DEVELOPMENT OF A ROAD MORTALITY HOTSPOT MODEL FOR AMPHIBIANS AND REPTILES IN RHODE ISLAND

Abstract

Road mortality poses a serious threat to amphibians and reptiles and can lead to population declines and localized extirpation. Given road mortality tends to aggregate spatially at locations commonly referred to as road mortality hotspots, identifying the most appropriate locations for mitigation measures, such as road tunnels and culverts, is an important first step in reducing road mortality. However, the influence of imperfect detection (e.g., false absences) during road mortality surveys can lead to poor spatial patterns of road mortality hotspots and suboptimal implementation of mitigation measures intended to reduce road mortality. In this study, occupancy modeling was used to identify where large amphibian and reptile roadkill events (≥ 5 carcasses) occurred along transects in Rhode Island, USA. I defined occupancy as the likelihood of a large roadkill event occurring at a transect. I defined detection probability as the probability of a large roadkill event being detected during a survey assuming it occurs at a transect. From 2019 to 2021, road mortality surveys were conducted along 48, 200-m long transects on 50 occasions for a total 240 surveys. A total of 606 carcasses were observed during surveys representing 19 of Rhode Island’s native amphibian and reptile species. I developed and tested models to examine factors associated with large roadkill events. The most supported model estimated a higher occurrence of large roadkill events on roads with a high proportion of wetlands (forest and non-forested wetlands) within 100-m and water bodies (streams and lakes) within 500-m and on roads with low traffic volume (15‒400 vehicles/day). For detection probability, large roadkill events were more likely to be detected as daily precipitation increased and less likely to be detected on roads with low traffic volume (15‒400 vehicles/day). Large roadkill events were observed at 40% of transects during surveys. However, model results indicated that large roadkill events likely occurred at 64% of transects and there was a 17.5% chance of detecting a large roadkill event at transects where at least one occurred. The low detection probability suggest that imperfect detection influences the ability to detect large roadkill events, and many are likely to be missed with regularity using our methodology. To better identify road morality hotspots and target locations for mitigation measures, I recommend surveys be conducted on low traffic roads near wetlands. In addition, I recommend surveying during or after a rain event, as the probability of detecting a large roadkill event increases with