State–space Models Link Elk Movement Patterns to Landscape Characteristics in Yellowstone National Park

Explaining and predicting animal movement in heterogeneous landscapes remains challenging. This is in part because movement paths often include a series of short, localized displacements separated by longer-distance forays. This multiphasic movement behavior reflects the complex response of an animal to present environmental conditions and to its internal behavioral state. This state is an autocorrelated process influenced by preceding behaviors and habitats visited. Movement patterns depending on the behavioral state of an animal represent the broad-scale response of that animal to the environment. Quantifying how animals respond both to local conditions and to their internal state reveals how animals respond to spatial heterogeneity at different spatial scales. We used a state–space statistical approach to model the internal behavioral state and the proximate movement response of elk (Cervus elaphus) to available forage biomass, landscape composition, topography, and wolf (Canis lupus) density during summer in Yellowstone National Park, USA. We analyzed movement paths of 16 female elk fitted with global positioning system (GPS) radio collars that recorded locations at 5-h intervals. Habitat variables were quantified within 175 m radii (onehalf of the median 5-h displacement) centered on the beginning location of each interval. Stepwise model selection identified models that best explained the movement distances of each animal. The behavioral state changed very slowly for most animals (median autocorrelation r 1⁄4 0.93), and all animals responded strongly to time of day (with more movement in the crepuscular hours). However, the spatial variables included in the best-fitting models varied substantially among individual elk. These results suggest that strong patterns of habitat selection observed in other studies may result from frequent visits to preferred areas rather than a reduction of movement in those areas.