Results in Engineering (Dec 2024)
Spatial analysis of daytime and nighttime crash severity on horizontal curves of mountainous rural highways: A case study in Northern Iran
Abstract
According to statistical analysis, many road crashes, especially on rural two-lane undivided (RTU) highways passing through mountainous areas, occur on horizontal curves. Spatial analysis of these crashes, while incorporating the influence of lighting conditions, can significantly improve safety on rural highways. This study proposes a method based on spatial clustering combined with a hybrid approach of random forest and binary logistic regression modeling framework to examine the severity of crashes occurring on RTU horizontal curves in mountainous areas. In contrast to prior studies that primarily focused on either statistical modeling or spatial analysis in isolation, this research integrates these methodologies to provide a more comprehensive understanding of the factors influencing crash severity. To assess the influence of various factors on crash severity, this study applied the proposed approach to all curves and crash-prone curves identified through spatial autocorrelation analysis, considering four separate datasets, four models defined by temporal conditions (daytime/nighttime), and spatial aspects (all curves/crash-prone curves) (RB1: daytime crashes on all curves, RB2: nighttime crashes on all curves, RB3: daytime crashes on crash-prone curves, and RB4: nighttime crashes on crash-prone curves). This innovative approach distinguishes this study from previous research by enabling a detailed investigation into how various crash factors manifest differently under distinct spatio-temporal conditions, thus facilitating a nuanced analysis of their interplay. Based on the analysis, collision type, collision angle, and collision reason emerged as significant factors influencing crash severity across all models, regardless of lighting conditions and spatial aspects. Additionally, head-on and rear-end collisions were found to increase the risk of severe injury crashes. Furthermore, the modeling results identified geometric features of curves, such as type, radius, and length, as critical factors in identifying crash-prone curves. These findings advance existing research by highlighting the significant role of spatial factors and lighting conditions in determining crash severity, while the comprehensive evaluation of both all curves and crash-prone curves through spatial analysis enhances the understanding of the interaction between spatial attributes and crash dynamics. This study provides valuable insights for the engineering community, offering practical applications in safety interventions, including targeted signage, curve design modifications, and driver education programs, which can help mitigate crash severity on RTU horizontal curves in mountainous areas.
