🤖 AI Summary
This study addresses the critical phenomenon of asynchronous evacuation and return among household members during natural disasters, using the 2019–2020 Australian bushfires as a case. Methodologically, it pioneers a large-scale, spatiotemporally tagged survey paradigm leveraging social media, integrating qualitative coding, quantitative regression, and cross-validation with multi-source data (survey responses + mobile signaling traces). The analysis systematically identifies four key drivers: risk perception, caregiving responsibilities, information sources, and transport accessibility. This approach overcomes the limitations of conventional mobile signaling data in elucidating behavioral motivations. Empirically, the study is the first to document “intra-household evacuation timing divergence”—a novel behavioral pattern—and delivers high-ecological-validity estimates of evacuation behavior distributions. These findings advance emergency response modeling and policy design by enabling more precise, human-centered interventions.
📝 Abstract
Evacuation in response to natural disasters is a complex process involving multiple decision-makers at the personal, household, community, and government levels. Consequently, many disparate factors influence who evacuates, when, and how to respond to a nearby disaster. In this paper, we leverage a novel method of data collection through social media to explore the evacuation response decisions of people in areas affected by the 2019-2020 Australian bushfires. We explore the validity of this data collection method for generating plausible estimates of evacuation and its ability to supplement cell phone location data using survey responses. Ultimately, we identify several key factors influencing household decisions on evacuation, specifically focusing on the phenomenon of household members evacuating or returning from evacuation at different times.
