Matthew Gould, Government Agency

Approximately 3–5 million visitors recreate annually in Yellowstone (YNP) and Grand Teton (GTNP) National Parks, USA, primarily near developed areas, campsites, and trails from May-October. This period overlaps with the active season for grizzly bears (Ursus arctos), increasing potential for human–bear encounters. Whereas the human-bear interface and associated mitigation strategies in developed areas are well studied, the potential for human-bear interactions along trails is less understood. Identifying where and when bears cross or use trails is critical for reducing risk. GPS collars provide detailed location data but connecting consecutive fixes with straight lines can misrepresent crossings and use intensity. Variable GPS fix rates and movement between fixes add uncertainty, particularly given high mobility of grizzly bears. Continuous trajectories with quantified uncertainty provide a more realistic and useful representation of movement paths. Treed Gaussian Process (TGP) models address this by combining Gaussian process interpolation with Bayesian treed partitioning to capture distinct behavioral shifts. Each shift in movement behavior is fit with an individual Gaussian process, improving accuracy for heterogeneous movement. Posterior predictive sampling produces continuous trajectories with credible intervals, enabling robust inference. Derived movement metrics can be computed and linked to environmental covariates. We applied TGPs to grizzly bear GPS telemetry data (2009–2025) to reconstruct continuous trajectories, quantify trail interactions, and identify crossing hotspots and trail use in YNP and GTNP. Results will inform National Park Service management strategies to enhance visitor safety on recreational trails while providing for long-term conservation of grizzly bears