Emergency Management

This study focused on household evacuation decisions and departure timing for Hurricane Ike. The data were consistent with an abbreviated form of the Protective-Action Decision Model in which female gender, official warning messages, hurricane experience, coastal location, and environmental and social cues were hypothesized to produce perceived storm characteristics, which in turn, would produce expected personal impacts. Finally, the latter, together with perceived evacuation impediments, would determine evacuation decisions and departure timing. However, there were fewer significant predictors of perceived storm characteristics and more significant predictors of expected personal impacts and evacuation decisions than hypothesized. Also contrary to hypothesis, female gender, perceived storm characteristics, official warnings, and hurricane experience predicted departure times. However, as expected, evacuation rates declined with distance from the coast; unlike Hurricane Rita 3 years earlier, there was a very low level of shadow evacuation in inland Harris County. Finally, most households evacuated 2 days before landfall, between the time of the National Hurricane Center hurricane watch and warning, and evacuated overwhelmingly during the daytime hours.

To provide people with more specific information about tornado threats, the National Weather Service has replaced its county-wide warnings with smaller warning polygons that more specifically indicate the risk area. However, tornado warning polygons do not have a standardized definition regarding tornado strike probabilities (p s) so it is unclear how warning recipients interpret them. To better understand this issue, 155 participants responded to 15 hypothetical warning polygons. After viewing each polygon, they rated the likelihood of a tornado striking their location and the likelihood that they would take nine different response actions ranging from continuing normal activities to getting in a car and driving somewhere safer. The results showed participants inferred that the p s was highest at the polygon's centroid, lower just inside the edges of the polygon, still lower (but not zero) just outside the edges of the polygon, and lowest in locations beyond that. Moreover, higher p s values were associated with lower expectations of continuing normal activities and higher expectations of seeking information from social sources (but not environmental cues) and higher expectations of seeking shelter (but not evacuating in their cars). These results indicate that most people make some errors in their p s judgments but are likely to respond appropriately to the p s they infer from the warning polygons. Overall, the findings from this study and other research can help meteorologists to better understand how people interpret the uncertainty associated with warning polygons and, thus, improve tornado warning systems.