A major goal of landscape genetics is to understand how landscapes structure genetic variation in natural populations. However, landscape genetics still lacks a framework for quantifying the effects of landscape features, such as habitat type, on realized gene flow. Here, we present a methodology for identifying the costs of dispersal through different habitats for the California tiger salamander (Ambystoma californiense), an endangered species restricted to grassland/vernal pool habitat mosaics. We sampled larvae from all 16 breeding ponds in a geographically restricted area of vernal pool habitat at the Fort Ord Natural Reserve, Monterey County, California. We estimated between-pond gene flow using 13 polymorphic microsatellite loci and constructed GIS data layers of habitat types in our study area. We then used least-cost path analysis to determine the relative costs of movement through each habitat that best match rates of gene flow measured by our genetic data. We identified four measurable rates of gene flow between pairs of ponds, with between 10.5% and 19.9% of larvae having immigrant ancestry. Although A. californiense is typically associated with breeding ponds in grassland habitat, we found that dispersal through grassland is nearly twice as costly as dispersal through chaparral and that oak woodland is by far the most costly habitat to traverse. With the increasing availability of molecular resources and GIS data, we anticipate that these methods could be applied to a broad range of study systems, particularly those with cryptic life histories that make direct observation of movement challenging.