Discretization of Road Networks for Safety Evaluation with Consideration of Intersection Impact Zones

Abstract

The goal of safety management is to reduce the number and severity of crashes in transportation networks. One of the methods of safety management is identification of roads that experience an excessive number of crashes by comparison of the actual number of crashes with exposure-based predictions. The identification task typically includes division of the road network into intersections and segments in such a way that independent prediction of safety at these components can be defended. However, evidence provided by recent research shows that the assumption of independence can be violated and lead to biased safety predictions. This study proposed a refined road network discretization and an accounting for the safety impact of intersections on nearby short segments. The latter was accomplished by inclusion of variables for the distance to the intersection in the safety performance functions for segments. Three network discretization methods, including the one proposed, are presented and discussed on the basis of their application to the Indiana state road network. Considerable differences between the average safety impact zones of signalized and unsignalized intersections were detected. Rural signalized intersections tended to have very short impact zones, whereas rural unsignalized intersections were influential over greater distances. In urban areas, unsignalized intersections had a short impact zone and signalized intersections were influential over greater distances. Potential sources of these differences are discussed, and future work is detailed.