Xuesong Wang,Yu Jin,Mohamed Abdel-Aty,Paul J.Tremont
Abstract:Traffic safety is beginning to receive increasing attention at the stage of transportation planning.Although road network features are an essential aspect of transportation planning, studies of the safety effects of network patterns still remain limited. In this study, macro level safety models were developed to explore the relationship between crash occurrence and several underlying variables including demographic, land use, and road network variables. Prior efforts to model network structures have been hampered by difficulties quantifying network properties. In this study, different indices (e.g., Meshedness Coefficient, Closeness Centrality) of network structures were developed to examine the network structure effects on zonal level safety. In many cases, a large percentage of the crash locations (especially for arterial crashes) were not related to the Traffic Analysis Zones (TAZs) where drivers lived. To ensure proper linkage of crashes and zonal level features, we propose to model crashes of each TAZ for non-state maintained arterials (i.e., off-system), and state-maintained arterials (i.e., on-system), separately. We developed several Conditional Autoregressive (CAR) Bayesian models that incorporated the spatial correlation of nearby zones. Estimation results showed that crashes occurring on non-state maintained roads correlate more closely to the zonal network structure, and the demographic characteristics inside the TAZ, compared to crashes occurring on state-maintained arterials that correlate more closely with the traffic and road features of the major roads. The categorical variable generated from the Meshedness Coefficient performed well in capturing the nature of network patterns as they relate to off-system road crashes. This study shows that both the roadway type and the structure of the road network should be considered when developing TAZ level safety models.