Spatiotemporal Evaluation of Injury Severity of Expressway Rear-End Crashes in China: Insights Using Random Parameters Approaches

Abstract

Rear-end crashes have become an urgent problem in China, resulting in severe casualties and massive property damage. This study aims, first, to discover the determinants affecting injury severity in expressway rear-end crashes and to examine the corresponding changes over time and space, and, second, to identify the sources of unobserved heterogeneity shifts, which could have substantial implications for efficient and effective crash prevention. Using three-year crash data (2017 -2019) from two expressways with speed limits of 120 km/h and 100 km/h (G2 and G25) in Jiangsu and Guangdong provinces in China, four random parameter logit model (RP-LM) approaches were used in this study to analyze the contributing factors: fixed parameter LM, RP-LM, RP-LM with heterogeneity in means, and RP-LM with heterogeneity in means and variances. With three injury severity consequences (severe injury, minor injury, and no injury), the characteristics of the driver, vehicle, roadway, environment, and others were proposed as possible determinants. The temporal and spatial stability were then explored using transferability tests. The marginal effects were computed to explore further potential heterogeneity and spatiotemporal variations. The estimated results indicate the superiority of the proposed model over its base counterparts, with very good ${\mathbf{\rho }}^2$ values all over 0.64. Saturday, early morning, and winter indicators were identified as significant random parameters, and several variables were observed to manifest heterogeneity in means and variances. Speeding behavior, early morning, and heavy truck indicators increased the likelihood of minor and severe injury. These findings could significantly improve expressway safety related to rear-end crashes.