Safety Impacts Study of Roadway Network Features on Suburban Highways

In order to investigate the safety impacts of roadway network features on suburban highways, macro-level safety analysis method was applied and Poisson log-normal conditional autoregressive (PLN-CAR) models were developed for suburban highways at the traffic analysis zone (TAZ) level. To investigate the spatial correlation between adjacent TAZs, PLN-CAR models were established on three types of weight matrixes of 0-1 first order, common boundary length and total lane number of connecting suburban highways. Both highway level variables (signalized intersection density, access density, etc.) and macro variables at TAZ level (road network pattern, land use type, etc.) were included in these models. The results show that the model with the weight matrix of common boundary length performs the best. The overall safety impacts of access density and signalized intersection density on both road segments and intersections of suburban highways can be taken into consideration more integrally by applying macro-level safety analysis method, compared with micro-level modeling. Specifically, length of suburban highway, access density and signalized intersection density are significantly positive correlated with highway crashes. When other variables held constant, crashes on highways in the TAZ with tree-like pattern road network are 88.3% more than those in the grid pattern road network, which indicates that the highway safety can be improved efficiently by enhancing highway access management, regulating the spacing of signalized intersections and constructing collector roads parallel to highways. In addition, more crashes tend to occur in commercial and residential land than in industrial land.