Evaluating methods for identifying large mammal road crossing locations: black bears as a case study

Context Roads have several negative effects on large mammals including restricting movements, isolating populations, and mortality due to vehicle collisions. Where large mammals regularly cross roads, driver safety is also a concern. Wildlife road crossing structures are often proposed to mitigate the negative effects on wildlife and human safety. However, few studies have compared the efficacy of different methods in identifying wildlife road crossing and mitigation locations. Objectives We examined five commonly used methods to identify road crossing locations for wildlife to determine which method best captured empirical crossing locations. Methods We used GPS collar data on black bears in Massachusetts, USA, to estimate road crossing locations with a road crossing resource selection function, factorial least-cost paths, resistant kernels, Circuitscape, and an individual based movement model. We evaluated model performance on each road class and all road classes combined with a hold out road crossing data set. Results All methods performed well, but Circuitscape consistently outperformed the other methods, both for each road class and all road classes combined. Conclusions Road mitigation efforts for wildlife are often costly and have a long-term footprint on the landscape, therefore, it is important to select locations for these efforts that provide functional connectivity for wildlife. We are the first to compare different road crossing models for large mammals across a large study area and on different road types. Our results provide information to assist researchers and managers in selecting an analytical method for identifying potential road mitigation locations for large mammals.