Urban morphologies determining road-travel route choices – from the rush hours to the wee hours

Urban data science offers us a variety of methodologies, perspectives, and insights from real-world data analysis that have greatly advanced our understanding of cities. One established strand of research uses network data, such as road maps and traffic networks, to quantify the geometric structure of cities. For example, an indicator that quantifies the griddedness of a road network by measuring the entropy of the road azimuth angle [1] has been proposed. However, while road networks often consist of a combination of roads radiating out from the center of a city and multiple ring roads around the city, there has been no indicator to measure the degree of such a structure. Some studies in urban data science pursue not only transport network structures but also their functions. For example, ref. [2] studied how travel routes between a given pair of locations change during congestion and normal times, and how that change varies depending on the road network structure. In this study, we carried out the following two steps: (I) We proposed a simple measure to capture the "circularity and radiality" of the transportation network, and used both the indicators and the griddedness measure [1] to capture the characteristics of the transportation network. (II) We analyzed how travel routes between two points change over the course of a day depending on the structure of the transportation network, and how the pattern of change varies with distance from the center of the network for ten cities worldwide. First, for step (I), the traffic network data for each city was obtained from Open Street Map, and the following values were calculated for each road that constitutes the traffic network (origin i and destination j, with the origin being the point farthest from the center of the city): C_ij= |L_i-L_j |/d_ij , (1) Where L_i (or L_j) are the Euclidean distances between i (or j) and the city center coordinates, and d_ij is the Euclidean distance between i and j. Note that 0