No street is an Island: Street network morphologies and traffic safety

Network morphological analysis has emerged as a tool to quantify street network structures, providing a nuanced foundation for evaluating their impacts on traffic safety. Yet, there is a lack of disaggregate-level evidence on the spillover effects and spatial heterogeneity of these impacts. This research conducts a comprehensive, disaggregate-level, multi-scale examination on the overall impacts of street network morphologies on traffic safety. Our study focuses on the frequency of traffic injury collisions over a five-year period across more than 190,000 street links in Greater London. We characterise street-link morphologies at local (0-1 km), meso (0-3 km), and city (0-8 km) scales using a spatial design network analysis. For each spatial scale, we apply extended auto-negative binomial models to examine the overall impact of street-link morphological characteristics on the injury collision frequency, considering both the link being investigated and other surrounding links determined by the spatial scale. We find significant spatial heterogeneity in the overall safety impacts of street-link morphologies. At the local scale, higher farness of a street link corresponds to an overall increase in injury collisions, whereas at the meso and city scales, it indicates an overall decrease. At the local and meso scales, higher betweenness of a street link is associated with an overall increase in injury collisions, but at the city scale, it correlates with an overall decrease. Independent of the spatial scale, a larger diversion ratio of a street link is linked to an overall decrease in injury collisions. These findings are similar to those on killed and seriously injured-only collisions. Our findings suggest that encouraging compact street network structures, which aligns well with New Urbanism and the Compact City policy, may not necessarily be effective for an overall reduction in injury collisions across an entire city.