A stochastic movement simulator improves estimates of landscape connectivity.

Conservation actions often focus on restoration or creation of natural areas designed to facilitate the movements of organisms among populations. To be efficient, these actions need to be based on reliable estimates or predictions of landscape connectivity. While circuit theory and least-cost paths (LCPs) are increasingly being used to estimate connectivity, these methods also have proven limitations. We compared their performance in predicting genetic connectivity with that of an alternative approach based on a simple, individual-based "stochastic movement simulator" (SMS). SMS predicts dispersal of organisms using the same landscape representation as LCPs and circuit theory-based estimates (i.e., a cost surface), while relaxing key LCP assumptions, namely individual omniscience of the landscape (by incorporating perceptual range) and the optimality of individual movements (by including stochasticity in simulated movements). The performance of the three estimators was assessed by the degree to which they correlated with genetic estimates of connectivity in two species with contrasting movement abilities (Cabanis's Greenbul, an Afrotropical forest bird species, and natterjack toad, an amphibian restricted to European sandy and heathland areas). For both species, the correlation between dispersal model and genetic data was substantially higher when SMS was used. Importantly, the results also demonstrate that the improvement gained by using SMS is robust both to variation in spatial resolution of the landscape and to uncertainty in the perceptual range model parameter. Integration of this individual-based approach with other developing methods in the field of connectivity research, such as graph theory, can yield rapid progress towards more robust connectivity indices and more effective recommendations for land management.