RovQuant : estimating density, abundance and population dynamics of bears, wolverines and wolves in Scandinavia

Background Reliable estimates of population status are a pre-requisite for informed wildlife management. However, abundance estimates can be challenging to obtain, especially for species that are highly mobile, rare and elusive. For nearly two decades, management agencies in Norway and Sweden have been monitoring populations of three large carnivores – brown bear (Ursus ursus), wolf (Canis lupus), and wolverine (Gulo gulo) – using non-invasive genetic sampling (NGS). DNA extracted from fæces, urine, and hair can be used to identify the species, sex, and individual from which each sample originated. Samples thus become evidence of the presence of an individual carnivore in space and time. Project RovQuant was initiated in 2017 with the objective to develop statistical methods that allow a comprehensive assessment of population status and dynamics using NGS data and other sources of information collected by the national monitoring programs in Sweden and Norway. Approach We developed a Bayesian open-population spatial capture-recapture (OPSCR) model that, using a combination of NGS and recoveries of dead carnivores, jointly estimates 1) the spatial variability in the probability of genetic detection, 2) the spatial distribution and interannual movements of individuals and 3) population size and dynamics. We fitted this model to the extensive individual-based monitoring data for bears, wolverines, and wolves, which had been compiled in the Scandinavian large carnivore database Rovbase 3.0 between 2012 and 2019. Results The OPSCR model yielded annual density maps both total and jurisdiction-specific population sizes for each species. The estimated number of bears for April 1, 2018 was 2 757 (95% credible interval, CrI: 2 636 - 2 877), of which 2 615 (CrI: 2 499-2 732) were located in Sweden and 142 (CrI: 124-162) in Norway. The estimated number of wolverines for December 1, 2018 was 1 035 (CrI: 985 - 1 088), of which 660 (CrI: 619-703) were located in Sweden and 375 (CrI: 353-397) in Norway. The estimated number of wolves for October 1, 2018 was 375 wolves (CrI: 352 - 402), of which 297 (CrI: 274-322) were located in Sweden and 79 (CrI: 72-86) in Norway. In addition to density and abundance estimates, the OPSCR models also yielded estimates of survival, recruitment, and space use parameters for each species. Six additional tasks linked to the development of OPSCR model were implemented as either prerequisite technical developments or to address persistent challenges in monitoring and management of large carnivores in Scandinavia. Although this report focuses on the main results from the OPSCR model, findings related to these additional tasks are briefly described as well. Conclusions The unique Scandinavian data set combined with a novel OPSCR model allowed RovQuant to quantify the population status of three large carnivore species at an unprecedented spatial scale (up to 593 000 km2). The approach used here has several advantages over proxy-based approaches for obtaining estimates of population size. The OPSCR model directly estimates annual abundance from NGS and dead recovery data while accounting for spatial and temporal variation in detection probability of individuals. The resulting estimates are spatially explicit, allowing extraction of abundance estimates and associated measures of uncertainty for any spatial extent desired by the user (e.g. management unit). Annual cause-specific mortality and recruitment are also estimated, which are both useful metrics of the population’s status and trajectory. Importantly, this approach efficiently exploits the data (NGS and dead recoveries) currently collected annually by Swedish and Norwegian management authorities at the population level. Although the OPSCR model has been extensively tested, it constitutes a novel approach and is still under development. The ability of the model to produce trustworthy estimates relies on several statistical assumptions and on the suitability of the input data. For example, although the model was able to produce annual density maps and abundance estimates for bears throughout Scandinavia, the current patchy sampling for this species in Sweden means that confidence in the reliability of the results for bear is substantially lower than for the other two species. We discuss the strengths and limitations of our approach and suggest areas for further study and development in order to increase the reliability of the OPSCR model and the cost-efficiency of large carnivore monitoring in Scandinavia