Predicting connectivity from movement trajectories and step selection

Predicting and mapping connectivity are essential for ecology and conservation in the face of environmental change. Yet connectivity modeling is challenging because it requires understanding how organisms successfully move across complex landscapes. GPS tracking increasingly produces detailed data on movement trajectories and fine‐scale resource selection, yet it remains unclear how to reliably link such data to connectivity models without relying on individual‐based simulations. Currently, most analytical connectivity models are not built from movement parameters that are estimable from GPS‐tracking data, and none capture directional persistence in movement that is often observed in movement trajectories, such as directed movement during dispersal and migration. We address these gaps by advancing new models using the connectivity framework of spatial absorbing Markov chains (SAMC) to directly link step selection and directional persistence in movement to connectivity. We unite these concepts by altering transition matrices used in connectivity models and by extending the SAMC to accommodate correlated random walks through a re‐parameterization using spatial line graphs. We illustrate these advances with simulations and an example using the Florida panther Puma concolor coryi . We find that the SAMC captures expectations from individual‐based simulations of step selection and can be several orders of magnitude faster. Modeling connectivity of the Florida panther with directional persistence that captures correlated random walks predicts faster movement across the landscape, leading to greater expected connectivity. Our results demonstrate that common current practices do not correctly relate step‐selection functions to resistance used in connectivity models, and that incorporating directionality into connectivity efforts leads to greater realism for modeling but at a computational cost. We provide guidance on when this realism may be needed and how to efficiently link step selection with connectivity. Taken together, we offer a generalizable and powerful framework to reliably link movement trajectories to connectivity across landscapes.