Non-linear homogenization of small mammal communities following habitat loss in a biodiversity hotspot

Habitat loss has pervasive effects on alfa and gamma diversity, but its impacts on beta diversity are less clear, with possible outcomes ranging from biotic heterogenization (increase in beta diversity) to biotic homogenization (decrease in beta diversity). A powerful tool to understand beta diversity is Generalized Dissimilarity Modelling (GDM), which allows comparing different drivers of beta diversity and testing for non-linear responses to habitat loss. Here we applied GDM, for the first time, to test for the occurrence of non-linear changes in beta diversity following habitat loss, while controlling for spatial and methodological factors, as well as geographic distance. We used a comprehensive empirical dataset with 30 sites sampled for non-volant small mammals over a 10-year period within the Atlantic Forest biodiversity hotspot. Forest cover was an important driver of total beta diversity and turnover, but not nestedness. A 20 % forest cover threshold was found for total beta diversity and turnover, with little compositional change below this threshold, suggesting the occurrence of biotic homogenization. This result was corroborated by a more traditional and categorical method, as multiple-site turnover was significantly lower for sites with low forest cover (<20 %) compared to both intermediate- (20–60 %) and high-cover sites (>60 %). The detection of a non-linear biotic homogenization can be useful to guide conservation initiatives, as it suggests the need to maintain >20 % of forest cover in landscapes to maintain biodiverse communities.