A piecewise linear modeling approach for testing competing theories of habitat selection: an example with mule deer...

Habitat selection fundamentally drives the distribution of organisms across landscapes; density-dependent habitat selection (DDHS) is considered a central component of ecological theories explaining habitat use and population regulation. A preponderance of DDHS theories is based on ideal distributions, such that organisms select habitat according to either the ideal free, despotic, or pre-emptive distributions. Models that can be used to simultaneously test competing DDHS theories are desirable to help improve our understanding of habitat selection. We developed hierarchical, piecewise linear models that allow for simultaneous testing of DDHS theories and accommodate densities from multiple habitats and regional populations, environmental covariates, and random effects. We demonstrate the use of these models with data on mule deer (Odocoileus hemionus) abundance and net energy costs in different snow depths within winter ranges of five regional populations in western Idaho, USA. Regional population density explained 40% of the variation in population growth, and we found that deer were ideal free in winter ranges. Deer occupied habitats with lowest net energy costs at higher densities and at a higher rate than compared to habitats with intermediate and high energy costs. The proportion of a regional population in low energy cost habitat the previous year accounted for a significant amount of variation in population growth (17%), demonstrating the importance of winter habitat selection in regulating deer populations. These linear models are most appropriate for empirical data collected from centralized habitat patches within the local range of a species where individuals are either year-round residents or migratory (but have already arrived from migration).