An Estimate of

Spiridon Lake Sockeye Salmon

Commercially Harvested Within the

Southwest Afognak Section and Northwest

Kodiak Management Area, 1996

By

Patricia A. Nelson

and

Charles O. Swanton

Regional Information Report[1] No. 4K97-44

Alaska Department of Fish and Game

Commercial Fisheries Management and Development Division

211 Mission Road

Kodiak, Alaska 99615

August 1997

ACKNOWLEDGMENTS

The Kodiak Regional Aquaculture Association provided funding for this project. Kodiak area field staff were responsible for data collection. Spiridon Lake project leader Steve Schrof and field crew collected terminal fishery data. Special thanks to Chris Hicks for scale age classification and measurement assistance and Chris Anderson for his efforts in Kodiak catch sampling. Lucinda Neel contributed publication expertise. Additional editing was provided by Dave Prokopowich, Kevin Brennan, Nick Sagalkin, and Steve Schrof.

TABLE OF CONTENTS

Page

LIST OF TABLES...... i

LIST OF FIGURES...... ii

LIST OF APPENDICES...... iii

INTRODUCTION...... 1

METHODS...... 2

Study Area...... 2

Stock Selection...... 2

Escapement Sampling...... 2

Commercial Catch Sampling...... 3

Age Designation of Catches and Escapements...... 3

Scale Pattern Analysis and Stock Composition Estimation...... 3

Spiridon Catch Assignment...... 5

Freshwater Growth Comparison...... 5

RESULTS...... 6

Stock Selection...... 6

Age Composition Estimates of Selected Catches...... 6

Stock Separation Model...... 6

Stock Composition Estimates...... 7

Estimated Spiridon Sockeye Salmon Catch and Run Timing...... 7

Freshwater Growth...... 7

DISCUSSION...... 8

LITERATURE CITED...... 10

TABLES...... 13

FIGURES...... 16

APPENDIX...... 24

LIST OF TABLES

TablePage

1.Classification accuracy, by stock, of the age 1.2 scale pattern analysis model...... 13

2.Stock composition estimates and estimated harvest of Spiridon sockeye salmon by area and week, 1996 14

3.Estimated number of Spiridon Lake sockeye salmon harvested by area and week, 1996 15

LIST OF FIGURES

FigurePage

1.Map of the Kodiak Management Area showing fishing districts and major sockeye systems including the Spiridon Lake 16

2.Map identifying the approximate boundaries of the Spiridon Lake Terminal Harvest Area in Telrod Cove, 1996 17

3.Photograph of the Spiridon Lake Terminal Harvest Area at Telrod Cove, 1996 ...18

4.Estimated age 1.2 component of selected escapements by system and week, 1996..19

5.Estimated Spiridon sockeye salmon commercially harvested in the SW Afognak Section and NW Kodiak District, 1996 20

6.Estimated percentage of Spiridon Lake sockeye salmon harvested in the SW Afognak Section, and the NW Kodiak, and SW Kodiak Districts by catch area and year, 1994 - 1996.. 21

7.Estimated Spiridon sockeye salmon harvested by area and week, 1996...... 22

8.A comparison of the 1996 Spiridon run timing based on harvest estimates (in number of fish) and Upper Station late run escapement timing (1974-1996 average percent). 23

LIST OF APPENDICES

Page

Appendix A: Age Composition Estimates

A.1. Estimated age composition of Karluk sockeye salmon escapement by week, post 14 July 1996 25

A.2. Estimated age composition of Ayakulik sockeye salmon escapement by week, post 14 July 1996 26

A.3. Estimated age composition of Upper Station sockeye salmon escapement by week, post 14 July 1996 27

A.4. Estimated age composition of Frazer sockeye salmon escapement by week, post 14 July 1996 28

A. 5. Estimated age composition of Southwest Afognak Section (251-10 and 20) sockeye salmon catch by week, post 14 July 1996 29

A. 6. Estimated age composition of Uganik harvest area (253-11to 35) sockeye salmon catch by week, post 14 July 1996 30

A.7.Estimated age composition of Uyak harvest area (254-10 to 40) sockeye salmon catch by week, post 14 July 1996 31

A.8. Estimated age composition of Spiridon Lake Terminal Harvest Area (254-50) sockeye salmon catches by week, 1996 32

Appendix B: Scale pattern analysis

B.1.Descriptive statistics for age 1.2 scale variables from Spiridon, and Upper Station stocks, 1996 33

B.2.Scale pattern of age 1.2 sockeye salmon collected at Spiridon Lake Terminal Harvest Area (SLTHA), 4 August 1996 34

B.3.Scale pattern of age 1.2 sockeye salmon collected at Upper Station weir, 30 July 1996 35

INTRODUCTION

Accurate assignment of salmon catches to stock or system of origin is necessary for evaluation of productivity trends, estimating smolt-to-adult survival, and preseason forecasting. For the purpose of estimating stock specific production within the Kodiak Management Area, a comprehensive salmon catch and escapement sampling program was initiated in 1985. This run reconstruction program initially focused on four major Kodiak sockeye salmon Oncorhynchus nerka stocks, including Karluk, Ayakulik, Upper Station, and Frazer (Barrett and Nelson 1995; Swanton 1992). During 1994, it was anticipated that an enhancement project at Spiridon Lake would be realizing it’s first substantial sockeye return (>100,000 fish), which were expected to be harvested in traditional westside fishing areas. In order to maintain consistency in run reconstruction programs for wild stocks as well as quantify returns from the Spiridon Lake enhancement project, a novel stock identification program was implemented (Nelson and Barrett 1994).

Scale pattern analysis along with visual identification of freshwater growth patterns have been employed since 1994 to estimate the contribution of Spiridon Lake sockeye salmon to the Northwest, and Southwest Kodiak Districts’ commercial catches. The objective of using scale pattern analysis (SPA) is to develop a statistical model that accurately identifies individual scales from known stocks within mixed stock fisheries or unknown samples. The freshwater scale pattern of Spiridon Lake sockeye salmon has remained consistent and easily distinguishable, lending itself to the application of this approach.

During 1994, a Spiridon stock contribution of 263,750 fish was estimated using visual freshwater scale pattern identification (Nelson and Barrett 1994). Scale pattern analysis incorporated in conjunction with visual scale pattern identification in 1995 resulted in a Spiridon sockeye salmon harvest estimate of 96,705 (Nelson and Swanton 1996a). Over 90% of these harvests occurred in the Northwest Kodiak District. Estimates were considered to be minimum values as this stock may have contributed to other Kodiak fisheries outside the Northwest Kodiak and Southwest Kodiak Districts, specifically the Southwest (SW) Afognak Section.

The migration and run timing of the Spiridon Lake stock appear to be very similar to that of the Upper Station late run brood stock source (Nelson and Swanton 1996a). Tagging experiments conducted in 1981 as well as catch estimates from 1994 and 1995 suggest that a majority of this run migrates from north to south along Kodiak Island’s westside peaking in mid to late August (Nelson and Swanton 1996a, Nelson and Barrett 1994, Tyler et al. 1986).

Unlike the migration timing, the age composition of Spiridon Lake sockeye salmon does not appear to be consistent with the brood stock source. During both 1994 and 1995, age 1.2 was the dominant age class of sockeye salmon sampled in the Spiridon Lake Terminal Harvest Area (SLTHA) while the majority of Upper Station late run sockeye salmon were classified as age 2.2 fish.

This report serves as the third in a series of annual run reconstruction estimates of Spiridon Lake sockeye salmon (Nelson and Swanton 1996a, Nelson and Barrett 1994). The objectives of this report are threefold: first to estimate the number of Spiridon sockeye salmon harvested in the SW Afognak Section and Northwest (NW) Kodiak District; second, to monitor run timing based on commercial harvest estimates; and finally, to compare and quantify annual freshwater growth of Spiridon sockeye salmon from scales collected during 1994, 1995, and 1996.

METHODS

Study Area

Spiridon Lake, located in the Central Section of the NW Kodiak District, is the third largest lake (9.6 km long, 1.6 km maximum width) on Kodiak Island (Figure 1). Prior to 1991, a series of barrier falls prevented access to the lake outlet and precluded the presence of anadromous fish (Kyle et al. 1990). Introduction of sockeye salmon fry to this system coupled with the construction of a smolt bypass system resulted in an artificial run. In 1993, the State Board of Fisheries (BOF) adopted the Spiridon Lake Sockeye Salmon Management Plan (5 AAC 18.366). In accordance with this plan, the Spiridon Lake sockeye run is intended to be harvested primarily in traditional commercial fishing areas of the NW Kodiak District during openings directed on local stocks (ADF&G 1996). The remainder is to be taken in an exclusive purse seine and beach seine terminal harvest area in Telrod Cove within Spiridon Bay (Figure 2).

The catch areas and time frame considered for this study (SW Afognak Section and NW Kodiak District; post-11 July) were based on spatial and temporal Spiridon harvest estimates from 1994 and 1995 (Nelson and Barrett 1994; Nelson and Swanton 1996a). The Southwest Kodiak District was not included in the 1996 study based on negligible estimates of Spiridon fish observed in this district during 1994 and 1995.

Stock Selection

Selection of the sockeye stocks to include in this analysis was based on historic run timing within the commercial catch areas of interest, and evaluation of the 1996 escapement age composition estimates post-14 July. All major Kodiak sockeye systems with a westside run potential of greater than 50,000 fish during July and August combined, and an age 1.2 escapement component of greater than 10% within any given week were considered (Barrett and Nelson 1994).

Escapement Sampling

Sockeye salmon escapements were sampled weekly for age (scales) at weir sites in the Kodiak Management Area (KMA; ADF&G 1995). These systems include Karluk, Ayakulik (Red Lake), Frazer, and Upper Station (Figure 1). Terminal catches were sampled weekly in the Spiridon Lake Terminal Harvest Area at Telrod Cove (statistical area 254-50; Figures 2 and 3) and were assumed to represent Spiridon escapement. The targeted sample size was 240 fish per system per week (Nelson and Swanton 1996b) .

Commercial Catch Sampling

During July and August, commercial sockeye salmon catches were sampled for age with a targeted sample size of 600 fish per week from the following areas:

Afognak District (Figure 1)

Southwest Afognak Section (Statistical Areas 251-10 through 251-20);

Northwest Kodiak District (Figure 1)

Central Section

Uyak Catch Area (Statistical Areas 254-10 through 254-40), and

Uganik Catch Area (Statistical Areas 253-11 through 253-35);

Age Designation of Catches and Escapements

Scales were collected from the preferred area following procedures outlined in INPFC (1963), mounted on gum cards, and impressions were made on cellulose acetate (Clutter and Whitesel 1956). Fish ages were assigned by examining scale impressions for annual growth increments using a microfiche reader fitted with a 48X lens following designation criteria established by Mosher (1968). Ages were recorded on sampling forms using European notation (Koo 1962). Age composition estimates of stock specific escapements and catches by area were assigned based on samples collected. Catch-at-age by area and day was estimated by multiplying the daily age composition of a particular sample by the daily catch from the corresponding catch area. Age composition of the catch from days not sampled was estimated using linear interpolation between sampling events (Blackburn 1993). Estimated age composition of escapements from major Kodiak systems followed similar procedures.

Scale Pattern Analysis and Stock Composition Estimation

Maximum sample sizes of 200 scales from the dominant age class per stock were selected for establishing standards (Cook 1982). The Spiridon Lake standard was constructed using terminal area catch samples collected from 29 July through 9 September. The Upper Station late run standard was constructed from weir escapement samples collected from 23 July to 30 August.

Mixed stock fishery samples (“unknowns”) meeting the selection criteria had a minimum desired weekly sample size of 30 and a maximum of 100 scales by harvest area. Scale measures were obtained by starting with the first scale of the selected age class within the sample and continuing until all scales suitable for measurement had been exhausted or the sample size was met.

Scale measurement data were collected using the Biosonics[2] optical pattern recognition system (OPRS), which integrates a compound microscope, ocular lens, frame grabber, digitizing tablet, and microcomputer.

Scale-data collection procedures consisted of:

(1) establishing a horizontal reference line below the scale focus through the reticulated region;

(2) identifying the center of the scale focus or measurement initiation point;

(3) measuring incremental distances from scale focus to each circuli within the first freshwater annular zone off an axis perpendicular to the reference line (Narver 1963);

(4) saving measured data to a unique computer file.

All scale measurements were specific to a single age class (age 1.2) utilizing 200X magnification. Scales with poorly defined images and those collected from a non-preferred region (Clutter and Whitesel 1956) were not measured.

Raw OPRS scale measures were transformed into individual variable format for both standard and unknown files using a BASIC program, REFORM1 (written by Larry Greer, ADF&G, Kodiak, AK). Variables constructed were circuli counts (C.C.) and incremental distances (I.D.) which start at the scale focus and end with the last circulus of the first freshwater annulus. These variables reflect the growth that occurred during the freshwater phase (lake residence) of each stock’s life history. The maximum number of variables available for model development was constrained to the fewest number of circuli counted on any of the stocks included (e.g., if a stock had one scale with only 10 circuli, then the maximum number of potential variables describing the freshwater growth of that stock would be 11; one circuli count variable and 10 incremental distance variables).

A linear discriminate function (LDF; Fisher 1936) was employed for classifying unknown mixed stock fishery samples to stock of origin (Dillon and Golstein 1984). Assumptions associated with using both SPA and the LDF were (1) all probable stocks contributing to the commercial fishery samples were represented in the model; and (2) scale variables from each stock were multivariate normal; and (3) variance-covariance matrices between groups were equal. Evaluation of univariate normality was assessed using frequency histograms for all variables of each stock considered. All variables assumed normal in distribution were included in the discriminant model. Accuracy of a model in correctly classifying individuals to actual stock of origin was determined by the “leaving-one-out” approach of Lachenbruch (1967). Stock composition estimates for unknown samples (by area and time period) were corrected for misclassification error using the matrix correction approach of Cook and Lord (1978), with 90% confidence coefficients calculated using the variance formula of Pella and Robertson (1979). Confidence coefficients for the two stock model were generated assuming a chi-square distribution. All discriminant modeling was completed using PROC DISCRIM (SAS Institute 1987).

During weeks when stock contribution estimates were not available (no catch sample collected) we employed linear interpolation between adjacent weekly stock composition estimates. To derive stock composition estimates for a maximum of two weeks following the last sample obtained (e.g. sample obtained on 15 July with no sample on 21 July) then the prior weeks stock contribution estimate was assumed static and applied. The harvest by area and week was then apportioned based on these stock composition estimates.

Spiridon Catch Assignment

Sockeye salmon catch numbers by area were obtained from the Alaska Department of Fish and Game (ADF&G) fish ticket database on 28 January 1997. Apportionment of the age 1.2 component of the commercial catch by week within the aforementioned commercial fishing areas was accomplished by multiplying the estimated weekly age 1.2 component of the catch by the weekly stock composition estimate of Spiridon Lake sockeye salmon. The expansion of this estimate to include other age classes was based on the weekly ratio of age 1.2 fish to all other age classes in the Spiridon stock using the following formulae:

(1)

(2)

Freshwater Growth Comparison

A single-factor ANOVA was computed (Microsoft Excel[3] 5.0, 1985-1994) to compare annual freshwater scale growth of age 1.2 Spiridon sockeye salmon between years (1994, 1995, 1996). A total of 200 age 1.2 scales (100 per year) were measured from samples collected from SLTHA during 1994 and 1995 while age 1.2 scales measured for the 1996 Spiridon standard were used to represent this years annual growth. Scale measures were obtained by starting with the first age 1.2 scale within the sample and continuing until 100 age 1.2 scales suitable for measurement had been met. All measurements were collected by a single reader and scale data collection followed procedures described previously. The studentized range test (T Method; Devore 1995) was used to identify significant differences in a multiple comparison analysis between yearly freshwater growth while maintaining an  value of 0.05. This test was performed only if the value of the computed F statistic in the ANOVA was significant.

RESULTS

Stock Selection

Stocks meeting the run timing criteria included Karluk late run, Ayakulik, Upper Station late run, and Frazer. However, based on age composition estimates, Upper Station late run was the only stock included in this analysis (Appendix A1-4 , Figure 4). Although Karluk late run is a substantial contributor to westside Kodiak catches during July and August (Barrett and Nelson 1995), based on escapement age composition estimates, it was not contributing to the age 1.2 component of the catch (Appendix A1). Age composition estimates of Ayakulik (Red River) sockeye salmon escapement post-14 July consisted primarily of age 2.2, and 2.3 fish with an overall age 1.2 component of less than 1% (Appendix A2). The Upper Station post 14 July escapement consisted of three age classes (0.2, 1.2, and 2.2) representing 83% of the escapement (Appendix A3). Age 1.2 fish represented 10.0% (range 3.7 %-17.2 %) of the escapement post-14 July. The dominant age classes contributing to the Frazer Lake escapement were age 2.1, 2.2, and 2.3 (Appendix A4).

Age Composition Estimates of Selected Catches

Based on commercial catch sampling efforts, over 30% of the overall commercial harvest in each of the three catch areas were estimated to be age 1.2 fish (Appendix A5 - 8). A total of 1,157 sockeye salmon were sampled from the SW Afognak section post 14 July with 1.2 and 2.2 fish representing an estimated 74% of the catch (Appendix A5). Of the 3,201 scales collected from the Uganik area post-14 July, age 1.2 was the dominant age class (39.3%) and three age classes combined (age 1.2, 1.3, and 2.2) composed greater than 87% of the catch (Appendix A6). For the Uyak commercial catch area, 2,388 fish were sampled during the study period, and age 1.2 and 2.2 fish represented 77.5% of the commercial sockeye salmon catch post-14 July (Appendix A7). Harvests within the SLTHA were sampled from 26 July through 12 September. Age 1.2 fish contributed an estimated 79% of the catch followed by age 2.2 fish which represented an additional 14.3% (Appendix A8).