data itself to create post-hoc strata or grouping
categories. Remember that in most cases, plot allocation is done on a spatial basis. That is, samples
are allocated to discrete areas that are identifiable and have meaning. If, for example, it is decided that
all (randomly allocated) burned plots will be aggregated and that all unburned plots will be similarly
aggregated, the results may be informative and descriptive, and variances will probably be reduced.
However, if the burned and unburned areas cannot be represented by geographic boundaries ( i.e., are
337
not mapped), then the results can be extrapolated to burned and unburned areas only, and not to
parcels delineated on a map.
7.9
Linking RTLA and Remote Sensing Data
Data for ground truthing remotely-sensed data is often required in many applications utilizing
remotely-sensed data. Such applications include change detection, classification, and classification
accuracy assessment. In many cases, monitoring (e.g., RTLA) data can be utilized as ground-truth
data for classification or accuracy assessment.
7.9.1
Assess Land Condition and Trends
Monitoring data has the potential to be used to detect changes in resource condition (Senseman et al.
1995). Since monitoring information is often based on a random sample, it alone may be inadequate
for detecting the locations and amount of change that occur across the landscape. This problem may
be compounded by inadequate sample sizes. However, by using plot data and remotely-sensed data
such as multi-spectral satellite imagery, it may be possible to identify and quantify change where no
field data were collected through the geographic extrapolation of sample data.
A general approach for linking monitoring data with remotely-sensed data for land condition and
trend analysis is to determine the relationship between the field data and the remotely-sensed data.
Developing a vegetation index is a typical example (Senseman et al. 1995). In effect, the remotelysensed data is used as an indirect means to document resource condition, thus permitting land
condition analysis where no monitoring data were directly collected. By repeating this procedure in
subsequent years, it is possible to determine if and where change (i.e. trends) has occurred.
7.9.2
Classify and Ground Truth Remotely-Sensed Images
Whether using aerial photography or multi-spectral satellite imagery, monitoring data provides an
opportunity for ground truthing remotely-sensed data. Monitoring data can be used in landcover
mapping and its subsequent accuracy assessment. Splitting the RTLA data into two data sets makes it
possible to use the data for both applications. The same plot data cannot be used to both develop and
assess a map because of the high correlation that would be expected at those points regardless of the
overall quality of the map or classification.
Remotely-sensed data is often used to derive information related to landcover, such as a vegetation
map. Multi-spectral satellite imagery is often used to map general landcover types. Plot and other
monitoring data can be used in the supervised classification of this remotely-sensed data. In this type
of application, the sample data would be used to “train” the supervised classification.
Accuracy assessment of vegetation maps is another potential application. A vegetation map may be
derived from the processing of multi-spectral satellite imagery of from traditional aerial photography
interpretation. In either case, RTLA or other site-based data can be used to calculate statistics for an
accuracy assessment of the vegetation map. If a sufficient number of RTLA plots is available, then
the data may be used as both training data for the supervised classification of satellite imagery and the
subsequent accuracy assessment of the resultant map. The RTLA data must be split into two data sets
for the dual use of the data. Accuracy assessment procedures are described in detail by Senseman et
al. (1995).
338
An excellent general resource to application of remote-sensing is Bright and Getlein (2002).
This document provides a comprehensive overview of remote sensing applications for land
management.
7.9.3
Accuracy Assessment of Classified Vegetation and Imagery
It is very common for an installation to have a vegetation map produced from remotely sensed data.
Applying an image classification algorithm to the sensed data identifies vegetation types. Often this
map is used in plot allocation. For the information derived from this map to be useful in decision
making, its accuracy must be assessed. One way to achieve this is perform a site-specific error
analysis, which compares the remotely sensed data against a "true" map of the area, or reference map.
A reference map can be derived from sample data of the area. The RTLA plots were allocated using a
stratified random method, which is an appropriate sampling method for accuracy assessment.
7.9.3.1
Data Needs
The data necessary for the analysis consists of the plot number, vegetation type from remotely sensed
data image classification, and the plant community as determined from field surveys or plant
community classification of field data. Any valid plant community classification can be used.
An error matrix is then constructed using the data mentioned above. An error matrix is derived from a
comparison of a reference map to the classified map. Calculated plant communities represent the
reference map and form the columns. The classified data form the rows. The error matrix is shown in
Table 7-22.
Table 7-22. Error matrix for classification and reference data.
Reference Data
Classified Data
Dense Woodland
Open Woodland
Grassland
Sparse/Barren
Column Marginals
Dense
Woodland
30
3
0
0
33
Open
Woodland
0
27
0
0
27
Grassland
0
0
30
0
30
Sparse/
Barren
0
0
0
20
20
Row
Marginals
30
30
30
20
110
The row marginals in the error matrix are simply the sum of the row values and the column marginals
are the sum of the column values. The row marginals represent the number of plots in each classified
category. The values in each cell across a row represent the number of plots in the category that fall
into the reference data category. For example, the open woodland classified category contains 30
plots, 3 of which were classified as dense woodland using the plant community classification. The
remaining 27 plots were classified as open woodland.
7.9.3.2
Determining Sample Adequacy for Accuracy Assessment
The first step is to determine if there are a sufficient number of plots, reference points, for an overall
accuracy assessment of the classification. It has been shown that a minimum sample size of 20 per
class is required for 85 percent classification accuracy, while 30 observations per class are required
for 90 percent accuracy (Senseman et al. 1995). Notice in the table above, there are sufficient samples
for a 90 percent accuracy assessment for three of four categories. Because the sparse/barren category
339
only contains 20 plots, an 85 percent classification accuracy is used. The equation below computes
the ideal number of points to sample as reference points:
N = (4 (p) (q~)) / E2
where
N = total number points to be sampled
p = expected percent accuracy
q~ = 100 - p
E = allowable error.
For this example:
N = (4 (85) (15))/ 7.52 = 90.667 = 91 samples
The example in Table 7-22 has 110 total plots, which is sufficient for an overall accuracy assessment.
7.9.3.3
Percentage of Pixels Correctly Classified
This is one of the most commonly used measures of agreement and is easy to calculate. Simply divide
the number of points correctly classified by the total number of reference pixels. The equation is
shown below.
r
∑x
i =1
r
∑x
i =1
ii
i+
The numerator represents the number of points correctly classified. This value is calculated by
summing the diagonal entries from the error matrix. The diagonal values, from upper left to bottom
right represent the number of points correctly identified in the classified image as compared to the
reference data, calculated plant community. The denominator is the total number of reference pixels.
This is the sum of the row marginals or the total number of points.
From the error matrix above:
(30 + 27 + 30 + 20) / 110 = .972727 or 97.3 % of the points were correctly classified.
It is also possible to determine if the percent of correctly classified points exceeds a pre-determined
minimum classification accuracy. See Senseman et al. 1995 for further information.
7.9.3.4
Errors of Omission
Errors of omission refers to points in the reference map that were classified as something other than
their "known" or "accepted" category value. In other words, points of a known category were
excluded from that category due to classification error.
340
Errors of omission for each category are computed by dividing the sum of the incorrectly classified
pixels in the non-diagonal entries of that category column by the total number of pixels in that
category according to the reference map (i.e., the column marginal or column total). The values in the
non-diagonal cells represent points that were classified differently in the reference map compared to
the classified map.
Look down the column of values for dense woodland (Table 7-22). Notice the value 3 in the second
row under this column. This number represents the number of plots classified as dense woodland,
using the plant community classification, that were classified as open woodland on the classified
image. The cells in the third and forth rows contain zero. The sum of incorrectly classified points is
therefore 3. The value 30 in the first row represents the number of correctly classified points. The
error of omission for dense woodlands is computed as:
3 / 33 = .0909 or 9.1% error of omission.
The remaining values were calculated and are shown in Table 7-23.
7.9.3.5
Errors of Commission
Errors of commission refer to points in the classification map that were incorrectly classified and do
not belong in the category in which they were assigned according to the classification. In other words,
points in the classified image are included in categories in which they do not belong. Errors of
commission are calculated by dividing the sum of incorrectly classified points in the non-diagonal
entries of the category row by the total number of points in that category according to the classified
map (i.e., the row marginal or row total).
Read across the row for open woodland. Notice the 3 under the first column. This number represents
the number of plots classified as open woodland in the classified map that were classified as dense
woodland using the plant community classification. The value in the second column represents the
number of plots correctly classified and is not used here. The remaining values in the row are zero,
making the sum of incorrectly classified plots 3. The error of commission for open woodland is
computed as:
3 / 30 = .10 or 10 percent error of commission.
The remaining values were calculated and are shown in Table 7-23.
7.9.3.6
Kappa Coefficient of Agreement
The final measure of agreement discussed is the Kappa Coefficient of Agreement. The Kappa
Coefficient provides a measure of how much better the classification performed in comparison to the
probability of randomly assigning points to their correct categories. The equation for the Kappa
Coefficient of Agreement is:
r
)
k=
r
N ∑ x ii − ∑ ( xi + * x i + )
i =1
i =1
r
N 2 − ∑ ( x i + * x +1 )
i =1
341
where:
r = the number of rows in the error matrix
xii = the number of observation in row i and column i
xi+ = the marginal totals of row i
x+i = the marginal totals of column i
N = the total number of observations.
From the error matrix (Table 7-22), the Kappa Coefficient is calculated as:
(110 * 107) − ((33 * 33) + ( 27 * 27) + (30 * 30) + ( 20 * 20))
= .96133
12100 − ((33 * 30) + ( 27 * 30) + (30 * 30) + ( 20 * 20))
It is also possible to calculate a measure of agreement for each class by using the Conditional Kappa
Coefficient of Agreement. This is calculated as:
Ki =
( N )( pii ) − pi + p +i
( N )( pi + ) − pi + p +i
where:
Ki = Conditional Kappa Coefficient of Agreement for the ith category
N = the total number of observations
pii = the number of correct observations for the ith category
pi+ = the ith row marginal
p+i = the ith column marginal.
The Conditional Kappa Coefficient of Agreement for open woodland is:
(110)( 27) − ( 27 * 30)
= .8674
(110)(30) − ( 27 * 30)
Table 7-23. Summary table for accuracy assessment.
Category
Dense Woodland
Open Woodland
Grassland
Sparse/Barren
% Commission
0
10
0
0
% Omission
9.090909091
0
0
0
Conditional Kappa
1
0.86746988
1
1
Kappa Coefficient 0.96133333
Observed Correct Total Observed
107
110
% Observed Correct
0.972727273
By examining the measures of agreement in Table 7-23, it is concluded that the classification
performed well. 97.27 percent of the plots, or 107 out of 110, were classified correctly. Looking at the
values for each of the individual categories it can be stated that each performed well. The open
woodland category was the only one that had plots incorrectly identified in the classification. Three
342
of the open woodland plots were actually classified as dense woodland, using the plant community
classification, resulting in a 10 percent error of commission. This means these plots were included in
the classified category of open woodland when they do not belong there. Notice that the dense
woodland category has a 9.09 percent error of omission. This suggests that three plots were classified
as something other then their known or accepted category value. In other words, these plots were
excluded from dense woodland due to a classification error.
Looking at the Conditional Kappa for each category it is concluded that all categories, with the
exception of open woodland, were accurate. The open woodland was fairly accurate with a value of
.8674. The remaining categories were classified exactly correct.
7.10
Additional Analyses
7.10.1
Biodiversity Indices
7.10.1.1
Diversity as a Management Concern
Biological diversity or biodiversity can be defined as the diversity of genes, species, communities,
and ecosystems. Biodiversity is a simple general concept that rapidly becomes complex with attempts
at measurement and comparison. Each level of biodiversity has three components: compositional
diversity, structural diversity, and functional diversity. Compositional diversity is examined most
often.
Considerable evidence suggests that biodiversity is being lost at a rapid rate. Most management
approaches to minimize loss focus on species, often when an organism is nearing extinction. This
species approach, however, can be inefficient and expensive, often focusing on symptoms rather than
the underlying causes. Habitat management and protection is essential to species population stability
and survival. Therefore, a successful management program should attempt to maintain an array of
representative ecosystems. Also, both species and ecosystem-level approaches are necessary because
ecosystem classification systems are often not comprehensive enough to encompass every species.
By examining the spatial and temporal distribution of different ecological communities, managers can
evaluate the influence of management activities on community processes, species dispersal and
migration (e.g., habitat linkages and migration corridors), or loss of habitat and artificial habitat
fragmentation. For example, trends in herbaceous plant diversity can be examined following the
introduction, cessation, or change in grazing regimes or burning prescriptions. Impacts to woody
vegetation can be similarly examined following forest management activities. Additional examples of
how diversity analyses can be applied and integrated with resource management include how
neotropical migrant birds are affected by management activities over time and how land use activities
adversely impact endangered species habitat. Structural diversity (e.g., foliar height diversity), a
function of the number of vertical layers present and the abundance of vegetation within them, has
been strongly linked to bird species diversity in woodland environments (MacArthur and MacArthur
1961). Caution should be exercised when inferring cause and effect from only several years of data.
Long-term data may be required to reveal trends, especially in arid environments where year to year
variability can be high.
343
7.10.1.2
Using Monitoring Data to Evaluate Diversity
The application of monitoring data and selection of a diversity statistic can be difficult because the
measures of species richness, evenness, and diversity are themselves diverse and cannot be applied
universally. Different installations may prefer different measures because of the distribution of
habitats or relative abundance of species. The choice of statistic may also be influenced by the
methods chosen by other land management agencies in the region in order to facilitate comparison of
results.
Species diversity measures can be divided into three main categories. Species richness indices are a
measure of the number of species in a defined sampling unit. Secondly, species abundance models
describe the distribution of species abundance. The third group of indices (e.g. Shannon, Simpson) is
based on the proportional abundance of species and integrates richness and evenness into a single
number.
Table 7-24 summarizes the performance and characteristics of a range of diversity statistics showing
relative merits and shortcomings. The column headed “Discriminant ability” refers to the ability to
detect subtle differences between sites or samples. The column headed “Richness or evenness
dominance” shows whether an index is biased towards species richness, evenness, or dominance
(weighted toward abundance of commonest species).
Table 7-24. Performance and characteristics of diversity statistics. Reprinted from Magurran (1988).
Sensitivity
Richness or
Discriminant
to sample
evenness
Widely
Diversity Statistic
ability
size
dominance
Calculation
used?
α (log series)
Good
Low
Richness
Simple
Yes
λ (log normal)
Good
Moderate
Richness
Complex
No
Q statistic
Good
Low
Richness
Complex
No
S (species richness)
Good
High
Richness
Simple
Yes
Margalef index
Good
High
Richness
Simple
No
Shannon index (H’)
Moderate
Moderate
Richness
Intermediate
Yes
Brillouin index
Moderate
Moderate
Richness
Complex
No
Good
Moderate
Richness
Intermediate
No
Moderate
Low
Dominance
Intermediate
Yes
McIntosh U index
Simpson index
Berger-Parker index
Poor
Low
Dominance
Simple
No
Shannon evenness
Poor
Moderate
Evenness
Simple
No
Brillouin evenness
Poor
Moderate
Evenness
Complex
No
McIntosh D index
Poor
Moderate
Dominance
Simple
No
To calculate diversity statistics from RTLA plot data, plots should first be grouped by desired criteria.
The chosen diversity statistic is then calculated for each plot, and then averaged by group. To
optimally interpret patterns of diversity, plant life forms such as woody and herbaceous, trees and
shrubs, should be considered separately in diversity studies (Huston 1994). While some actual or
theoretical situations may cause commonly used diversity statistics to give contradictory results, for
most sample data from natural communities the values for all diversity statistics are highly correlated
(Huston 1994).
Following are some of the more common diversity statistic equations.
344
S = total number of species recorded
N = the total number of individuals summed for all S species (combined)
Margalef's diversity index (DMg)
DMg =
(S − 1)
ln N
Berger-Parker diversity index (d)
d=
N max
N
where:
Nmax = number of individuals for the most abundant species
To ensure the index increases with increasing diversity the reciprocal form of the measure is usually
adopted (1/d).
Simpson's index (D)
( − 1)
D = ∑ ni ni
N(N − 1 )
where:
ni = number of individuals in the ith species
To ensure the index increases with increasing diversity the reciprocal form of the measure is usually
adopted (1/D).
Shannon diversity index (H’)
H′= −
where:
∑ p (ln p )
i
i
pi = proportional abundance of the ith species (ni / N)
345
Shannon evenness (E)
E=
H′
ln S
where:
H' = Shannon diversity index
The following general guidelines for diversity analyses are provided by Magurran (1988) and
Southwood (1978):
(a) Ensure where possible that sample sizes are equal and large enough to be representative.
(b) Calculate the Margalef and Berger-Parker indices. These straightforward measures give a quick
measure of the species abundance and dominance components of diversity. Their ease of calculation
and interpretation is an important advantage.
(c) If one study is to be directly compared with another, the same diversity index should be used.
7.10.2
Similarity Coefficients
7.10.2.1
General Description
Similarity coefficients evaluate the relatedness of sites, communities, training areas, etc. Used in
rangeland ecology to compare a single site to a desired condition or status, the analysis can be
applied to any two sites, or groups of sites. There are a number of variations for the calculation of
similarity coefficients. The USDA Forest Service (1996) recommends the Sorensen coefficient
(Shimwell 1972):
2w
a+b
where a is the number of the constant (e.g., Cover-Frequency Index) of the first group, b is the
number of the constant of the second group, and w is the number both have in common (i.e., the
lowest of the two values for a specific condition). For example, if the point of separation between
similarity and the lack of similarity is at 65%, there is similarity for values from 65-100% and a lack
of similarity for values from 0-64%. However, the 65% point of separation is an arbitrary value and
professional judgment should be used increase or decrease of the cut-off value. If, for example, a
comparison is made between a pristine site and a utilized site, based on the presence of the same
dominant species, a cut-off of 60% may be more appropriate. Likewise, 70% may not be considered
similar because of the species present and the species desired for the site. An understanding of
composition and the ecology of a community type is necessary.
Another calculation for similarity is Jaccard's Coefficient of Similarity (Shimwell 1972):
346
w
× 100
a+b−w
While similar to Sorensen's coefficient, Jaccard's calculation tends to be a lower value, and 50% is the
general cut-off. Other indices of similarity are based on species presence, species dominance, and the
combination of species present. The quality of one index describing similarity over another is hard to
quantify.
7.10.2.2
Applicability
Similarity coefficients are used for comparing the degree of likeness. In general, cover and frequency
data are used in combination as the constant for comparison; however, other descriptors can be used.
7.10.2.3
Advantages and Limitations
Similarity coefficients are easy to calculate; however, an understanding of the groups being compared
is necessary to determine an applicable cut-off point. Also, there should be greater similarity within
than between the groups being compared (i.e., greater similarity within a plant community type than
between plant community types).
7.10.2.4
Example
The following example compares the community classification of five plots in a single training area
to five plots in a "pristine" area within the Mojave desert. Both groups of plots are representative of
the Ambrosia dumosa/Larrea tridentata (AMDU2/LATR2) vegetation type. The question is -- How
similar are plots in a disturbed area compared to an undisturbed, or control, area?
1. Compile a data set to test species similarity between two locations with the same vegetation
type. This can be two training areas with different uses. In the example, vegetation in a control
site (treatment 1) is compared to that in a training area (treatment 2).
2. Calculate canopy cover (%) (Table 7-25). Display canopy cover by species, plot, and treatment
(Table 7-26). Create a table displaying occurrence (presence or absence) of species by plot and
treatment (Table 7-27). The value will be 1 or 0.
3. Calculate frequency (%) (Table 7-28). Frequency is the proportion or percent of plots in which
a species occurs.
4. Calculate the average canopy cover (%) by species (Table 7-29).
5. Calculate Cover-Frequency Index (Table 7-30).
6. Determine similarity (Table 7-30).
7. Calculate Sorensen's and Jaccard's similarity indices (Table 7-30).
347
Canopy cover data are used in this example. Canopy cover consists of all plant life forms in a
community and provides an indication of species dominance. Cover-Frequency Index values are
typically the constant used in the description of similarity:
Cover-Frequency Index = Average % Canopy Cover x % Frequency
Table 7-25. Canopy cover data from 5 plots in Training Area B.
Data are used to determine the similarity of 5 plots in Training Area B to 5 plots in a control area adjacent to
the installation. All plots are representative of the Ambrosia dumosa/Larrea tridentata vegetation type. The
number of canopy intercepts/ point/ species and the total number of canopy intercepts/line are given. The
percent canopy cover for each species is shown as the number of intercepts for a species (a) divided by the total
number of intercepts of all species (b) x 100.
PlotID
Location
VegID
Count
Total Intercepts
% Canopy Cover
4
4
4
25
25
27
27
27
27
28
28
28
28
54
54
54
54
54
54
54
8
8
8
18
18
18
18
108
108
114
132
132
348
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
TrArea B
AMDU2
ERBO
LATR2
AMDU2
LATR2
AMDU2
BRTE
GRSP
LATR2
AMDU2
GRSP
LATR2
LYAN
AMDU2
CHPA12
EPNE
ERFAP
LATR2
SAME
THMO
AMDU2
ERIN4
LATR2
AMDU2
EULA5
GRSP
LATR2
AMDU2
LATR2
STSP3
LATR2
LYAN
28
5
8
8
5
5
12
4
26
9
1
6
1
9
7
1
3
5
14
4
1
1
3
2
2
3
8
1
2
1
2
1
41
41
41
13
13
47
47
47
47
17
17
17
17
43
43
43
43
43
43
43
5
5
5
15
15
15
15
3
3
1
3
3
68.3
12.2
19.5
61.5
38.5
10.6
25.5
8.5
55.3
52.9
5.9
35.3
5.9
20.9
16.3
2.3
7.0
11.6
32.6
9.3
20.0
20.0
60.0
13.3
13.3
20.0
53.3
33.3
66.7
100.0
66.7
33.3
Table 7-26. Species canopy cover (%) by plot in Ambrosia dumosa/Larrea tridentata vegetation types
in the control area and in Training Area B.
Percent Canopy Cover
Control Area
Species Code
AMDU2
BRTE
CHPA12
EPNE
ERBO
ERFAP
ERIN4
EULA5
GRSP
LATR2
LYAN
SAME
STSP3
THMO
plot 4
68.3
plot 25
61.5
plot 27
10.6
25.5
Training Area B
plot 28
52.9
plot 54
20.9
plot 8
20.0
plot 18
13.3
plot 108
33.3
13.3
20.0
53.3
66.7
plot 114
plot 132
16.3
2.3
12.2
7.0
20.0
19.5
38.5
8.5
55.3
5.9
35.3
5.9
11.6
60.0
66.7
33.3
32.6
100.0
9.3
Table 7-27. Presence of species on plots in the Ambrosia dumosa/Larrea tridentata vegetation types
in a control area and in Training Area B.
Occurrence
Control Plot #s
Training Area B Plot #s
Species Code
4
25
27
28
54
8
18
108
AMDU2
BRTE
CHPA12
EPNE
ERBO
ERFAP
ERIN4
EULA5
GRSP
LATR2
LYAN
SAME
STSP3
THMO
1
1
1
1
1
1
1
1
1
1
1
1
1
114
132
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
349
Table 7-28. Species frequency (%) in Ambrosia dumosa/Larrea tridentata vegetation types in a
control area and in Training Area B. Percent frequency is the number of plots where the species
occurred divided by the number of plots surveyed X 100.
Species Code
Percent Frequency
Control
Training Area
AMDU2
BRTE
CHPA12
100
20
20
60
0
0
EPNE
20
0
ERBO
ERFAP
ERIN4
EULA5
GRSP
LATR2
LYAN
SAME
STSP3
THMO
20
20
0
0
40
100
20
20
0
20
0
0
20
20
20
80
20
0
20
0
Table 7-29. Average percent cover of species in Ambrosia dumosa/Larrea tridentata vegetation
types in a control area and in Training Area B.
Species Code
AMDU2
BRTE
CHPA12
EPNE
ERBO
ERFAP
ERIN4
EULA5
GRSP
LATR2
LYAN
SAME
STSP3
THMO
350
Average Canopy Cover (%)
Control
Training Area
42.9
25.5
16.3
2.3
12.2
7.0
0.0
0.0
7.2
32.0
5.9
32.6
0.0
9.3
22.2
0.0
0.0
0.0
0.0
0.0
20.0
13.3
20.0
61.7
33.3
0.0
100.0
0.0
Table 7-30. Cover-Frequency Index (CFI) and similarity between the control area and Training Area B in
Ambrosia dumosa/Larrea tridentata vegetation types.
Sorensen's and Jaccard's similarity indices are shown. Cover-frequency index values are canopy cover
(Table 7-29) X frequency (Table 7-28). Similarity is the amount of commonness of the CFI values by
species (the lowest of the two values).
Control
a
Species Code
CFI
Training Area B
b
Similarity
w
AMDU2
4286.8
1333.3
1333.3
BRTE
510.6
0.0
0.0
CHPA12
325.6
0.0
0.0
EPNE
46.5
0.0
0.0
ERBO
243.9
0.0
0.0
ERFAP
139.5
0.0
0.0
ERIN4
0.0
400.0
0.0
EULA5
0.0
266.7
0.0
GRSP
287.9
400.0
0.0
LATR2
3204.3
4933.3
0.0
LYAN
117.6
666.7
0.0
SAME
651.2
0.0
0.0
STSP3
0.0
2000.0
0.0
THMO
186.0
0.0
0.0
10000.0
10000.0
1333.3
Sum
Sorensen
(2w/a+b)*100
13.3
Jaccard
(w/a+b-w)*100
7.1
Results. Each equation gives a different value. Whether there is dissimilarity or similarity depends on
the cut-off level. Sorensen uses 65% and Jaccard uses 50%. Both tests indicate dissimilarity between
the two vegetation types. That is, given the control area and Training Area B were similar prior to
military and other possible uses, they have changed with use.
7.10.3
Importance Values
7.10.3.1
General Description
Importance values are the summation of a number of measures describing characteristics of a species
on a plot, in a training area, in a community, or on an installation. A single measure (e.g., cover,
frequency, or density) may inadequately describe the role of a species. An importance value is a
comprehensive index, generally consisting of 1) relative frequency (the frequency of a species as a
percent of the total frequency value of all species within the sampling unit), plus 2) relative density
(the density of a species as a percent of the total density of all species), plus 3) relative dominance, or
cover, of a species (the cover of a species as a percent of the total area measured). Other measures can
be included in determining a importance values, such as production or volume.
351
7.10.3.2
Applicability
Importance values can be used as the sum or as the average of two or more descriptive characteristics
of species to describe the significance of a species to other species present.
7.10.3.3
Advantages and limitations
Care must be exercised when choosing the attributes used to calculate importance values. Importance
values can end-up being arbitrary and not truly descriptive of a species role. Combined measures
should be used critically (Greig-Smith 1983).
7.10.3.4
Example
The following example compares woody plant data collected on 210 transects. Data for the two
dominant species, Ambrosia dumosa and Larrea tridentata are shown and compared. Both species are
components of community types in the Sonoran desert. Ambrosia is noted for its density (Figure
7-23A) and Larrea for its visual dominance (Figure 7-23B). To calculate the importance values for
these two species:
1) Calculate species density: Density = number of individuals/sample area
2) Calculate relative density
Relative Density = (the density for a species / total density of all species) X 100
3) Calculate species frequency
Frequency = number of plots where a species occurs / number of plots sampled
352
A.
4000
Number of Plants
3500
3000
AMDU2
2500
LATR2
2000
1500
1000
500
0
0.5
1.0
1.5
2.0 3.0
Height Classes (m)
Volume (m3)
B.
1000
900
800
700
600
500
400
300
200
100
0
AMDU2
LATR2
0.5
1.0
1.5
Height Classes (m)
2.0 3.0
Figure 7-23. A. The distribution of Ambrosia dumosa and Larrea tridentata in the Sonoran desert. B.
The distribution of Ambrosia dumosa and Larrea tridentata volume.
4) Calculate the relative frequency
Relative Frequency = frequency of a species /total frequency of all species X 100
5) Calculate species dominance, in this case using volume 13
Dominance = volume of a species / area sampled
6) Calculate the relative dominance
Relative Dominance = (dominance of a species / total dominance of all species) x 100
13
In this example, plants are considered to be spheres. Therefore, volume was calculated as height X Πr2;
where r = 0.5 x height. (See Bonham 1989 for a detailed discussion on plant shapes and volumes).
353
7) Calculate the importance value (IV)
Relative Density + Relative Frequency + Relative Dominance / number of components (in this
case 3)
In the example shown below, the number of components is three.
Relative Density
Relative Frequency
Relative Dominance
Importance Value
(IV)
AMDU2
49.1
15.6
5.0
23.2
LATR2
33.1
11.2
43.2
29.2
Results -- Ambrosia is more abundant (Relative Density), occurs on more plots (Relative Frequency),
but is much smaller (Relative Dominance) than Larrea. Because of number and the greater
occurrence of Ambrosia, even with its smaller size, Ambrosia is only slightly less important than
Larrea.
7.11
Software for Statistical Analysis
There are a number of software packages available for statistical analysis. They are typically divided
into four basic categories: (1) spreadsheets, (2) spreadsheet add-ins, (3) pseudo-spreadsheet and
menu-driven packages, and (4) command-line, programmable packages. The relative advantages and
disadvantages of these four types of packages are shown in Table 7-31. This section provides a
summary of the different statistical packages that are currently available. Versions, features, and
prices change rapidly; this discussion does not represent an endorsement of any particular product.
354
Table 7-31. Relative advantages and disadvantages of statistical software package types.
Type of
package
Advantages
1. Easy to use.
Spreadsheet
Spreadsheet
Add-in
Pseudospreadsheets
1. Limited capabilities.
2. Short learning curve for implementing simple functions. 2. Difficult and tedious to implement for complex
data models.
3. Programming capabilities.
3. May be tedious to repeat complex analysis on
several datasets.
4. Will not do most types of multivariate analysis.
1. Fairly easy to use.
1. Somewhat limited capabilities.
2. Short learning curve for implementing simple functions. 2. Additional expense on top of spreadsheets.
3. Programming capabilities.
3. Will not do most types of multivariate analysis.
4. Fairly easy to repeat complex analyses on several
datasets.
1. Fairly easy to use.
2. Some programming capabilities.
3. Fairly easy to repeat complex analyses on several
datasets.
4. Will perform many types of multivariate analysis.
1. Very powerful.
2. Good documentation of statistical techniques.
Command line
Disadvantages
3. Will perform virtually all types of multivariate analysis.
1. Moderate learning curve.
2. Requires a more thorough understanding of
statistical techniques and theory.
1. Substantial learning curve involved
2. Requires a more thorough understanding of
statistical techniques and theory.
3. Can be expensive, but depends on the
package you choose.
4. Programmable with a flexible, very powerful
programming language.
7.11.1
Spreadsheets and Add-ins
The spreadsheet program most commonly used by military land managers is MS Excel. MS
Excel has a number of basic statistical capabilities built into program, such as descriptive
statistics, t-tests, correlation, regression, and analysis of variance (ANOVA) (Table 7-32). Addin packages can greatly expand the statistical capabilities of MS Excel. Examples of add-in
packages include WinStat, Analyse-It, and XLStat-Pro (Table 7-32).
355
Table 7-32. Statistical tests/functions for spreadsheets and spreadsheet add-ins packages.
Statistical tests/functions
Descriptive statistics
t-test
Correlation
Regression
Analysis of variance (ANOVA)
Bootstrapping
Canonical correlations
Cluster analysis
Correspondence analysis
Discriminant analysis
Factor analysis
Multiple comparisons
Nonparametric tests
Principal component analysis
Power analysis
Repeated measures analysis
Programmable level1
Spreadsheets
Spreadsheet add-ins
MS Excel
WinStat Analyse-It XLStat-Pro
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
3
3
x
3
x
x
x
x
x
x
x
x
x
x
x
x
3
Current version
2003
2001.1 not listed
7.5
The numbers 1-3 indicate three levels of programmability. A “1” indicates that the package uses its own
command line-type program language and highly complex and interactive programs can be written. Learning
and understanding the programming language is usually necessary to use the package effectively. A “2” means
that the package allows “scenarios” to be set that can be applied to other datasets with similar structure. This is
useful if repeating the same tests on a series of different datasets. A “3” indicates that the package has its own
program language available. However, the programming environment is quite different from that defined by the
number 1. Rather than being command line driven, code is associated with the cells within the spreadsheet or is
tied to buttons or forms that are inserted into the spreadsheet. An example is Microsoft Visual Basic for
Applications, which is used with MS Excel.
1
7.11.2
Command-Line and Pseudo-Spreadsheets
The term “command-line” refers more to the origin of a program than to how it is currently executed.
Packages such as SPSS, SAS, and S-Plus began as programs installed on mainframe computers
running UNIX or some other non-DOS or Windows operating system. These programs have evolved
and still can be run in a “command-line” environment. This means that a user can execute commands
from a command prompt. However, most of these packages also have various levels of interactivity
built into them (i.e., menu driven or graphical user interface, GUI).
For example, the SAS language allows one to insert commands and functions directly from a series of
help tools. SPSS now integrates the command line environment with a GUI that makes it more a
hybrid statistics tool. Capabilities of command-line packages are summarized in Table 7-33.
The term “pseudo-spreadsheets” refers to how the user organizes and views data; it does not
refer to how data are entered or edited. These packages allow a user to input or import data in
rows and columns and view data in a manner very similar to a spreadsheet. However, most do
356
not allow the user to manipulate the data with the same ease as spreadsheets. The data are not
technically stored in “cells” as they are in spreadsheets, so one cannot reference data in a specific
cell. Capabilities of pseudo-spreadsheets such as Systat, SigmaStat, SigmaPlot, JMP, and
Minitab are summarized in Table 7-33. SigmaStat can also be used as an add-in to SigmaPlot to
expand statistical capabilities. Further information about command-line and pseudo-spreadsheet
packages can be obtained from the company websites listed in
Table 7-34.
Table 7-33. Statistical tests/functions for command-line and pseudo-spreadsheet statistical software
packages.
Statistical tests/functions
Command-line
Pseudo-spreadsheets
SPSS SAS S-Plus Systat SigmaStat SigmaPlot JMP Minitab
Descriptive statistics
t-test
Correlation
Regression
Analysis of variance (ANOVA)
Bootstrapping
Canonical correlations
Cluster analysis
Correspondence analysis
Discriminant analysis
Factor analysis
Multiple comparisons
Nonparametric tests
Principal component analysis
Power analysis
Repeated measures analysis
Programmable level1
Current version (December
2004)
x
x
x
x
x
x
x
x
x
x
x
x
x
x
1
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
1
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
1
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
2
x
x
2
13.0
9.1
6.0
11.0
3.1
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
2
2
2
9.0
5.12
14.0
x
x
x
x
x
x
x
See Table 7-32 for explanation.
1
7.11.3
Graphics Capabilities
Graphical capabilities are available for all of the aforementioned packages. Some are highly
interactive (e.g., MS Excel, SigmaPlot) whereas others are extremely powerful but not very userfriendly (e.g., SAS). MS Excel meets most users needs for producing a variety of graphs.
Furthermore, the output of many statistical packages is easily imported into an MS Excel spreadsheet.
SigmaPlot offers more graph types and greater flexibility than MS Excel and is more oriented towards
science applications.
357
Table 7-34. Website addresses for further information on statistical packages discussed in this
section.
Package
Spreadsheets
MS Excel
Further information
http://office.microsoft.com/en-us/FX010858001033.aspx
Spreadsheet add-ins
WinStat
Analyse-It
XLStat-Pro
http://www.winstat.com/
http://www.analyse-it.com/
http://www.xlstat.com/
Command-line
SPSS
SAS
S-Plus
http://www.spss.com/spss/
http://www.sas.com/
http://www.insightful.com/
Pseudo-spreadsheets
Systat
SigmaStat
SigmaPlot
JMP
Minitab
7.11.4
http://www.systat.com/products/Systat/
http://www.systat.com/products/SigmaStat/
http://www.systat.com/products/SigmaPlot/
http://www.jmp.com/
http://www.minitab.com/
Selecting a Package
The selection of a statistical package should be based on specific applications and needs. Most
companies offer free trial versions of software, allowing a user to evaluate a package relative to their
specific needs. The cost for software licenses varies by product, and can also depend on the buying
programs in place at different installations. Check with your purchasing group to see if pre-negotiated
discounts exist for any of the packages.
Users interested in programming and advanced statistics should consider any of the three command
line systems. Users are most likely to achieve success with advanced packages if they have access to
good customer support or an experienced users-group. In the absence of any local users, you may
want to base your choice on the quality of technical support and documentation provided by the
company.
For less-advanced users, any of the pseudo-spreadsheet packages should prove adequate. When a
number of packages have similar features, the preferred software is often the one the user is most
familiar and has the best technical support. Finally, if descriptive statistics, t- tests, ANOVA, and
regression are the principal uses, MS Excel with or without add-inns should suffice.
7.11.5
Stand-Alone Sample Size and Power Analysis Software
Sample size adequacy and power analysis are important when planning or evaluating a monitoring
program. Sample sizes, statistical power, and minimum detectable change sizes can be calculated
using tables, charts, calculators, and spreadsheets (see Chapter 3 Introduction to Sampling. However,
computer software may improve the accuracy and ease of these calculations. Sample size calculators
358
and power analysis has been incorporated into several of the statistical packages previously discussed
(Table 7-33). There are also several stand-alone packages available for conducting power analysis
(Table 7-35).
Thomas and Krebs (1997) provide a critical review of stand-alone statistical power analysis software
(http://www.zoology.ubc.ca/~krebs/power.html). For beginner to intermediate users, they recommend
one of the commercial general purpose power packages such as Nquery Advisor or PASS (Table
7-35). Additional commercial packages include Power and Precision, Ex-Sample, and SPSS
SamplePower (Table 7-35). The authors concluded that most general purpose statistical programs
reviewed were inadequate in one or more respects. A number of freeware and shareware programs are
also available, including G*Power, DSTPLAN, and MONITOR (population trend analysis). Elzinga
et al. (1998) reviewed a number software programs and recommended DSTPLAN (freeware) and PC
SIZE: Consultant (shareware). G*Power was not recommended for vegetation monitoring
applications due to limited software documentation and a high level of assumed knowledge (Elzinga
et al. 1998). A statistical power capability is included in SYSTAT software.
Table 7-35. Sample size and statistical power software packages and website addresses for further
information.
Package
Current version
Further information
nQuery Advisor
5.0
http://www.statsol.ie/nquery/nquery.htm
PASS
2002
http://www.ncss.com/pass.html
Power and Precision
2.0
http://www.power-analysis.com/
Ex-Sample
Not listed
http://www.ideaworks.com/MToolchest.shtml
SPSS SamplePower
2.0
http://www.spss.com/samplepower/
G*Power
Not listed
http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/
DSTPLAN
Not listed
http://charlotte.at.northwestern.edu/dce/bull/dstplan.html
MONITOR
7.0
http://www.mbr-pwrc.usgs.gov/software/monitor.html
A webpage with power analysis information and on-line calculators is found at:
http://members.aol.com/johnp71/javastat.html#Power.
7.12
Data Analysis using MS Access, MS Excel, Systat, and ArcGIS
The technical reference document “Introduction to Data Analysis using MS Access, MS Excel,
Systat, and ArcGIS” describes steps for analyzing and displaying RTLA data (Bern 2004). The
document describes some simple queries and data analyses that can be performed in MS Access, as
well as how to move the data to MS Excel for further analysis. In Excel, the use of pivot tables is
explained, along with some basic statistical analyses such as descriptive statistics and ANOVA.
Procedures to conduct the same analyses in Systat are also described, as are steps to conduct repeated
measures analysis. Finally, steps for exporting, displaying, and analyzing data in ArcGIS are
described. The document is available at: http://www.cemml.colostate.edu/cemmlpub.htm.
359
7.13
7.13.1
Guidelines for Reporting Monitoring Results
Purpose and Types of Reports
The purpose of a report is to present information and facts, examine relationships, and present
conclusions based on the information analyzed. A report provides readers with background
information and an understanding of the state of the natural resources on an installation. In some
cases, this could be the first comprehensive document describing or summarizing an installation's
natural resources. Some specific installation questions cannot be directly addressed by the data set or
other available resources. However, working with the data will help recognize additional data needs.
Conclusions made should deduced from the results and data presented in the report.
The types of reports that precede the one to be written may define the level of detail. An initial report
may contain much more detail than subsequent reports. Once a foundation is established, only new
information should be addressed. The intent is to refresh memories and to direct interested parties to
earlier reports for more detailed information. An initial report may inform the geologist who finds the
occurrence of an endangered plant on a specific geological formation interesting, or the trainer who is
curious about expanding training activities into an under-utilized area. If trend analysis is important,
then it is essential to present results over the available time period.
A report addresses a problem or question in specific terms. Essential facts are classified and organized
with an emphasis on processes, causes, and results. The facts are evaluated and interpreted. Some
reports may emphasize recommendations which serve as the basis for future action.
Recommendations should exclude personal opinion, interest, and bias (Jones 1976). In many
situations, a report is used as a tool to help managers and funding agencies assess the success of a
project or program (Pneena 1986). Reports should therefore address management and monitoring
objectives upon which data collection and analysis are based.
The type of question and the audience defines a report's structure and content. Typically, a report that
will be reviewed and used by colleagues is more detailed than one required by those at higher
organizational levels. Regardless of the depth of analysis and presentation, all statements must meet
scientific evaluation standards. Report writing requires a number of tasks that contribute to the overall
product. For example, the following tasks are typically associated with producing a comprehensive
monitoring report, addressing a variety of issues and including both qualitative and quantitative data
summaries:
Assemble Information: this task may require inventories of available background information
including descriptive (i.e., plans, reports) and spatial (i.e., map and GIS) data. A literature search
using library and installation documents is essential to the validity of the report by providing
additional substance and building on the work of others. A literature search should provide important
background and reference information regarding ecology, monitoring approaches, and management
issues.
Data Preparation: requires the assembly, organization, and evaluation of existing data to be
analyzed in the report. Missing or invalid data should be found or corrected, respectively. Data may
have to be reformatted depending on the analysis tools. This is sometimes a very time-consuming
effort, depending on quality control and data management efforts up to that point. Report preparation
and writing requires the database be complete, organized, and in condition for general use.
360
Data Summary and Statistical Analysis: Once the objectives of the report are determined, data can
be summarized and analyzed. Analysis typically involves extracting summaries from raw data and
subsequent analysis.
Prepare Graphics and Write: Once summaries are generated, graphics can be developed for the
report. Graphics are often prepared as a framework for the presentation and discussion of results.
Submit to Others for Review and Perform Final Edits: The draft report should be reviewed by
both members of the intended audience and others who have a good understanding of the subject
matter. Once edits are completed, the report is ready for reproduction, binding, and distribution.
In addition to comprehensive reports that contain details of all project aspects, less detailed reports
can be prepared for other audiences. For example, a Directorate of Plans and Training (DPT) or
Training and Range (Trainers’ Report) may consist of reduced version of comprehensive report. A
trainers report should contain information regarding objectives, an implementation summary, a
summary of results, graphics, and maps (data values and corresponding red, amber, green) of results
relative to training objectives. Training-related objectives might include training support,
quality/realism, training constraints, sustainability, and planning. Furthermore, a command briefing
intended for briefing upper level trainers (e.g., DPT) and installation commanders and staff can also
be an important way to disseminate results. A command briefing should be a concise presentation
(e.g., PowerPoint) of implementation and results relative to all program areas and objectives.
7.13.2
Generic Report Organization
Reports generally follow scientific writing protocols and include an introduction, methods and
materials, results and discussion, conclusions, and recommendations. Report writing requires both
structured analysis and creativity.
7.13.2.1
Preliminary Sections
Title page, table of contents, list of tables, list of figures, and funding source make up the preliminary
pages of a report. Additional pages may include acknowledgments, executive summary, abstract, or
preface. These pages are numbered with lower case Roman numerals. These initial pages allow the
reader to find specific information and quickly achieve a sense for the organization, rationale, and
findings of the report.
The title page includes the title of the report, authorship and affiliation, for whom the report was
prepared, and the date. The title should be concise and informative. Authorship is typically given to
those individuals who have substantially contributed to the report. Individuals who have advised or
given technical assistance as part of their normal duties are not included as authors. All authors
should review and approve the final draft (O'Connor and Woodford 1975, National Bureau of
Standards 1980).
The table of contents typically includes primary and secondary section headings. Additional levels of
organization can be included if so desired. Descriptive headings will lead readers to sections of
interest. A list of figures and tables should follow the section locations. Concise and informative titles
to figures and tables are helpful to readers.
361
The executive summary is a condensed version of the report. Limited to one or two pages, an
executive summary presents the rationale for the report, the findings, conclusions, and
recommendations in a non technical way (Shelton 1994). An abstract is similar to an executive
summary but is generally shorter, often a single paragraph. Acknowledgments recognize individuals
or groups who have made a significant contribution to the report, but who cannot be regarded as
authors.
7.13.2.2
Introduction
The introduction sets the stage for the report. The background, justification, and the scope of the
project or program are described. The project goals and objectives are presented in a clear, concise
manner. The introduction may include a general description of the site location and an installation's
mission. Background information relevant to the ecology and history of the project area aids in
justifying methods and objectives. Land uses unique to the installation may be described, especially if
they affect the interpretation of results.
7.13.2.3
Study Area or Site Description
A summary of an installation's natural resources should include geologic development and features,
soils, climate, vegetation, wildlife, physiography, hydrography, and special concerns. Any attribute
reviewed should aid the project information that will be presented later in the report.
Natural Resources -- The natural resources are the setting upon which an installation's mission(s)
takes place. Often mission is defined by the resources available. How these resources are used and
preserved are an important part of mission continuation. A description of the natural resources
provides a framework to anchor the field data.
Geological Development and Features -- The geology of an area affects the types of soils present as
well as the associated vegetation. A description of the geology can outline a number of physical
limitations to training. Geology can also help explain why certain areas are more heavily used, and
why other sites should not be used.
Soils -- An installation's soils are related to geology, landform, relief, climate, and the vegetation
(Cochran 1992). The information identifies the potential constraints to training, vegetation patterns,
and the likelihood of erosion problems. Not all installations have specific soil information available.
For installations with completed soil surveys, the Natural Resources Conservation Service (NRCS)
can provide needed information. In addition, the NRCS can provide the R-value used in calculating
soil erosion potential with the Universal Soil Loss Equation.
Climate -- Many installations have weather stations on site. Often these stations are at airfields. Data
are available from the National Climatic Data Center, Asheville, North Carolina, for established
stations. These data are also available on CDROM from EarthInfo, Inc, Boulder, Colorado
(http://www.earthinfo.com/). Compact disks (CDs) containing these data may be available from
university libraries. Some data are available on the world wide web. For example, see: On Line
Climate Data page; http://www.ncdc.noaa.gov/ol/climate/climatedata.html.
Vegetation -- Putting the vegetation into a historical framework helps readers understand the
associations present and responses to natural and human-induced events. Severe disturbances may
simplify or add to the vegetative complexity. In a historical context, the sensitivity, resistance, and
362
resilience of the vegetation may become more apparent. This information may help in planning
prescribed burns, forestry practices, or agricultural leases.
Wildlife -- As with vegetation, a brief background of the wildlife present is helpful to the reader.
Physiography -- Landform, topographic position, and aspect can be important determinants of plant
communities. This information may be derived in part from elevation layers using a GIS. Field data
collection and soil surveys also provide information.
Hydrography -- Streams and water body locations are available from digital and paper maps. Stream
flow data are usually maintained by universities and/or state and federal agencies, and are often
available via the USGS, http://water.usgs.gov/.
Areas of Special Interest or Concern -- Areas of special interest may be descriptions or locations of
species of concern, natural features or plant communities, wetlands, soils, etc. A brief description of
the areas, their importance, and ongoing measures for protection help the reader understand the
natural resources of an installation.
7.13.2.4
Methods
The methodology of a project should be explained with enough detail for the project to be repeated,
achieving similar results. Detailed methodologies can be presented in other documents such as a
monitoring protocol. Any modifications or methods unique to the project need to be detailed. General
treatments and sample size(s) must be provided, as well as the duration and time of fieldwork.
Statistical analyses and justifications can be addressed in detail in this section and then mentioned in
the results.
7.13.2.5
Results and Discussion
The results and the discussion can be either one or two sections. Placing them together is easier to
write and easier on the reader, especially if the study is complex. If the two sections are separate, be
sure not to discuss results in the results section and to address all of the results in the discussion
section. Make the results section comprehensible and coherent on its own. Describe the purpose, the
significance, and the relevance of the information, but do not discuss the results extensively. Refer to
tables and figures to illustrate the findings and support conclusions. Excessive description of data
already presented in graphics should be avoided. However, no table or figure should be included that
is not directly cited and relevant to the discussion. Descriptions of and references to tables and figures
should be straightforward and explicit.
Develop the discussion in the same order as the results were presented. The discussion is an
elaboration and an assessment of the results section. The results are related to previous studies and the
implications of the results are discussed. Do not conceal negative results and discrepancies. Instead,
try to explain, or admit your inability to do so (O'Connor and Woodford 1975).
7.13.2.6
Conclusions and Recommendations
The conclusions tie the objective(s) to the results and the discussion. The emphasis is on what was
found in light of the stated objectives of the report. The conclusions re-address the important findings
of the project. If someone were to only read the introduction and the conclusion, they should have a
good idea of the contents of the report. Recommendations are based on the technical evidence and
363
the author's professional expertise (Shelton 1994). When data and expertise do not answer the
objectives of a project, recommendations should address alternative methods or approaches.
7.13.2.7
Literature Cited and Bibliographic References
All published works cited must be referenced. Unpublished works, obscure documents, and personal
communications are not included in the literature cited, but should be referenced in the text or placed
in a footnote if more detailed information is necessary. The format of the Literature Cited section
should be consistent and have a logical organization.
7.13.3
Style and Format
All communication is imperfect, because the ability to understand information depends on both the
sender and the receiver (Pneena 1986). Style is a subtle method of encouraging someone to read the
written word. Style includes everything from page layout to grammar to choice of words. Style is
largely a personal characteristic but authors should always strive for clarity, conciseness, and
consistency.
Vigorous writing is concise. A sentence should contain no unnecessary words, a paragraph no
unnecessary sentences, for the same reason that a drawing should contain no unnecessary lines and a
machine no unnecessary parts… -- (Strunk and White 1979)
The format of a document is an invitation to read the document or put it down. Pick a font, type size,
and line spacing that make reading easy. Use descriptive headings to help readers locate specific
topics and to break up the page. Illustrations provide information in an alternative format and give the
reader a momentary diversion. Bulleted items alter the overall look to a page and provide a change in
rhythm. Do not, however, add so many distractions and increase margins and line spacing to make the
reader wonder why so much paper is being used. Additional writing references include Sabin (1993)
and Style Manual Committee (1994).
7.13.3.1
Tips for Effective Writing
Effective writing is a skill learned over time, which comes naturally to some, but is laborious to most.
Some writing critiques and editorial revisions are based on style differences and not grammatical
problems. Some errors seen by others would be evident to the writer if there was time to put the
document aside for awhile. Pneena (1986) presents some simple writing techniques to keep in mind
while writing and proof reading a report:
1. Avoid long, convoluted sentences and paragraphs.
2. Avoid jargon.
3. Limit the use of prepositions - the number of prepositional phrases in a sentence makes reading
and understanding laborious.
4. Whenever possible, use precise writing in favor of comfortable writing. Examples include:
364
COMFORTABLE WRITING
was low in frequency
PRECISE WRITING
less frequent
was of greater importance
was more important
in order to
to
the preparation of reports
preparing reports
the targeting of erosion
targeting erosion
to be used in place of
to replace
are of importance
are important
5.
Limit the use of the word 'and' to join phrases. Instead:
• Use short sentences.
• Use other connectives (e.g., thus, often, in addition).
• Connected items must parallel each other (i.e., the same structural or grammatical form
for all parts of a series -- preparing documentation, reporting findings, and addressing
goals).
6.
Limit clutter words and crutch phrases (e.g., there are, it is apparent that, it is important to
note, etc.).
7.
Limit repeating expressions (e.g., Invasive species abundance in Training Area 2A
increased by 12%, which was a much larger increase in invasive species abundance
compared to Training Area 2B.)
8.
Limit unnecessary adjectives and adverbs (e.g., it seems, likelihood, sufficiently,
frequently, etc.).
9.
Use ordinary words.
10.
Limit verbs such as is and occurred.
11.
Conduct which hunts. Which is used to introduce nonessential clauses: “Increased
training loads, which were forecast in the Restationing EIS, have not been accompanied
by habitat loss.” That ordinarily introduces essential clauses: “The report that we
prepared for the Colonel was helpful.”
12.
Follow which hunts by searches for the words that, it, and of. Sometimes these are the
best words, but often they are not.
13.
Use the active verb tense rather than the passive. Using the passive verb tense is common
and accepted in scientific writing, but should be minimized.
365
14.
7.13.4
Put related but non-essential information into an appendix. Information contained in
appendices should be cited in the text; otherwise, another document may be a more
appropriate place for the information. Do not put excessive raw data in an appendix
unless it is useful.
Tables
Every table must have a purpose and convey a message (Table 7-36). A table should be able to stand
alone from the text. The table title should be descriptive. If a number of tables contain similar
information, the legend of the first should be inclusive of all necessary information. In subsequent
tables, the previous table can be cited for the specifics.
A table's format needs to be logical. If the amount of data is voluminous, use a summary table in the
text and place the detailed table in an appendix. Put control or baseline values near the beginning of
the table. Columns with comparative data should be next to each other.
The number of significant figures (i.e., decimal places) should be consistent and indicative of the
level of precision used in measurement. When possible, align numbers using decimal points. The
number of samples, the standard error or standard deviation of the mean, the probability, and the type
of statistical analysis should be stated.
9 Able to stand alone
9
9
9
9
9
9
Format logical
Limited amount of information
Note sample size
Measure of statistical variability shown
Statistical test noted
Terms explained
Table 7-36. Example of tabular data that includes essential components.
366
7.13.5
Graphics
A graphic, like a table, needs to be interpreted independent from the text. Most graphics are titled as
Figures in a report. Types of graphics include maps; histograms, line graphs, pie charts, and other
representations of qualitative or quantitative information; flow charts; organizational charts,
illustrations, and photographs. The title or figure caption should be informative and contain
information to explain the graphic. If a number of similar graphics occur together, the legend of the
first graphic should be complete to avoid unnecessary replication. Subsequent legends can then refer
to the earlier figure for specific information. Examples of properly labeled and captioned graphics are
presented in Figure 7-24.
All axes must be labeled and measurement units displayed if appropriate. Standard errors or standard
deviations of the mean, sample size, and the type of statistical test used should be included. Symbols
and lettering must be defined. Also, do not extrapolate beyond sample data without an explanation or
a caution to readers.
Although color figures can add to the cost of report reproduction, they can significantly enhance the
appearance of results. If color figures are used, choose colors that are pleasing and reproduce well in
black and white. Size figures so the information is clear and easily discerned. Most figures can be
presented in the text instead of using an entire page.
Zyzzyc River Sub-Watershed
Live Fire Training Area
9 Title describes graphic
Range 12
Zy
9 Legend includes all boundaries
and symbols
zz
yc
R.
D ig
g in
gS
it
e
9 The graphic can stand alone
Pond
Water Detention Basin
Watershed Boundary
Ra
n ge
21
North
Direction of Water Flow
Boundary of Training Area
0.0
0.5
1.0
2.0 km
Figure 3. Water flow in the Zyzzyc River Sub-Watershed located in the Live Fire Training Area of Camp USA.
9 The graphic is understandable
in black and white
9 Map figures should include a
North arrow and scale bar
25
20
Percents
Legend
15
10
5
0
6% Cryptogams
Grass
Forb
Shrub & Trees
Categories
18% Plants
17% Bare Ground
1% Rock
58% Litter
Figure 13. Ground cover by category on 110 plots at Camp USA. Means and their standard errors shown
for plant cover types.
367
9 Title describes graphic
9 All symbols are identified
9 The colors are distinguishable in black and white
9 The title legend includes the type of statistical test used
Figure 7-24. Examples of graphics showing necessary components.
7.13.6
Brief Format
As stated above, professional colleagues tend to expect more detail and a higher level of reporting
compared to inexperienced or non-technical staff. A comprehensive report, reviewing a number of
years of data, can consist of over 100 pages of text, and with appendices, a report can approach 150
pages. While the complexity of a report is necessary, a brief review of the contents is often
appreciated. One method is to develop a computer presentation. The objectives as they pertain to a
specific audience can be highlighted. The methods, data, results, discussion, and conclusions are
developed for the intended audience, and illustrated as bulleted items. Photographs illustrating the
data, rather than tables and charts provide a more dynamic document and a document supported by
scientific protocols. Results can be presented in an abbreviated format (Figure 7-25) or condensed to
the level of information common to presentation slides (Figure 7-26).
368
Figure 7-25. Examples of abbreviated reporting approach.
369
DISTURBANCE
• Note tracking
evidence
• Tracking increased
43% between surveys
• Heavily tracked
areas showed signs of
recovery when training
was withheld
• Soil erosion rates
increased 15%
Figure 7-26. Example of information slide prepared for oral presentation.
7.13.7
Suggested Range and Training Land Assessment (RTLA) Report
Outlines for Various Audiences
Suggested outlines are presented below for three different readers or audiences.
1) ITAM and Natural Resources (Comprehensive Technical Report)
•
detailed summary of objectives, approaches, accomplishments, and results (data values,
graphics, maps) relative to all objectives.
2) DPT Training and Range (Trainers’ Report)
•
objectives, implementation summary, summary of results, graphics, maps (data values and
corresponding red, amber, green) of results relative to training objectives (training support,
quality/realism, training constraints, sustainability, and planning).
•
reduced version of comprehensive report with minimal additional preparation.
3) Command Briefing
•
intended for briefing upper level trainers (e.g., DPTM) and installation commanders and staff
•
concise presentation (e.g., PowerPoint) of implementation and results relative to all program
areas and objectives.
370
7.13.7.1
Outline for Comprehensive Technical RTLA Report (30-50 pages)
TABLE OF CONTENTS
INTRODUCTION
Background, scope, rationale
RTLA organization
General RTLA objectives
(include program history in initial report)
IMPLEMENTATION SUMMARY
Level of effort, resourcing, execution
GENERAL SITE DESCRIPTION (brief)
Geographic Location and Size
Installation Land Use History
Installation Mission
Other Land Uses
PHYSICAL and BIOLOGICAL ENVIRONMENT (very brief unless more detail is essential –
reference other documents such as INRMPs for more lengthy descriptions)
Climate
Geology
Topography and Hydrography
Soils
Flora
Fauna
TYPES AND IMPACTS OF TRAINING (emphasis depends on program objectives)
Military Units and Usage - Training Loads
Unit types, frequency and duration of use, training distribution and annual cycles
Impacts of Training
OTHER MANAGEMENT ISSUES AND CONCERNS
Species/communities of concern
Weeds
Fire
Soil erosion
Other
SPECIFIC RTLA PROGRAM OBJECTIVES
METHODS - For each method, describe why the method was selected (e.g. advantages over other
methods), the population/area of interest, sampling design, and work accomplished. Methods could be
nested with the section below so each objective would be stand-alone. This would be appropriate if
different methods were employed to address different objectives.
Method A Description
Method B Description
371
Method C Description
Method D Description
Method … Description
RESULTS AND DISCUSSION- Organize by RTLA Program Area or Monitoring Objective
Monitoring Objective A
Data analysis approach
Sources of Variability
Tabular, graphic, and/or spatial presentation of results
Discuss results relative to monitoring objectives
Management implications of results
Assessment of approach/method relative to objectives
Monitoring Objective B
Data analysis approach
Sources of Variability
Tabular, graphic, and/or spatial presentation of results
Discuss results relative to monitoring objectives
Management implications of results
Assessment of approach/method relative to objectives
Monitoring Objective …. Z
Data analysis approach
Sources of Variability
Tabular, graphic, and/or spatial presentation of results
Discuss results relative to monitoring objectives
Management implications of results
Assessment of approach/method relative to objectives
OTHER APPLICATIONS (NR, training, ATTACC, etc.) or PRODUCTS
CONCLUSIONS AND RECOMMENDATIONS
ACKNOWLEDGEMENTS
Recognize individuals that directly or indirectly support RTLA efforts but who are not co-authors of
the report
LITERATURE CITED
APPENDICES
Keep to a minimum or include as electronic documents only. Other documents should be referenced
when possible.
Tabular data summaries (more detailed than in text), where appropriate
7.13.7.2
Outline for Training and Range RTLA Report (5-10 pages)
INTRODUCTION (include in all reports)
RTLA background, scope, rationale
372
RTLA implementation history for the installation/state
General RTLA program objectives
RTLA organization
TYPES AND IMPACTS OF TRAINING (emphasis relative to objectives (next section))
Military Units and Usage - Training Loads
Unit types, frequency and duration of use, training distribution and cycles
Impacts of Training
RTLA OBJECTIVES RELATED TO RANGES AND TRAINING
RESULTS AND DISCUSSION- Organize by Monitoring Objective
Monitoring Objective A
Implementation summary
Tabular, graphic, and/or spatial presentation of results, with addition of red, amber, green
summary
Discuss results relative to monitoring objectives
Implications of results to range management and training
Monitoring Objective B
Implementation summary
Tabular, graphic, and/or spatial presentation of results, with addition of red, amber, green
summary
Discuss results relative to monitoring objectives
Implications of results to range management and training
Monitoring Objective ….
Implementation summary
Tabular, graphic, and/or spatial presentation of results, with addition of red, amber, green
summary
Discuss results relative to monitoring objectives
Implications of results to range management and training
OTHER POTENTIAL TRAINING APPLICATIONS
CONCLUSIONS and RECOMMENDATIONS
ACKNOWLEDGEMENTS
Recognize individuals that directly or indirectly support RTLA efforts but who are not co-authors of
the report
LITERATURE CITED
7.13.7.3
Outline for Command RTLA Briefing
This brief should no more than 10-20 minutes in length, using one-two slides per bullet/sub-bullet
listed below. The information can be excerpted and condensed from the comprehensive technical
report or training and range report.
373
• RTLA Purpose and Scope
• Installation RTLA Objectives
• Implementation Summary for Current Year (or bring up to speed from initial
implementation or previous command brief)
• Program Costs and Execution
• Results and Management Implications:
o
Program Area/Objective A
o
Program Area/Objective B
o
Program Area/Objective ….
• Other Applications or Products
• Dissemination of Results
7.13.8
Additional Guidelines
Reporting of monitoring results should be done in a timely and efficient manner. Data should be
converted to electronic format, if appropriate, summarized, and evaluated before the beginning of the
next data collection cycle. To ensure continued program support, monitoring programs must generate
reports that are useful, address specific management concerns or issues, and widely applicable. The
precision of the data should be known and specifically stated in summaries and reports. Lastly,
reports should be prepared and distributed on a regular basis using a format that is straightforward
and appropriate to the user community, including range operations personnel/military trainers, land
managers, and public land agencies (e.g., where BLM, National Forest, and State lands are used for
training).
7.14 References
ACITS 1997. Repeated Measures ANOVA Using SAS PROC GLM. Stat-40, The University of
Texas at Austin. http://www.utexas.edu/cc/docs/stat40.html.
Bern, C.M. 2004. Introduction to Data Analysis using MS Access, MS Excel, Systat, and ArcGIS.
CEMML TPS 04-14. Center for Environmental Management of Military Lands, Colorado State
University, Fort Collins, CO. 30 pp.
Bonham, C.D. 1989. Measurements for Terrestrial Vegetation. John Wiley, New York.
Bright, T.A. and S. Getlein. 2002. Remote Sensing Users’ Guide, Version 2.5. SFIM-AEC-EQ-TR200053. U.S. Army Environmental Center and U.S. Army Corps of Engineers. Available online at:
http://www.fas.org/irp/imint/docs/remote.pdf.
Cochran, C.C. 1992. Soil Survey of the U.S. Army Yuma Proving Ground, Arizona -- Parts of La Paz
and Yuma Counties. USDA Soil Conservation Service.
Cochran, W.G. 1977. Sampling Techniques, 3rd Ed. John Wiley and Sons, New York.
D'Agostino, R.B. and M.A. Stephens. 1986. Goodness-of-fit techniques. M. Dekker, New York. 560
pp.
374
Devore J. and R. Peck. 1986. Statistics, the exploration and analysis of data. 1983. West Publishing
Company, St. Paul, MN. 699 pp.
Elzinga, C.L., D.W. Salzer, and J.W. Willoughby. 1998. Measuring and Monitoring Plant
Populations. BLM Technical Reference 1730-1. USDI Bureau of Land Management, National
Applied Resource Sciences Center, Denver, CO.
Fowler, N. 1990. The 10 most common statistical errors. Bulletin of the Ecological Society of
America 71: 161-164.
GraphPad Software, Inc. 1998. InStat guide to choosing and interpreting statistical tests.
http://www.graphpad.com/instatman/instat3.htm. San Diego, CA. Accessed 16 November 1998
Green, R.H. 1979. Sampling Design and Statistical Methods for Environmental Biologists. John
Wiley and Sons, New York. 257 pp.
Greig-Smith, P. 1983. Quantitative Plant Ecology. Third Edition. Blackwell Scientific, London. 359
pp.
Hill, M.O. and H.G. Gauch Jr. 1980. Detrended correspondence analysis: an improved ordination
technique. Vegetatio 42: 47-58.
Hill, M.O. 1979. TWINSPAN - A FORTRAN program for arranging multivariate data in an ordered
two-way table by classification of the individuals and attributes. Ecology and Systematics, Cornell
University, Ithaca.
Hinds, W.T. 1984. Towards monitoring of long-term trends in terrestrial ecosystems. Environmental
Conservation 11(1): 11-18.
Huck, S.W. 2000. Reading Statistics and Research. 3rd edition. Adison Wesley Longman, Inc., New
York, NY. 688 pp.
Huston, M.A. 1994. Biological Diversity: The coexistence of species of changing landscapes.
Cambridge University Press, Cambridge.
Jones, W.P. 1976. Writing Scientific Papers and Reports, 7th Edition. William Brown Publishers,
Dubuque, Iowa. 366 pp.
Jongman, R.H.G., C.J.F. ter Braak, and O.F.R. van Tongeren, (eds). 1987. Data Analysis in
Community and Landscape Ecology. Pudoc, Wageningen, The Netherlands. (Now available in a 1995
edition by Cambridge University Press).
Ludwig, J.A. and J.F. Reynolds. 1988. Statistical Ecology: a Primer on Methods and Computing.
Wiley, New York.
MacArthur, R.H. and J.W. MacArthur. 1961. On bird species diversity. Ecology 42: 594-598.
Magurran, A.E. 1988. Measuring Biological Diversity. Princeton University Publishing, Princeton,
N.J.
375
Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and Methods of Vegetation Ecology. Wiley,
New York.
National Bureau of Standards. 1980. NBS Communications Manual for Scientific, Technical, and
Public Information. C. W. Solomon and R. D. Bograd (eds.). U.S. Department of Commerce,
National Biological Survey.
O'Connor, M. and F.P. Woodford. 1975. Writing Scientific Papers in English. Elsevier, Excerpta
Medica, North Holland. 108 pp.
Pneena, S. 1986. Helping Researchers Write … So Managers Can Understand. Battelle Press,
Columbus. 168 pp.
PSU (Penn State University). 2004. http://www.stat.psu.edu/~rho/stat200/chap12-p1.pdf.
Jongman, R.H.G., C.J.F. ter Braak, and O.F.R. van Tongeren, (eds). 1995. Data Analysis in
Community and Landscape Ecology. Pudoc, Wageningen, The Netherlands.
Rice Virtual Lab. 1998. HyperStat. http://davidmlane.com/hyperstat/
Sabin, W.A. 1993. The Gregg Reference Manual, 7th Ed. MacMillan/McGraw Hill Publishers, New
York. 502 pp.
SAS Institute 1996. SAS/STAT Guide.
Senseman, G.M., C.F. Bagley, and S.A. Tweddale. 1995. Accuracy Assessment of the Discrete
Classification of Remotely-Sensed Digital Data for Landcover Mapping. USACERL Technical
Report EN-95/04.
Shelton, J.H. 1994. Handbook for Technical Writing. NTC Business Books, Lincolnwood, Illinois.
210 pp.
Shimwell, D.W. 1971. The Description and Classification of Vegetation. University of Washington
Press, Seattle. 322 pp.
Snedecor, G.W. and W.G. Cochrane. 1980. Statistical Methods. Seventh Edition. Iowa State
University Press, Ames, Iowa.
Sokal, R.R. and F.J. Rohlf. 1981. Biometry: The Principles and Practices of Statistics in Biological
Research. W.H. Freeman and Company, New York, 859 pp.
Southwood, T.R. 1978. Ecological Methods. Chapman and Hall, London.
Strunk, W.Jr. and E.B. White. 1979. The Elements of Style, 3rd Ed. MacMillan Publishing Co. Inc,
New York. 92 pp.
Style Manual Committee, Council of Biology Editors. 1994. Scientific Style and Format: The CBE
Manual for Authors, Editors, and Publishers, 6th Ed. Cambridge University Press. 825 pp.
376
Tazik, D.J., V.E. Diersing, J.A. Courson, S.D. Warren, R.B. Shaw, and E.W. Novak. 1990. A
Climatic Basis for Planning Military Training Operations and Land Maintenance Activities.
USACERL Technical Report N-90/13. Champaign, IL.
ter Braak, C.J.F. 1987. CANOCO- a FORTRAN Program for Canonical Community Ordination by
Partial Detrended Canonical Correspondence Analysis, Principal Component Analysis and
Redundancy Analysis. TNO, Wateningen.
The Nature Conservancy (TNC). 1997. Vegetation Monitoring in a Management Context Workbook. Workshop coordinated by The Nature Conservancy and co-sponsored by the U.S. Forest
Service, held in Polson, Montana, September 1997.
Thomas, L. and C.J. Krebs. 1997. A review of statistical power analysis software. Bulletin of the
Ecological Society of America 78(2): 128-139.
USDA Forest Service. 1996. Rangeland Analysis and Management Training Guide. USDA Forest
Service, Rocky Mountain Region, Denver, CO.
Walter, H. 1985. Vegetation of the Earth and Ecological Systems of the Geo-Biosphere. SpringerVerlag, Berlin. 318 pp.
Wendorf C. A. 1997. Manuals for Multivariate Statistics.
http://www.uwsp.edu/psych/cw/statmanual/index.html
Woolf, C.M. 1968. Statistics for Biologists: Principles of Biometry. D. Van Nostrand Company, Inc.,
Princeton. 359 pp.
Yoccoz, N.G. 1991. Use, overuse, and misuse of significance tests in evolutionary biology and
ecology. Commentary, Bulletin of the Ecological Society of America 72:106-111.
Zar, J. H. 1996. Biostatistical Analysis. Prentice Hall, Upper Saddle River, New Jersey. 633 pp.
377
7.15
Appendix Statistical Reference Tables
Table 7-37
Table 7-38
Table 7-39
Table 7-40
378
Critical values of the two tailed Student’s t-distribution
Critical values for correlation coefficients
Binomial (percentage) confidence limits table
Critical values of the chi-square distribution
Table 7-37. Critical values of the two-tailed Student’s t-distribution. Reprinted with permission from
Sokal and Rohlf (1981), Table 12.
To look up the critical values of t for a given number of degrees of freedom, look up v = n-1 df in the
left column of the table and read off the desired values of t in that row. If a one-tailed test is desired,
the probabilities at the head of the table must be halved. For example, for a one-tailed test with 4 df,
the critical value of t = 3.474 delimits 0.01 of the area of the curve.
379
Table 7-38. Critical values for (product moment) correlation coefficients. Reprinted with permission from
Sokal and Rohlf (1981), Table 25.
To test the significance of a correlation coefficient, the sample size n upon which it is based must be
known. Enter the table for v = n-2 degrees of freedom and consult the first column of values headed
“number of independent variables”.
380
Table 7-39. Confidence limits for percentages (for sample sizes up to n=30) based on the binomial
distribution. Reprinted with permission from Sokal and Rohlf (1981), Table 23.
381
Table 7-39 Continued. Confidence limits for percentages based on the binomial distribution, for larger
sample sizes (n=50, 100, 200, 500, and 1000). Reprinted with permission from Sokal and Rohlf (1981),
Table 23.
382
Table 7-39 Continued.
383
Table 7-39 Continued. Large sample sizes (31%-50%).
384
Table 7-40. Critical values of the chi-square distribution. Reprinted with permission from Sokal and Rohlf
(1981), Table 14.
To find the critical value of X2 for a given number of degrees of freedom, look up v df in the left
column of the table and read off the desired values of X2 in that row.
385
Table 7-40. Continued.
386