Background/Question/Methods Dispersal and movement are key factors in the spatial distribution of genetic variation. Individual dispersal ability is ultimately determined by a combination of species’ biology and landscape structure. Understanding connectivity within a landscape is impeded when individuals are not readily observable or when samples sizes are low. Combining high-resolution molecular estimates of population structure with consideration of landscape attributes allows more detailed interpretation of movement patterns from indirect observations of gene flow. Understanding the influence of landscape on movement can increase inference on the ecology of poorly understood organisms, including identifying dispersal barriers or corridors and relating habitat features to meta-population dynamics or sex-biased dispersal. I used highly variable, neutral microsatellite markers to estimate genetic distance and fine-scale population structure of a secretive species of dwarf crocodile in forests of Central Africa. I rasterized habitat and stream maps to construct a set of migration models hypothesizing crocodilian habitat use. Hypothesized least-cost routes were generated by conferring friction values to habitat and stream types and calculating the travel paths that minimize the aggregate product of map cell distance and friction cost. Correlations from partial Mantel tests were used to select those movement models that best explained the genetic data at two spatial scales.

Results/Conclusions

Moderate levels of genetic population differentiation were observed between and within three national parks in Gabon. Mantel tests detected significant isolation by distance, suggesting crocodiles are limited in their dispersal ability. At regional scales (reflecting generational migration patterns), the best supported model implied dwarf crocodiles avoid coastal routes and open water bodies such as lagoons and lakes, while using larger-order streams. At a local scale, 57% of individuals from 12 sites were genetically assigned to populations other that their population of capture (i.e. migrants). Migration rates, estimated from genetic assignment tests, were significantly correlated with gene flow (linearized F_st) and Euclidian geographic distance. Alternate models predicting movement patterns based on landscape structure, however, explained a greater proportion of variance in genetic data than did simple Euclidian measures. Dwarf crocodiles appear to travel via medium-sized forested streams and inland lagoons, using littoral zone rather than moving across open water. Asymmetrical migration rates also suggest more frequent downstream than upstream travel. Landscape-based dispersal models agree with observations of behavior, habitat use, and a novel application of Bayesian Structure analysis.