Safety Evaluation of High-Occupancy Toll Facilities Using Bayesian Networks

Abstract

High-occupancy toll (HOT) lanes have increasingly been adopted as a strategy to reduce congestion. While numerous studies have focused on the operations of HOT facilities, little is known about their safety performance. This study used a Bayesian network model to evaluate the safety performance of HOT facilities by identifying factors contributing to single-vehicle (SV) and multiple-vehicle (MV) crashes at these facilities. The study utilized 3 years (2012–2014) of data from four HOT facilities in California. Concrete barrier separation, wet road surface condition, nighttime condition, and weekend are major contributing factors for SV crashes. MV crashes are associated with pylon separation, weekdays, and daytime conditions. The maximum possible probability (79%) of a SV crash is expected to occur over the weekend, during nighttime, and on a wet road surface located in a rolling/mountainous terrain having double solid white line separation. Meanwhile, the maximum probability (93%) of a MV crash is expected to occur over the weekend, during the daytime, and on a dry road surface located in rolling/mountainous terrain having pylon separation. The study results can assist transportation officials in implementing policies that will improve the safety performance of HOT facilities.