TORNADO WATCH PROBABILITIES IN PRACTICE: PERFORMANCE, TRUST, AND USE AMONG EMERGENCY MANAGERS

Abstract

Probabilistic information in forecasts was long thought to be too confusing for the public and other key stakeholders. As such, it was not included in many forecasts. However, research over the last twenty years has shown that including probabilistic information increases trust and understanding in forecast information, especially among emergency managers who use probabilistic information as thresholds for decision making in their operations. Since April 28th, 2006, the Storm Prediction Center (SPC) has officially included two probabilities in tornado watches: the probability of two or more tornadoes occurring in a tornado watch, and the probability of one or more strong (EF2+) tornadoes occurring in a tornado watch. Despite these probabilities being in place for almost twenty years, little to no research has been conducted to understand how these probabilities verify, or how stakeholders, such as emergency managers, use these probabilities in their operations. Utilizing tornado watch and tornado report data, we show that the tornado probabilities associated with tornado watches are overall very reliable, with some slight variations in time and area. Through the Extreme Weather and Emergency Management survey, we find that emergency managers use these probabilities in their operations, and many take increased action as tornado probability values increase. Additionally, while most emergency managers correctly identified the reference class for tornado probabilities within tornado watches, many confused this reference class with the Storm Prediction Centers convective outlook reference class. Severe weather communicators should ensure that the reference class for each product is explicitly defined, and the difference between commonly used products is stressed when communicating this information with stakeholders such as emergency managers.