Complexity drives resource selection of two sympatric ungulate species

For decades, mule deer Odocoileus hemionus populations have appeared to shrink across their range in North America, while white‐tailed deer O. virginianus populations have increased, and their range has expanded. The underlying drivers of these patterns are unclear, and a combination of factors may be at play, including land use changes, climate change, and direct or indirect competition between the two species. Understanding how the two species differentially use the landscape and select habitat would help shed light on the underlying drivers of mule deer decline and highlight opportunities to change population trajectories through habitat management. Using an extensive GPS dataset on mule deer (n = 345) and white‐tailed deer (n = 345) in their sympatric range, we used resource selection functions to investigate the relative probability of selection of different resources available to the two species. We estimated population‐level relative probability of selection in winter and summer and demonstrated that a combination of factors, including nutrition, thermal cover, and human impact described resource selection better than any single factor on its own, despite annual variation in these metrics. In keeping with long‐established patterns, we verified that white‐tailed deer had more consistent selection than mule deer for agriculture, higher human modification, road density and tree cover. Across seasons and years, individuals of both species consistently selected for shrubs, perennial forbs and grasses, and steeper slopes. Using our large dataset, we performed out‐of‐sample validation and demonstrated that although the resource selection functions predicted space use well at the population level, they performed poorly at predicting resource selection between individuals or even within an individual over time. Thus, while we successfully identified key habitat characteristics that were selected differentially by individuals from the two deer species, inferring individual habitat selection and future resource use patterns proved challenging.