Landscape genetics is an emerging discipline that utilizes environmental and historical data to understand geographic patterns of genetic diversity. Niche modelling has added a new dimension to such efforts by allowing species–environmental associations to be projected into the past so that hypotheses about historical vicariance can be generated and tested independently with genetic data. However, previous approaches have primarily utilized DNA sequence data to test inferences about historical isolation and may have missed very recent episodes of environmentally mediated divergence. We type 15 microsatellite loci in California mule deer and identify five genetic groupings through a Structure analysis that are also well predicted by environmental data. We project the niches of these five deer ecotypes to the last glacial maximum (LGM) and show they overlap to a much greater extent than today, suggesting that vicariance associated with the LGM cannot explain the present-day genetic patterns. Further, we analyse mitochondrial DNA (mtDNA) sequence trees to search for evidence of historical vicariance and find only two well-supported clades. A coalescence-based analysis of mtDNA data shows that the genetic divergence of the mule deer genetic clusters in California is recent and appears to be mediated by ecological factors. The importance of environmental factors in explaining the genetic diversity of California mule deer is unexpected given that they are highly mobile species and have a broad habitat distribution. Geographic differences in the timing of reproduction and peak vegetation as well as habitat choice reflecting natal origin may explain the persistence of genetic subdivision.