248 



Fishery Bulletin 101(2) 



a single locus to provide accurate estimates of 

 stock composition varied considerably among 

 loci and among populations. For example, the 

 mean error in estimation of stock composition 

 of a sample of pure Birkenhead River chinook 

 salmon was less than 1% when only Ogo4 

 was used to estimate stock compositions but 

 ranged as high as 40% when a sample of pure 

 Quesnel River chinook salmon was evaluated. 

 Clearly, not all loci were equally effective in 

 stock identification, and the usefulness of the 

 loci varied among populations. The accuracy 

 of the estimates generally improved with an 

 increasing number of loci used in the estima- 

 tion procedure, but the increase in accuracy 

 was generally marginal after eight loci were 

 used to estimate stock compositions. In the case 

 of the Birkenhead River, additional loci did not 

 increase the accuracy over that observed with 

 only Ogo4. The precision of the estimates gen- 

 erally increased with an increasing number 

 of loci used, but the increase in precision was 

 marginal when the least effective five loci were 

 added to the estimation procedure. 



On average, the number of alleles present at 

 a locus was related to the power of the locus to 

 provide accurate estimation of stock composi- 

 tion. For example, the mean bias of estimated 

 stock composition for loci with fewer than 20 

 observed alleles {Ogo2, Oke-4. Ots2, Ots9) was 

 45% per locus, whereas the mean bias for loci 

 with 20 or more alleles was 26% per locus 

 (Table 3). Loci with fewer than 20 observed 

 were generally less valuable for stock iden- 

 tification applications than loci with greater 

 numbers of alleles. 



Population estimation of 

 stock composition 



We evaluated whether the degree of genetic 

 differentiation observed among Fraser River 

 populations included in the baseline was suf- 

 ficient for mixed-stock analysis in which the 

 objective was estimation of specific population 

 contributions to fishery samples. Three simu- 

 lated fishery mixture samples were developed, 

 representing an early, middle, and late-timing 

 return to the lower Fraser River With 14 pop- 

 ulations present in a simulated spring fishery 

 sample, the mean error in population specific 

 estimated stock composition with a 50-popula- 

 tion baseline was 1.4% (Table 4). Similar mean 

 population-specific error rates were observed 

 in the simulated summer fishery sample con- 

 taining fish from 10 populations ( 1.2% ), and in 

 the simulated fall sample containing fish from 

 seven populations (1.1%). Regional estimates 

 of stock contributions were all within 2% of 

 the actual value. We concluded tliat accurate 



