Home About ATSIP Program Awards Venue Sponsors/Exhibitors

Wisconsin Roadway Departure Crashes Network Screening

Abstract: Wisconsin Strategic Highway Safety Plans have consistently identified roadway departure crashes and cross median crashes as one of the highest priority issue areas. The objective of this research was to identify sites with promise for treatment. Data sources consisted of roadway segments Meta-Manager data that was created and maintained by the Wisconsin Department of Transportation. This dataset focuses on the Wisconsin State Trunk Highway Network (STHN). The Meta-Manager dataset was first dissolved by highway names in ArcGIS. Each highway was divided into approximately 1-mile long segments along the entire length of the designated segment. The period of analysis was between 2009 and 2016. The database included the following variables: segment length, AADT, truck percentage, number of lanes, barriers, divided/undivided, horizontal curves, shoulders, median type and width, pavement type, rumble strips, and interchanges. Network screening Safety Performance Functions (SPF) were developed for Wisconsin specific roadway departure crashes. The SPFs were developed by roadway segment functional classification and crash type. Model development consisted of regression modeling using the Negative Binomial. Measures of model goodness of fit were the log-likelihood, overdispersion, and Cumulative Residuals (CURE) plots. SPFs provide predicted number of road departure crashes per year. The model SPFs were used to perform network screening using the Excess Expected Average Crash Frequency with Empirical Bayes Adjustments. The excess is the difference between the adjusted expected crashes with the long term average predicted crashes from the model to account for regression to the mean. Sites with positive excess values were considered for ranking which consists of sorting the facilities from the largest to smallest excess. Lists of flagged segments with ranking and visualization maps were generated. Segments flagged and ranking provide information to identify locations for further safety analysis and determine specific contributing factors and appropriate countermeasures. From the explored predictor variables, the AADT, segment length, percentage of trucks, pavement type, barriers, horizontal curves, and rumble strips were found to have significant contributions to the prediction of roadway departure crashes.

Boris Claros - Postdoctoral Research Associate - University of Wisconsin - Madison



Session Material
Quick Links


Chris Osbourn


fb_logo   linkedin