Get reliable rangeland science

Most habitat selection studies explicitly acknowledge issues of spatial scale, but many do not fully consider temporal scale. Even when organisms exhibit clear diel behavioral patterns that coincide with changes in habitat preference, studies often ignore selection patterns finer than seasons. Using elk (Cervus canadensis), a species with crepuscular diel behavioral patterns, as an example, we sought to determine if modeling habitat selection at a diel temporal scale improved understanding of habitat selection more than simply modeling at a seasonal scale. We fitted 23 female elk in the Jemez Mountains of New Mexico with GPS collars and recorded location fixes every 5 - 6 hours from December 2012 to February 2015. To create biologically-valid temporal delineations for modeling, we used climatic data to partition seasons, and we observed elk behavior to determine major shifts in foraging behavior throughout diel periods. To model habitat covariates, we distributed circular sampling units systematically from a random origin and populated them with vegetation type, fire history, aspect, distance to water sources, and distance to roads, and predicted forage biomass. We modeled elk habitat selection with a bootstrapped resource selection probability function, using a negative binomal generalized linear model, by seasons (winter, spring, summer, fall) and by diel intervals (dawn-dusk, midday, night) within seasons. Diel models detected more detailed, novel, and some patterns contradictory to seasonal models. During dawn-dusk intervals, elk tended to select for grassland vegetation and low or high canopy cover. At midday, elk selected for forest vegetation types, recently burned areas, and areas with intermediate levels of herbaceous biomass. During the night, elk selected for areas with highest levels of herbaceous biomass, areas closer to water, against canopy cover, and recently burned areas. Our findings highlight the importance of accounting for fine temporal scales and behavioral patterns when modeling habitat selection.