The impact of probabilistic tornado warnings on risk perceptions and responses

This study investigates whether the inclusion likelihood information affects actions and choices, such as shelter decisions, ahead of a tornado. This is because many warnings issued to members of the public are deterministic in that they do not include event likelihood information. This is true of the current polygon-based tornado warning used by the American National Weather Service. Two experimental studies and one in-person interview study were conducted by the researchers. The experimental studies compared five probabilistic formats, two with color and three with numeric probabilities alone, to the deterministic polygon. 

The results showed that, across both experiments, probabilistic formats led to better understanding of tornado likelihood and higher trust than the polygon alone. By contrast, color-coding led to several misunderstandings. When the polygon boundary was drawn at 10% chance, those using probabilistic formats made fewer correct shelter decisions at low probabilities and more correct shelter decisions at high probabilities compared to those using the deterministic warning. However, when the polygon boundary was drawn around 30%, participants with probabilistic forecasts had higher expected value. The interview study revealed that, although tornado-experienced individuals would not shelter at 10% chance, they would take intermediate actions, such as information-seeking and sharing.