Thinking Outside the Polygon: Using Signal Detection Theory to Study Tornado Warning Reception

We are working with the FACETs (Probability of What?) team to develop and establish baseline measures of the extent to which members of the U.S. public receive, comprehend, and respond to NWS warnings, watches, and advisories under the current WWA system. In this presentation, we highlight preliminary research on one of these components: warning reception. Like other detection processes (e.g., warning issuance/verification), we suggest that warning issuance/reception is subject to the following four possibilities: hit (warning is issued and received), miss (warning is issued but not received), false alarm (warning is not issued but is received), and correct rejection (warning is not issued and not received). Though useful, extant research on tornado warning reception generally occurs “within the polygon” and, by extension, focuses exclusively on hits and misses—when a warning is issued for a given area, how many people in that area receive it? This practice ignores the possibility that some people may (mistakenly) receive warnings that are not issued for their local area. To address this possibility, we use survey data from the Meso-Scale Integrated Socio-geographic Net (M-SISNet) to study warning reception following two relatively memorable storm systems that produced tornadoes in Oklahoma (April 29th and May 9th, 2016). Findings indicate that misses are relatively rare but false alarms are surprisingly common. This is problematic because false alarms may cause confusion, complacency, and unnecessary (costly) reactions to non-existent warnings. We close with a brief discussion about the implications of these findings for the FACETs initiative.