Generalizing animal movement predictions across landscapes: a scalable framework grounded in empirical telemetry data
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Accurately predicting animal movement across broad ecological contexts is vital for effective wildlife management, yet most models are developed for specific locations and fail to generalize across diverse landscapes. New approaches are needed to scale individual-level movement processes in a way that accounts for ecological variation.
Methods
We developed a modeling framework to predict movement distance and habitat selection across large geographic areas using GPS telemetry data from 564 wild pigs collected in 25 studies across the United States. Movement distances were modeled using gamma distribution parameters, and habitat selection was modeled using integrated step selection analysis. We used gradient-boosted regression models with nested cross-validation to optimize predictions and reduce overfitting. To account for spatial non-stationarity in habitat preference, individuals were grouped by habitat availability.
Results
Our framework captured broad-scale spatial and temporal variation in movement behavior. Mean daily movement distances ranged from 152 to 1,404 meters, with the greatest seasonal variation observed in northern populations. Habitat selection models revealed consistent preference for wetlands and avoidance of croplands, with variation in other land types depending on the local ecological context. Leave-one-out sensitivity analyses indicated greater model reliability in subsets with more individuals, sites, and land types.
Conclusions
This framework offers a scalable approach to predicting animal movement across heterogeneous landscapes. It enables estimation of movement parameters in regions lacking local telemetry data, supporting more informed decision-making for invasive species management, disease control, and conservation planning.
