Understanding urban traffi c-fl ow characteristics: a rethinking of betweenness centrality

In this study we estimate urban traffi c fl ow using GPS-enabled taxi trajectory data in Qingdao, China, and examine the capability of the betweenness centrality of the street network to predict traffi c fl ow. The results show that betweenness centrality is not a good predictor variable for urban traffi c fl ow, which has, theoretically, been pointed out in existing literature. With a critique of the betweenness centrality as a predictor, we further analyze the characteristics of betweenness centrality and point out the ‘gap’ between this centrality measure and actual fl ow. Rather than considering only the topological properties of a street network, we take into account two aspects, the spatial heterogeneity of human activities and the distance-decay law, to explain the observed traffi c-fl ow distribution. The spatial distribution of human activities is estimated using mobile phone Erlang values, and the power law distance decay is adopted. We run Monte Carlo simulations to generate trips and predict traffi c-fl ow distributions, and use a weighted correlation coeffi cient to measure the goodness of fi t between the observed and the simulated data. The correlation coeffi cient achieves the maximum (0.623) when the exponent equals 2.0, indicating that the proposed model, which incorporates geographical constraints and human mobility patterns, can interpret urban traffi c fl ow well.