856 
Fishery Bulletin 96(4), 1 998 
lower Fraser River populations are characteristic of 
coho salmon derived from a refuge located on, or to 
the south of, Vancouver Island, Puget Sound coho 
salmon populations might be expected to be of simi- 
lar origin. This is consistent with the inclusion, based 
primarily on genetic data, of lower Fraser and south- 
eastern Vancouver Island coho salmon populations 
in a Puget Sound and Strait of Georgia ESU 
(Weitkamp et ah, 1995). Further sampling of 
Vancouver Island populations will be necessary to 
define the boundaries of the historical groups of coho 
salmon that likely converge there. 
The geographic basis for population structure of 
coho salmon revealed in this study is remarkably 
similar to that described for sockeye salmon based 
on allozyme and mtDNA data (Wood et al., 1994; 
Bickham et ah, 1995). A genetic discontinuity be- 
tween chinook salmon originating from the Colum- 
bian and Beringial glacial refugia (Gharrett et ah, 
1987; Cronin et ah, 1993) also lies in British Colum- 
bia, and dispersal from a coastal refuge(ia) can be 
traced in the microsatellite data for chinook salmon 
as well (Beacham, unpubl. data). Thus, the phylo- 
geographic reconstruction of postglacial dispersal 
based on freshwater fish distributions in British 
Columbia (McPhail and Lindsey, 1970, 1986; Lindsey 
and McPhail, 1986) has proven to be taxonomically 
robust and also provides the foundation for the 
genetic architecture of anadromous Pacific salmon 
species. 
The mixed-stock fishery analyses demonstrated 
the utility of microsatellite DNA variation for coho 
salmon stock identification. We obtained accurate 
and precise estimates both of population contribu- 
tions in mixed-stock samples from a single drainage 
and of population and regional contributions in 
mixed-stock samples drawn from several regions. An 
important feature of the microsatellite data set is 
the strong regional structuring of the observed ge- 
netic variation, which means that contributions from 
populations present in a mixture sample, but not in 
the baseline, will be identified correctly to region. 
Identifying individual fish to correct populations is 
more difficult than estimating percentage contribu- 
tions to a stock mixture, because only characteris- 
tics of individual fish are used in the classification. 
In general, the microsatellite loci of this study pro- 
vided a similar level of accuracy in the classification 
of individual coho salmon to population and region 
as did minisatellite DNA markers (Beacham et al., 
1996). Within the Fraser River drainage, identifica- 
tion of individual fish was more accurate with 
microsatellite data (correct identification of 54% of 
lower Fraser and 85% of upper Fraser Rive coho 
salmon) than with minisatellite data (correct identi- 
fication of 30% and 60% of the respective groups). 
The identification of individual fish is an important 
enforcement tool, and may improve as more 
microsatellite loci are added to the database. 
The results of this study are consistent with a de- 
piction of population structure in coho salmon as dis- 
tinct phylogenetic lines composed of geographically 
based metapopulations (McPhail, 1997). The more 
consistent regional grouping of coho salmon popula- 
tions than of sockeye salmon (Wood et al., 1994) may 
reflect greater, or more recent, gene flow among geo- 
graphically proximate coho salmon populations than 
among similar sockeye populations, or may reflect 
differences between allozyme- and microsatellite- 
based data sets. Moreover, the increasing power of 
genetic methods applied to coho salmon data, as dem- 
onstrated in this and other (Weitkamp et al., 1995; 
Beacham et al., 1996; Van Doornik et al., 1996; Miller 
et al., 1996; Miller and Withler, 1997; Small et al., 
1998) studies, enable us not only to delineate regional 
(metapopulation) structure in coho salmon but also 
to identify the regional contributions of coho from 
different metapopulations in mixed-stock fishery 
harvests. Thus, the challenge for metapopulation- 
based coho salmon management may lie not so much 
in the delineation of metapopulation structure 
( McPhail, 1997 ) as in the evaluation of the biological 
and social costs associated with the loss, even if only 
temporary in historical terms, of the less productive 
components of metapopulation structure during pe- 
riods of overall low abundance. 
Additional aspects of metapopulation theory, as 
applied to Pacific salmonids, need to be addressed 
before this model of population structure will sup- 
port practical management decisions. Managers need 
to know, at any given point in history, how the adap- 
tive genetic diversity of a metapopulation is likely to 
be distributed among its subpopulations, and how 
many subpopulations can be lost before evolution- 
ary potential is compromised. This depends on, 
among other things, which model of metapopulation 
structure is adopted. Are salmonid metapopulations 
of the “source-sink” variety in which gene flow is 
basically unidirectional from large source subpopu- 
lations to ephemeral sink subpopulations (Pulliam, 
1988; Pulliam and Danielson 1991)? Or are salmo- 
nid metapopulations of the “balanced exchange” type, 
in which gene flow is bidirectional and migration 
rates are inversely proportional to subpopulation size 
(McPeek and Holt, 1992; Doncaster et al., 1997)? 
Current models of metapopulation structure have 
been more extensively investigated with respect to 
population dynamics (extinctions and recolonizations 
among subpopulations) than with respect to popula- 
tion and evolutionary genetics (the spatial and tern- 
