Neotropical land-bridge avifaunas: new approaches to null hypotheses in biogeography

The construction of null hypotheses and statistical tests for non-random avian distributions has been vigorously debated by island biogeographers. Less attention has been given to the underlying quality of the data, which often consist of species lists from islands and adjacent mainland areas. We believe three major problems with past analyses make it difficult to interpret the results. 1. Source pools for statistical tests are constructed unrealistically. Arbitrary geopolitical units are used to define source pools. Archipelagos are analyzed as "closed" systems with equivalent source pools for all islands. 2. Habitat preferences and availability are not explicitly incorporated in null models. 3. Estimates of colonization potential based on species incidence functions may not be satisfactory. Incidence functions do not incorporate available habitat and distance effects. We propose new organism-based methods for the analysis of birds on islands that incorporate geometrically standardized source pools, habitat availability, and geographic ranges of source pool species. We apply these methods in an analysis of seven Neotropical land-bridge island avifaunas. With few exceptions, land-bridge island communities appear to be a random subset of the mainland "habitat" pool, at the family level. A comparison of "total" and "habitat" pool shows that the habitat pool is a superior predictor of species richness in each family. Finally, species with widespread mainland ranges are disproportionately common on islands. Our results suggest that habitat availability and area of a species range are responsible for some of the differences between island and mainland communities.