214 



Fishery Bulletin 97(2), 1999 



some individuals stray to streams other than their 

 natal one, and those streams may be inhabited or 

 uninhabited by other conspecifc populations (Ricker, 

 1972; Quinn, 1993). Indeed, straying constitutes the 

 process by which salmonids colonize new habitats 

 (Milner and Bailey, 1989; Wood, 1995). Individuals 

 that stray during the spawning migration may thus 

 serve as the mechanism for dispersal between salmo- 

 nid populations on both an evolutionary and ecologi- 

 cal time scale. 



Reliable estimates of the magnitude of straying are 

 rare and span a wide range of values across and 

 within species (Quinn, 1993; Pascual and Quinn, 

 1994). Quinn and Fresh ( 1984) documented a stray- 

 ing rate of 1.4% in their study of wild chinook salmon 

 (Oncorhynchus tshawytscha ) from the Cowlitz River 

 Hatchery, Washington. Quinn et. al(1991) estimated 

 hatchery straying rates ranging from 9.9 to 27.5% 

 for five populations of autumn chinook on the Co- 

 lumbia River. Heard (1991) estimated that, in gen- 

 eral, nearly 10% of wild pink salmon (Oncorhynchus 

 gorbitscha) stray from their natal streams. Labelle 

 (1992) estimated that approximately 4.7% of indi- 

 vidual coho salmon strayed between nine separate 

 streams along the coast of Vancouver Island, British 

 Colombia, but that straying could be greater than 

 40% for some streams in some years. Genetic stud- 

 ies such as that of Gall et al. ( 1992) suggest that the 

 average number of migrants exchanging genes per 

 generation (Nm) in west coast Chinook salmon popu- 

 lations is on the order of 5-15 individuals. 



Because the incidence of straying is common and 

 the magnitude of straying is so variable, it is quite 

 likely that metapopulation structure could exist for 

 at least some salmonid populations. In fact, the Na- 

 tional Research Council's report on Pacific Northwest 

 salmonids recognizes that ". . . maintaining a 

 metapopulation structure with good geographic dis- 

 tribution should be a top management priority to 

 sustain salmon populations over the long term" 

 (NRC, 1996, p. 8). Given the geographic scale of the 

 straying documented in the previous studies, com- 

 pared with the range of an ESU (note that the entire 

 West Coast comprised only six ESUs), it is possible 

 that a single ESU may even contain multiple 

 metapopulations, as would be expected because ESUs 

 are explicitly evolutionary constructs, whereas 

 metapopulations are explicitly ecological constructs. 



In this paper, we investigate the implications of 

 metapopulation structure for conservation efforts 

 given a variety of spatial scal^s. In particular, we 

 identify the problems such structure could cause for 

 managers if it left undetected. If one is concerned 

 strictly with the risk of extinction for a species, 

 metapopulation structure may be quite beneficial 



(Levins, 1970; Hanski and Gilpin, 1991; Hanski, 

 1994; Ruxton, 1996; Ruxton and Doebeli, 1996). Be- 

 cause the metapopulation occurs in patches (each of 

 which contains a deme with its own probability of 

 extinction) and because these demes are connected 

 through dispersal, if any single deme becomes ex- 

 tinct, then there is a nonzero probability that the 

 patch will be recolonized by individuals from another 

 deme. Over time, an individual patch may therefore 

 experience multiple extinctions and recolonization 

 events. These events result in the metapopulation 

 as a whole persisting far longer than any one of its 

 individual demes. Potential problems arise when one 

 is concerned not just with the risk of extinction but 

 with the management (and therefore monitoring) of 

 these populations. 



In the most simple metapopulation model, one as- 

 sumes that all demes, and the patches they inhabit, 

 are identical (Levins, 1970). This, however, need not 

 be the case, and in the real world, is likely not to be 

 the case. One metapopulation model that takes such 

 variation into account is the source-sink meta- 

 population model (Pulliam, 1988). In this model, sink 

 habitats are patches where local mortality exceeds 

 local reproduction (so that R^^<l or r<0). In other 

 words, without individuals immigrating to the patch, 

 a sink population cannot sustain itself. Source 

 patches, on the other hand, are patches where local 

 reproduction exceeds the sum of local mortality and 

 emigration (so that /?q>1 or r>0). Populations in 

 source habitats can persist without the populations 

 in the sink habitats, but the opposite is not true. 

 There are no assumptions regarding the relative 

 abundance of individuals between these source and 

 sink patches. In fact, it is quite possible for the sink 

 patches to have larger populations than source popu- 

 lations (Pulliam, 1988). For example, if competitively 

 dominant individuals hold territories of fixed size (as 

 is the case with some bird species), a source habitat 

 would be highly productive, yet would have a con- 

 stant population size because all subdominant indi- 

 viduals would be forced to disperse. If this dispersal 

 rate into the sink habitats were greater than the 

 natural rate of decline (the difference between births 

 and deaths) in the sinks, then sink habitats could 

 contain more individuals than the source habitats. 

 In such a case, undetected metapopulation structure 

 could lead managers astray. 



When metapopulation structure (especially source- 

 sink dynamics) exists, the abundance of a species in 

 an area can be disconnected from the specific survi- 

 vorship and fecundity rates of that area owing to the 

 effects of immigration. If ignored, this disconnection 

 poses two problems for managers, both of which are 

 made worse if the jurisdiction of the manager does 



