| Conservation of wildlife populations on managed landscapes requires planning at the appropriate spatial scale, since scale dramatically affects results and thus interpretation. We examined multi-scale habitat relationships at Pacific marten (Martes americana) rest structures in Lassen National Forest using fine-resolution vegetation data (30-meter airborne Light Detection and Ranging [LiDAR]). Using a moving-window framework to compare selection, we optimized 14 covariates at 12 spatial scales (30 meters-990 meters) centered on each rest structure. We monitored martens from 2009-2012 and 2015-2017 (n = 312 resting structures, 31 martens), and then compared used versus randomly sampled locations (n-rand = 624) to develop multivariate habitat selection models. Our top model included trees per acre (990-meter scale) and elevation (900 meter), suggesting that martens select for increased tree cover at higher elevations at the home range scale. Increased structural complexity and stand density surrounding rest structures (270 and 180 meters, respectively) increased probability of selection. Because martens selected locations with vegetation characteristics optimized at 180-270 meters, 270 meters may be an appropriate scale to consider for management, for instance, establishing leave islands or focal areas for restoration. We provide the first evaluation of marten habitat using LiDAR, which can be broadly and accurately extrapolated for prioritizing management planning and restoration. |