648 



Fishery Bulletin 101(3) 



and September (32%, SD=9%). The Oregon coastal region 

 contributed about 18% (SD=5%') to the June mixture and 

 21% (SD=7% ) to the September sample. The estimated con- 

 tribution of Puget Sound fish to the pooled ocean samples 

 was much greater in September ( 17%, SD=7%) than it was 

 in June(3%, SD=2%'). 



Genetic mixed-stock analysis of ocean-caught hatchery 

 fish with CWTs provided a direct comparison of genetic 

 estimates and a mixed-stock sample of known origins 

 (Brodziak et al. 1992). Only 41 fish had CWTs (Table 6). 

 No fish with CWTs appeared in the 1998 sample. Most of 

 the fish with CWTs in 1999 and 2000 originated from Co- 

 lumbia River (68%, n=28) and Oregon coastal (12%, n=5) 

 hatcheries. In the genetic analysis of the 41 fish, Columbia 

 River hatcheries were estimated to contribute about 22 fish 

 (53% , SD=2 1% ). Approximately 7 fish ( 16% , SD= 17% ) were 

 estimated to originate from Oregon coastal hatcheries. 



Of the 730 juveniles sampled during the study, 501 

 (69% ) bore hatchery marks ( clipped adipose fins ). The per- 

 centage of unmarked fish in the September sample (35%) 

 was greater than that in June (29% ). Genetic mixed-stock 

 estimates for hatchery-marked fish alone indicated that 

 69% (SD=6%) originated from the Columbia River and 

 14% (SD=4%) from Oregon coastal hatcheries (Table 7). 

 The sample of unmarked fish, which contained a mixture 

 of wild and unmarked hatchery fish, was estimated to 

 have a much smaller proportion of Columbia River fish 

 (20%, SD=8% ) but a larger proportion of coastal Oregon 

 (36%, SD=9%) and northern Washington (25%, SD=7%) 

 fish (Table 7). About 30% of unmarked fish in the pooled 

 ocean sample originated from hatcheries (Eq. 1) and 70% 

 from wild populations. Estimated contributions from 

 hatchery and wild populations of all ocean juveniles 

 sampled (marked and unmarked) were 78% and 22%, re- 

 spectively. Coho salmon originating in the Columbia River 

 were estimated to comprise 54% of the total sample, but 

 only 1% consisted of wild fish. Oregon coastal rivers con- 

 tributed 21% to the total ocean sample, and nearly equal 

 proportions were contributed from hatcheries and wild 

 populations. 



Discussion 



Usefulness of coho salmon allozyme data 

 for mixed-stock analysis 



Although the gene diversity analysis indicated that the 

 level of allele-frequency differentiation among populations 

 within regions was similar to that between regions, further 

 analyses showed that the magnitude of regional differen- 

 tiation in the baseline was sufficient to provide accurate 

 mixed-stock estimates. First, we found several genetically 

 discrete population groups of coho salmon over an area 

 extending from California to southern British Columbia. 

 Most of the samples in the MDS plot clustered with nearby 

 samples, and the north-south arrangement of neighboring 

 population groups indicated that isolation by distance is 

 an important component of genetic population structure 

 on this geographic scale. As with other species of Pacific 

 salmon, natal homing to spawning areas is an important 

 isolating mechanism between populations of coho salmon. 

 Second, the analysis of simulated stock mixtures also 

 demonstrated that regional differences were sufficient to 

 provide reliable estimates of coho salmon stock composi- 

 tions. Accurate estimates were obtained from simulated 

 sample sets composed of 100% contributions from each re- 

 gion (Table 3). Third, a more rigorous test of the adequacy 

 of the baseline was made by comparing genetic estimates 

 with direct determinations based on CWTs. These esti- 

 mates were reasonably accurate, especially for the largest 

 contributing regions (Table 6), given the small sample of 

 only 41 fish bearing CWTs. Both the simulation and CWT 

 mixture results are consistent with the findings of Wood et 

 al. ( 1987 ) that estimation accuracy decreases substantially 

 when mixture sample sizes are small and when genetic 

 separation among stocks is limited. Lastly, the analyses 

 of ocean-caught mixture samples themselves appeared to 

 provide reasonable composition estimates (Table 5). Ad- 

 ditionally, estimates for samples pooled over years tended 

 to be intermediate between the two annual estimates, as 

 would be expected from pooling. 



