Exploring the Spatial Heterogeneity in Nonmotorized Vehicle Crashes on Urban Roads

Abstract

The number of crashes involving non-motorized vehicles (NMVs), such as bicycles and electric bicycles, on urban roads is high, so it is crucial to explore the factors influencing NMV safety. However, to date, most studies have relied on limited models that fail to capture the spatial heterogeneity in the effects of factors (e.g., bike lane and bike lane width) on NMV crash frequency. This study investigated the spatial heterogeneity in NMV crashes on urban roads using a geographically weighted random forest model. Three categories of data including roadway, bicycle facility, and traffic operation characteristics on 585 segments in Shanghai, China were collected for the development of the model. The results indicated significant spatial variations in the effects of key factors in the three categories of data on NMV crashes on urban roads. Roadway factors were important, especially segment length and the number of accesses, which were positively correlated with NMV crashes. The importance of access in some segments of the northwestern and northeastern areas was high. As for bicycle facilities, greenbelts and guardrails had substantial effects on NMV crashes, and greenbelts were negatively correlated with NMV crashes. The importance of the absence of physical separation was high in the northwestern area. The importance of average speed was also high, as it negatively affected NMV crashes and had a greater effect on NMV crash frequency in the northwestern area. These findings could help policy makers of crash-prone segments implement targeted practical engineering solutions and safety countermeasures to improve NMV safety.