Transportation infrastructure impacts wildlife by restricting movement and increasing mortality through wildlife-vehicle collisions (WVCs). WVCs cost Californians $250 million each year in property damage, personal injuries, and wildlife and driver fatalities. Roadways typically feature medians that separate opposing lanes of traffic and include: concrete barrier, metal guardrail, cable, vegetated strip, gravel strip, and pavement marking. Transportation planners select a median type based on safety concerns (i.e., crashes) and other conditions. Different median types may influence animal behavior. Physical barriers may inhibit crossing, while permeable medians may facilitate it. Vegetation in medians may attract large and small grazers (e.g., deer, mice), potentially increasing WVC risk. At Caltrans’ request, we assessed the potential impacts of median treatment types on species-specific and total WVC rates in two ways: 1) using Google Street View to determine the median type for WVC incidents of a small and large mammal species, and 2) using Google Earth and Street View to identify adjacent median treatment “pairs” (e.g., concrete barrier median adjacent to cable guard median) and compare WVC rates within each pair.

Geocoded data from the California Roadkill Observation System were used to identify WVC sites along highways for mule deer (Odocoileus hemionus) in Caltrans District 9 (n=1069) and western gray squirrel (Sciurus griseus) in Caltrans District 2 (n=332) between January 2015 and April 2024. Comparison spatial datasets of randomly-generated points along highways were also created for each district. Each site was categorized by median type using Google Street View imagery. Chi-square tests found that median types at both mule deer (p<2.1e-15) and western gray squirrel (p<5.7e-14) WVC sites were statistically different from the distribution of median types at random sites. These results suggest that median type affects the spatial density of WVCs for both species.

A supplemental analysis was conducted to account for potential influences of land cover and use. Google Earth imagery was used to identify stretches of highways where median types transition from one to another, resulting in a dataset of paired highway transects (n=73). Rates of WVCs (n=939) between January 2015 and April 2024 were estimated for 1-mile transects. Wilcoxon signed-rank tests were performed for each transect pair type to evaluate differences in WVC rates between median types. The mean WVC rate was found to be 1.95 times higher for metal guardrails than the adjacent vegetated strips (p=0.045). No other differences were significant among other median-type pairs. Future analysis should consider additional environmental variables that may also influence WVCs.

These findings suggest that roadway features like median treatments have an effect on WVCs, which reinforces the importance of assessing wildlife impacts when planning transportation infrastructure.