256 
Fishery Bulletin 107(2) 
were not distinct clusters observed between northern 
coastal British Columbia populations and those from 
southeast Alaska. Populations immediately south of 
the Skeena River in the Grenville Channel area clus- 
tered separately from those further south in the central 
coastal region of British Columbia. Yet farther south, 
populations from Rivers Inlet and Smith Inlet clustered 
together in geographically based groups, and this result 
was confirmed by 100% boostrap support observed for 
Smith Inlet populations (Fig. 2). 
In southern BC, five geographically based groups of 
populations were revealed. East coast and west coast of 
Vancouver Island populations were regionally separate 
from each other, and also from other regional popu- 
lations in southern BC. On the mainland, Johnstone 
Strait populations were separate from those in southern 
coastal areas, and the demarcation point between the 
two groups is Bute Inlet, at the northeast limit of the 
Strait of Georgia. Fraser River populations were dis- 
tinct from other regional groups in southern BC. 
In Washington, regional structuring of chum popula- 
tions was observed. The most distinct regional group 
comprised populations from the outer Pacific coast, 
with populations clustering together with 100% boot- 
strap support. In more inside waters, populations from 
north Puget Sound were generally distinct from those 
in south Puget Sound, Hood Canal, and the Strait of 
Juan de Fuca. South Puget Sound populations were 
distinct from those in Hood Canal and the Strait of 
Juan de Fuca. 
Discussion 
The survey of microsatellite variation included an exami- 
nation of variation at 14 loci encompassing approximately 
800 alleles, with 26 to 149 alleles recognized per locus. 
The number of fish surveyed per population ranged from 
12 to 597 individuals (Beacham et al. 1 ). With a vari- 
able number of individuals surveyed per population, 
there was a potential for sampling error in estimated 
allele frequencies and in obscuring genetic relation- 
ships among related populations, particularly if sample 
sizes were small for some populations in a lineage. For 
example, for the Primorye populations from Russia, pop- 
ulation sample sizes ranged from 17 to 49 individuals, 
and it was possible that estimates of genetic distances 
among populations were not determined satisfactorily for 
populations of smaller sample size, particularly for those 
loci with larger numbers of alleles. However, Kalinowski 
(2005) reported that loci with larger numbers of alleles 
(higher mutation rates) produced estimates of genetic 
distance with lower coefficients of variation than loci 
with fewer numbers of alleles, without requiring larger 
sample sizes from each population. Population structur- 
ing based upon geographic differences were observed for 
populations from Primorye, and all populations clustered 
together in 67% of dendrograms evaluated. Therefore, it 
seems likely that variation in the number of individuals 
surveyed within a population in our study did not gener- 
ally result in misidentification of genetic relationships 
among populations. 
Size homoplasy of microsatellite alleles may have 
some effect on the estimate of genetic differentiation ob- 
served among populations. Inferences about the genetic 
relationships of populations surveyed in our study were 
dependent upon accurate determination of population 
allele frequencies. Microsatellite alleles differ in size, 
but alleles of the same size at a locus in geographically 
disparate populations may not have the same origin 
as a result of size homoplasy. Convergent mutations in 
different lineages may produce alleles of the same size, 
with the result that there may be greater differentiation 
among lineages than revealed by analysis of size varia- 
tion. However, with approximately 800 alleles observed 
across all loci in the study, the large amount of varia- 
tion present at these loci largely compensates for size 
homoplasy (Estoup et al., 2002). 
In this study, population allele frequencies were es- 
timated by combining all samples collected over time 
for a population, regardless of the length of time that 
occurred between samples. In practice, the maximum 
length of time between samples for a population was 
21 years, and up to six annual samples were combined 
for a population. Analysis of the distribution of genetic 
variation indicated that differentiation among regions 
and populations within regions was approximately 18 
times greater than that of annual variation within 
populations, indicating that pooling of annual samples 
over time is a practical approach to estimate population 
allele frequencies. Relative stability of microsatellite al- 
lele frequencies over time is not unique to chum salmon; 
similar relative stability has been reported for sockeye 
salmon (O. nerka) (Beacham et al., 2006a) and Chinook 
salmon (O. tshawytscha ) (Beacham et al., 2006b). 
Surveys of genetically based population structure 
in chum salmon were initially conducted with allo- 
zymes. Okazaki (1982b), in a study evaluating allozyme 
variation in Asian and North American populations, 
concluded that there were 11 geographically based re- 
gional groups of populations across the Pacific Rim. The 
regional groups consisted of adjacent river populations 
that were genetically similar within one region. Many 
allozyme-based studies of regional population structure 
were subsequently reported. For example, Winans et al. 
( 1994) provided additional details concerning population 
structure of Asian populations, Wilmot et al. (1994) 
compared population structure of chum salmon from 
western Alaska and northeast Russia, Kondzela et al. 
(1994) compared population structure of chum salmon 
from southeast Alaska and northern British Columbia, 
Beacham et al.( 1987) evaluated population structure 
of chum salmon in British Columbia, and Phelps et al. 
(1994) evaluated population structure in the Pacific 
Northwest. Seeb and Crane (1999) again investigated 
chum salmon population structure throughout the Pa- 
cific Rim by examining variation at 40 allozymes, and 
reported that two major lineages of populations were 
observed. The northern lineage occurred in areas north 
of the Alaska Peninsula and into Russia and Japan, 
