Realized habitat selection and space use by brown bears reintroduced to the Italian Alps

Reintroductions are often difficult and can result in increased dispersal and mortality of released animals. Especially landscape fragmentation poses a challenge to conservation biologists and wildlife managers. Connecting habitat patches that support occupancy is key to ensuring long-term population recovery for umbrella species such as brown bears (Ursus arctos). In 1999 the reintroduction project ‘Life Ursus’ was initiated and 10 brown bears were translocated from Slovenia to the Eastern Italian Alps (Adamello-Brenta) where the bear population was considered extirpated by the end of the 20th century. Here, we describe the habitat selection and space use patterns of reintroduced brown bears in Trentino, Italy, with the ultimate goal of identifying corridors between preferred habitat patches. We built annual resource selection functions (RSF) using generalized linear mixed-effects models with data from 6 bears fitted with GPS collars between 2006 and 2012 at the landscape and home-range scales. Then, integrating habitat selection, movement and landscape features we employed least cost path (LCP) analyses to empirically define regional movement corridors. Finally, we used Kappa-statistics to compare the realized habitat selection at the landscape scale with a habitat suitability model (HSM) developed to predict potential habitat prior to reintroduction. At the landscape scale bears selected home-ranges at intermediate elevations in steep terrain. Habitat selection inversely correlated with road occurrence and human use landcover types, i.e. settlements, pastures and agricultural lands. Bears selected landcover types that provide forage, such as shrublands, deciduous forests, but also orchards. Within their home-ranges bears also avoided landcover types related to human use (especially bike trails, settlements and pastures). K-fold cross-validation indicated a high predictive capacity of our RSFs. We identified road crossings of movement corridors between preferred habitat patches throughout the study area; with higher densities in the south. A comparison of our LCP model with confirmed road crossings and locations of 5 collisions indicated good predictive capacity. When bears crossed roads they preferred shrublands or wetlands. The comparison between the predictive pre-reintroduction HSM and the realized habitat selection showed fair agreement with highest agreement in settlements and agricultural lands. Spatial discrepancies resulted from the HSM predicting lower suitability in orchards and shrublands and higher suitability in mixed forests than the RSF. Lastly, we provide recommendations for the management and conservation of bears in the Alps. The combined process-based approach of the RSF and LCP analyses aids to identify both potential habitat for brown bear settlement and corridors that will allow for settlement in the future. This approach also provides information on the biggest constraints to bear movements, such as the barrier created by the Adige-Valley. We make recommendations for urbanization, road design and improvements (e.g. overpasses), human recreation and the habitat suitability in different parts of the valley. We suggest verifying and monitoring the use of movement corridors with e.g. camera traps or DNA monitoring. Increasing awareness for key brown bear habitats and corridors and raising local acceptance of this species, especially in the interface between human dominated landscapes and wilderness areas, is necessary for brown bear conservation.