How do graph properties of road networks influence vehicle miles traveled? A study of 486 urban areas in the United States

This study delivers a comprehensive investigation of the road networks of 486 urbanized areas in the United States. This study considers a region-to-region macroscopic perspective to discuss the role of road networks in passenger vehicle miles traveled (VMT). The main goal of this study is to quantify the relationship between VMT and region-level road network characteristics as measured by 25 graph-theoretical metrics. Principal Component Analysis (PCA) was performed to reduce the dimensionality of graph metrics and to combine highly correlated measures (e.g., node density and edge density) into uncorrelated principal components. PCA generated five graph-related principal components to describe five aspects of road networks: network size, density, connectivity, complexity, and vulnerability. Random-parameter models were built to relate road network graph properties to VMT per capita and VMT per household. All models showed a significant relationship between passenger VMT and road network variables, including original graph metrics and graph-based principal components. Results implied residents in large urban areas travel more than people in small urban areas; a denser road network could reduce VMT; and more dead ends or 3-way intersections can lead to increased VMT in a region. The findings quantify the impact of roadway characteristics on the VMT variation of US urban areas. This study advances the state-of-the-art by conducting a comprehensive regional-level VMT analysis across all urban areas in the U.S. and identifying key roadway network features associated with high or low VMT per capita and per household. The findings provide valuable policy insights for roadway development, including new regional expansions and improvements to existing road networks (e.g., transportation improvement plans), to support urban planning that fosters a low-carbon transportation system.