A Review on Surrogate Safety Measures Using Extreme Value Theory

Over the past few decades, the observed accident data has served as the primary source for road safety analyses. To avoid the shortcomings of crash data, traffic conflict techniques have been promoted as an alternative approach to analyze safety from a wider perspective than relying solely on crash data. Still, the application of surrogate safety measures, and the validity of conflict techniques is a great concern. The Extreme Value Theory (EVT) offers Peak Over Threshold (POT) and Block Maxima (BM) approaches to provide a robust modeling framework for relating surrogate safety measures (SSMs) to crash frequency. This study is a comprehensive review of the development and use of EVT in combination with surrogate measures for safety evaluation. A number of formulated research questions on the use of EVT for SSM analysis are also identified and discussed in a precise manner. The key finding of the study demonstrates that bivariate extreme value modeling approaches with a combination of surrogate measures, time to collision (TTC), and Post Encroachment Time (PET), are more useful to estimate safety than uni-variate extreme value modeling approaches with individual safety measures. Finally, the authors identified several research gaps to assist researchers and practitioners with recommendations for potentially useful avenues of future research.