Moving in a crowded world: ecological and human-related factors affecting brown bear (Ursus arctos) space-use patterns

Under the current scenario of human expansion and land-use change, one resource emerges as being particularly disputed between humans and other wildlife species: space. The spatiotemporally detailed and real-time nature of Global Positioning System (GPS) tracking data supports the use of tagged animals as in situ sensors of the environment, to document how ongoing changes are affecting species’ distribution and ecological function. Large carnivores are particularly susceptible to disturbance from infrastructure development, and can represent a promising case study to investigate the effects of human expansion on species spatial ecology at multiple levels, spanning from patch- to landscape scales. I investigated space-use patterns in a south-eastern European population of brown bears (Ursus arctos), whose distribution is shared among more than five countries, from Slovenia to Northern Greece, with its core area located between Slovenia and Croatia. In the first chapter, I investigated ecological and human-related effects on home range size and inner configuration of Dinaric brown bears (Ursus arctos) contrasting two areas, one located in the North (n= 5 bears, 1 females, 4 males) and one in the South (n= 5 bears, 2 females, 3 males), which differed in terms of road and human density, as well as in the availability of supplementary feeding sites. I used Brownian bridge movement models (BBMMs) to estimate circadian and seasonal home ranges and I used linear mixed-effect models (LMM) to investigate the effects of gender, time of the day, season and study area on home range size. Using an individual- based method, I also depicted seasonal core areas and used Environmental Niche Factor Analysis (ENFA) to assess if the internal configuration of seasonal home ranges changed among study areas. Although we failed to find a sex effect on home range size, time of day was an important predictor of home range size, with nocturnal home ranges larger (103.3 km2 72.8) than diurnal ones (62.3 km2 16.6). Then, I detected a seasonal effect on home range size, but this was limited to areas where bears had relatively lower accessibility to artificial feeding sites. Where densities of roads, human settlements, and artificial feeding sites were the highest, selection of core areas by bears was characterized by avoidance for anthropogenic features. Overall, this study revealed that even within the same population range, bears can show behavioural plasticity and adapt to local conditions of human disturbance. Although behavioural plasticity may contribute to ensure carnivore persistence in human-dominated areas, the changes in home range patterns that we detected can also be a warning sign of environmental degradation. The aims of chapter II were i) to classify individual movement patterns of brown bears indicating frequency, period, duration and length of inter-seasonal range movements; ii) to assess the main environmental descriptors within the pre-migratory and post-migratory ranges, as compared to the average habitat conditions in our study area; iii) to identify main differences in bear habitat use during pre-migratory and post-migratory phases. I classified individual movement patterns by means of non-linear Net Squared Displacement (NSD) models. By means of canonical Overlying Mean Index analysis (OMI), I identified habitat descriptors of pre- migratory and post-migratory ranges at the landscape scale. I then quantified the strength of variation in bear habitat use at summer versus fall ranges using a latent selection difference analysis (LSD), at both the population and the individual level. My findings revealed that 6 out of 12 individuals showed facultative and partial seasonal migrations between disjointed seasonal ranges, in contrast to the other bears that remained resident or nomadic throughout the year. Migratory patterns were markedly seasonal, with all departures occurring between mid-September and mid- October (median= 19th September), and returns occurring before the wintering period (median= 18th November). Migratory movements connected seasonal ranges up to >40 km apart (mean = 28.9 km). Most bears migrated from areas characterised by coniferous and mixed forests to lower areas with high proportion of deciduous forest, forest edge and shrubs. Compared to pre-migratory ranges, within migration ranges bears increased both their distance to anthropogenic structures (i.e. paved roads, settlements, artificial feeding sites, cultivated lands) and their selection for highly productive areas (i.e. deciduous forest, forest edge and shrubs). Due to lack of data on fine-scale forest productivity and on human disturbance within bear ranges, the ultimate cause that triggered the observed bear migrations remains to be assessed. However, our findings represent a remarkable contribute that improved our understanding of the species ecology, as migration patterns such as those observed in our study have never been observed before in any other bear population in Europe. In chapter III, I investigated habitat selection on a seasonal basis for 11 individuals (4 females and 7 males), focussing the analysis on bears that did not perform seasonal migrations. In particular, I investigated seasonal changes in bear use/avoidance of human infrastructures such as highway, paved roads, railway, forest roads, human settlements and supplementary feeding sites. The general working hypothesis of this chapter was that bears might display stronger avoidance towards our proxies of human disturbance during periods of increased hunting pressure (i.e. spring and fall). To confirm my hypothesis, we focussed on non-migrating individuals to first exploring general patterns of seasonal habitat selection using k-select analysis, aimed at identifying the principal components of bear seasonal habitat preferences accounting for individuals. In a second phase, I have used resource selection functions (RSF) in a use/available design to compare habitat conditions that were available to bears within their home ranges to habitat that bears used (i.e. at GPS locations). In agreement with my hypothesis, the k-select analysis showed a common pattern of habitat selection during spring and fall for the quasi-totality of individuals, characterized by avoidance to forest roads and selection for areas more hardly reachable by humans (i.e. slopes). Conversely, during summer, bears showed more heterogeneous patterns of habitat selection, reflecting higher individual variability in habitat selection for this period of the year. The results of the RSF analysis confirmed these patterns more thoroughly, with distance to forest roads (spring= 0.33 0.01; fall = 0.30 0.04) and use of slopes (spring= 0.12 0.01; fall = 0.20 0.02) playing a major role in defining bear habitat selection during both periods when hunting was open, and a significant use of supplementary food (i.e. decreased distance from artificial feeding sites) during hyperphagia season (fall = -0.30 0.04). Overall, my findings suggest that hunting might be perceived by bears as a form of predation risk that forces them to increase their concealment, influencing within- home-range habitat use and distribution of the individuals. In chapter IV, I developed a modelling approach to simulate bear movements among suitable resource patches, integrating classical habitat selection studies and cutting-edge movement algorithms. To this aim, I have used a movement-based modelling approach to project potential corridors connecting summer and fall habitat patches. To model habitat patches associated to intensive habitat use during summer and fall, I used Resource Selection Functions (RSF) based on bear relocations representing stationary behaviour (i.e. feeding or resting). Based on bear trajectories representing bear travelling, I then studied bear response to both natural elements (e.g. terrain topography) and human structures (e.g. roads) during their movements using Step Selection Functions (SSF). After quantifying the effects of each tested landscape feature on bear movements, I modelled a map of landscape permeability to bear movement. Finally, I used a Randomised Shortest Path (RSP) algorithm to project potential bear corridors between summer and fall habitat patches within our study area. According to my findings, bears can successfully travel across sub-optimal habitat to reach suitable patches in the fall, although the presence of anthropogenic structures such as highways, main paved roads, railways, and cultivated fields strongly decreased the probability of bear traveling. Model predictions correlated moderately with the frequency of bears killed by vehicle collisions (range r=0.68-0.79). Overall the degree of inter-patch connectivity was rather high across the study area, with many potential bear paths connecting suitable summer and fall habitat patches. However, we recommend that improvement of mitigation measures should be evaluated at the intersection between modelled corridors and linear infrastructures. Our modelling approach might be applied also to other ecological contexts where habitat fragmentation represents a major threat to the long-term persistence of wide- ranging mammalian populations. Overall, this case study represented an intriguing occasion to assess how recent trends of human development might affect the life history traits of wide-ranging brown bear population.