Incorporating inconsistency patterns on road networks into crash modeling

Abstract

This paper expands the scope of geometric design inconsistency analysis from corridor scales to network-wide perspectives, exploring the impact of inconsistencies’ spatial-patterns on crashes, which remains largely under-explored. We define spatial-patterns of segment-level inconsistencies, focusing on their spread, contiguity, frequency, density, and magnitude. We devise a new method to measure inconsistency-contiguity and inconsistency-frequency based on adjacent segment-triplets within regions. Through micro–macro integrated models, we reveal the scalable influence of inconsistency which remain significant at the segment-level but gets modulated by spatial-patterns at the regional-level. The integrated models consistently outperform their non-integrated counterparts, emphasizing the importance of this integrated approach. This study highlights that regions with rare inconsistency occurrences demonstrate higher crash counts, while regions with uniform inconsistency occurrences exhibit lower crash rates, unveiling insights into the road conditions’ impact on driver behavior. Finally, we also propose a novel tool - vulnerability contours on frequency-hyperplane to map regions’ relative safety.