Discrepancies between wolf distribution estimates from opportunistic mortality records and probabilistic field-based surveys

The gray wolf (Canis lupus) has expanded its distribution in Europe, where some countries implemented population monitoring schemes through structured surveys. However, increasing wolf populations has also resulted in a growing number of wolves being found dead each year. Although these opportunistic reports are promising for population monitoring, their accuracy in reflecting wolf distribution is still unclear. We compared the spatial distribution of the density of wolf carcasses found in Central Italy (n = 983), with robust occupancy estimates at a 100 km2 resolution, obtained from a structured population monitoring based on spatially-balanced sampling over an area of 50,200 km2. We fitted a zero inflated Bayesian GLM to estimate the density of carcasses, according to covariates explaining both wolf presence and detectability. The density of carcasses was higher in cells with low landscape naturalness and a low density of deers, higher for cells with a low abundance of the wild boar and higher for cells with high landscape anthropization. Model estimates for the density of carcasses also differed from occupancy estimates from the national monitoring program, for most of our study area. Conditional effects, overdispersion in model residuals, and discrepancy between occupancy estimates based on the national structured survey and the distribution of carcasses indicate that estimating the spatial distribution of wolves from the latter can be misleading. Bias can stem from the systematic persecution of wolves at particular hotspots, as well as from undetected or unreported collisions with vehicles. Authorities responsible for wildlife monitoring should therefore improve the standardization and centralization of records of wolf carcasses and engage experts to understand the causes of their bias before using them for population monitoring in place of surveys adopting probabilistic sampling.