Roads, Soil, Snow, and Topography Influence Genetic Connectivity: A Machine Learning Approach for a Peripheral American Badger Population

Effective management and conservation of peripheral populations require an understanding of the landscape conditions inhibiting dispersal and spatially explicit predictions of connectivity. Here, we modeled landscape resistance and genetic connectivity for the western population of an American badger subspecies (Taxidea taxus jeffersonii) across ~170,000 km² in southern British Columbia, Canada, using 116 genetic samples genotyped at 14 microsatellite loci. We used gradient boosting machine models in a corridor-based approach to predict genetic distances between pairs of individual badgers as a function of landscape variable data. Spatial genetic autocorrelation tests and our top model predicted that genetic similarities of T. t. jeffersonii were present up to ~110 km. Gene diversity was lowest in the Cariboo region in the northwest portion of the study area and highest in the Okanagan region in the southeast. Our analyses suggest that the genetic connectivity of T. t. jeffersonii was impeded by colluvial soil parent material, geographic distance, steep slopes, and major roads, but was facilitated by organic and fluvial soil parent materials, and areas with relatively little snow cover during winter. Our predictive maps of landscape resistance and connectivity can help guide management actions such as habitat protection and underpass placement on major roads to promote genetic connectivity.