By Maya Situnayake
Background: Tolerance for brown bears (Ursus arctos) in Sweden has had a long and complicated history, with ever-changing attitudes on human-bear coexistence in the last century. Currently, the population size is managed by establishing an annual quota for recreational hunters every year. The efficiency of hunters is impressive, and the quota is typically met within the first week of the hunting season. The hunt has a striking demographic impact, and the average lifespan of bears within the study site is now only five years. There are, however, several individuals (up to 22 years old) that appear to cope well with the hunting pressure, and have survived several successive hunting seasons. It remains unclear how these individuals adjust their behavioural strategies for avoiding being killed by hunters, what they are doing differently compared to individuals that are killed at younger age, and how that contributes to their survival. The bears in question are those occupying the wood-production forests of Dalarna and Gävleborg counties of central Sweden. Whether or not these bears display a habitat selection shift with the onset of the hunting season is established first. Following this, space-use between survivor and non-survivor bears is compared.
Methodology: To determine if a habitat shift occurs with the onset of the hunting season, resource selection functions were applied comparing used (recorded GPS) locations with available (randomly generated) locations. Models were created both at the population level (using a binomial GLMM with use vs. available as the response variable, and bear-year as a random effect) as well as for each individual bear-year (binomial GLMs with use vs. available as the response variable). The covariate ―Hunt seasonality‖ is a binary variable, categorising locations by whether they were recorded before the hunting season began, or recorded during the hunting season. This variable was included as an interaction term for all landscape variables in the models, and sensitivity to this interaction term was evaluated by p-value of significance (chi-square test for GLMM, F-test for GLM). To then evaluate the magnitude of the habitat shift, parameter estimates for all landscape variables from individual bear-year models were entered into a principal component analysis. The distance between paired observations (before and after hunt began, for the same bear-year), were computed, and survivors were compared with non-survivors.
To analyse space-use amongst survivors and non-survivors, habitat use was estimated for each group using logistic regression with a binary response term of survivor (=0) and non-survivor (=1). Here, both hunt seasonality and bear-age were included as interaction terms. Principal findings: Bears changed their resource selection in response to hunting (seasonality). This was evident for their habitat selection both at the population-level (e.g. “Habitat” variable, chi sq. test, p value < 0.01), and at the individual-level (e.g. 52% of individuals depicting shift in “Habitat” variable selection, and well over 5% of bear-years shifted selection for all other spatial features). The magnitude of this shift in behaviour did not differ significantly between survivors and non-survivors, and both groups appeared equally sensitive. In terms of habitat-use, when compared to non-survivors, survivors showed kept further away from most anthropogenic features such as minor roads and buildings, occupied locations with more rugged and elevated terrain, closer to major roads, and used habitats that offered more cover. Conclusion: Most bears notice the onset of the hunting season and adjust behaviour accordingly. These adjustments have various degrees of success, and three distinct space-use strategies have been gleaned from successful survivors: 1) The use of impenetrable habitats; 2) The use of locations making bears less-detectable; and 3) Protection from the human-shield effect, by using areas close to major roads. Deeper exploration into the role that vegetation density and hunter distribution play is needed to further confirm the proposed strategies.
Keywords: brown bear; coexistence; collar data; GPS; habitat selection; hunting; positional information; Sweden; resource selection functions; Ursus arctos