Modeling eighteen years of community science data reveals extensive recolonization of bobcats in Illinois, USA

Context: Many terrestrial mammals have undergone substantial distribution changes in recent decades; yet collecting broad-scale occurrence data for carnivores is difficult due to their low densities and cryptic behaviors. Carnivore observations from community (i.e., citizen) science programs can be a potentially valuable approach for understanding changes in carnivore distributions over long time periods. Objectives: We used 18 years of bobcat (Lynx rufus) observations collected by archery deer hunters (i.e., participants) across Illinois, USA, to estimate spatiotemporal patterns in occurrence and determine how landscape features influenced patterns of recolonization. Methods: We developed Bayesian spatial and non-spatial multi-scale dynamic occupancy models to estimate county-level occupancy, persistence, and colonization and participant-level occupancy. We modeled county-level parameters as a function of multiple a priori landscape covariates and compared model predictive performance using cross-validation. Results: Our non-spatial occupancy model had greater predictive support than our spatial occupancy model. Mean annual statewide county-level occupancy increased from approximately 0.43–0.83 while mean annual participant-level occupancy increased from approximately 0.07–0.28. Bobcats were primarily restricted to southern Illinois during the early 2000s but by 2018 occurred throughout western and southern Illinois. Landscape covariates had relatively weak effects on model parameters. Conclusions: Our study illustrates how community science observations analyzed with hierarchical occupancy models can be used to model spatiotemporal changes in species distributions. Bobcats have recolonized much of Illinois, but this colonization was not strongly mediated by county-level landscape features at the scales we measured.