A Statistical Analysis of the Effect of Wet-Pavement Friction on Highway Traffic Safety

This study focuses on analysis of statistical association between wet-pavement friction and roadway crashes aimed at answering these questions: (1) Is wet-pavement friction a significant factor for explaining variation in crash history? (2) Is this factor more relevant at locations with high expected braking frequency, such as sharp curves and intersections? The data set includes “found data” locations with previously measured wet-pavement friction, and “random data” locations at which the friction was measured specifically for this project. Including the random data locations was necessary to overcome bias in the found data because they were selected due to suspected safety issues. Road characteristics were collected and incorporated into the data set. Negative binominal regression was used to estimate models with coefficients for the main factors and interactions fitting the data. The locations where improving the wet-pavement friction will most reduce crashes include sections with nonisolated curves on undivided roads and sections with driveways or mild curves on divided roads. A counterintuitive finding of higher crash frequency is associated with increases in friction in urban areas, especially on divided roads.