Development of Safety Performance Functions for Undivided Rural Two-Lane Roadways

Abstract

The objective of the research was to develop a methodology for creating safety performance functions specific to the state of Wisconsin, using undivided rural two-lane roadways. The methodology was designed to complement the practices of Wisconsin Department of Transportation (WisDOT) and work with the database WisDOT uses. The considered variables included curve presence, lane width, right shoulder total width, right shoulder paved width, truck percent, international roughness index, and posted speed. Negative Binomial regression models were developed using a dataset of 7,911 segments. Cumulative residual plots identified significant points in the AADT range where the SPFs could be split into two. Results show using two models, split by AADT at 7,500 vehicles per day (vpd) for all crash types and 7,000 vpd for fatal and injury crashes, were more accurate models than using a single grouping of AADT. Weather-related crashes were found to reduce the dispersion parameter and Akaike information criterion when comparing models including weather-related crashes and models excluding weather-related crashes. SPFs were developed for weather-related crashes: one with AADT ≤ 6,000 vpd and one with AADT > 6,000 vpd. Six SPFs were developed for undivided rural two-lane roadways, including all crash types and fatal and injury crashes.