Safety performance functions for crash severity on undivided rural roads

The objective of this paper is to explore the effect of the road features of two-lane rural road networks on crash severity. One of the main goals is to calibrate Safety Performance Functions (SPFs) that can predict the frequency per year of injuries and fatalities on homogeneous road segments. It was found that on more than 2000 km of study-road network that annual average daily traffic, lane width, curvature change rate, length, and vertical grade are important variables in explaining the severity of crashes. A crash database covering a 5-year period was examined to achieve the goals (1295 injurious crashes that included 2089 injuries and 235 fatalities). A total of 1000 km were used to calibrate SPFs and the remaining 1000 km reflecting the traffic, geometric, functional features of the preceding one were used to validate their effectiveness. A negative binomial regression model was used. Reflecting the crash configurations of the dataset and maximizing the validation outcomes, four main sets of SPFs were developed as follows: (a) one equation to predict only injury frequency per year for the subset where only non-fatal injuries occurred, (b) two different equations to predict injury frequency and fatality frequency per year per sub-set where at least one fa tality occurred together with one injury, and (c) only one equation to predict the total frequency per year of total casualties correlating accurate percentages to obtain the final expected frequency of injuries and fatalities per year on homogeneous road segments. Residual analysis confirms the effectiveness of the SPFs.