Angle-accident Risk at Signalized Intersections

This paper demonstrates how a new modeling methodology can be applied to intersection angle-accident risk evaluation with data collected from 81 signalized intersections in the Tokyo Metropolitan area. Angle collisions between through traffic flow and opposite right-turn (please note that vehicles travel along the left side of the road in Japan) traffic flow are the second most common type at signalized intersections in Tokyo, accounting for 25.6% of total vehicle-to-vehicle accidents. A risk model for such angle accidents was developed with the occurrence mechanism considered in this study. Unlike most existing accident models, human factors, which caused about 95% of all the traffic accidents, can be considered quantitatively in this model. Very specific flow data, regulation data, geometric data and accident observations of each approach were applied for calibrating the model using a modified negative binomial regression. Nineteen explanatory variables were found significantly affecting angle-accident risk at signalized intersections. Such estimation results may help to improve traffic safety at signalized intersections in metropolitan areas.