The spatial scaling and individuality of habitat selection in a widespread ungulate

Animal-habitat relationships tend to manifest at specific spatial scales. Accurately identifying these scales and accounting for the variance in habitat selection across them is crucial for linking habitat selection patterns to the ecological processes giving rise to them. Although this fundamental issue has long been recognized, it has been seldom addressed empirically in habitat selection studies. In this study, we investigated how spatial scale influences the outputs of habitat selection analyses. Furthermore, we examined whether the effect of spatial scale varies among individual animals and whether these effects could be predicted via intrinsic or extrinsic factors. We used a dataset collected from 485 GPS-collared white-tailed deer (Odocoileus virginianus) across three study sites in Missouri, USA to model habitat selection at 65 spatial scales from 900 m2 to 15 km2 using integrated step selection functions. To investigate potential drivers of spatial scaling we used multiple linear regression to model how scale of effect, defined as the spatial scale at which model AIC was minimized, could be predicted by intrinsic (age, sex, and home range size) and extrinsic factors (study site, season, mean percentage forest in home range, mean distance to nearest road in home range). Scale of effect varied substantially among individuals, and individual variation in scale of effect was predicted by home range size, study site, and proportion of forest within a home range. In contrast, other intrinsic and extrinsic factors had little to no relationship with scale of effect. Parameter coefficients for forest cover and distance to nearest road varied strongly with opposing directionality of responses across spatial scales, revealing that spatial scale may bias habitat selection analyses. Coefficients were both positive and negative at different scales for an average of 63.2