Chaotic Population Dynamics Favors the Evolution of Dispersal

Dispersal-movement between populations-is a central feature in the biology of most organisms. There is an enormous literature on the ecology and evolution of dispersal (e.g., Swingland and Greenwood 1986). Many theoretical studies have explored factors favoring the evolution of dispersal, including competition among kin and inbreeding effects (e.g., Hamilton and May 1977; Comins 1982; Frank 1986; Taylor 1988; Wiener and Feldman 1991), the influences of extrinsically generated, spatiotemporal heterogeneity (Gadgil 1971; Roff 1975; Metz et al. 1983; Levin et al. 1984; Cohen and Levin 1991), and the interplay of withinpopulation and between-population selection (Kuno 1981; Olivieri et al. 1995). It has been demonstrated (Hastings 1983; Holt 1985) that if individuals disperse at fixed per capita rates between sites with local density dependence, then, without emporal heterogeneity, spatial heterogeneity inabundance alone is unable to select for dispersal (see also Liberman and Feldman 1989). The reason is that if habitats vary in carrying capacity, K, there is an asymmetric flow of individuals from high-K to low-K patches (Holt 1985). Such flow, in turn, reduces density in high-K patches (increasing fitness there) while increasing density in low-K patches (depressing fitness there). Because dispersal is basically moving individuals down gradients in fitness, on average, dispersal is disfavored in spatially (but not temporally) heterogeneous environments. These theoretical results highlight the importance of temporal heterogeneity infavoring dispersal. In a metapopulation ofopen patches, if the rank order of fitness among patches varies through time, dispersal can be selectively advantageous (Bull et al. 1987). Dispersal in effect provides an evolutionary strategy that permits individuals to exploit spatiotemporal variation in fitness. The theoretical expectation of a relation between temporal variability and dispersal matches ome data from natural populations (Roff 1990). We previously examined a simple two-patch, discrete-generation model in which individuals dispersed at constant rates between two patches and experienced density dependence in each patch (McPeek and Holt 1992). We showed that temporal variation in density-independent growth rates, partially uncorrelated across patches, favored dispersal. Moreover, a polymorphism indispersal rates could be stably maintained if the two patches were heterogeneous (with different carrying capacities). Other authors have also shown that dispersal polymorphisms may be maintained in temporally and spatially heterogeneous environments (Frank 1986; Cohen and Levin 1991; Karlson and Taylor 1992). In this note, we demonstrate that chaotic population dynamics in otherwise