Beacham et al : Population identincation of Oncorhynchus tshawytscha by variation at microsatellite loci 



247 



obtained by using expected genotypic frequencies for all 

 loci. A comparison of estimated stock composition of a CWT 

 sample with all loci in expected genotypic frequencies in 

 the baseline populations compared with observed genotypic 

 frequencies at Otsl02 resulted in more accurate estimates 

 with expected genotypic frequencies at all loci. Similar 

 results were observed by Beacham et al. (2001) in mixed- 

 stock analysis of coho salmon. Stock compositions from the 

 1995 test fishery at Albion were estimated with five loci 

 iOtslOO, OtslOl, Otsl02, Otsl04, and Ssal97), those from 

 the 1997 net fisheries with six loci (previous five loci plus 

 Otsl07), and those from 1998 and 1999 fisheries with 13 

 loci. The change in the number of loci used in estimation of 

 stock composition over time was reflected in the number of 

 loci available in the baseline populations. The initial analy- 

 sis of baseline populations started in 1994 and three loci 

 were included the population survey (Nelson et al., 2001). 

 By early 1996, at the time of analysis of the first test fish- 

 ery samples, five loci were routinely scored in the baseline 

 populations and mixed-stock fishery samples, and by 1997 

 six loci were routinely scored for both baseline and mixed- 

 stock fishery samples. With the acquistion of an automated 

 sequencer in our laboratory in 1998, an additional seven 

 loci were added to the routine survey of baseline population 

 and mixed-stock fishery samples. 



Reported stock compositions for the CWT and actual 

 fishery samples are the point estimate of each mixture ana- 

 lyzed, with variance estimates derived from 100 bootstrap 

 simulations. Each baseline population and fishery sample 

 was sampled with replacement in order to simulate ran- 

 dom variation involved in the collection of the baseline and 

 fishery samples. Reported stock composition for simulated 

 mixtures was the bootstrap mean, along with the standard 

 deviation of the mean. 



Results 



Comparison of individual loci 



Determination of the relative power of individual loci for 

 either population or regional discrimination is of prime sig- 

 nificance for practical stock identification applications. In 

 simulations comparing the relative power of the microsat- 

 ellite loci to estimate stock compositions of representative 

 single-population samples, there were only minor differ- 

 ences in the relative power of the best nine loci, with "best" 

 defined as those loci resulting in the minimum bias in the 

 estimated stock compositions. The mean error of the esti- 

 mates ranged between 20% and 31% (Table 3). The power of 



