Defining Mountain Goats habitat in the Mallard

1
Defining Mountain Goats habitat in
the Mallard Larkins Primitive Area
Abstract
Mountain goat (Oreamnos americanus), are
a species of concern for the state of Idaho and an
iconic animal of the sub-alpine environment of the
western United States. Mountain goats assist the
State of Idaho in bringing a high amount of outdoor
recreational revenue. In this study I will be
investigating the habitat of mountain goats in the
Mallard Larkins Primitive Area by assessing the
relationship between escape terrain used by
mountain goats, slope and vegetation. This will
assist me in understanding the importance of
vegetation as a factor in habitat selection. Defining
slope within the study area as well as a buffered
zone away from escape terrain will define the
habitat that is suitable for mountain goat.
Quantifying vegetation types in the suitable habitat
(as defined by distance to escape terrain) will give
me a better understanding of whether or not
vegetation is a factor in habitat selection by
mountain goats.
Introduction
Mountain goats are a listed as a species of
concern for the State of Idaho. This is due to the
fact that they are a very niche specific ungulate,
with habitat being broadly defined as sub-alpine
regions of the state [7]. They also are a highly
controlled game species in the State of Idaho. This
is to help control the harvest amount per year and
make adjustments on the number of animals
harvested on a year-to-year basis, which helps with
population
management
and
sustainability
statewide. In the study area this population is used
for extracting mountain goats to start new
populations around the state, thus making this study
important for the future of the mountain goats
statewide. Mountain goats tend to have a rather long
life span, with the average age being 12-15 years
and weights range from 125-180 lbs for adults [7].
During the summers, the vegetation chosen in
mountain goats’ diet is mainly shrub and forbs, with
winter foliage being grasses and brows [8].
Mountain goats also utilize natural and artificial
mineral licks during spring and summer months.
This is to provide them with key nutrition and
digestion components such as sodium, magnesium
and sulfur, and buffering compounds such as
carbonates and clays. [9]
I will be looking at the areas that are classed
as escape terrain ET and the relationship that
mountain goats display with these features. As of
yet, there have been limited studies conducted in
Colorado and Western Washington. During these
studies they strongly suggest that ET is the number
one factor of habitat selection for mountain goats
and that vegetation does not play a role.
The habitat range for mountain goats in
previous studies was ≤ 258 m from ET [1]. This is
of importance to help them handle with predation in
their habitat. Due to the very steep terrain defined as
habitat for mountain goats, a lot of this area is
vertical cliffs and unstable rocky terrain. This
assessment will give us a better understanding of
how habitat is selected, and whether it is solely
dependent on ET or if vegetation is a factor as well.
Currently there is very little data available
precisely defining habitat for mountain goats in the
state of Idaho. Thus, further understanding of the
ecology of mountain goats in Idaho would assist
scientists and biologists statewide in an enhanced
understanding of the relationship between escape
terrain ET and vegetation selections for mountain
goats.
Objectives
To define ET in the study area I used a 10 m
digital elevation model (DEM) and defining slope
values of ≥ 25º, 33º, and 35º [1, 2]. I then created
buffer rings for every 10 m around the ET, using a
Distance from Escape Terrain model and the
observation data I defined suitable habitat away
from ET for mountain goats.
The second objective is to assess the
accuracy of the National Land Cover Data in the
study area. This is tested against National
Agriculture Imagery Program (NAIP) image tiles
with fifteen manually selected and interpreted test
points that were randomly selected and evenly
distributed around the study area. The 9 cover types
in this study are water, barren land, deciduous,
evergreen forest, mixed forest, shrubs, herbaceous,
woody wet, and emergent herbaceous wet.
The third objective is to quantify the
vegetation cover types in habitat used by mountain
goats. This will allow me to obtain a value that will
give me in better understanding if vegetation is a
factor in habitat selection for mountain goats.
Hypothesis
2
I believe that escape terrain is a defining
factor in habitat selection. Furthermore, I
hypothesize that vegetation plays a key role in their
selection of habitat as well.
Methods
The study area is the Mallard Larkins
Primitive Area, 100 km east of Moscow, Idaho.
Splitting the Clearwater and St Joe National Forests,
and spanning 1052 km2, this region is defined as a
roadless area in the State of Idaho. There are 38
sub-alpine lakes and 10 sub-alpine peaks, ranging
from 1934-2157m, with total elevation ranges from
488-2157m [3].
I used a shapefile provided by the United
State Forest Service via Friends of the Clearwater to
define the border of the study area. Then computing
the topographic variable slope, which defines ET.
This was done using a 10 m DEM provided by
United States Geological Survey via Inside Idaho.
Land cover classification was pre classed by
National Land Cover Data, there were 9 classes and
the resolution was at 30 m. Finally, I used
observational point data of mountain goats in the
Mallard
Larkins
provided by Idaho
Fish and Game.
Figure 1
In the red we have displayed the Escape Terrain (ET) and the blue is the
distance in meters
A large portion in the northwestern portion
of study area is missing from the provided
shapefile. This was manually corrected by digitizing
northward up the Little North Fork of the
Clearwater River, and eastward up Sawtooth Creek,
continuing eastward to the St Joe National Forest
Rd 395.
The observation data was provided to me in
a note pad file, which could easily be place into and
Excel format, there was an issue with the coordinate
system that was used by the GPS to gather the
observation data. The data was gathered using
Garmin decimal degree and to project the point into
ArcMap I needed to convert them to decimal
degree. This was done using a formula provided by
Anna Giesmann [11]. (Figure 1)
The next step is to compute the topographic
features of slope ≥ 25°, 33° and 35° [1, 2], using the
clipped DEM and converting these to polygon files.
This created layers for the three defined slope
values. Using the observational data and a spatial
join to each of the individual buffered increments I
𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛𝑠 𝑖𝑛 𝐵𝑢𝑓𝑓𝑒𝑟
create the value of P1= 𝑇𝑜𝑡𝑎𝑙 𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛𝑠 . To
defined the random spatial model of
𝐴𝑟𝑒𝑎 𝑜𝑓 𝐵𝑢𝑓𝑓𝑒𝑟𝑒𝑑 𝐼𝑛𝑐𝑜𝑟𝑚𝑒𝑛𝑡𝑠
P2 =
I buffered in 10 m
𝐴𝑟𝑒𝑎 𝑜𝑓 𝑆𝑡𝑢𝑑𝑦 𝐴𝑟𝑒𝑎
increments away from ET. Once these values were
acquired we took P1-P2 to display the percent of
improvement that the ET model produces over the
random spatial model. (Figure 2, 3)
The steps explained above gives us an
estimate of the suitable habitat selection for
mountain goats in the study area, the next portion of
the study is to assess the accuracy of the vegetation
data pre classed by National Land Cover Data.
Doing so I used 9 of the original 12 classes
provided by NLCD due to the fact that classes of
snow/ice and hay/pasture combine were less than
1% of the entire study area. The 9 classes used were
water, evergreen forest, deciduous forest, mixed
forest, shrubs, barren land, woody wet, emergent
herbaceous wet and herbaceous. Using these classes
I tested the accuracy of the data, using National
Agriculture Image Program (NAIP) data as
reference data. To test this I created 15 manually
placed and interpreted ground points, that were
place in the center of the NLCD 30 m pixels. The
coordinates for each point where place in a Excel
file and the X,Y coordinates were displayed in
ArcMap. Then the values were extracted as a
shapefile. This was then placed over the NAIP
image to determine the true land cover type. Each of
the 15 point for the 9 classes, 135 total, were values
implemented in to a confusion matrix to provide us
with a value for the Kappa statistic which is a
measurement of accuracy between the remotely
sensed classification (NLCD) and the reference
image (NAIP)[12, 13].
Figure 2: Kappa statistic used for accuracy assessment. Where N= 135
3
Results
By defining slope and the values for P1-P2,
there was a peak in the topographic variable of
slope >33° which gave me the defined ET in this
study. The distance from ET used by mountain
goats in this study area was 0-30 m away from ET.
This value gave me a 22.81% improvement over the
random spatial model. The second closed was slope
>35° defined as ET and the distance from that being
0-10 m, this showed a 22.61% improvement over
the random spatial model.
Figure 3: chart depicting P1-P2 values. Highlighted area is highest percent of
improvement over the random spatial model
While looking at the accuracy of the NLCD
data I showed 52.5% accuracy for this data in this
area, while seeing a 1/3 increase in the foliage in
mountain goat diet of shrubs and herbaceous land
cover types [8]. The increase suggests that
vegetation is a factor in habitat selection for
mountain goats.
helicopter and was a capture recapture study over a
3-month period with 73 observations said to be
approximately 58% of the total population. “This is
an issue because the mountain goats tend to flee to
their ET hearing the helicopter,” said Nelson. The
limited data is a factor that can be altered with
summer fieldwork to gather more, due to recent
helicopter crashes in this area the capture recapture
study has been terminated. Due to the remote
locality of the study area week long trips in during
summer months the only way to gather said data.
Secondly the low accuracy in the land cover data
provided by USGS in the form of NLCD. In looking
at the matrix provided in the appendix (figure 4)
you will see that classing emergent herbaceous wet
and wood wet into one class of riparian would have
increased the accuracy up to the high 60%. You will
also see a large misclassification in water being
mistaken for barren land with eight of the test points
and barren land being mistaken for herbaceous with
10 of the test points. Seeing such a low accuracy
has led me to the next step of testing other land
cover data such as LANDFIRE and GAP data to
determine if they are more accurate. The 30 m
NLCD data was tested against 1 m NAIP images,
this was done off measuring out a 30 m grid on the
NAIP image and estimating the land cover in that
grid. This could cause some of the inaccuracy if the
pixel from the NLCD and grid on the NAIP didn’t
match up exactly.
= (12+7+6+13+9+9+3+12+7)=78
=(22*15)+(17*15)+(7*15)+(20*15)+
(14*15)+(16*15)+(10*15)+(20*15)+(9*15)=2025
135(78)−2025
8505
Therefor (135∗135)−2025= 16200=52.5%
[13]
Figure 4: Break down of Kappa
Discussion
Looking at the result several major questions
that arose. Why is the defined habitat much closer
to ET than the value provided in the Gross study?
And why is my percent of improvement only at
22.81% compared to 50%? There are two reasons
for this, the amount of observations and the way
they were gathered. The way the data was gathered,
in the study conducted by Gross, the observations
were gathered over a six year period that contained
5,343 observation points and were gathered from a
road that was driven twice a week. The data provide
by Idaho Fish and Game was gathered via
Figure 5: Confusion matrix computed for NLCD orange represents highest
error.
Having stated these factors, study was
effective because it gave us an estimation of the
suitable habitat used by mountain goats in this area.
With a few minor recalculations and the chance to
obtain more observation data points I predict that
we will see that percent of improvement in the
distance from ET model and the accuracy of the
land cover data increase greatly.
Acknowledgements
This work was supported by: Idaho
EPSCoR, and the Idaho Fish and Game.
4
Apendix
Garmin decimal degree: N46 53.894, W115 36.287
Conversion: 46+(53.894/60)= 46.89823333, 115+(36.287/60)*-1=-115.6029667
This was done in excel files
Bibliography
1. Gross, J.E.; Kneel, M.C.; Reed, D.F; Reich, R.M. GIS-based habitat models for
mountain goats. J. Mammal. 2002, 83, 218-228.
2. Wells, A.G.,Wallin, D.O.,Rice, C.G.& Chang,W-Y. 2011: GPS bias correction
and habitat selection by mountain goats. - Remote Sensing 3(3): 435-459.
3. http://www.summitpost.org/mallard-larkins-pioneer-area/288752
4. Idaho Fish and Game (observation data)
5. Friends of the Clearwater (border map provider)
6. United States Forest Service (border map creator)
7. http://animaldiversity.ummz.umich.edu/accounts/Oreamnos_americanus/
8. Brandborg, S.M, Life History and Ecology of the Mountain Goat in Idaho and
Montana. 1950.
9. http://www.fs.fed.us/database/feis/animals/mammal/oram/all.html#FoodHabits
10. http://www.qmrg.org.uk/files/2008/11/10-logit-models.pdf
11. Anna Giesmann, Garmin Decimal format to Decimal Degree.
12. Fitzpatrick-Lins, K., 1981, Comparison of Sampling Procedures and Data Analysis
for a Land-use Land cover Map, “Photogrammetric Engeineering & Remote
Sensing”, 47(3): 343-351
13. Consgalton, R.G., 1991,A Review of Assessing the accuracy of Classification of
Remotely Sensed Data, “Remote Sensing of Environments”, 37:35-46