Modeling broad-scale patterns of elk summer resource selection in Montana using regional and population-specific models

Understanding animal distribution is important for management of populations and their habitats. Across the western United States, elk (Cervus canadensis) provide important ecological, cultural, and economic benefits and the sound management of their habitats is of vital importance. In western Montana, National Forest lands are managed in part to provide and protect elk habitat needs, and summer elk habitat is managed with consideration to motorized routes. We evaluated the relative importance of nutritional resources, access routes, and other landscape attributes on elk summer resource selection at multiple spatial scales, and compared resource selection among nine different southwestern Montana elk populations to determine the applicability of generalized regional models for informing habitat management recommendations. First, we developed nine population-specific and two regional summer resource selection models. Second, we evaluated the predictive performance of each model within and among elk populations using cross-validation scores to identify the best model. We found that in all populations nutritional resources, best represented using normalized difference vegetation index (NDVI) metrics, were the most important factors associated with elk summer resource selection. Access routes affected resource selection in all populations; however, the influence of access routes was relatively modest as compared with nutritional resources. Of the access route covariates we considered, density of all routes (i.e., routes open and closed to motorized use) explained most variation in summer elk resource selection. Validation of population-specific resource selection models among populations revealed that in many cases model predictions extrapolated to areas outside of the development area had modest to poor predictive performance, especially as distance from the modeled population increased. Thus, caution should be used when extrapolating resource selection models based on a single study population to other populations. Regional models of resource selection predicted resource selection across populations better than population-specific models, particularly when constructed by pooling data from multiple populations, and we recommend these types of models be used to inform regional habitat management policies. Our results suggest that managers should expand any current management paradigm for elk summer habitat that is focused on limiting access route density to also consider nutritional resources as an important component of elk summer habitat.