Toward Data-Driven Urban Walkability: Developing an Empirical Understanding of Walking Behavior and Urban Form

With an increasing urban population and a growing demand for sustainable mobility, there is a need to understand what drives walkability and walking behavior. Despite numerous studies exploring urban walkability, a notable challenge lies in substantiating these investigations with empirical data. In this paper, we address this research gap by leveraging open-source data on pedestrian counts, urban form features, and weather data. We use random forest regression to predict pedestrian counts using temporal, urban form, and weather as well as population and visual features, and we quantify the importance of each feature using the percent increase in mean square error when the feature is randomly permuted. We observe distinct temporal patterns in walking behavior and diverse contributions of urban form features. Ultimately, this work is a first step toward a data-driven and empirical understanding of the associations between walking behavior and the urban environment, which, in the end, will provide engineers, designers, and planners with improved tools for decision-making on advancing sustainable mobility.