Many organisms occupy heterogeneous landscapes that contain both barriers to movement as well as corridors that facilitate dispersal. The extent to which such features determine population connectivity will depend on the mechanisms utilized by organisms to disperse. Here we examined the interaction between landscape structure and dispersal in the endemic aquatic snail, Fonscochlea accepta, in the fragmented artesian spring ecosystem of arid central Australia. We used frequentist and Bayesian analyses of microsatellite data to identify population structure and immigration for 1130 snails sampled from 50 springs across an entire spring complex. We introduce a modified isolation-by-distance analysis to test hypotheses about how populations are clustered and to distinguish the most likely dispersal pathways within and between those clusters. Highly significant differences in FST values and significant isolation-by-distance patterns were detected among springs across the entire complex, while Bayesian assignment tests revealed the presence of two hierarchical levels of spring clustering. Clusters were defined by the spatial aggregation of springs, dynamic aquatic habitat connections between springs and the ecology of the snails. Bayesian immigrant identification and our modified isolation-by-distance analysis revealed that dispersal occurs at two geographical scales via two very different mechanisms. Short range dispersal (usually ≤ 300 m) occurs via active movement facilitated by aquatic connections among springs while long-range dispersal (≥ 3 km) is likely facilitated by an animal vector (phoresy). These results underline the importance of both dispersal mode and landscape structure in influencing connectivity rates and patterns among populations.