Mapping Human Mobility Variation and Identifying Critical Services During a Disaster Using Dynamic Mobility Network

Accurately analyzing human movement patterns is of importance for building community resilience during a disaster. Conventionally, surveys and call detail records are commonly used in examining human movement behaviors. However, such data sources often fall short in capturing the dynamics of human mobility at higher spatial and temporal resolutions. Recently, emerging human mobility data of high-granularity unravels unique opportunities to understand the underlying human activities. In this paper, we propose a framework to characterize spatiotemporal movements of people that would help to examine the accessibility of critical facilities in face of a natural disaster. Specifically, we construct a time-varying mobility network that describes the footprint of people from their home census block groups (CBGs) to essential points-of-interest (POIs), leveraging the Maximum Entropy method and Lagrangian Relaxation technique. We then demonstrate the applicability of our framework by forming a dynamic CBG-POI network in Harris County (TX) to analyze how people prepared for and responded to 2021 Winter Storm Uri. Our results show that the movements to POIs are being disproportionally affected across CBGs. Additionally, as expected grocery stores are found to be critical POIs before the disaster enabling people to stockpile foods, while gas stations and religious services are found to be essential during the disaster. The mobility flows to each POI category could underscore the importance level of essential services at different periods of the disaster, which can be used for aiding informed decision-making for resource allocation and humanitarian assistance.