Small et ai.: Population and stock indentification of Oncorhynchus kisutch 
853 
correctly identified to region. Regions with no popu- 
lations present in the mixture were allocated less 
than 5% of the fish, except in the mixture composed 
of Vancouver Island populations (mix 1, Table 5). In 
that case, 10% of the fish were allocated to the lower 
Fraser River. The regional misclassification of Van- 
couver Island coho salmon as fish of lower Fraser 
origin in mix 1 was due largely to the allocation of 
4% of the mixture to Chilliwack River, the lower 
Fraser population most genetically similar to the 
northern Vancouver Island populations (Fig. 2). Simi- 
larly, Vancouver Island and central coast fish were 
underestimated in mix 5, and some of them attrib- 
uted to the lower Fraser rather than to other coastal 
island, mainland, or lower Skeena and Nass popula- 
tions as might be expected from the dendrogram 
(Table 5; Fig. 2). Conversely, when lower Fraser coho 
salmon formed a high proportion of the mixture (mix 
3) , they were underestimated and misidentified fish 
tended to be attributed to Vancouver Island popula- 
tions (Table 5). 
In contrast, the genetic distinctiveness of Thomp- 
son River coho salmon resulted in their accurate iden- 
tification in mixtures in which they were present 
(mixes 3 and 5) and little misrepresentation in mix- 
tures from which they were absent (mixes 1,2, and 
4) . Nass and Skeena populations were well-separated 
(mix 4), and contributions from lower Skeena (Ce- 
dar and Clearwater) populations were identified as 
well as those from the upper Skeena-Nass popula- 
tions. Given the lack of separation of Skeena and 
Nass in the NJ dendrogram (Fig. 2), this result needs 
to be confirmed by more extensive sampling and fur- 
ther mixture analysis for populations within these 
two watersheds. Individual classification of Nass and 
Skeena fish (see below), and the lack of significant 
allele frequency differences between two Nass- 
Skeena pairs of populations, indicated that additional 
genetic markers may be required for accurate differ- 
entiation of Nass and Skeena coho salmon. Never- 
theless, the results of these preliminary analyses 
indicate that a mixed-stock sample of coho salmon 
collected in the field can generally be resolved into 
its regional components by using microsatellite ge- 
netic markers. 
Identifying individuals 
Accuracy of the classification of individual fish to 
region with discriminant analysis varied among re- 
gions (Table 6), but individual classification was gen- 
erally much less accurate than was estimation of 
stock composition. An average of 48% of fish, rang- 
ing from 85% of Thompson to 20% of Nass individu- 
als, was correctly classified to region. Misclassified 
Table 4 
Accuracy and precision of estimated Fraser River coho 
salmon population contributions in a simulated mixed- 
stock sample from a 16-population Fraser River baseline. 
A 200-fish mixture was generated with replacement from 
the baseline 50 times and the population composition was 
estimated for each mixture. The mean percentage of the 
mixture allocated to each population is given, followed by 
the standard deviation in parentheses. Regional totals are 
given in the rows marked TR (Thompson River) and L Fr 
(lower Fraser River). 
True 
Mean 
SD 
Coldwater 
30 
26.4 
(4.4) 
Salmon 
0 
1.3 
(1.6) 
Eagle 
20 
15.7 
(3.4) 
Spius 
0 
3.1 
(3.5) 
Lemieux 
0 
1.0 
(1.5) 
Dunn 
0 
0.6 
(1.3) 
Louis 
0 
1.0 
(1.7) 
Deadman 
0 
0.5 
(0.9) 
Bridge 
0 
0.7 
(1.3) 
TR 
50 
50.2 
(1.4) 
Chilliwack 
20 
16.8 
(4.0) 
Chehalis 
0 
1.4 
(1.8) 
Stave 
15 
12.6 
(3.3) 
Upper Pitt 
0 
1.8 
(2.1) 
Nicomen 
15 
12.2 
(3.4) 
Norrish 
0 
2.9 
(2.5) 
Inch 
0 
2.0 
(2.4) 
L Fr 
50 
49.8 
(1.4) 
fish were most commonly attributed to the most ge- 
netically similar region according to relationships 
depicted in the NJ dendrogram. The Nass and Skeena 
populations were interspersed in the NJ analysis 
(Fig. 2), and an essentially equal proportion of Nass 
River fish (20%) were identified as Nass and Skeena 
fish (Table 6). Skeena River fish were identified more 
accurately than Nass River fish, with 53% and 11% 
of the Skeena River fish classified as being of Skeena 
and Nass origin, respectively. If only the Fraser River 
populations were included in the baseline, Fraser 
River coho salmon were assigned to region accurately, 
with 93% correctly identified to either the lower 
Fraser River or the Thompson River (data not 
shown). 
Discussion 
Microsatellite DNA analysis shows great promise for 
the elucidation of population structure in coho 
salmon. Very strong regional structure was appar- 
ent in the British Columbia coho salmon populations 
