Using Conflict Prediction Models to Evaluate the Impacts of Conflicting Volumes on Opposing Left Turn Conflicts at Signalized Intersections

Abstract

In this study, conflict prediction models were developed to evaluate the impact of conflicting volumes on opposing left-turn conflicts at signalized intersections. Using data collected from five signalized intersections in Kunming areas in China, we divided conflicting volumes into four scenarios based on the v/c ratio of both through traffic and left-turn traffic. Conflict prediction models were developed for different scenarios. The model specification followed the normal procedure developed for crash prediction models where Poisson and NB models were used to relate the number of traffic conflicts to conflicting volumes. The model specification results showed that the impact of conflicting volumes on traffic conflicts varied under different traffic conditions. As compared to traditional safety performance functions, which usually relate crash counts to AADT, the conflict prediction models developed in this study more accurately reflect the impacts of traffic volumes on drivers’ behavior. Thus, conflict prediction models can be developed as a supplement to traditional crash prediction models to help us better understand the impacts of various contributing factors of safety performance on traffic facilities. In addition, traffic conflicts estimated by conflict prediction models can be used as surrogate safety measures for safety assessment when crash data are not available.