224 



Fishery Bulletin 97(2), 1999 



success assumption) or an entire stream (where the 

 uniform success assumption may be less accurate). 

 If a manager had the entire time series data for all of 

 the demes, the problem (though not the solution) would 

 be evident; all demes would have declined significantly. 

 Suppose, however, the manager is currently in year 

 30, and has only 25 years of data (years 5-30). Over 

 that time period, all demes in sink habitats would be 

 relatively stable or decline only slowly, even though 

 none could sustain themselves without immigrants. In 

 fact, deme 1 had been increasing until year 30. The 

 demes in the source habitats, however, showed the ef- 

 fects of habitat degradation. Seeing these trends, the 

 manager could perhaps stabilize these demes and thus 

 inadvertently save the entire metapopulation. Still, the 

 true risk for the sink populations would be unknown 

 to the manager. The sink populations would appear 

 stable not because they were in good condition but 

 rather because of their interconnectedness with one 

 another and the source populations. 



Rieman and Mclntyre ( 1995 ) suggested that a simi- 

 lar process is occurring with bull trout {Salvelinus 

 confluentus) populations in Idaho. Although bull 

 trout were found to use small streams, they do so 

 only at a very low frequency. Rieman and Mclntyre 

 (1995) concluded that the presence of trout in these 

 streams may be influenced by habitat preference but 

 that these populations depend on dispersal of indi- 

 viduals from larger streams for their sustainability. 



The addition of a varying environment clouds the 

 picture even more (Fig. 7). The manager must now 

 distinguish between decreases due to environmen- 

 tal factors and decreases due to factors that can be 

 controlled. Imagine a manager starting in year 1. 

 How can one recognize that the metapopulation is 

 collapsing due to anthropogenic effects and not sim- 

 ply to environmentally induced effects? When would 

 the alarms sound? Probably not until sometime af- 

 ter year 30, when even the most historically produc- 

 tive demes do not begin to increase. Seven out often 

 of the demes could not maintain themselves without 

 immigration, there is neither observation error nor 

 stochasticity, and it would still take over 25 years of 

 habitat degradation before the problem was noticed 

 and the alarms sounded. 



These problems become even more serious when 

 management boundaries do not coincide with 

 metapopulation boundaries or when only a portion 

 of the metapopulation is used as index streams, as 

 is likely the case. In such a situation, a manager will 

 be concerned with, have jurisdiction over, and maybe 

 have data on some subset of the metapopulation. 

 Returning to Figure 3 (where managers do not have 

 to deal with fluctuating environments), imagine if a 

 manager had responsibility, and therefore data, for 



only demes 6, 7, 8, 9, and 10, which are nevertheless 

 a significant portion of the metapopulation. Between 

 years 5 and 30, all demes were relatively stable, de- 

 spite the fact that without immigration, even the 

 most productive deme (deme 9) would decrease at a 

 rate of 5% per year. After year 30, all these demes 

 begin to decline. The manager would, of course, be- 

 gin looking at these demes to try and see what was 

 causing this decline; a good manager would try to 

 find what had changed. In fact, nothing has changed 

 with these demes; they have exactly the same rates 

 of per-capita reproduction and straying over the en- 

 tire simulation. Only the number of immigrants has 

 changed. Without looking beyond their own jurisdic- 

 tion or at streams other than the index streams, 

 managers would not find the true cause for the 

 change in dynamics. As the number of demes that 

 are used as index streams or that lie within a 

 manager's jurisdiction decreases, the likelihood of 

 such a problem increases. Considering that over 90'7r 

 of the sink populations increased in size during some 

 portion of the simulation time horizon (i.e. their 

 maximum size was reached sometime after year 1, 

 Fig. 4B), the fate of sink populations has as much, if 

 not more, to do with the health of the demes in other 

 habitats as with the quality of their own habitat. 



Another problem associated with undetected 

 metapopulation structure arises directly from this 

 last example. A manager may see a change in the 

 dynamics in the populations and look for the cause. 

 However, the cause of this change is outside the 

 manager's jurisdiction or data set; it is not local. If 

 the manager looks only for local causes for this 

 change, she or he is likely to find some variable that 

 is correlated with this cause and possibly infer local 

 causality. Having attributed causality to a local vari- 

 able, the manager would likely start funding projects 

 to fix the perceived problem in the correlated vari- 

 able. If one is lucky, very lucky, this variable might 

 have some relationship to the per-capita reproduc- 

 tive rate and thus improve the situation a bit. This 

 might, then, slow the rate of decline, but because it 

 is not the true cause of the decline, the situation 

 would likely continue to deteriorate. How much of a 

 manager's limited resources might be spent on such 

 activities before the true causal relationship was dis- 

 covered? The study by Gowan and Fausch (1996) 

 suggests that this problem may be occurring with 

 habitat enhancement projects for trout in Colorado; 

 managers promote the addition of woody debris in 

 streams because it increases trout density, even 

 though it does not seem to improve the demographic 

 parameters of the population. 



Including density dependence (as in Ricker stock- 

 recruitment relationships) and observation or pro- 



