Identifying Risk Factors for Deer-Vehicle Collisions

Deer-vehicle collisions are a significant safety hazard on Minnesota roads. While MnDOT strives to employ safety measures on roads at high risk for these collisions, identifying these areas can be challenging. Numerous variables impact where deer are more likely to be present near roads, and many collisions go unreported. A new tool estimates the risk levels—based on road type, geographical features, deer population and other characteristics—for each road segment in the state. In addition, a new method of estimating reporting rates will help MnDOT understand the extent of deer-vehicle collisions in a specific area.

Collisions between deer and vehicles not only harm or kill the animals, but may cause property damage, human injuries and fatalities. On average, over 1,200 crashes per year were reported to the Minnesota Department of Public Safety (MnDPS) between 2016 and 2020 in the state. But auto insurance claims suggest that actual crash numbers are much higher. Costs related to human injuries, fatalities and property damage resulting from collisions involving deer could be $220 million annually in Minnesota, with additional costs for social and environmental factors.

Deer-vehicle collisions are ubiquitous across the state. MnDOT uses fencing, culverts or bridges as safety measures to keep deer off the roads and provide a corridor for them to pass over or under a highway within their habitat. Where safety measures or warnings would be most effective, however, depends on multiple factors such as deer and traffic volume, vehicle speeds, road type and surrounding land use.

Given that deer are present throughout the state, MnDOT wanted to identify the areas that are at highest risk for deer-vehicle collisions and the characteristics of these areas. To understand the true extent of deer-vehicle collisions, the agency needed to estimate the rates of crash reporting.

What Did We Do?

Numerous factors informed the project analyses, including previous studies on deer-vehicle collision reporting rates; common factors related to crashes; methods for analyzing crash concentrations; and data from over 36,000 collisions in Minnesota since 2005.

To estimate the percentage of collisions that are reported in Minnesota, researchers observed the aftermath of deer-vehicle collisions within a defined area for more than 18 months, including two fall seasons.

“This project was a big first step in helping MnDOT prioritize highest risk areas for deer-vehicle collisions. The mapping tool will help us—on an ongoing basis—to identify areas in greatest need of safety measures,” said Christopher Smith, Protected Species Program coordinator, MnDOT Office of Environmental Stewardship.

Driving 1,000 miles per month along roads with varying characteristics in the Duluth area, they collected data of observed likely crashes, including pictures of deer carcasses and roadside features, and compared the number of crashes with annually reported incidents tracked by MnDPS from 2006 to 2020.

A layered mapping exercise involved fusing several data sets with the geographic distribution of deer-vehicle collisions. Average annual daily traffic and posted speed limits combined with land cover and roadway factors such as width began to point to the factors that correlated with higher incidences of collisions. Portions of the Twin Cities metropolitan area and select other roadways were excluded to reduce data inconsistencies and address data gaps.

Researchers then produced a model to estimate collision risk using statistical analyses and machine-learning algorithms.

What Did We Learn?

Previous studies showed that fewer than one in three deer-vehicle collisions is generally reported. Comparing collision site observations with MnDPS records indicated that between 0% and 38% of crashes were reported, depending on the roadway, with an average of 10%.

While the results suggested substantial variability, roads with the highest reporting levels tended to be highways with high speeds or municipal roads with speeds under 30 mph. The methodology used to estimate reporting levels is repeatable and can be applied across the state.

Modeling suggested that traffic volume was a significant variable in predicting the probability of deer-vehicle collisions on a road segment. But when the model was controlled for traffic volume to indicate a per-driver risk, higher deer densities and wider roads positively correlated to higher crash risk.

Using the data-driven machine-learning model, researchers created an interactive map to predict the relative risk of deer-vehicle collisions on most segments in Minnesota based on road and habitat features such as proximity to streams.

What’s Next?

MnDOT will consider how to operationalize project results to prioritize where deer-vehicle collision prevention efforts may be needed.

MnDOT environmental staff is already using the tool on a project-by-project basis to identify areas at high risk for deer-vehicle collisions for consideration in road design and other safety measures. Researchers provided guidance on how to periodically update the mapping tool.