Challenges in estimating wolf population size using spatial count models and camera trap data

Accurate abundance estimates are essential for wildlife conservation but remain difficult to obtain for wide-ranging and elusive species. Camera-traps (CT) are one of the most popular tools to study wildlife. However, the usefulness of CT data to estimate population sizes for species without individual marks remains debated. Several statistical methods showed promise in simulation studies, but researchers urged caution, and large-scale applicability remains unclear. This work aims to evaluate the use of Bayesian Spatial Count (SC) models to estimate the wolf ( Canis lupus ) population size in the Italian Alps using CT data collected through an opportunistic sampling design. A total of 755 CT were deployed across the entire range of occurrence during 2020/2021. Thanks to the availability of independent abundance estimates during the same period, we were able to investigate the effect of including prior information on the quality of large-scale population size estimates. SC models failed to provide accurate population size estimates, largely overestimating abundance, especially when using non-informative priors. While including informative priors on the effect of density covariates slightly improved estimates, providing informative priors for the spatial scale parameter led to no improvement, contrary to previous suggestions. Our findings highlighted the importance of using an adequate sampling design and accounting for individual heterogeneity in detectability within the population. Overall, SC models based on CT data seem not yet suitable for large-scale monitoring of unmarked, socially structured species like wolves, but improved sampling designs and methodological advances may increase their applicability for cost-effective population assessment.