852 
Fishery Bulletin 96(4), 1998 
in all 500 trees used to construct the consensus tree. 
The branch containing all lower Fraser River popu- 
lations was also well supported. Populations from 
the northern Skeena and Nass rivers were inter- 
spersed on the third branch, but only the upper 
Skeena and Nass populations formed a well-sup- 
ported group (Fig. 2). Lower Skeena and Nass popu- 
lations were more similar to a diverse central clus- 
ter of Vancouver Island and mainland coastal coho 
salmon populations. Within the coastal group, east 
coast (Big Qualicum and Quinsam), west coast 
(Robertson and Nitinat), and north coast (Cluxewe, 
Waukwaas and Stephens) Vancouver Island popula- 
tions were as well distinguished from each other as 
they were from the more northern mainland coastal 
populations or the Queen Charlotte Island popula- 
tion of Pallant Creek. 
Estimation of stock composition 
In the mixed-stock fishery simulated within the 
Fraser River, estimates of stock composition were 
accurate and precise, with an aver- 
age of 2.3% of the mixture incorrectly 
assigned to each population (Table 4). 
In general, misassigned portions of 
the mixture were assigned to closely 
related populations. For instance, the 
30% contribution of Coldwater River 
fish to the mixture was underesti- 
mated as 26.4%, but 3.1% of the mix- 
ture was assigned to the genetically 
similar Spius River population, al- 
though no Spius fish were included 
in the mixture. Fish misassigned to 
population were almost invariably 
assigned to the correct region. Less 
than 1%' of fish were misclassified 
between the Thompson River and 
Lower Fraser regions (Table 4). Thus, 
within the Fraser River drainage, 
populations contributing to a mixed- 
stock sample can be identified to re- 
gion and to individual population, or 
population group within region, with 
a high degree of accuracy and precision. 
Accuracy and precision of stock 
composition estimates were also high 
in the simulated mixed-stock samples 
to which populations from different 
regions contributed (Table 5). Aver- 
age error of individual population 
contributions to the mixed fishery 
was 1.5% (3% for populations actu- 
ally contributing to the mixture), and 
average error of regional contribu- 
tions was 2.2% (higher for regions 
actually contributing). In most simu- 
lations, the contributions of popula- 
tions present in the mixture were un- 
derestimated because a small propor- 
tion of the mixture was allocated to 
baseline populations not present in 
the mixture. In general, misassigned 
fish were allocated to genetically 
similar populations and thus were 
Figure 2 
Unrooted neighbor-joining tree relating 34 British Columbia coho salmon 
populations. The tree was constructed from a consensus of Cavalli-Sforza and 
Edward’s chord distances. Bootstrap values at the tree nodes were computed 
over 500 replications by resampling the allele frequency matrix. Bootstrap 
values indicate the percentage of trees in which the populations beyond the 
node occurred together. All names correspond to population names given in 
Tables 2 and 3 with the exception of Big Qualicum River which is shortened 
to Big Qualic. The number of the population (from Fig. 1) is next to the name. 
