Dynamic Incident Progression Curve for Classifying Secondary Traffic Crashes

Abstract

The classification of secondary crashes is a useful performance measure of incident management systems. Previous classification methodologies used a static threshold for classifying secondary crashes. Such a threshold represents the spatial and temporal influence of a primary incident, such as 3.2 km upstream (2 mi) and 2 h after the incident. The dynamic methodology described herein improves upon existing static methodology by marking the end of the varying queue throughout the entire incident using incident progression curves. The four steps in the development of incident progression curves are: (1) processing of intranet incident reports; (2) filling in of incomplete incident reports; (3) nonlinear regression of incident progression curves; and (4) merging of individual incident progression curves into a master curve. The result from a 640 sample validation set shows that the dynamic methodology reduces Type I error by 24.38% and Type II by 3.13%. The application to a 5,514 freeway crash data set shows the results from using dynamic versus static methodology can differ by more than 30%.