Identifying mechanisms of population change in two threatened grizzly bear populations

<p>Identifying the mechanisms causing population change is essential for conserving small and declining populations. Substantial range contraction of many carnivore species has resulted in fragmented global populations with numerous small isolates in need of conservation. Here I investigate the rate and possible agents of change in two threatened grizzly bear (Ursus arctos) populations in southwestern British Columbia, Canada. I use a combination of population vital rates estimates, population trends, habitat quality analyses, and comparisons to what has been described in the literature, to carefully compare among possible mechanisms of change. First, I estimate population density, realized growth rates (λ), and the demographic components of population change for each population using DNA based capture-recapture data in both spatially explicit capture-recapture (SECR) and non-spatial Pradel robust design frameworks. The larger population had 21.5 bears/1000km2 and between 2006 and 2016 was growing (λPradel = 1.02 ± 0.02 SE, λsecr = 1.01 ± 4.6 x10-5 SE) following the cessation of hunting. The adjacent but smaller population had 6.3 bears/1000km2 and between 2005 and 2017 was likely declining (λPradel = 0.95 ± 0.03 SE, λsecr = 0.98 ± 0.02 SE). Estimates of apparent survival and recruitment indicated that lower recruitment was the dominant factor limiting population growth in the smaller population. Then I use data from GPS-collared bears to estimate reproduction, survival and projected population change (λ) in both populations. Adult female survival was 0.96 (95% CI: 0.80-0.99) in the larger population (McGillvary Mountains or MM) and 0.87 (95% CI: 0.69-0.95) in the small, isolated population (North Stein-Nahatlatch or NSN). Cub survival was also higher in the MM (0.85, 95%CI: 0.62-0.95) than the NSN population (0.33, 95%CI: 0.11-0.67). This analysis identifies both low adult female survival and low cub survival as the demographic factors associated with population decline in the smaller population. By comparing the vital rates from these two populations with other small populations, I suggest that when grizzly bear populations are isolated, there appears to be a tipping point (de Silva and Leimgruber 2019) around 50 individuals, below which adult female mortality, even with intensive management, becomes prohibitive for population recovery. This analysis provides the first detailed estimates of population vital rates for a grizzly bear population of this size, and this information has been important for subsequent management action. To determine whether bottom-up factors (i.e. food) are limiting population growth and recovery in the small isolated population I use resource selection analysis from GPS collar data. I develop resource selection functions (RSF) for four dominant foraging seasons: the spring-early summer season when bears feed predominantly on herbaceous plants and dig for bulbs, the early fruit season where they feed on low elevation berries and cherries, the huckleberry season and the post berry season when foraging behaviours are most diverse but whitebark pine nuts are a relatively common food source. The differences in overall availability of high-quality habitats for different food types, especially huckleberries, between populations suggests that season specific bottom-up effects may account for some differences in population densities. Resource selections are a very common tool used for estimating resource distribution and availability, however, their ability to estimate food abundance on the ground are usually not tested. I assessed the accuracy of the resulting RSF models for predicting huckleberry presence and abundance measured in field plots. My results show that berry specific models did predict berry abundance in previously disturbed sites though varied in accuracy depending on how the models were categorized and projected across the landscape. Finally, I combine spatially explicit capture-recapture methods and models developed from resource selection modelling to estimate the effect of seasonal habitat availability and open road density, as a surrogate for top-down effects, on the bear density in the two populations. I found that population density is most strongly connected to habitats selected during a season when bears fed on huckleberries, the major high-energy food bears eat during hyperphagia in this area, as well as a large baseline difference between populations. The abundance of high-quality huckleberry habitat appears to be an important factor enabling the recovery of the larger population that is also genetically connected to other bears. The adjacent, smaller and genetically isolated population is not growing. The relatively low abundance of high-quality berry habitat in this population may be contributing to the lack of growth of the population. However, it is likely that the legacy of historic mortality and current stochastic effects, inbreeding effects, or other Allee effects, are also contributing to the continued low density observed. While these small population effects may be more challenging to overcome, this analysis suggests that the landscape can accommodate a higher population density than that currently observed.</p>