Comparative Analysis and Mapping of Human Mobility during Various Natural Disasters based on Geotagged Photos

Abstract:

The present study aims at comparatively analyzing and mapping the impacts of several past natural disasters that occurred in the Tokyo metropolitan area (TMA) on human mobility patterns based on Flickr geotagged data. First, we picked 8 extreme events of different types that occurred between 2008 and 2019 based on the number of Flickr users and taken photos. Second, based on weather information relevant to each type of disaster, we delineated steady and unsteady phases (before, during, after). Third, we analyzed mobility patterns through multiple indicators namely displacement, the radius of gyration, and the mean square displacement. Additionally, we developed a transfer learning-based convolutional neural network (CNN) model to classify images into indoor and outdoor to reconstruct movement trajectories by type of environments and duration of stay. Preliminary results show differences between movement patterns during the Tohoku earthquake and wet weather-related disasters although the probability distribution of displacements and radius of gyration mostly follows a truncated power-law. This study is expected to stimulate future research on human mobility patterns of tourists during natural disasters given the popularity of Tokyo as a touristic hotspots and yet as a metropolitan area of recurring disasters with scarce studies on tourists' mobility behavior.