Analysis of Run-Off-Road Crashes in Relation to Roadway Features and Driver Behavior

The state of Wisconsin has been collecting crash information for the entire state trunk highway system that covers over 11,000 miles of roadways. This study examined crashes that occurred between 1998 and 2002, located based on a state-developed linear referencing system. A method (PRECIS), originally developed to systematically identify crashes on undivided state trunk highways (STH) and compute crash rates, crash densities (crashes/mile), and other safety statistics at any given point along a STH using a floating highway segment, was utilized. The study expanded the PRECIS application to divided highways and established relationships between driver actions that lead to a run-off-road crash and roadway information collected from the State Trunk Network log database. Information such as shoulder width, pavement condition, and roadside features indexed by mile point was examined. The crash database houses over 250,000 crashes that occurred over the analyzed five years, allowing a meaningful analysis of rural low-frequency crash types such as run-off-road crashes. The study merged crash and roadway databases and provided results on a linear highway- and mile point-indexed tabulation system. The results were also transferable to Geographical Information Systems for presentation. The present analysis describes three tasks that demonstrate applications of the PRECIS algorithm and database: (1) development of average and 95th percentile crash rates for targeted crash subsets (e.g., rural, undivided, two-lane highways with 10 ft. shoulders); (2) identification of highway sections for safety improvements where crash rates exceed a set threshold value; (3) identification of the point of diminishing safety returns for highway improvements.