摘要：Landscape genetics is a field of research that seeks to understand the drivers of the observed spatial distribution of genetic variation in a species of plant or animal using methods from population genetics, landscape ecology, geography, and spatial statistics. One of the important research areas in landscape genetics is to identify landscape barriers to genetic flow. Barriers can limit interaction of organisms and hence lead to genetic structure (i.e. frequency of genotypes) in a population that becomes increasingly spatially structured over time. Here, we investigate model-based spatial methods to assess the relationship between landscape and genotype distributions of cougars ($Puma concolor$) in Western North America. Previous research has assessed genetic differentiation in cougars in North America using a non-spatial Bayesian clustering model and found evidence of genetic population structure in cougars, with suggestive but indistinct spatial boundaries between subpopulations. To determine if including spatial information on samples in a classification model would refine the observed spatial signal within the genetic population structure, we applied Bayesian classification models to microsatellite loci data with associated spatial locations. The spatial model revealed two clearly differentiated cougar subpopulations, in contrast to the two overlapping subpopulations suggested by methods not accounting for space. We also explored through geographic information systems and generalized linear models whether resulting genetic population structure boundaries aligned with landscape features. The spatial correspondence of genetic subpopulations and a major river and road is suggestive of possible landscape barriers to cougar movement. This study demonstrates that the use of explicit spatial information and Bayesian classification models adds novel insight when investigating genetic population structure in a species.
