Random placement models explain species richness and dissimilarity of frog assemblages within Atlantic Forest fragments.

Abstract

Understanding the effects of random versus niche-based processes on biodiversity patterns is a central theme in ecology, and an important tool for predicting effects of habitat loss and fragmentation on biodiversity. We investigated the predictive power of random processes to explain species richness and species dissimilarity of amphibian assemblages in a fragmented tropical landscape of the Atlantic Forest of South America. We analyzed a large database of amphibian abundance and occupancy, sampled in 21 forest fragments ranging in size from 1.9 to 619ha. We compared observed species richness and species dissimilarity with the outcomes of two null (random placement) models: 1- the traditional Coleman's area-based model and 2- an abundance-based model (based on the number of individuals observed in each fragment). We applied these models for all species combined, and separately for forest-dependent and habitat-generalist species. The abundance-based model fitted the observed species richness data better than the area-based model for all species, forest-dependent species, and generalist species. The area-based and the abundance-based models were also able to significantly explain species dissimilarity for all species and for generalists, but not for forest dependent species. The traditional area-based model assigned too many individuals to large fragments, thus failing to accurately explain species richness within patches across the landscape. Although niche-based processes may be important to structuring the regional pool of species in fragmented landscapes, our results suggest that part of the variation in species richness and species dissimilarity can be successfully explained by random placement models, especially for generalist species. Evaluating which factors cause variation in the number of individuals among patches should be a focus in future studies aiming to understand biodiversity patterns in fragmented landscapes.