Science impact: experimental warn-on-forecast system yields 75-minute lead time on violent tornado

A stock photo of a tornado over an open field in the Great Plains. Credit: Adobe Stock

On May 21, 2024, severe thunderstorms struck central Iowa, producing flash flooding, straight line wind damage and at least nine tornadoes. One of those tornadoes impacted the community of Greenfield, resulting in five fatalities. The tornado was rated an EF4 by the NOAA National Weather Service (NWS), with an estimated wind speed of up to 185 miles per hour. 

NOAA National Weather Service forecasters issued excellent forecasts and warnings for the storms. Just hours before they occurred, researchers inside the NOAA Hazardous Weather Testbed in Norman, Okla., were using an experimental new tool called Warn-on-Forecast to provide additional information to NWS forecasters. Warn-on-Forecast uses computer models and machine learning to project tornadoes at long lead times. 

For the Greenfield tornado, Warn-on-Forecast yielded strong confidence in the probability of extremely strong near-ground rotation in the Greenfield area more than an hour before the tornado touched down. These results were shared with NWS forecasters in Iowa to assist in their work issuing warnings.