Incident-Induced Diversion Behavior: Existence, Magnitude, and Contributing Factors

Abstract

The need to understand the effects of diverting traffic is emphasized by growing congestion and delays. This paper examines incident-induced diversion behavior by using loop-detector data and incident records on a freeway in Virginia. This work diverges from previous studies by (1) addressing both existence of diversion and its magnitude, (2) relying on field data rather than surveys, and (3) statistically relating diversion behavior and magnitude to quantifiable incident characteristics and traffic conditions. A dynamic programming-based procedure is used to identify diversions by isolating transient level shifts, and the diversions are associated with incident and traffic characteristics and variable message sign (VMS) displays through a binary logit model. The magnitude of the diversion is statistically related to traffic conditions via a linear regression model. The models indicate that the probability of triggering a diversion increases when an incident lasts longer, more general-purpose lanes are blocked, and speeds are lower. The results on the effects of trip purpose/time and information availability are consistent with previous studies. The magnitude of the diversion, measured by diversion rate, is related to instant traffic flow characteristics, general traffic demand considerations, and the incident characteristics.