Protective action for tornado threats: Probabilistic vs. Deterministic hazard information

摘要

The current operational warning system in the U.S. issues deterministic warnings for severe weather threats, like tornadoes. National Severe Storms Laboratory has been conducting research on issuing Probabilistic Hazard Information (PHI). Unlike the current system that uses polygons to depict the area under the threat, PHI uses swaths for that purpose. The goal of this study is to predict people's protective action at each closeness level to a tornado with or without uncertainty information about the threat. Images of polygons and color-coded swaths were presented to 50 participants, who were instructed to choose one of three levels of protective action at different closeness levels to the tornado. At each closeness level, an ordinal logistic regression was conducted to model how presence or absence of uncertainty information about a particular tornado, factor of "design", and presence or absence of probabilistic information about formation of any tornado in the area, factor of "background", affect probability of taking different levels of protective action. The results show that the effect of "design" on probability of taking action is significant in scenarios that there is less than 20 minutes to get impacted by the tornado. However, "background" does not have a significant effect on people's responses.