Modeling bike crash risk helps identify effective safety improvements
Protecting vulnerable road users such as bicyclists is a top priority for both transportation managers and the public. However, a lack of data on bike crash risk makes implementing effective safety measures difficult.
To address this information need, U of M researchers have developed new methodologies and tools for estimating bicyclist exposure to risk. Their work also illustrates, through case studies, how these new measures can be used to assess crash risk and be incorporated into planning-level studies to improve safety.
“In our community, we cannot rely on bicycle and pedestrian crash data to prioritize infrastructure investments, as our number of incidents is very small,” says Maren Webb, Safe Routes to School/SHIP coordinator in Grand Marais/Cook County. “However, anecdotal evidence and other data suggest that improvements are needed. This research will assist communities like ours to use locally collected bicycle and pedestrian count data to estimate and assess safety and crash risk, and to inform local Safe Routes to School planning and other infrastructure investments.”
The goal of the research was to provide tools and methodologies that support a data-driven approach, says Greg Lindsey, professor in the Humphrey School of Public Affairs and the principal investigator.
Researchers began by conducting a literature review summarizing recent advances in demand modeling, estimating exposure to risk (defined as bicyclist demand, or traffic volumes), and assessing crash risk. Next, they developed new bicycle demand models using a large database of afternoon peak-period bicycle counts in Minneapolis. Using this information, they then developed additional models to demonstrate how estimates of exposure correlate with the probability of crashes in Minneapolis.
“The demand models correlate the count data with different road and built-environment characteristics, enabling estimation of the number of bicyclists through any given segment in a network,” Lindsey explains. “By combining these demand models with crash data, it becomes possible to predict and understand the risk of bicyclist-car crashes at various locations and times. It also becomes easier to plan where to put countermeasures.”
Lindsey’s team used the exposure estimates to assess the need for countermeasures at 184 roadway-trail crossings in Minneapolis. “We discovered most locations that potentially warrant traffic signals or pedestrian hybrid beacons already have them,” he says. “However, as many as 17 crossings warrant site-specific analyses to determine whether additional safety countermeasures may be needed.”
“This research will add to our collective knowledge of bicycle use and inform bicycle safety improvements in Minneapolis,” says Simon Blenski, a transportation planner with the City of Minneapolis.
Expanding their work beyond the Twin Cities, the researchers examined bike data in Duluth and Bemidji. In Duluth, they developed count-based models of bicyclist exposure and tested the correlation between measures of exposure and crashes. In Bemidji, researchers completed exploratory analysis of available count and crash data.
“While MnDOT is a multimodal agency working to connect people and places, our roads and facilities are often barriers for people walking or biking to their destination,” says Michael Petesch, bicycle and pedestrian data coordinator with MnDOT’s Office of Transit & Active Transportation. “This study, along with several ongoing MnDOT efforts to characterize risk, will be useful in developing proactive approaches for planning and programming safety countermeasures on projects throughout Minnesota.”
The models will also be useful for developing performance indicators and measuring progress toward them, Lindsey notes.
The study was sponsored by the Roadway Safety Institute (RSI), a federally funded University Transportation Center. Moving forward, Lindsey has received RSI funding to conduct a follow-up study, this time focused on equity. “Do we provide equal service to all parts of the community, and do all parts of the community experience the same levels of risk? Is crash risk higher in poor neighborhoods than in wealthier neighborhoods? These are some relevant questions we aim to address,” Lindsey says.