1372021 - Extreme Value Theory Estimation of Vehicle-Pedestrian Safety Surrogates at Intersection with Right-Turn Flashing Yellow Arrow Indication

Traffic conflicts and surrogate safety measures (SSM) have been used as an alternative to crash-based methods to study roadway safety. Extreme value theory (EVT) offers a modeling framework that can be used to expand the use SSM to conduct proactive safety evaluations. This study explores the EVT modeling approach to analyze vehicle-pedestrian interactions and compare safety risks between a site with a right-turn flashing yellow arrow (RT FYA) indication and sites with a permissive circular green indication. Using trajectory data extracted from video using a frame-by-frame analysis approach, post-encroachment-time (PET) values were determined along with an obstructed right turn time (ORTT) measure which is defined as the time it takes a vehicle to complete a right turn maneuver when a conflicting pedestrian is present.

At-site univariate and bivariate extreme value theory models were developed using the block maxima (BM) approach and the peak over threshold (POT) approach. Additionally, joint-site univariate and bivariate Bayesian hierarchical models were developed for each approach. Using the resulting estimates, the number of crashes was estimated for each model and compared to the observed crashes. Results showed that models using ORTT produced a better fit model with their covariates indicating that ORTT helps describe traffic interactions objectively. The number of crashes estimated from the Bayesian hierarchical models was found to also be closer to the observed number of crashes than those from other models. Particularly, Bivariate Bayesian hierarchical models outperformed the at-site models (univariate and bivariate) and the univariate joint-site model in terms of crash estimation.