Correcting under-reporting in the U.S. tornado database

What is it about?

Tornadoes that occur away from densely populated cities often go unreported, causing tornado frequency estimates to be too low. We adopt a sophisticated statistical technique to estimate tornado reporting rates and actual tornado counts over the central U.S. during 1975-2016. We find that the actual number of tornadoes over the analysis domain was on average more than double the number recorded.

Featured Image

Why is it important?

The tornado reporting rates and corrected tornado counts estimated by this method could be of great value to climate studies; tornado and other hazard models for forecasters, planners, and insurance/reinsurance companies; and the development and verification of storm-scale prediction systems. The substantial tornado under-reporting highlighted by this study indicates that many more tornadoes occur without warning than currently thought, but also that fewer tornado warnings are false alarms. The substantial under-reporting also indicates that supercells and quasi-linear convective systems - the storms that spawn most of the tornadoes in the U.S. - are more efficient tornado producers than currently thought. The method is directly extendable to other events subject to under-reporting, for example, severe hail and wind.