Wolverines (Gulo gulo) are one of the rarest carnivores in the contiguous United States. Effective population sizes in Montana, Idaho, and Wyoming, where most of the wolverines in the contiguous United States exist, were calculated to be 35 (credible limits, 28 52) suggesting low abundance. Landscape features that influence wolverine population substructure and gene flow are largely unknown. Recent work has identified strong associations between areas with persistent spring snow and wolverine presence and range. We tested whether a dispersal model in which wolverines prefer to disperse through areas characterized by persistent spring snow cover produced least-cost paths among all individuals that correlated with genetic distance among individuals. Models simulating large preferences for dispersing within areas characterized by persistent spring snow explained the data better than a model based on Euclidean distance. Partial Mantel tests separating Euclidean distance from spring snow-cover-based effects indicated that Euclidean distance was not significant in describing patterns of genetic distance. Because these models indicated that successful dispersal paths followed areas characterized by spring snow cover, we used these understandings to derive empirically based least-cost corridor maps in the U.S. Rocky Mountains. These corridor maps largely explain previously published population subdivision patterns based on mitochondrial DNA and indicate that natural colonization of the southern Rocky Mountains by wolverines will be difficult but not impossible.