There are approximately 290 million registered vehicles in the US, traveling over 3 trillion miles per year on over 4 million miles of rural and urban roads. Vehicles traveling through wildlife habitat can strike wildlife, resulting in a wildlife-vehicle collision (WVC). WVC acts as both an artificial predator, causing direct mortality, and a barrier to population connectivity, eventually limiting gene flow (because dead animals don’t contribute to connectivity). Although the idea of WVC is well-characterized and studied, the total population impact of WVC on any species, or wildlife in general, is poorly understood. We describe here a method for estimating total roadkill impact per species at the US state scale, using mule deer (Odocoileus hemionus). We used 43,021 observations of deer-vehicle collisions (DVC) on CA roads between 2016 and 2023. We used an R package called GenEst (USGS) that was originally developed for estimating mortality of birds and bats at wind turbines but is readily generalizable to vehicular mortality (Dalthrop et al., 2018). GenEst fits models for variables such as carcass persistence, search efficiency, and effort and was used recently to model giant anteater roadkill totals (Ascensão et al. 2021). We estimated rates of carcass observation, persistence, and effort from our own data collections systems and used search efficiency values from the literature. The model was constrained by a habitat suitability model (CA Department of Fish and Wildlife, 2016). Observations were divided by: 1) traffic volume (average annual daily travel); 2) bioregion (5 coastal or inland regions); and 3) year and season, resulting in 34 separate categories, and rates of DVC were estimated for each category. The rates were then extrapolated within each category to all other roadways with similar characteristics. Rates of DVC/km varied among regions, between 0-1.4 DVC/km (South Coast) and 0.4-5 DVC/km (Sierra Nevada/Modoc). There has been speculation that there is a peak in WVC at traffic volumes near 10,000 vehicles/day (Seiler and Helldin, 2006). Visually, we observed a peak in DVC between 5,000 and 15,000 vehicles/day. We also observed a second peak between 50,000 and 200,000 vehicles/day. This second peak may represent very high-volume, un-fenced highways adjacent to wildlife habitat. We estimated that 33,943 (2022) to 53,036 (2016) mule deer were killed per year, with an average of 43,027 DVC/year (2016-2023). This represents about 10% of the estimated statewide population of mule deer and thus may represent an unsustainable rate of DVC, which may be an important factor in CA’s steadily declining mule deer population. Our approach could be used for any species with sufficient data collection, which we are carrying out for other common CA species.
Estimating Total Roadkill Rate for a US State
ICOET 2025
Abstract