Snow enshrouds up to one third of global land mass and exerts a major influence on the demography and movements of animals that occupy these snowscapes. The snow-covered season is often limiting for many terrestrial mammals as cold temperatures, restricted resources, and constrained mobility can increase physiological stress relative to summer conditions. The accelerated pace of warming in arctic and boreal systems is altering the phenology and composition of snowscapes, with profound implications for widlife. Yet, linking animal movement and resource selection to snowscape properties is challenged by a paucity of snow products that reflect spatiotemporal heterogeneity at scales relevant to biological data. A lack of knowledge regarding how animals respond to snowscapes represents a critical gap in understanding vulnerability and resilience of natural systems to climate change. We used a step-selection framework to evaluate the influence of snowscape properties, as characterized by remotely sensed products (MODIS Snow Cover and MODSCAG) and a snowpack evolution model (SnowModel), on scale-dependent movement and resource selection of Dall sheep (Ovis dalli) in Lake Clark National Park, Alaska. We evaluated the relative performance of each snow metric using competing conditional logistic models ranked by quasilikelihood under the independence model criterion (QIC) and 5-fold cross-validation.
Results/Conclusions
We monitored movement and resource selection in 30 sheep (12 rams and 18 ewes) over the course of three winters (2006 – 2008), defined as January through May of each year. We defined eight scales of selection characterized by steps taken over seven-hour (our shortest fix interval) through 38-day intervals. We generated snow depth and density outputs from SnowModel at resolutions of 90-m with wind processes through 10-km without wind and used daily snow cover products at 500-m resolutions. Snow depth and density at 90-m with wind consistently performed best at almost all scales of selection except for movements made over the longest step intervals, indicating that many readily available snow products may be insufficient for describing animal movements at all but the largest spatiotemporal scales. Although SnowModel was developed primarily for hydrological applications, we demonstrate its utility in describing relevant variation in sheep resource selection at finer scales during winter. However, our findings highlight a critical need for further development of fine-scaled, process-driven snow metrics for use in animal movement modeling, particularly in light of increased climatic variability at higher latitudes and subsequent impacts on animal fitness and demography.