Beacham et al: Population structure and stock identification of Oncorhynchus keta 
251 
, 0 . 002 , 
ioor 
— ! Narva 
'2 Ryazanovka 
3 Avakumovka 
Primorye 
100 
€ 
6 Naiba 
7 Udarnitsa 
8 Kalininka 
47 
100 
5 Tym 
“4 Amur | Amur 
9 Tugur 
” 1 0 Okhota 
‘ 1 1 Magadan Magadan 
3 Ola 
— 12 Tauy 
14 Oklan 
1 5 Penzhina 
Sakhalin 
Northern Sea of Okhotsk 
100 
33 Anadyr 
r 34 Kanchalan 
Northeast 
21 Utka I West Kamchatka 
4 : 
— 32 Oiutorsky Bay 
3 1 Apuka 
— 23 Plotnikova 
~ 22 Bolshaya 
16 Hairusova 
20 Kikchik 
19 Pymta 
18 Kol 
17 Vorovskaya 
24 Zhypanova 
27 Nerpichi 
East Kamchatka 
West Kamchatka 
“25 Kamchatka 
— “ 28 Karaga 
" 30 Dranka 
26 Ivashka 
29 Ossora 
East Kamchatka 
Figure 2 
Neighbour-joining dendrogram of the Cavalli-Sforza and Edwards (1967) 
chord distance for 34 populations of chum salmon (Oncorhynchus keta) 
analyzed at 14 microsatellite loci. Bootstrap values at major tree nodes 
indicate the percentage of 500 trees where populations beyond the node 
clustered together. Numbers for and locations of populations are indicated 
in Table 1. 
multiple regions were within 3% of actual regional 
contributions (Table 6, mixture 4). Accurate regional 
estimates of stock composition should be obtained when 
the current baseline is applied to mixed-stock samples 
of chum salmon taken from Russian coastal waters, 
provided that the individuals in the mixture originate 
entirely from Russian populations. 
Discussion 
Population structure 
The range of populations sampled in the study required 
a concerted sampling effort, and in some locations col- 
lection of appropriate samples proved to be difficult. 
The number of fish sampled from a site or population 
ranged from 17 to 338 individuals. Estimated popula- 
tion-level allele frequencies will be subject to relatively 
larger sampling error at smaller population sample sizes, 
particularly for loci with large numbers of alleles such 
as Onelll. Small sample size may have contributed to 
errors on allele-frequency estimates for some popula- 
tions. Sampling error may obscure genetic relation- 
ships among related populations, or conversely genetic 
relationships among some populations may be falsely 
inferred. However, the available evidence indicates that 
variation in population sample sizes did not obscure 
relationships among related populations. For example, 
we analyzed chum salmon from three sampling sites in 
Primorye with sample sizes ranging between 17 and 49 
fish per site,. Although these sample sizes were small, 
