RULE-BASED MODELS FOR EVALUATING MECHANISMS OF DISTRIBUTIONAL CHANGE

Viability models assume that particular mechanisms (e.g., decreased pr obability of colonization in isolated habitats) drive the pattern of p opulation invasion and extinction. Although authors often provide evid ence supporting the inclusion or exclusion of a mechanism, there has b een no means of rigorously comparing different factors for a given spe cies. We present a method of evaluating alternative mechanisms of dist ributional change that relies on two or more surveys of species presen ce-absence across a large number of sites, a dataset often available ( or at least attainable) for many taxa. The key idea of our approach is to use hypotheses to specify rules for dividing sites into classes. O ur model estimates class-specific probabilities of invasion and extinc tion and then uses these rates to assign invasions and extinctions to sites in multiple stochastic simulations. The output of the model is a frequency distribution of mistakes generated by comparing predicted d istributions with the actual distributions. As a first attempt to appl y the approach we modeled invasions and extinctions of 14 amphibian sp ecies across 32 ponds in Michigan. We compared hypotheses that amphibi an distributions changed as a result of the spatial arrangement of sit es, as a result of succession, or randomly. Overall, the spatial appro ach provided a poor explanation of distributional changes, performing no better than the random model for all species. In contrast, the succ ession model, based on temporal changes in breeding sites, performed b etter than the null model for at least three species. These results we re surprising as spatial effects are thought to be important to the dy namics of pond-breeding amphibians. Our results say little about the g eneral importance of fragmentation and other spatial effects, but do s uggest that alternative mechanisms of change can be important. Because of their ability to assess the importance of different mechanisms, ru le based models could provide useful input into the design of biodiver sity management strategies.