646 



Fishery Bulletin 101(3) 



then subtracted P^,^ for each region from the genetic esti- 

 mate of the region's contribution to the sample of unmarked 

 fish. The sum of the remaining values estimated the pro- 

 portion of wild fish in the sample of unmarked fish. When 

 Pijfj for a region was greater than the genetic estimate of 

 the region's contribution to the sample of unmarked fish, 

 the percentage of wild fish from that region was considered 

 to be zero. 



We estimated regional proportions of hatchery and wild 

 coho salmon in the all-fish marine sample that included 

 both marked and unmarked coho salmon. Regional hatch- 

 ery contributions to the all-fish sample were made by sum- 

 ming each region's estimated contribution to the sampled 

 marked and unmarked fish, weighted by the proportion of 

 each of these sample types in the total sample. Regional 

 proportions of wild coho salmon in the all-fish sample 

 were made by multiplying a region's estimated proportion 

 of wild coho salmon in the unmarked sample by the propor- 

 tion of unmarked fish in the total sample. 



Results 



Baseline genetic data and population structure 



Although coho salmon generally have low levels of genetic 

 variability in relation to other Pacific salmon, a sufficient 

 number of polymorphic loci were detected to distinguish 

 many populations and regional population groups. Of 59 

 loci screened in all 89 populations, 56 were polymorphic, 

 and 29 of these were at the Pq 95 level of polymorphism 

 in at least one population (Table 2). Allelic frequencies 

 are reported in an appendix that can be retrieved at 

 the Northwest Fisheries Science Center website [http: 

 //www.nwfsc. noaa.gov]. Twenty of the 56 polymorphic loci 

 had two alleles per locus, 24 had three alleles per locus, 

 nine had four alleles, two had five alleles, and one had six 

 alleles. Two loci (BGALA* and PEPC* ) varied in all popula- 

 tions studied. Three loci (GAPDH-5*, LDH-C*, and TPI-2*) 

 were monomorphic in all populations. Observed genotypic 

 proportions for polymorphic loci in 128 samples departed 

 significantly (P<0.05) from expected Hardy- Weinberg pro- 

 portions in 75 of 1476 tests (5.1% ). There were no consis- 

 tent trends by population or locus. Because the number of 

 significant tests is close to the number expected by chance 

 for this rejection level, we did not attach any biological 

 significance to these departures. 



The percentages of Ppgr^ loci and average heterozygosi- 

 ties over 56 loci for each population appear in Table 1. The 

 percentage of P^^^ loci ranged from only 5.4% in Lewis 

 River hatchery early run (population 41) to 17.9% in the 

 Mad River hatchery (4). Average heterozygosities ranged 

 from 0.021 in Iron Gate hatchery (5) and Elk River (9) to 

 0.046 in Sandy River hatchery (46 ). Gene diversity analysis 

 of the 89 populations resulted in a total gene diversity (//•,.) 

 of 0.035 and an average sample diversity (H^) of 0.033. 

 Thus, 94.5% of the total genetic diversity was attributable 

 to within-sample variability and 5.5% was attributable to 

 variability among samples. About 2.9% of the total gene 

 diversity was due to variability among populations within 



regions, and 2.6% was due to variability among the nine 

 regions. 



Genetic relationships among populations of coho salmon 

 as revealed by two-dimensional MDS analysis showed that 

 genetic differences among populations were geographically 

 structured (Fig. 2). The first axis in the plot separated pop- 

 ulations in coastal Oregon and California from northern 

 populations. Several populations, including two from the 

 Rogue River in southern Oregon (numbers 7 and 8) and 

 Big Qualicum hatchery (85) on Vancouver Island, were po- 

 sitioned near the convergence of the southern and northern 

 population groups. The Iron Gate hatchery sample (5) from 

 the Klamath River, California, clustered with the northern 

 population group. Several genetically discrete groups ap- 

 peared on smaller geographical scales. However, samples 

 from Iron Gate hatchery (5), Yaquina River (27), Nehalem 

 hatchery (33), Willapa Bay area (50, 51, and 52), Dungeness 

 hatchery (64), McGovern Creek (74), upper Cascade River 

 (81), and Ennis Creek (82) did not cluster with nearby 

 populations. The single population in our study from the 

 upper Fraser River region — Spius hatchery (89) of the 

 Thompson River-was the most genetically distinct in the 

 MDS analysis (.r=-2.3,y=-0.9) and was positioned beyond 

 the scaling shown in Figure 2. The Little River (2) popula- 

 tion also fell outside the area of the plot {.v=5.1,y=2.0), but 

 was genetically most similar to other California coastal 

 populations (1, 3, and 4). 



Genetic estimates of simulated stock mixtures 



One demonstration of discreteness among regional groups 

 is the correct allocation in a mixed-stock analysis of simu- 

 lated samples from baseline populations to their stock of 

 origin. We used simulated sample sizes of 100, 300, and 

 500 taken from one region at a time; therefore the results 

 represent the accuracy of reallocation back to the region of 

 origin. Table 3 presents the average values of 100 bootstrap 

 resamplings of both the baseline and the mixture samples. 

 For simulated sample sizes of 100, reallocation accuracy 

 ranged from 81% (coastal northern Washington) to 98% 

 (upper Fraser River population) and averaged 88.7% over 

 the nine regions. Average accuracy increased to 92.9% with 

 an increase in the size of the simulated sample to 300. Only 

 marginal improvement (93.6%r accuracy) was achieved by 

 increasing the simulated sample size to 500. 



We also used mixed-stock analysis of simulated samples 

 to examine the accuracy of composition estimates for Cali- 

 fornia, Puget Sound, and British Columbia regions when 

 fish from these areas were not present in mixtures. Average 

 values for sample sizes of 100 ranged from 0%^ (California 

 coast, upper Fraser River) to 4% (Oregon coast) and aver- 

 aged 1.8% over the five regions (Table 4). Increased sample 

 sizes of 300 and 500 resulted in small improvements in 

 average accuracy (1.4%^ and 1.0 %). 



Stock compositions of ocean-caught coho salmon 



Genotypes for 56 loci were scored for 730 juvenile coho 

 salmon captured in ocean trawls in 1998-2000 (Table 5). 

 About 65% of the 455 fish in June trawls were sampled 



