Collecting Longitudinal, Perishable Social Science Observations during Hurricanes

Abstract A hurricane is highly dynamic, evolving over minutes, hours, and days. It can change in intensity, track, and translational speed, all of which can change the hazards from wind, tornado, storm surge, and inland flooding. A sophisticated meteorological observation network exists to observe hurricanes over time, to monitor, understand, and forecast their evolution. It is reasonable to think that people also are dynamic, evolving in response to evolving hurricane risks. Yet, very little is known about how people perceive and respond to hurricane risks over time because developing such knowledge requires collecting social science observational data repeatedly from the same set of individuals in real time during a real-world hurricane event. We describe the methodology we developed and implemented to collect such longitudinal, perishable social science observations for three hurricane events: Laura and Marco in 2020, Henri in 2021, and Ian in 2022. With these data, we examined whether, when, and how people’s amount of forecast information obtained and risk perceptions changed during each of these events. A key finding is that people were dynamic, updating their perceptions and behaviors over time as the hurricanes evolved rather than anchoring on early knowledge and assessments. Moreover, the ways people were dynamic varied based on their predicted exposure to different hurricane hazards and across the hurricane events. Based on the novel insights this research yields, we advocate scaling and expanding this methodology to collect such data for additional hurricanes and for different high-impact and extreme weather hazards. Significance Statement This research illustrates how to collect and analyze surveys repeatedly from the same set of people who are at risk from a given hurricane as it approaches landfall. These data are novel, and they are essential to robustly understand how people’s perceptions and responses evolve with evolving hurricane risks. Such event-based, longitudinal, perishable data should be seen as a type of “social science observations,” akin to meteorological observations, that should be collected for all hazardous weather.