Beacham et al : Microsatellite DNA variation and estimation of stock composition of Oncorhynchus nerka 



23 



Successfial application of microsatellite loci to esti- 

 mation of stock composition in mixed-stock fisheries 

 requires that loci be chosen that highlight differences 

 among stocks to be separated and that adequate num- 

 bers of fish in the baseline stocks be surveyed to pro- 

 vide reliable estimates of allele fi-equencies, and thus 

 genotypic fi-equencies used in the conditional maxi- 

 mum likelihood analysis. Microsatellite loci can con- 

 tain a large number of alleles, and baseline sample 

 sizes need to be of sufficient size to ensure that alleles 

 present in fish fi-om a stock in the mixture have also 

 been observed in the baseline samples. Binning low- 

 frequency similar-size alleles (Small et al., 1998) is 

 also a strategy to consider in practical applications. 



Although simulated mixtures can provide insights 

 into the expected performance of the mixture analy- 

 sis, the stock contribution estimates for actual fishery 

 samples can only be evaluated by corroboration with 

 data fi-om other sources. Two supportive sources of 

 independent information occur: time of return of the 

 Henderson Lake stock and the typical catch composi- 

 tion for Barkley Sound that was previously derived 

 from parasites. In Barkley Sound, the time of return 

 of Henderson Lake sockeye salmon has been reported 

 to be later than that of either Sproat Lake or Great 

 Central Lake fish (Steer et al., 1988). For example, in 

 1984, Henderson Lake sockeye salmon were evident, 

 on the basis of parasite analysis, in the commercial 

 fishery prior to 27 June but increased in relative abun- 

 dance after that time. The current analysis indicated 

 that Henderson Lake sockeye salmon were absent 

 from, or at low abundance in, the 1997 commercial 

 fishery prior to the week of 4 July. Analysis of the 

 gillnet test fishery and purse-seine samples indicated 

 that the proportion of Henderson Lake sockeye salmon 

 in those catches was low until mid-July but thereafter 

 was substantial, consistent with a later time of arrival 

 of the Henderson stock in Barkley Sound. In a typi- 

 cal return year, about 60% of the Barkley Sound sock- 

 eye salmon catch is derived fi-om Great Central Lake, 

 30% fi-om Sproat Lake, and 10% fi-om Henderson Lake 

 (Steer et al., 1988). Estimated stock compositions for 

 the 1997 fishery catches are in reasonable agreement 

 with the expected stock contributions. Results of the 

 simulation analysis indicated that more precise, but 

 not necessarily more accurate, estimates of the stock 

 contributions (especially that from Henderson Lake) 

 could have been obtained for the fishery catches if 

 sample sizes had been larger than 50 (for the gillnet 

 test fisheries) or approximately 100 (for the purse- 

 seine test and commercial gillnet fisheries). 



Differences in estimated stock composition were 

 obtained for the purse-seine and gillnet test fisheries 

 in July samples, where higher proportions of Great 

 Central Lake sockeye salmon were observed in the 



gillnet fishery samples. Although the fishery samples 

 came ft-om different areas (the purse-seine samples 

 were collected farther inland in Albemi Inlet than 

 were the gillnet samples), the most likely explana- 

 tion of the difference in estimated proportions of stock 

 composition between the two gears is a difference in 

 size selectivity. Sockeye salmon caught in purse seines 

 in Barkley Sound are generally more variable in size 

 and of smaller mean size than those caught in gill nets 

 (Steer et al., 1986). Probably gill nets were more selective 

 for Great Central Lake sockeye salmon than for Sproat 

 Lake salmon. Thus, it is important to estimate stock con- 

 tributions to a fishery catch based on samples collected 

 with the type of gear employed in the fishery. Further- 

 more, the analysis of catch samples to estimate the stock 

 composition of fish present in an area (as opposed to 

 those caught in an area) vriH be biased to the degree that 

 the sampling gear nonrandomly catches the fish that 

 are present. The results of this study and other analyses 

 (Beacham et al., unpubl. data) indicate that for salmo- 

 nids, different gear types sample the various stocks in a 

 stock mixture with very different efficiencies. 



Differentiation among local spawning populations 

 that are relatively stable over time provides the basis 

 for applying biological markers to problems of salmo- 

 nid fisheries management. This study confirmed our 

 expectation that the level of differentiation observed 

 at microsatellite loci among sockeye salmon of the 

 three major lake systems draining into Barkley Sound 

 is sufficient and stable to assess stock composition of 

 the fishery catches. The abundance of highly polymor- 

 phic microsatellite loci in salmonid fish, the relative 

 ease of nonlethal sample collection, and the moderate 

 cost per fish for laboratory analysis combine to provide 

 a technology that will become increasingly used in the 

 assessment and management of salmonid fisheries. 



Acknowledgments 



We would like to acknowledge all those involved in 

 sampling collections irom sockeye salmon spawning 

 grounds. The collection of test fishery samples was 

 supervised by Bruce Patten and Jim Mitchell, and 

 scale samples from the 1997 fisheries were provided 

 to us by D. Gillespie of the Ageing Laboratory at the 

 Pacific Biological Station. J. Candy assisted in some of 

 the data analysis and figure preparation. Funding was 

 provided by the Department of Fisheries and Oceans. 



Literature cited 



Angers, B., L. Bematchez, A. Angers, and L. Desgroseillers. 

 1995. Specific microsatellite loci for brook charr reveal strong 



