1. sports
2. fishing

INFORMATIONAL L E A FL ET NO. 175
STOCK SEPARATION STUDIES OF ALASKAN SALMON
BASED ON SCALE PAITERN ANALYS I S
By
Paul V. Krasnowski
and
Michael L. Bethe
STATE OF ALASKA
Jay S. Hammond, Governor
DEPARTMENT OF Fl SH AND GAME
Ronald 0. Skoog. Commissioner
Subport Building, Juneau 99801
May 1978
STOCK SEPARATION STUDIES OF ALASKAN SALMON
BAS ED ON SCALE PATTERN ANALYSIS
By:
Paul V. Krasnowski
and
Michael L. Bethe
Statewide Salmon Stock Separation Project
Division of Commercial Fisheries
Anchorage, Alaska
TABLE OF CONTENTS
Page
.........................
LIST OF APPENDIX TABLES AND FIGURES . . . . . . . . . . . . .
ABSTRACT
............................
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . .
MATERIALS AND METHODS . . . . . . . . . . . . . . . . . . . .
Scale Collection and Processing . . . . . . . . . . . . .
Scale Examination . . . . . . . . . . . . . . . . . . . .
Digitizer and Linear Encoder . . . . . . . . . . . . . . .
Statistical Techniques . . . . . . . . . . . . . . . . . .
RESULTS AND DISCUSSION
...................
Bristol Bay .Sockeye Salmon
..............
Cook Inlet .Sockeye Salmon
..............
Cook Inlet .Coho Salmon . . . . . . . . . . . . . . . .
Cook Inlet .Chinook Salmon
..............
Kodiak and South Peninsula .Sockeye Salmon . . . . . .
Yukon River .Chum Salmon . . . . . . . . . . . . . . .
Norton Sound and Kotzebue Sound .Chum Salmon . . . .
Samples Grouped by Geographical Area
.........
.......................
LITERATURE CITED
............................
APPENDIX
LIST OF FIGURES
i
ii
iii
1
5
5
5
6
7
12
12
16
16
16
18
LIST OF FIGURES
Page
Figure 1 .
Figure 2 .
Figure 3 .
Figure 4.
Figure 5.
Age 42 sockeye salmon scale showing location of the
various scale c h a r a c t e r i s t i c s used in discriminant
analysis
...............................................
3
The Bristol Bay area. Shaded locations delineate
fishing d i s t r i c t s . .....................................
14
The upper Cook I n l e t area showing locations where scale
samples were collected f o r stock seperation studies ....
17
The Kodiak-South Peninsula area showing locations where
sockeye salmon scales were collected f o r stock
separation studies
.....................................
19
The Arctic-Yukon-Kuskokwim region showing areas where
scales were collected f o r stock separation studies .....
20
LIST OF APPENDIX TABLES
Page
Appendix Table 1 . Scale characteristics measured, by species .......... 2 6
LIST OF APPENDIX FIGURES
Appendix Figure
1.
Hydraulic scale press used to make impressions
of salmon scales on acetate plastic cards .........2 9
Appendix Figure 2.
Leitz microprojector and scale impression
card ..............................................29
Appendix Figure
3.
Microprojector, mirror assemby and table
surface f o r projection of scale images and
marking of scale characteristics ..................30
Appendix Figure 4.
Scale drawing sheet with pre-printed axis
lines. Short lines intersecting axes indicate
positions of annuli ...............................31
Appendix Figure
Linear digitizing rule (Rouchi rule) and
remote d i g i t i z e r controls . . . . . . . . . . . . . . . . . . . . . . . . . 3
5.
2
Appendix Figure
6.
Digitizer electronics. Numbers visable in
lower windows are fixed sample information and
are controlled by resettable thumbwheels ..........33
Appendix Figure
7.
Configuration of Ro,uchi rule, d i g i t i z e r and
ASR-33 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4
Appendix Figure 8.
Configuration of papertape reader, computer
terminal and acoustic coupler .....................35
Appendix Figure 9.
Diagrammatic representation of a section of
a scale within 0.1 mm (10.0 mm magnified image)
of the selected axis, showing "breakage" and
"branching" of c i r c u l i and indicating the
c r i t e r i a for inclusion of circuli in counts and
measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 6
Appendix Figure 10.
Sample hardcopy output of digitized scale
measurement data from ASR-33 ......................3 7
Appendix Figure 1 1 .
Data shown in Figure 10 a f t e r editing, sorting,
compacting and conversion from cumulative to
incremental measurement... ........................3 7
ABSTRACT
The Statewide Stock Separation Project was initiated i n July, 1 9 7 6
to research and apply new techniques of s t o c k identification for u s e i n
mixed-stock salmon fisheries. S c a l e s collected from sockeye, chum,
coho and chinook salmon have been examined using a projection microscope a t lOOx magnification. Numbers of circuli and d i s t a n c e s from t h e
focus t o annuli and supplementary checks were t h e commonly measured
characteristics Actual stock identification was based on pattern recognition procedures using discriminant function a n a l y s i s of s c a l e characteristics.
These techniques have successfully applied t o sockeye salmon (Bristol Bay,
Cook Inlet, Kodiak) , chum salmon (Norton Sound, Kotzebue Sound, Yukon
River), coho salmon (Cook Inlet) and chinook salmon (Cook Inlet). Application of stock identification techniques b a s e d on s c a l e pattern recognition
t o mixed-stock fishery management is logistically and s t a t i s t i c a l l y feasible.
.
STOCK SEPARATION STUDIES OF ALASKAN SALMON
BASED ON SCALE PATTERN ANALYS IS
Paul V. Krasnows ki , Research Proj e c t Leader
and
Michael L. Bethe, Fishery Research Biologist
Statewide Salmon Stock Separation Project
Division of Commercial Fisheries
Anchorage, Alaska
INTRODUCTION
The Statewide Salmon Stock Separation Project of t h e Alaska
Department of Fish and Game, Division of Commercial F i s h e r i e s , was
first funded for Fiscal Year 19 77 (beginning July 1 , 1976) I t s object i v e s a r e t h e r e s e a r c h , development and application of new techniques
of s t o c k identification which will permit determination of stock compos ition for Pacific salmon (Oncorhynchus s p .) harvested in a r e a s where
f i s h from more than o n e system a r e present.
.
For purposes of this report, stock is defined a s a somewhat disc r e t e group of f i s h which originates from t h e s a m e river system. A "stock"
may include more than one spawning group or population b u t , although
there may be more genetic similarity within a stock than between neighboring s t o c k s , t h e term is mostly a matter of convenience and does not imply
a strictly genetic b a s i s for identifying or separating t h e s e groups (Larkin
1972). Unless otherwise indicated a specific stock will refer t o a l l t h e
component sub-populations of a particular river drainage including a l l l a k e
b a s i n s and tributaries
.
Generally, t h e management of Alaskan salmon, operating on t h e
principles of optimum sustained yield, is done on t h e b a s i s of d i s c r e t e
stocks , t h a t i s , by river system. In some a r e a s , escapement enumeration
projects and s t a t i s t i c a l c a t c h allocation techniques have provided sufficient
data for development of spawner-recruit models. More commonly, only
escapement estimates and rough c a t c h figures or c a t c h per unit of effort
(CPUE) a r e available for determination of escapement requirements. It is
on t h e b a s i s of this information that escapement goals must be s e t .
Consequently, decisions t o open or c l o s e a fishery are only a s good
a s the catch and escapement data upon which they a r e based. Obviously,
i f t h e commercial harvest is operating on mixed s t o c k s , t h e ability t o
harvest t h e surplus of t h e healthy stocks and protect t h e stocks that may
fall below escapement requirements, must be based on some stock identification technique. It i s , therefore, e s s e n t i a l t o determine the proportion of each stock in the commercial harvest.
The development of techniques t o identify individual stocks of fish
would enable the design of sampling programs to describe the movements
of each stock through time i n the areas of concern t o the fishery managers.
Management decisions c a n then b e implemented which, by opening and
closing various district and sub-districts, optimize the harvest based on
stock composition.
Although investigations into the application of x-ray fluorescence
spectroscopy and protein electrophoresis for identification of Alaskan
salmon stocks have been underway for several y e a r s , t h e Division's stock
separation project has been directed toward s c a l e characteristic analysis.
There are several advantages of s c a l e characteristic methods. Scale
sampling for a g e determination is already a n integral part of the research
and management programs in many a r e a s . Scale collection is a quick,
logistically simple, and inexpensive operation even when handling live
fish. Scales d o not require s p e c i a l preservation and preparation time and
expense a r e minimal. Finally, where s c a l e a n a l y s i s techniques are applicable t o existing collections of catch and escapement s c a l e s , it may b e
possible t o allocate t o t h e systems of origin t h e numbers of fish taken in
past commercial harvests and provide t h e background data on total return
which is necessary for t h e determination of spawner-recmit relationships.
Scale development begins when salmon fry reach approximately 2 5 40 mm in length, depending on the s p e c i e s . Magnified images of salmon
s c a l e s appear a s a s e r i e s of concentric rings called circuli. The different
s p e c i e s of Pacific salmon c a n b e recognized by their different s c a l e patterns
(Koo 1962; Bilton e t . a l . 1964) In a l l salmon s p e c i e s , there is an overall
correlation of growth of t h e f i s h with radial growth of t h e s c a l e (Clutter and
Whites e l 1956) Changes in growth rate due t o environmental and/or
physiological conditions, therefore, a r e reflected in changes in the spacing of t h e circuli (Major and Craddock 1962; Bilton 1972; Bilton and Robins
1971a, b , c) Salmon that spend a significant portion of their life cycle i n
fresh water have a central portion of the s c a l e within which the spacing of
the circuli is more compact than that in the outer portion. This is referred
t o a s the freshwater growth zone (Figure 1). Outward from t h e freshwater
zone, the circuli a r e typically widely s p a c e d , probably reflecting t h e rapid
.
.
.
Figure 1 .
Age 4 sockeye salmon scale showing location of the various
scale 2 characteristics used in discriminant analysis ( a f t e r
Biiton, 1964).
growth which occurs early i n t h e marine l i f e s t a g e . Ocean circuli a r e
generally more broadly spaced than freshwater circuli s i n c e most of the
rapid growth occurs i n t h e marine environment. Additionally, bands of
very closely s p a c e d circuli, many of which a r e branched or broken, occur
i n both t h e freshwater and ocean zones of s c a l e s . The c l o s e spacing of
circuli in t h e s e annular rings or annuli is t h e result of extreme environmental
changes due t o winter weather and their effect on growth (Bilton and Messinger
1975) Determination of f i s h a g e by reading annular rings is common practice in many a r e a s .
.
Growth i n general, and s c a l e growth specifically, a r e genetically
moderated, environmentally influenced and a r e recorded i n t h e s c a l e pattern.
Differences i n s c a l e patterns between f i s h from different systems have been
noted for various s c a l e characteristics. Commonly, the number of circuli
and t h e width of t h e s c a l e for each y e a r ' s growth h a v e been examined.
Much of t h e research b a s e d on s c a l e pattern a n a l y s i s h a s been aimed a t
allocation of t h e harvest of salmon by the Japanese high seas fishery t o
continent of origin (Pearson 1964; Mason 1967; Mosher 1963, 1972; Anas
and Murai 1969). These researchers have examined s c a l e patterns of pink
(0. gorbuscha), chinook (Q.tshawytscha), and sockeye salmon (Q.
nerka).
Research conducted in Canada h a s described recognizable differences
between Canadian and North American s t o c k s of pink and sockeye salmon
(Bilton 19 70, 19 71; Bilton and Messinger 1975) In addition, researchers
have attempted t o identify s t o c k s harvested within inshore fisheries (Wright
1965) a s well a s attempting identification of sub-stocks occurring within
complex river systems s u c h a s the Skeena (Bilton and Smith 1969) and t h e
Fras er River (Henry 19 6 1)
.
.
In general, most research h a s involved groupings of s t o c k s from
large geographical a r e a s . Applications of t h e s e techniques t o inshore
fisheries i n Alaska h a s been attempted in some a r e a s (Wright 1965; Bergander
19 77) However, i n most c a s e s t h e differences between populations a r e
not sufficient t o u s e with standard measuring and analytical techniques.
In s u c h c a s e s , larger data b a s e s must b e constructed and more involved,
multivariate s t a t i s t i c a l methods utilized t o recognize p o s s i b l e s c a l e pattern
differences between s t o c k s
.
.
The development i n Canada of a reasonably-priced, high resolution
projector and semi-automated measuring and d a t a encoding equipment
designed for u s e on f i s h s c a l e s h a s provided the necessary means t o gene r a t e t h e required d a t a . Availability of high s p e e d data processing through
t h e University of Alaska Computer Network has provided t h e capability for
rapid a n a l y s i s of s c a l e characters
.
The Statewide Salmon Stock Separation Project presently c o n s i s t s of
two permanent fishery biologists and from one t o four temporary technicians.
From July 1976 through December 1977, more than 25,000 s c a l e s from
s o c k e y e , chum (0. keta) , coho (0. kisutch) , and chinook salmon have
been processed. Geographical a r e a s of concern have included Morton
Sound, Kotzebue Sound and the Yukon River, Kodiak and t h e Alaska Peni n s u l a , Cook Inlet, and Bristol Bay.
The following report covers t h e sainpling , sample processing,
measurement and a n a l y s i s of t h e data and includes general summaries of
t h e results by area and s p e c i e s .
MATERIALS AND METHODS
S c a l e Collection and Processing
An attempt was made t o have a l l s c a l e s collected from a preferred
a r e a on t h e left s i d e of t h e body below t h e insertion of t h e dorsal fin and
two or three rows above the lateral l i n e (INPFC 1961) S c a l e s were mounted
on gum cards and impressions of t h e s c a l e surface were made on cellulose
a c e t a t e cards using methods similar t o t h o s e described by Clutter and Whitesel
(1956) Initial examination and aging was accomplished by using a portable
microfische reader. Ages a r e described using Gilbert-Rich notation.
.
.
S c a l e Examination
S c a l e s were projected onto a t a b l e surface utilizing equipment similar t o t h a t described by Bilton (1970) and later modified by Peter Ryan of t h e
Canadian Fisheries and Marine Service (Ryan and Christie 1975). Photographs of the s c a l e projection and measuring equipment a r e presented i n
Appendix Figures 1-8.
The b a s i c projector is a Leitz Micro-Promar projection microscope
equipped with a wide-field eyepiece and plano objective l e n s e s (Appendix
Figures 2-3). A high contrast image is achieved by u s e of a Prado Universal
250 watt quartz iodine lamp. The microprojector and l e n s s y s t e m , u s e d i n
conjunction with a n overhead adjustable mirror and frame assembly, is
designed t o deliver a f l a t , undistorted image t o t h e table surface. The t a b l e
surface is constructed of flat white formica t o enhance t h e contrast of the
projected image. High and low magnifications c a n b e achieved by s e l e c t ing different combinations of occular and objective l e n s e s and adjusting t h e
height of the overhead mirror.
All adult salmon s c a l e s were projected a t a magnification of 100x.
To ensure continuity of measurements, magnification a t t h e t a b l e surface
was calibrated a t frequent intervals using a gridded millimeter microscope slide.
After a s c a l e was s e l e c t e d for measurement, the image was projected onto a large s h e e t of white bond paper which is pre-printed with
nine a x i s lines (Appendix Figure 4). For s o c k e y e , a n a x i s l i n e was
oriented s u c h that i t intersected t h e center of t h e nuclear area and lay
along a radius which was 20° ventral t o t h e anterior - posterior a x i s of
the s c a l e (Clutter and Whitesel 1956). For other s p e c i e s , t h e longest
s c a l e a x i s was s e l e c t e d . Selection of t h e s e a x e s afforded examination
of t h e longest portion of t h e s c a l e that had t h e fewest broken or branched
circuli. Where each circulus crossed t h e a x i s , a tracing approximately
one inch long was made o n t h e paper. Only t h o s e circuli which continued
more or l e s s intact within a d i s t a n c e of 0.1 mm on both s i d e s of the 20°
ventral a x i s (i. e. , within a d i s t a n c e of 1.0 cm on t h e projected image)
were counted. Selection criteria were adapted from Bilton (1971) and a r e
further detailed in Appendix Table 1 .
Digitizer and Linear Encoder
To enable rapid generation of digital output of s c a l e measurements
in computer readable format, a digitizer and linear encoder were employed.
This equipment i s described in Ryan and Christie (1975) and shown i n
Appendix Figures 5-8.
The digitizer unit c o n s i s t s of electronics which format and output
s c a l e data i n computer readable, fixed format (ASCII). Sample identification
d a t a is input via a twelve digit code representing location of sampling,
s c a l e card number, length of f i s h , s e x , year sampled, and a g e . The linear
encoder is connected t o t h e digitizer and f a c i l i t a t e s entry of measurements
via a remote control. The encoder is oriented parallel t o a x i s l i n e s and t h e
sliding Rouchi rule is zeroed a t t h e nuclear portion of t h e s c a l e drawing.
Depression of t h e "record" button on t h e remote control panel initiates the
automatic recording and formatting of a l l data on t o a typed s h e e t and t o
punched paper t a p e on a n ASR33 teletype. The index of the rule is then
moved a c r o s s t h e s c a l e t o the end of t h e first annulus or t o t h e f i r s t characteri s t i c of concern while manually depressing t h e count button for e a c h circulus
p a s s e d . When t h e index i s aligned, the record button is depressed causing
the interval measurement and t h e circuli count t o be recorded on t a p e a s
described above. The digitizer is capable of recording s e v e n paired measurements (numbers of circuli and d i s t a n c e from focus) per record.
At present, data are either keypunched or paper t a p e s a r e processed
by the University of Alaska, Geophysical Institute i n Fairbanks for conversion
t o NOVA magnetic tape and subsequent data entry. On-line terminal computer data entry h a s been necessary during field operations. Acquisition of a paper t a p e reader will allow direct entry of raw data t o computer
file when interface of the equipment is complete this year. Format of data
output from the digitizer is shown in Appendix Figures 9-1 1
.
As described above, circuli counts and distance measurements t o
each point of interest a r e paired. The data output a r e cumulative, representing total circuli and distance from t h e s c a l e origin (focus). An editing
program has been developed which eliminates records with obvious errors
in the fixed data ( e .g , s e x code, district code, age) and provides general
editing criteria t o b e applied t o t h e s c a l e data. Additionally, t h e output of
this program converts the measurements from cumulative t o incremental and
compacts t h e d a t a . Examples are provided in the Appendix. These edited
data a r e in FORTRAN format (3F2.0, F3.0, F1.0, F2.0, 12F4 .O) where t h e
column assignments are:
.
Variable
Column Assignment
District Code
Card (AWL#)
Year
Length (mm)
Sex Code
Age
NC 1
ID1
NC2
ID2
NC3
ID3
NC4
ID4
NC5
ID5
NC 6
ID6
where NCi = the number of circuli along the selected axis for the i t h
characteristic
IDi = the interval distance along the selected axis for the ith
characteristic
Most of the research t o d a t e has dealt with s c a l e s from adult
sockeye salmon. In most c a s e s , measurements of sockeye s c a l e s have
been from the focus t o t h e outside of the first annulus (Al) and thence
from t h e outside edge of each annulus t o the outside of the succeeding
annulus. However, there a r e many different characteristics that c a n b e
used. The measurements a s used in the sockeye and described above,
a r e referred t o a s standard measurements. In other c a s e s ( e . g , sockeye
smolt and adults of other species) other characteristics have been u s e d .
These have been further described in Appendix Table 1.
.
Statistical Techniques
Linear discriminant function analysis was developed b e c a u s e of a
need t o distinguish statistically two or more groups, and is based upon
work done by R .A. Fisher (193 6) Applications of t h e technique t o biologic a l data have developed rapidly i n recent years due t o the advent of digital
computers. The utility of discriminant function analysis a s applied t o
stock separation based on s c a l e characteristics is based on the concept
that two groups (stocks of fish in our c a s e ) may differ slightly in t h e mean
and distribution of values for some measureable characteristics (e. g ,
circuli counts and radii). Characteristics taken singly and measured from
samples of two populations may be u s e l e s s for identifying group membership s i n c e , d e s p i t e slight differences in means, t h e degree of overlap of
values between the two groups is generally s o great a s t o render the individual characteristic u s e l e s s for discriminating between groups.
.
.
Because no single characteristic will allow identification of group
membership, discriminant analysis techniques attempt t o do t h i s by using
a multivariate approach which combines variables t o yield "discriminant
functions" which serve t o better identify membership. These functions may
b e linear or nonlinear. Since s c a l e measurements have been found t o generally satisfy t h e required assumptions for linear discriminant analysis
(Anas and Murai 1969; Cook, personal communication) and s i n c e software
for the University of Alaska Honeywell computer provides linear discriminant
function a n a l y s i s routines, t h e work described in this report utilizes t h e
linear methods.
The analysis first requires measurements from samples of known
group membership. These samples, a l s o called standards or learning
samples, provide t h e data required t o formulate t h e discriminant function,
essentially the discriminating model. The program s e l e c t s t h e discriminating variables in a stepwise fashion. The order of selection of variables
for inclusion into the analysis reflects the relative between-group variability of each characteristics, i . e. , their relative discriminating ability.
The discriminant functions a r e of the form:
.. .
where Di is the discriminant score for the ith s c a l e , d l , d2,
, dp are
weighting coefficients and z i l , z i f ,
, z ip are standardized values of
the measurements from t h e i t h s c a e . In other words,
.. .
where xij is the value of t h e jth measurement from the ith s c a l e , mj is the
mean of t h e jth measurement for a l l s c a l e s , and sj is t h e standard deviation
of t h e jth measurement, again for a l l s c a l e s .
The discriminant functions define p-dimensional hyperplanes which
cut across t h e intermixed clusters of points s o that a s many a s possible of
t h e members of one group have high values of Di and most of t h e other group
members have low values of Die The weighting coefficients d l , d2,
, dp
a r e calculated s o that t h e discriminant scores Di are standard normal varia b l e s , and the mean discriminant scores for a l l s c a l e s is zero, with a standard
deviation of one.
...
For each group taken singly, the mean of the discriminant scores for
a l l its members is called the centroid and describes t h e most probable location of that group i n discriminant function space. The distance between t h e
group centroids is an indication of t h e distance by which two groups a r e
separated (again, along t h e s e dimensions i n hyperspace). The midpoint
between two centroids (Do 5) serves a s a decision point and unknowns c a n
b e classified a s to probable group membership based on which s i d e of Do .5
they fall. In practice, the SPSS (Nie e t . a l . 1975) and BMD ( ~ i x o n1965)
programs u s e d i n t h e s c a l e analysis project a l s o output classification functions, one equation for each group, which are more convenient for classifying
unknowns.
In c a s e s where classification of more than two groups is required,
the problem of visualizing the discriminant functions becomes more difficult.
The number of discriminant functions generated is equal t o the number of
original discriminating variables (p) or t o one l e s s than the number of groups
(g-1) , whichever is l e s s . Generally, the (g- 1) limitation has been used i n
t h e s e a n a l y s e s . Each resulting discriminant function is orthogonal (at right
angles) t o t h e previous functions and the resulting discriminant scores a r e
taken t o b e (g-1) dimensional descriptions of the locations of t h e g groups
in discriminant function s p a c e .
To t e s t the efficacy of the a n a l y s i s , a trial classification is made
using the above mentioned classification function and the standards (learning
samples)
.
The classification equations are of t h e form
where the Ci a r e t h e classification s c o r e s which a r e the sum of the c j ' s
(classification coefficients) times t h e raw variable values (xijts) plus t h e
classification constant co. Since there is one classification function for
each group, there will be g classification scores for each c a s e c l a s s i f i e d .
An unknown c a s e is classified a s a member of the group in which i t h a s t h e
highest score. Since the a c t u a l group membership for each c a s e is known,
t h e results of t h i s trial classification c a n be summarized and tabulated a s
below:
Actual Group
Members hip
Group A
number
proportion
Group B
Classified
Group Members hip
Group A
Group B
Aa
Ab
Paa
pab
Na
1.O
number
Ba
Bb
Nb
proportion
Pba
Pbb
1 .O
where Aa and Bb a r e t h e numbers of their respective groups that were correctly classified and Ab and Ba a r e t h e numbers incorrectly identified.
Furthermore, t h e proportions correctly and incorrectly c l a s s i f i e d are taken
a s estimates of t h e probabilities of classification accuracy and c l a s sification error, i e . , Pa, is t h e estimated probability of correctly classifying
a n unknown sample that is actually a member of group A , whereas Pab is
t h e estimated probability of misclassifying a n unknown sample that is
actually a member of group A a s a member of group B . Given equal sample
s i z e s from each group and normal distributions of discriminant s c o r e s differing only i n mean v a l u e s , t h e apparent numbers i n groups A and B (i e. , t h e
sum of the correctly classified members plus t h e misclassified members of
t h e other group) should b e approximately equal. If not, both SPSS and BMD
have options t o make a priori adjustments which affect t h e probabilities of
group membership and can s e r v e t o equalize t h e misclassification errors.
However, a thorough understanding of t h e affects of manipulating a priori
probabilities is needed before attempting t h i s .
.
.
When classification of a n actual mixed sample of unknown composition (group membership proportions) is completed, t h e results represent t h e
apparent or observed numbers and frequencies of e a c h group (similar t o t h e
a and b above) and represent both t h e correctly c l a s s i f i e d members of
e a c h group plus the misclassified numbers of the other group. Since
t h e trial classification of t h e learning samples provides t h e estimates
of t h e probabilities a s s o c i a t e d with correct and erroneous c l a s s i f i c a t i o n ,
the observed frequencies can b e adjusted with t h e s e probabilities t o
estimate t h e a c t u a l proportions present i n the sample (Worlund and Fredin
1962) This adjustment procedure is accomplished through t h e solution of
a s e t of simultaneous equations. Since the observed number of e a c h group
in t h e mixed sample is t h e sum of t h e correct decisions for that group plus
t h e incorrect decisions for t h e other groups, and s i n c e i n each c a s e t h e s e
decisions a r e t h e product of classification probabilities (Pjk) and t h e a c t u a l
number of f i s h from that group i n the sample, a s e r i e s of equations c a n b e
constructed. For a three-group situation t h e s e equations are:
.
where
K a , Kb and Kc a r e t h e numbers of f i s h c l a s s i f i e d t o each system;
Na, Nb and Nc a r e t h e estimated numbers of fish from each
system in t h e mixed sample (unknown); and
Pjk are the proportions of f i s h from system k c l a s s i f i e d a s from
system j (known -- estimated from the training s e t s ) .
Confidence interval estimates for t h e two group classification model
a r e given by Worlund and Fredin (1962) However, t h e s e estimates a s s u m e
t h a t t h e Pjk a r e known without error and therefore t h e intervals a r e too
narrow. A method for calculation of confidence intervals for g groups which
t a k e s into account the variability of Pjk is i n the final s t a g e s of development and will b e u s e d with future classifications based on s c a l e characters.
.
Required sample s i z e s for both t h e a n a l y s i s and t h e classification
a s p e c t s a r e not y e t well worked out. As with most sample s t a t i s t i c s t h e
variance of estimates d e c r e a s e s with increased sample s i z e . Some inves tigators have recommended sample s i z e s of 50 minimum for each group standard.
Rod Cook (Fisheries Research Institute, personal communication) h a s u s e d
a s few a s 25. In general, t h e r e s u l t s reported here are b a s e d on a minimum
sample s i z e of 50 s c a l e s .
For classification of a mixed sample we recommend a minimum of
100 s c a l e s . The variance of t h e estimated frequencies will b e inversely
proportional t o the s i z e of t h e sample classified.
There is a b i a s associated with t h e estimates of classification
accuracy and error, t h e P j k , which is d u e t o t h e fact that the classification
functions are not t e s t e d with additional random samples of each group but
are tested with t h e same c a s e s that were u s e d t o generate functions.
Therefore, t h e s e data will fit t h e models slightly better than might b e
expected. Some initial research has indicated that t h i s self-clas sification bias might be a s much a s +4-6%, with t h e sample s i z e s we have been
using. The bias c a n b e avoided by collecting additional samples of knowns
which a r e used t o test t h e classification function but which a r e not u s e d t o
generate t h e functions. There are other methods t o avoid t h e b i a s , but we
have not yet incorporated them. If separate samples a r e u s e d , we recommend
equal sample s i z e s of 25 or more.
RESULTS AND DISCUSSION
.
To d a t e more than 25,000 s c a l e s have been measured, digitized,
edited and analyzed. The data have provided several hundred discriminant
a n a l y s e s which cover four of t h e salmon s p e c i e s (sockeye, chum, coho,
chinook) and five geographical areas of Alaska. Approximately s i x thousand
of t h e s e s c a l e s were collected in conjunction with "in-season" stock separation projects conducted in Cook Inlet and Bristol Bay during 1977.
Detailed a n a l y s e s which will provide information on timing, distribution,
and catch allocation of component s t o c k s are underway and will b e described
i n l a t e r reports. The results summarized below were derived from s c a l e s
collected a s routine samples from on-going management and research projects prior t o t h e development of t h e Stock Separation Project. Scale measurements of chum, coho and chinook were non-standard and are detailed
i n t h e Appendix.
Bristol Bay
- Sockeye Salmon
Prior to initiating in-depth discriminant a n a l y s e s , a s e r i e s of
a n a l y s e s of variances were performed on s c a l e characteristic measurements
from Bristol Bay sockeye salmon collected for routine escapement samples
i n 1970 through 19 75. In general, for individual river s y s t e m s , f i s h of t h e
same freshwater a g e and brood year but of different ocean a g e s have very
significant differences (p < 0 .0 1) for a l l variables Similarly within s t o c k s ,
there a r e very significant differences between fish of differing freshwater
.
and total a g e for a l l variables. In addition, there frequently a r e significant differences between s e x e s within years, a g e c l a s s e s , and river
s y s t e m s . The indications, then a r e that new standards (learning samples)
should b e developed for each s y s t e m , year and a g e c l a s s examined using
discriminant analysis Although identification of group membership would
probably b e further enhanced by examining each s e x independently, t h e
increased sample s i z e s required would make t h e small gain in accuracy
very costly.
.
For t h e Naknek and Kvichak Rivers (Figure 2) s c a l e s from a g e 42,
53, and 63 fish sampled i n 1970 through 1975 were examined. For most
y e a r s , overall classification accuracies were above t h e mid-80% level.
Discriminant a n a l y s i s of data from t h e s e s a m e years from Naknek, Kvichak,
and Ugashik a g e 53 fish (3-way analyses) produced classification accuracies
i n t h e low 70% range. Egegik and Ugashik, for a l l a g e c l a s s e s , appear separable with overall accuracies varying between 80% and 85%. A three-way
a n a l y s i s of Naknek , Kvichak and Egegik, produced overall accuracies that
were quite variable, ranging from about 60% t o 80% accuracy. Naknek,
Kvichak, Egegik and Ugashik (four-way discriminant analyses) produced
overall accuracies in the low 70% range with Naknek being t h e l e a s t separable (largest misclassification error), and Kvichak and Ugashik being t h e
most distinctive. In general, i t appears that i n most years there a r e enough
distinct differences between a l l systems on t h e e a s t s i d e of Bristol Bay and
for a l l t h e major a g e c l a s s e s (particularly a g e 53 and 63 fish) t o provide a n
effective tool for s t o c k identification.
In general, the systems on t h e w e s t s i d e of Bristol Bay produce f i s h
t h a t spend o n e y e a r i n freshwater (42 and 52). For Wood River and Nuyakuk
River, there were sufficient s c a l e s t o examine one year of data from each
a g e c l a s s . The data from a g e 42 and 52 (1972) fish yielded overall accura c i e s i n e x c e s s of 90%. Small samples from a g e 53 and 63 (1971) fish
produced overall accuracies in the low 70% range. Several years data from
Wood, Nuyakuk and Igushik Rivers were compared i n a three-way discriminant a n a l y s i s (age 52). The r e s u l t s were highly variable ranging from a low
of 41% t o highs of approximately 90% overall classification accuracy. Nuyakuk and Igushik s c a l e characters a r e frequently quite similar.
Despite the s i z e of the d a t a b a s e , there a r e only a few a g e
c l a s s e s and years i n which there a r e sufficient data to compare more than
four systems a t a time. Comparison of Naknek, Kvichak, Ugashik, Wood
and Nuyakuk (age 42) produced 68% overall accuracy. However, t h e Ugashik
and Wood f i s h were c l a s s i f i e d accurately only in 48% of t h e c a s e s , whereas
Nuyakuk showed no misclassification error. Samples from only one year
provided sufficient data t o compare a g e 52 and 63 fish from Naknek, Kvichak,,
Igushik, Wood and Nuyakuk Rivers. The data yielded 51% overall accuracy
BRIBTOL
AREA
I#?#
UV
ILINCR
OmAWU
Figure 2.
BAY
RIVIICD
I .
Ttie Bristol Bay area.
Shaded areas delineate fishing districts.
for t h e a g e S2 and 49% overall accuracy for t h e a g e 63 f i s h . For a g e 53
fish from 1975, a l l s e v e n of t h e major systems were analyzed and resulted
in overall classification accuracies in t h e high 50% range. Generally,
overall classification accuracy d e c r e a s e s with increasing numbers of discriminating groups.
Several y e a r s ' data from one and two check smolt were analyzed
for t h e Kvichak and Naknek Rivers. The results from s c a l e a n a l y s i s of
smolt which had spent two full years i n freshwater ranged from t h e low t o
mid-70% range. Data from one check smolt provided generally poor overall
a c c u r a c i e s . Data derived from smolt s c a l e s would b e of limited value i n
identification of future adult returns s i n c e subsequent measurements of t h e
freshwater portions of t h e adult s c a l e s from t h e s e same brood years indic a t e s a strong s e l e c t i v e pressure for t h e larger smolt during the marine
life s t a g e . In most i n s t a n c e s , the means for the same characteristics
measured on t h e returning adults were significantly larger than t h e corresponding measurements of smolt s c a l e s . In most c a s e s , t h e measurement
of adults were larger by a t l e a s t one standard deviation.
In t h e final s e r i e s of t e s t s , from t h e e a s t - s i d e systems (Kvichak,
Naknek, Egegik and Ugashik) and t h e west-side systems (Wood, Nuyakuk
and Igushik) were pooled by a g e group. An offshore t e s t fishing program
near Port Moller on the north s i d e of t h e Alaska Peninsula provides abundance estimates for t h e entrance of Bristol Bay sockeye b a s e d on CPUE,
approximately 6-8 days prior t o t h e arrival of the fish i n t h e districts.
Since s c a l e samples a r e routinely c o l l e c t e d , application of a n e a s t - s i d e
systems versus west-side systems pooled function would allow refinement
of t h e s e estimates t o provide additional timing and abundance information.
For a g e d2 f i s h , t h e overall accuracy range ( e a s t v s . west) was i n t h e high
70's; for 52 and 63 samples accuracies were i n t h e mid 80% range and for
a g e 63 f i s h , performance was i n t h e low 8 0 ' s .
Data from a l l years can b e pooled without significantly reducing
classification accuracy. However, there a r e significant between-year
variabilities within t h e groups which prevents t h e s e y e a r s ' pooled functions
from being u s e d a s a n "in-season" tool prior t o obtaining new standards
from t h e escapements.
In general for t h e e a s t - s i d e s y s t e m s , s c a l e measurements of t h e
freshwater zone provided the b e s t discriminating variables in most y e a r s .
However, s c a l e measurements from t h e first marine years were in many
i n s t a n c e s excellent discriminating characteristics. For f i s h collected in
t h e rivers in Nushagak Bay (west s i d e ) , marine characteristics were frequently chosen i n the s t e p w i s e procedure a s t h e variables showing the
largest between group variances.
Cook Inlet - Sockeye Salmon
The major sockeye salmon producing systems i n t h e Cook Inlet
area of southcentral Alaska (Figure 3) a r e the Susitna, Kenai and Kasilof
Rivers. Analysis of s c a l e s collected from fish wheel samples i n 1975
provided overall classification accuracies of about 80%. Kenai f i s h appear
more similar t o Susitna than t o t h e neighboring Kasilof system. Freshwater
variables were the most effective, followed by measurements from the first
marine year. Analysis of data from Cook Inlet sockeye collected in 1976
for a l l a g e c l a s s e s examined ( J 2 , 52 and 53) , proved quite s u c c e s s f u l .
The a g e 53 fish separated with overall accuracies i n the high 90% range.
In a two-way a n a l y s i s , Kenai and Kasilof, a g e 5 3 , sockeye a r e separable
with nearly 100% accuracy in that year. Analysis of t h e a g e 42 and 52 f i s h
from that year yielded overall accuracies i n t h e low 70% range with a considerable number of Kenai fish being misclassified a s Susitna f i s h . Howe v e r , t h e number of samples available from t h e Susitna was quite limited
and this generally c a u s e s a l o s s of accuracy. For 1976 s a m p l e s , t h e marine
characteristics were t h e most effective variables.
Analysis of data collected in 1977 is not yet complete. Preliminary
results indicate overall accuracies in t h e mid-70 percent range. Samples
from t h e multiple b a s i n Kenai system again show t h e greatest error of misclassification.
Cook Inlet
- Coho Salmon
Although there a r e many systems in the Cook Inlet area that produce
coho salmon, a s much a s 80 t o 90% of t h e production may b e attributed t o
the Kenai and Susitna Rivers. Approximately 80% overall accuracy h a s been
achieved i n separating Susitna from Kenai coho salmon based upon s c a l e
characteristics (data from 1975 and 1977 a g e 43) Measurements from t h e
marine portions of t h e s c a l e s were t h e most s u c c e s s f u l in separating t h e
two s t o c k s . Recent investigations b y t h e staff of t h e Sport Fish Division,
ADF&G (unpublished data) indicate t h a t mean f i s h weight may b e substantially different between t h e two s t o c k s . Future research will incorporate
fish length and weight with s c a l e measurement data t o determine their value
a s discriminant characteris t i c s
.
.
Cook Inlet
- Chinook Salmon
The major systems which produce significant runs of chinook salmon
i n Cook Inlet a r e t h e Susitna, Kenai, Ninilchik, and Anchor Rivers. S c a l e s
collected from a g e 52 f i s h from t h e s e systems (1977) were measured for
discriminant analysis
.
'
Figure 3.
The upper Cook I n l e t area showing locations where scale samples were
collected f o r stock separation studies.
Pairwise (2-way) analyses of t h e four major systems provided
overall accuracies ranging from 71% t o 83%. A four-way a n a l y s i s yielded
51% overall classification accuracy. As i n t h e sockeye a n a l y s e s , t h e
Kenai system showed the greatest variability and therefore t h e highest
misclassification error. The s c a l e measurements reflecting t h e first and
second marine summers growth were consistently t h e b e s t discriminating
variables.
Kodiak and South Peninsula
- Sockeye Salmon
Data from a g e 53 sockeye salmon s c a l e s collected in 1976 from
the Karluk and Frazer Rivers on Kodiak Island and from Chignik on t h e south
s i d e of the Alaska Peninsula provided 80% overall classification accuracy
d e s p i t e small sample s i z e s ( s e e Figure 4) Ninety-six percent of t h e Chignik
samples were correctly identified, whereas, there were 66% correct Karluk
decisions (26% of the Karluk f i s h were misclassified a s Frazer) and 74%
correct decisions of Frazer f i s h (all errors for Frazer were misclassification
a s Karluk) Fish length and freshwater s c a l e measurements were consistently s e l e c t e d a s the best discriminating variables. Discriminant a n a l y s i s
using only freshwater variables, provided overall accuracies i n t h e high
60% range. Karluk and Frazer examined i n a two-way a n a l y s i s yielded a n
overall accuracy of approximately 70% b a s e d primarily on freshwater
characteristics.
.
.
Yukon River
- Chum Salmon
Chum salmon s c a l e s were measured i n a non-standard manner. The
data include the measurement from t h e s c a l e focus t o a f a l s e check (transition or migration check) which occurs before the first winter check.
s c a l e samples from t h e Sheenjek River, a large tribAge 31 and
utary of t h e Porcupine system i n Northeast Alaska and t h e Toklat and Delta
Rivers, both Tanana River tributaries in t h e Central Interior were available
for 1976 (Figure 5 ) . Overall accuracy was 76% with individual group
accuracies of 89% for Sheenjek, 67% for Toklat and 72% for Delta River.
The two Tanana tributaries were most similar i n s c a l e measurements and
only a small proportion of t h e errors were misclassification a s Sheenjek.
Some additional samples of a g e 41 chum salmon collected i n 1974
from t h e Sheenjek and Toklat Rivers were a l s o examined but produced c l a s s i fication accuracies i n the low 60% range.
____------
Norton Sound
Figure 5.
Kotzebue Sound
Yukon River
The Arctic-Yukon-Kuskokwim re ion showing areas where chum
salmon scales were collected or stock separation studies.
7
-
20
-
Norton Sound and Kotzebue Sound
- Chum Salmon
Age 41 chum salmon s c a l e s collected i n 1977 from the Kobuk and
Noatak Rivers i n Kotzebue Sound (Figure 5) were measured using the same
s c a l e characteristics a s t h e Yukon River chums. Overall c l a s s i f i c a t i o n
accuracy was 69.3% with 20 .O% classification error for Noatak and 41.2%
classification error for Kobuk. The bimodal distribution of discriminant
s c o r e s from t h e Kobuk samples may indicate t h e presence of two discrete
s t o c k s in that s y s t e m , perhaps from t h e Squirrel River, a major tributary
and mainstream s t o c k s of t h e Kobuk. Additional sampling i n each tributary
may improve the a c c u r a c i e s . In Norton Sound (Figure 5) samples were
available from only three of t h e contributing systems: Kwiniuk, Nuikluk
and Kachavik. The three-way a n a l y s i s yielded 63% overall accuracy for
a g e 41 chum salmon. Whereas t h e separation of t h e Kotzebue Sound s t o c k s
was b a s e d primarily on characteristics from t h e second marine y e a r , t h e
Norton Sound samples had greater between group variability based on the
first year characteristics
.
Samples Grouped by Geographical Area
Throughout t h e various a r e a s and s p e c i e s studied there appears t o
b e a n overall tendency for variation i n s c a l e patterns t o b e partially a function
of geographic d i s t a n c e between s y s t e m s . For example, t h e Kotzebue Sound
and Norton Sound s t o c k s , although showing considerable s c a l e pattern
differences within the a r e a s , show substantial differences between t h e s e
a r e a s if each is considered a s a discriminant group. Discriminant a n a l y s i s
of t h i s pooled-area sample yields classification accuracy in the 80% range.
Also, t h e means from measurements of t h e Yukon River chums are considerably different from Norton or Kotzebue Sound fish. In the Kodiak d a t a ,
Karluk and Frazer fish have somewhat similar s c a l e characteristics. Thes e
l a k e s a r e both on Kodiak Island. The s c a l e s from Chignik, located on the
Alaska Peninsula west of Kodiak Island, a r e e a s i l y distinguishable from
t h e Kodiak Island s t o c k s . A s s t a t e d above, there is a general similarity
among the systems in Bristol Bay when pooled into groups of e a s t and westside systems.
Within some smaller a r e a s , some a n a l y s e s have combined systems
i n order t o reduce the number of discriminant groups. Since, i n general,
overall accuracy i n c r e a s e s with fewer groups, this may b e a valid technique where multiple group a n a l y s e s yield poor a c c u r a c i e s , and separation
into individual river systems is not e s s e n t i a l . For example, i n Bristol Bay,
comparison of Wood, Nuyakuk and Igushik River d a t a provide low accura c i e s i n some y e a r s . Pooling data from Wood and Nuyakuk and comparing
this pooled sample with Igushik d o e s , i n some c a s e s , substantially improve
the accuracy. However, preliminary research into t h e application of resulting
classification equations t o a mixed sample indicates that this may have
a secondary effect of increasing t h e variance of t h e group frequency
estimates (T .L Robertson, personal communication)
.
.
Since s c a l e measurements a r e reflections of f i s h growth through
various l i f e s t a g e s , similarity i n s c a l e patterns of fish from adjacent
systems might reflect t h e similar environmental influences affecting t h o s e
s t o c k s . However, in many i n s t a n c e s , t h e b e s t discriminating characteristics were measurements from growth i n t h e second and third marine y e a r s .
This would seem t o suggest t h a t either there were different environmental
influences affecting t h e s e s t o c k s on the high s e a s , or that genetic factors
affecting f i s h growth and subsequently s c a l e development a r e t o some
measurable extent responsible for t h e s e population differences. Taken
a s t e p further, if it is assumed that genetic similarity would b e greatest
among neighboring populations (systems), i t would explain why s c a l e
pattern differences are more pronounced with increasing geographical distance.
LITERATURE CITED
Anas, R.E. and S. Murai. 1969. Use of s c a l e c h a r a c t e r s and a d i s c r i m i n a n t f u n c t i o n f o r c l a s s i f y i n g sockeye salmon (Onchorhynchus
nerka) by c o n t i n e n t o f o r i g i n . B u l l . I n t . N o r t h Pac, F i s h . Comm.,
NO. 26: 157-192.
Bergander, F.E. 1977. Sockeye s t o c k c o n t r i b u t i o n and i d e n t i f i c a t i o n ,
S.E. Alaska. Alaska Dept. o f F i s h and Game Tech. Rpt.9 P r o j e c t
NO. AFC-51-2. 22p.
B i l t o n , H.T.
1970. Data and computations used i n r a c i a l analyses o f
age 1.2 and 1.3 N o r t h American sockeye on t h e b a s i s of t h e i r s c a l e
c h a r a c t e r s . F i s h . Res. Bd. Can., Tech. Rpt. No. 223, 57p.
.
1971. Data and computations used i n r a c i a l a n a l y s i s o f evenand odd-year N o r t h American p i n k salmon on t h e b a s i s o f t h e i r s c a l e
c h a r a c t e r s . F i s h . Res. Bd. Can., Tech. Rpt. No. 234, l o p .
.
1972. A comparison o f body and s c a l e growth o f young sockeye
salmon (Onchorhynchus nerka) r e a r e d under two l i g h t p e r i o d s and
i n t o t a l darkness. F i s h . Res. Bd. Can., Tech. Rpt. No. 330, 12p.
B i l t o n , H.T., D.W. Jenkinson and M.P. Shepard. 1964. A key t o f i v e species
o f P a c i f i c salmon (Genus Oncorh nchus) based on s c a l e c h a r a c t e r s
J. F i s h . Res. Bd. C C 2 6 2 6 8 .
..
B i l t o n , H. T. and H. B. Messinger. 1975. I d e n t i f i c a t i o n o f major
B r i t i s h Columbia and Alaska r u n s o f 1.2 and 1.3 sockeye f r o m t h e i r
scale characters. I n : B u l l e t i n , International North P a c i f i c Fisheries
Commission 32: 109-128.
B i l t o n , H.T. and G.L. Robins. 1971a. E f f e c t s o f f e e d i n g l e v e l on
c i r c u l u s f o r m a t i o n on s c a l e s o f young sockeye salmon (Oncorhynchus
n e r k a ) . J. F i s h . Res. Bd. Can., 28: 861 :868.
.
1971 b. E f f e c t s o f s t a r v a t i o n , f e e d i n g and l i g h t p e r i o d on
c i r c u l u s f o r m a t i o n on s c a l e s o f young sockeye salmon (Onchorhynchus
nerka) I b i d . 28: 1749-1755.
.
1971 c . Response o f young sockeye salmon (Oncorhynchus nerka) t o
prolonged p e r i o d s o f s t a r v a t i o n . I b i d . 28: 1757-1761.
B i l t o n , H.T. and D.H. Smith. 1969. Scale c h a r a c t e r i s t i c s o f sockeye
salmon (Oncorhynchus nerka) o r i g i n a t i n g from small n u r s e r y areas of
t h e Skeena R i v e r system. F i s h Res. Bd. Can. Tech. Rpt. No. 133, 33p.
C l u t t e r , R . I . and L.E. Whitesel. 1956. C o l l e c t i o n and i n t e r p r e t a t i o n
o f sockeye salmon scales. B u l l . I n t . Pac. Salmon F i s h . Comm., No. 9,
159p.
.
Dixon, W .J. (ed.)
1976. BMD Biomedical Computer Programs.
Press, Los Angeles, 773p.
Univ. Calif.
F i s h e r , R.A.
1936. The use o f m u l t i p l e measurements i n taxonomic problems.
Ann. Eugenics 7: 179-188.
Henry, K.A.
1961. R a c i a l i d e n t i f i c a t i o n o f F r a s e r R i v e r sockeye
salmon by means of s c a l e s and i t s a p p l i c a t i o n t o salmon management.
B u l l . I n t . Pac. Salmon F i s h . Comm. No. 12, 97p.
I n t e r n a t i o n a l N o r t h P a c i f i c F i s h e r i e s Commission.
1961: 167p.
1963.
Annual Report
Klecha, W.R.
1975. D i s c r i m i n a n t A n a l y s i s . I n : SPSS S t a t i s t i c a l Package
f o r t h e S o c i a l Sciences. (ed. N.H. NeiC
,H
..
H u l l , J.G. Jenkins,
K. S t e i n b r e n n e r and D .H. B r e n t ) . McGraw H i 11 , N .Y. pp 434-467.
Koo, T.S.Y.
1962. D i f f e r e n t i a l s c a l e c h a r a c t e r s among species o f
P a c i f i c salmon. I n : S t u d i e s o f Alaskan r e d salmon (Ed. T.S.Y.
Koo) Univ. Wash. Press, S e a t t l e . pp. 127-135.
L a r k i n , P.A.
1972. The s t o c k concept and management of P a c i f i c salmon.
In The s t o c k concept i n P a c i f i c salmon (ed. R.C. Simson and P. A.
L a r k i n ) Univ. B r i t . Columbia, pp 11-15.
Major, R.L. and D.R. Craddock. 1962. Marking sockeye salmon s c a l e s by
s h o r t p e r i o d s of s t a r v a t i o n . U.S. F i s h and W i l d l . Serv., Spec.
S c i . Rpt, Fish., No. 416, 12 p.
Mason, J.E.
1967. Scale s t u d i e s o f sockeye salmon.
F i s h . Comm., Ann. Rpt. 1966; 99-111.
I n t . N o r t h Pac.
Mosher, K.H. 1963. R a c i a l a n a l y s i s o f r e d salmon by means o f scales.
B u l l . I n t . N o r t h Pac. F i s h . Comm., No. 11, p 31-55.
.
1972. S c a l e features of sockeye salmon from Asian and N o r t h
American c o a s t a l r e g i o n s . F i s h e r y B u l l . 7 0 ( 1 ) : 141 -183.
Nie, N.H., C.H. H u l l , J.G. Jenkins, K. Steinbrenner and D.H. Bent. 1975.
SPSS S t a t i s t i c a l Package f o r t h e S o c i a l Sciences. McGraw-Hill,
New York, 675p.
Pearson, R.E.
1964. P i n k salmon s c a l e s t u d i e s . I n : Report on t h e i n v e s t i g a t i o n s by t h e U n i t e d S t a t e s f o r t h e ~ n t e r n a t i o n a lN o r t h P a c i f i c
F i s h e r i e s Commission - 1963. I n t . N o r t h Pac. F i s h . Comm., Annual
Rpt, pp 162-165.
Ryan, P. and M. C h r i s t i e . 1976. Scale r e a d i n g equipment. F i s h e r i e s and
Marine Service, Canada, T e c h n i c a l Report No. PAC/T. -75-8. 38p.
Worlund, D.D. and R.A. F r e d i n . 1962. D i f f e r e n t i a t i o n o f s t o c k s . I n :
Symposium on p i n k salmon. H .R. MacMi 1 l a n Lectures i n ~ i s h e r i z ,
Univ. B r i t . Columbia, Vancouver, pp 143-153.
Wright, A.T.
1965. The use o f s c a l e c i r c u l i spacings as a means of
s e p a r a t i n g races of P r i n c e W i l l i a m Sound p i n k salmon. Alaska Dept.
o f F i s h and Game, I n f o r m a t i o n a l L e a f l e t No. 66, 27p.
APPENDIX
Appendix T a b l e 1
.
Scale c h a r a c t e r i s t i c s measured, by species. Data
were recorded as p a i r e d measurements of NCi and
IDi where:
NCi = Number o f c i r c u l i i n i t h c h a r a c t e r i s t i c
IDi = I n t e r v a l d i s t a n c e o f i t h c h a r a c t e r i s t i c
= Annulus formed d u r i n g j t h w i n t e r growth
period
fw
= freshwater
oc
= ocean
Unless o t h e r w i s e noted, measurements t o o r f r o m an annulus ( A j ) i n c l u d e
t h e c i r c u l i and d i s t a n c e t o o r f r o m t h e l a s t c l o s e l y spaced c i r c u l u s
f o r m i n g a p a r t o f t h a t annulus.
Sockeye (age 42, 52)
i= 1
2
3
4
= f o c u s t o A1 t o Ag ( 1 s t fw y e a r )
= A1 t o A* ( 1 s t oc y e a r )
= A2 t o A3 (2nd oc y e a r )
= A3 t o A4 ( 3 r d oc y e a r
-
age 52 o n l y )
Sockeye (age 53, 63)
i= 1
2
3
4
5
= focus t o A
( 1 s t fw year)
= A1 t o A2 ( I n d f w y e a r )
= A2 t o A3 ( 1 s t oc y e a r )
= A3 t o A4 (2nd oc y e a r )
= A4 t o A5 ( 3 r d oc y e a r - age 63 o n l y )
Sockeye smolt (age I )
i = 1 = f o c u s t o f i r s t c i r c u l u s o f A1 ( 1 s t f w summer)
2 = w i t h i n Al ( 1 s t f w w i n t e r )
3 = A1 t o s c a l e margin ( f w p l u s growth)
Sockeye smol t (age II )
i= 1
2
3
4
5
= f o c u s t o f i r s t c i r c u l u s o f A1 ( 1 s t f w summer)
= w i t h i n Al ( 1 s t f w w i n t e r )
= A, t o f i r s t c i r c u l u s AZ (2nd fw summer)
= w i t h i n A2 (2nd f w w i n t e r )
= A2 t o s c a l e margin ( f w p l u s growth)
Appendix Table 1.
(cant.)
chum (age 3], 4 ] )
i= 1
2
3
4
5
= focus t o l a s t c i r c u l u s supplementary ( f a l s e ) check
= supplementary check t o f i r s t c i r c u l u s A1 ( 1 s t oc summer)
= w i t h i n A1 ( 1 s t oc w i n t e r )
= A1 t o f i r s t c i c u l u s A2 (2nd oc summer)
= w i t h i n A2 (2nd oc w i n t e r )
Coho (age 43)
i = 1 = focus t o A1 ( 1 s t fw y e a r )
= *1 to A2 (2nd fw y e a r )
3 = A2 t o A3 ( 1 s t oc y e a r )
Chinook (age 52)
i
= 1 = focus t o f i r s t c i r c u l u s A1 ( f w summer)
2
3
4
5
6
= w i t h i n A1 ( f w w i n t e r )
= A1 t o f i r s t c i r c u l u s A2 ( 1 s t oc summer)
= w i t h i n A2 ( 1 s t oc w i n t e r )
= A2 t o f i r s t c i r c u l u s A3 )2nd oc summer)
= w i t h i n A3 (2nd oc w i n t e r )
Append~xFigure
1,
Hydrat~lic s c a l e press used t o make impressions
of salmon s c a l e s on a c e t a t e p l a s t i c cards.
Apperldix Figure 2, L e i La niicro-projector and s c a l e impression c a r d ,
Appendix Figure 3.
Micro-projector, mirror assenibly and table surface f o r p r o j e c t i o n o f
scale inlages and marking of scale c h a r a c t e r i s t j c s ,
DATA
SYSTEM
YEAR
C A R D No
PAGI
Appendix Figure 4 .
---
DATE R E A D
READER
5CALES
Scale d r a w i n g sheet with pre-printed axis l i n e s ,
Short l i n e s i n t e r s e c t i n g axes indicate positions of
C I T C U ~I , longer l ines represent positions of annul i ,
(Actual paper s i z e = 2 2 i n , by 17 i n . )
Appendix Figure 5 .
Linear d i g i t i z i n g rule (Rouchi r u l e ) a n d remote d i g i t i z e r c o n t r o l s ,
Appendix Figure 6.
Digitizer electronics. Nunlbers visible in lower windows are f i x e d san-iplc!
informat-ion and are controlled by resettable xhumbwheels.
Appendix Figure 7 .
C o n f i g u r a t i o n o f Rouchi rule, d i g i t i z e r and ASR-33,
Append-ix Figure 8 .
Configuration o f paper tape reader,
computer
t e r m i n a l and acoustic coupler,
ONE
.NO
ONE
TWO
I '
I
I
I
I
f
-
i
1I *
II
NO
ONE
1
SELECTED
I
Appendix Figure 9. Diagrammatic representation of a section of a scale
within 0.1 mm ( 10.0 mm magnified image) of the selected axis, showing
"breakage" and "branching" of circuli and indicating the criteria for
inclusion of ci rculi in counts and measurements. "One" and "two"
indicate the number of circuli that would be included in counts and
for which positions would be marked for measurement. "No" indicates
that circulus would not be included in counts and that the position
of its image would not be marked for measurement. (after Bi 1 ton, 1971 )
Appendix Figure 10. Sample hardcopy output of digitized scale measurement
data from ASR-33.
Appendix Figure 1 1 . Data shown in Figure 10 after editing, sorting, compacting
and conversion from cummulative to incremental measurement.
The Alaska Department of Fish and Game administers all programs and activities free from discrimination
based on race, color, national origin, age, sex, religion, marital status, pregnancy, parenthood, or disability.
The department administers all programs and activities in compliance with Title VI of the Civil Rights Act
of 1964, Section 504 of the Rehabilitation Act of 1973, Title II of the Americans with Disabilities Act of
1990, the Age Discrimination Act of 1975, and Title IX of the Education Amendments of 1972.
If you believe you have been discriminated against in any program, activity, or facility, or if you desire
further information please write to ADF&G, P.O. Box 25526, Juneau, AK 99802-5526; U.S. Fish and
Wildlife Service, 4040 N. Fairfax Drive, Suite 300 Webb, Arlington, VA 22203 or O.E.O., U.S.
Department of the Interior, Washington DC 20240.
For information on alternative formats for this and other department publications, please contact the
department ADA Coordinator at (voice) 907-465-6077, (TDD) 907-465-3646, or (FAX) 907-465-6078.