Meso-Level Safety Analysis for Urban Arterials Considering Road Network Characteristics

Abstract

With the increase in the density of urban road networks, the problem of inappropriate road networks has become increasingly prominent. In particular, improper road connectivity has raised significant safety concerns on urban arterials. Traditional micro-level crash prediction models have explored the relationship among crash frequency, geometric design, and traffic operation characteristics by modeling intersections and road segments separately, and macro-level models analyze traffic safety in larger areas, considering socioeconomic and road network factors. However, neither micro-level models nor macro-level models are suitable for safety analysis of urban arterials. To investigate the safety effect of both road connections and network patterns, this study adopted a meso-level approach that combines the signalized intersections with adjacent segments into an analysis unit. The study introduced the function difference number (FDN) to quantitatively express the degree of functional matching in road connections and analyze the effects of both road connection and road network pattern on urban arterial safety. Using Guangzhou, China, as a case study and considering the spatial correlation among meso-level analysis units within individual arterials, a Bayesian negative binomial conditional autoregressive (NB-CAR) model was developed for urban arterials. The results showed that, compared with the Poisson lognormal and negative binomial models, the Bayesian NB-CAR model demonstrated superior performance. FDN was identified as a significant variable, proving its applicability for safety assessments of road network functional matching. Urban arterials with higher FDNs experienced more crashes than those with lower FDNs. Additionally, irregular grids and mixed network patterns were more strongly associated with crashes than the grid pattern.