Crash patterns at signalized intersections

Abstract

Traffic signals are often implemented to provide for efficient movement and to improve traffic safety. Nevertheless, severe crashes still occur at signalized intersections. This study aims to improve the understanding of signalized intersection safety by identifying crash types, locations and factors associated with signalized intersections. For this purpose, 1295 police-reported crashes at 87 signalized intersections are analyzed based on detailed crash descriptions, i.e. crash data and collision diagrams. The information of the collision diagrams is used to distinguish six different crash types and to create a crash location typology to divide the signalized intersection into 13 detailed and different typical segments. Logistic regression modeling techniques are used to identify relations between crash types, their crash location on certain signalized intersection segments, the crash severity and the different features that affect their crash occurrence. Four dominant crash types are identified: rear-end, side (i.e. left-turn + right-angle), head-on and VRU crashes. The results of the logistic regression models showed that the crash location of these crash types is related to specific signalized intersection segments. The results also reveal important signalized intersection features that affect the crash occurrence. As a result, connections between certain signalized intersection crash types, their crash location and signalized intersection design characteristics have been found. The combination of intersection features with detailed signalized intersection segments provides valuable insights in the nature of signalized intersection crashes and the safety impact of signalized intersection design.