Using fine scale resolution vegetation data from LiDAR and ground-based sampling to predict Pacific marten resting habitat at multiple spatial scales

Abstract

Pacific marten (Martes caurina) populations have become fragmented and constricted throughout their western range, often due to factors such as increased severity of wildfires and timber harvesting. Future population declines are predicted given decreasing snow packs and changes in vegetation communities. One element of marten habitat, rest and den structures, may be particularly vulnerable. These structures are used for protection from inclement weather and predation, as well as sites for parturition. Rest structures are often considered a limiting habitat element; characterizing their abundance, type, and distribution has been suggested as a way to evaluate habitat quality. We evaluated marten resting habitat, combining vegetation data from light detection and ranging technology (LiDAR) and ground-based surveys. From 2009–2013 and 2015–2017, we located 312 unique rest structures used by 31 martens (18 males, 13 females). With ground-based surveys, we examined selection of used structures by measuring the diameter at breast height for comparison of rest structures and trees located in random plots. For broader landscape-level predicted habitat, we paired used locations with 624 randomly-sampled locations, and optimized 14 habitat covariates at 12 spatial scales using case-controlled logistic regression. Each covariate’s optimized scale was used to develop a series of a priori hypothesized multi-scale habitat selection models. Martens selected woody structures that were larger than random structures (rest structures = 97.8 ± 31.0 cm; random = 52.7 ± 24.9 cm, x̅ ± SD, t = 21.6, p < 0.001). Marten habitat selection was also positively associated with increased canopy cover and structural complexity within 270 m radius of suitable rest structures and increased tree cover at the broadest scale evaluated (990 m). Our models revealed elevation was positively correlated with predicted marten resting habitat; average elevation at our used sites was 1940 m. Finally, our model depicted areas of predicted habitat near road systems, but we assumed this was an artifact from our sampling bias. Because both canopy cover and structural complexity were optimized at a 270 m radius, this may be an appropriate scale to consider for management activities such as establishing leave islands or focal areas of restoration. We provide one of the first evaluations of marten habitat incorporating the use of LiDAR, which can be broadly and accurately extrapolated for management planning and restoration prioritization.