American marten (Martes americana) are a conservation priority in many forested regions of North America. Populations are fragmented at the southern edge of their distribution due to suboptimal habitat conditions. Facilitating gene flow may improve population resilience through genetic and demographic rescue. We used a multi-scale approach to estimate the relationship between genetic connectivity and landscape characteristics among individuals at three scales in the northeastern United States: regional, subregional, and local. We integrated multiple modeling techniques and identified top models based on consensus. Top models were used to parameterize resistance surfaces at each scale, and circuit theory was used to identify potential movement corridors.

Regional gene flow was affected by forest cover, elevation, developed land cover, and slope. At subregional and local scales, effects were site-specific and included subsets of temperature, elevation, developed land cover, and slope. Developed land cover significantly affected gene flow in all sites examined at each scale. At finer scales, lack of variance in forest cover may have limited the ability to detect a relationship with gene flow. The effect of slope on gene flow was positive or negative depending on the site examined. Occupancy probability was a relatively poor predictor and we caution its use as a proxy for landscape resistance. Our results underscore the importance of replication and multi-scale approaches in landscape genetics. Climate warming and landscape conversion may reduce the genetic connectivity of marten populations in the northeastern United States and represent the primary challenges to marten conservation at the southern periphery of their range.