Evaluating connectivity for tropical amphibians using empirically derived resistance surfaces

Agricultural expansion continues to drive forest loss in species-rich tropical systems and often disrupts movement and distributions of organisms. The ability of species to occupy and move through altered habitats likely depends on the level of contrast between natural forest and surrounding land uses. Connectivity models, such as circuit theory models, are widely used in conservation biology, and their primary input consists of resistance surfaces representing movement costs associated with landscape features. Cost values are most frequently determined by expert opinion, which may not capture relevant levels of contrast among features. We developed resistance surfaces using experiments that represent different local mechanisms hypothesized to affect connectivity for two Neotropical amphibian species. Response ratios were calculated to translate experimental results to cost values used in connectivity modeling. We used relative abundance data in three land-cover types to generate resistance surfaces for evaluating independent support of models derived from experiments. Finally, we analyzed agreement among movement pathways predicted for each species and among three commonly used connectivity measures: Euclidean, least cost, and resistance distances. Experiments showed that extreme microclimates associated with altered habitats significantly increased desiccation and mortality risk for both species. Resistances estimated from microclimate experiments were concordant with those from survey data for both species. For one focal species, resistance estimates derived from predator encounter rates were also highly correlated with abundance-derived resistances. There was generally low agreement among the three alternative distance measures, which underscores the importance of choosing connectivity models that are most appropriate for the study objectives. Overall, similarity among linkages modeled for each species was high, but decreased with declining forest cover. Our results highlight the value of experiments for drawing inferences about processes in resistance modeling, as well as the need to consider model differences and species-specific responses when developing strategies to maintain connectivity.