Large-Scale Incidents Modeling and Network Simulation

Traffic incidents on urban freeways are a major source of congestion and travel time uncertainty. In particular, large-scale incidents have longer durations and need more incident response resources. As such, large-scale incidents deserve more attention by practitioners and researchers. The objective of this study is to analyze large-scale incidents and explore their correlates and implications for traffic operations. An innovative analysis method, based on a 2008 incident data set from the Hampton Roads Traffic Operation Center (HRTOC) in Hampton Roads, Virginia was developed (N=59176). This study defined large-scale incidents as having a two hour minimum duration, based on guidance from the Manual of Uniform Traffic Control Devices (MUTCD). The study discovers how the spatial and temporal patterns of large-scale incidents vary. To quantify associated key factors that include incident characteristics, roadway geometry, traffic flow and operational responses, rigorous statistical models were estimated. The results indicate that for large-scale incidents, their longer durations are associated with extreme events, e.g., being in a work zone incident, the presence of curvature on the segment where incident occurs, morning peak hours and occurrence of secondary incidents. The new findings provide insights regarding the understanding of large-scale incidents and have certain implications for effective incident management. Using a 2008 crash database obtained from Virginia Police Department (N=93776), traffic accidents were analyzed further in this study. Compared to other type of incidents, accidents are more likely to cause severe problems because of injuries (fatalities), longer durations and larger amount of economic loss. Accidents are divided into two categories: minor (PDO) and severe accidents (injury/fatal). Logistic models indicate that the following variables are associated with severe accidents: number of lanes, speed limit, rural area, non-divided facility, most collision type (excludes side swipe in same direction), alcohol involved, non-signalized intersections, and number of vehicles involved. To analyze network level impacts of large-scale incidents, a simulation model was developed using TransModeler software. A 2-hours incident was simulated and three scenarios were developed. The results suggest that the total delay in an accident scenario is 304% higher with no diversion and 262% higher with 10% diversion than free flow scenario.