Developing Traffic Conflict-Based Safety Performance Functions for Urban Intersections

Intersections are the traffic entities where conditions are complicated, accounting for more proportion of crashes and deserving attention. This paper developed traffic conflict-based safety performance functions (SPFs) for urban intersections using the pNEUMA dataset to investigate how various factors affect the frequency of traffic conflicts of different severity levels. The traffic conflict data originated from 77 intersections in Athens, Greece. Traffic volume, the placement of signal control, and the geometric shape of the intersection are incorporated as independent variables of the SPFs. Poisson regression was utilized to develop the models, together with a Bayesian estimation approach with the aim of inferring the parameters. The results indicate that all models show good fitting performance for each severity level and the developed models perform better in predicting conflicts with higher severity. It is also found that more traffic volume, the lack of traffic signal, and T-shaped intersection can lead to a greater number of traffic conflicts. The results of the current study can to some extent enrich the methods to develop SPFs and provide insights on the influencing factors of conflicts at signalized intersections.