Leveraging Geotagged Social Media to Monitor Spatial Behavior During Population Movements Triggered by Hurricanes
Yago Martín González
Advisor: Dr. Susan L. Cutter
In a world of increased mobility and interconnectedness, the study of spatial behavior becomes more relevant than ever. However, multiple researchers have highlighted that the understanding of these dynamic processes has reached a bottleneck derived from the rigidity of traditional spatial behavior inquiry methods and the unavailability of trustworthy and relevant information. These difficulties are even more prominent during emergencies and disasters as these events often create scenarios where spatial behavior does not follow regular and logical patterns and where conventional mobility datasets are often skewed or not existent. Thus, many scholars working within the spatial behavior sub-discipline are pursuing innovative data collection methods to deepen the understanding of human spatial behavior. Researchers see digital geospatial trace data, also known as passive citizen sensor data, as one of the most promising opportunities to develop and test new hypotheses on spatial behavior. Nevertheless, the application of these new methods has not been fully explored within the hazard/disaster discipline for spatial behavior purposes under stressed situations.
This dissertation investigates the suitability of geotagged social media (Twitter) as an innovative approach for the study of spatial behavior of people in stressed contexts and responds to three main research questions: 1) How well do geotagged social media estimate hurricane evacuation compliance? 2) To what extent is geotagged social media amenable for determining hurricane evacuation behavior? 3) How suitable is geotagged social media to evaluate post-disaster displacement and tourist flows? The dissertation therefore not only attempts to develop a new method to estimate the number of movements associated with the different stages of an emergency but also tries to answer long-standing questions about the response of different population sub-groups (residential status, gender, age, race/ethnicity) before, during, and after hurricanes. Results confirm the potential of geotagged social media to tackle some of the deficiencies of traditional approaches, particularly offering more timely, dynamic, and affordable information about the evacuation and post-disaster population movements. In addition, results demonstrate that the Twitter-based approach complements survey-based methods as it permits accessing underrepresented groups in traditional approaches such as the young, short-term residents, and racial/ethnic minorities. Although the representativeness of Twitter samples is still debatable and needs further research, this method to investigate emergency-triggered population movements can ultimately improve our understanding of the response and recovery phases of a disaster.