Validating the performance of occupancy models for estimating habitat use and predicting the distribution of highly-mobile species: A case study using the American black bear

Occupancy models have become a valuable tool for estimating wildlife-habitat relationships and for predicting species distributions. Highly-mobile species often violate the assumption that sampling units are geographically closed shifting the probability of occupancy to be interpreted as the probability of use. We used occupancy models, in conjunction with noninvasive sampling, to estimate habitat use and predict the distribution of a highly-mobile carnivore, the American black bear (Ursus americanus) in New Mexico, USA. The top model indicated that black bears use areas with higher primary productivity and fewer roads. The predictive performance of such models is rarely validated with independent data, so we validated our model predictions with 2-independent datasets. We first assessed the correlation between predicted and observed habitat use for 28 telemetry-collared bears in the Jemez Mountains. Predicted habitat use was positively correlated with observed use for all 3 years (2012: ρ = 0.81; 2013: ρ = 0.87; 2014: ρ = 0.90). We then predicted the probability of use within a cell where a bear mortality was documented using 2043 mortality locations from sport harvest, depredation, and vehicle collisions. The probability of habitat use at a mortality location was also positively correlated with observed use by the species (2012: ρ = 0.74; 2013: ρ = 0.89; 2014: ρ = 0.93). Our validation procedure supports the notion that occupancy models can be an effective tool for estimating habitat use and predicting the distribution of highly-mobile species when the assumption of geographic closure has been violated. Our findings may be of interest to studies that are estimating habitat use for highly-mobile species that are secretive or rare, difficult to capture, or expensive to monitor with other more intensive methods.