Patterns of home range use emerge from the interaction between an animal's movement process and a multivariate landscape that is perceived by the animal at multiple spatial and temporal scales. Modeling this interaction is difficult because while typical statistical analyses describe broad-scale patterns of home range use, they fail to characterize the underlying movement process. Likewise, correlated random walk analyses focus on the movement process, but often fail to capture broad-scale home range behaviour. Previous work has shown that adding a centralizing tendency, such as preferential movement towards a den or roost, into a correlated random walk model can give rise to stable home range patterns. However, many species do not exhibit the central-place attraction behavior that gives rise to stable home ranges in these models. Here, we test whether the addition of a latent memory process into mechanistic home range models is sufficient to capture home range behavior under a variety of landscape patterns. We then fit these models to movement data collected from 18 female elk (Cervus elaphus) in Yellowstone National Park and explore the consequences of using these models to predict habitat use and home range patterns of elk in other areas of Yellowstone.

Results/Conclusions

Adding a simple memory process to the movement simulation allows for the development of stable home ranges under a wide variety of movement models and landscape compositions. When we fit these models to the elk data, we found that although there was a high degree of individual variability in movement patterns, coarse features of the movement process emerged: a slow decay rate for the memory landscape, and a strong preference for landscapes that provide a mixture of high-forage grasslands and high-cover, young stands of coniferous forest (the relative mixture of these two landscape components depended strongly on season and time of day).The model was moderately successful at predicting the movement patterns of other individuals, indicating that there is considerable individual variability in the underlying movement rules. We discuss the implications and limitations of using mechanistic home range models in management contexts and propose alternatives that may improve the generality of this approach.