Why do we need to improve tornado warnings?
The most basic answer is to save more lives. The best way to do that is to realize that different users of weather information might have different needs. For an individual at home, a 15-minute warning of a tornado might be enough. But consider schools, athletic stadiums or even Nascar races. Where you have large groups of people concentrated in small areas, 15 minutes is not much time to move tens of thousands of people to safety. Warnings can be more tailored to specific users’ needs. We also need to understand better how people actually perceive and act on the described risk. If you’re the emergency manager of a small town and have longer tornado warning lead times, even if there is more uncertainty, it could help you to better manage and preserve your limited resources to respond to that tornado afterward.

What is Warn-on-Forecast?
Warn-on-Forecast aims to augment the current warning system by adding weather prediction models into the process. The intent is for the model to provide a forecast of an area’s risk over the next hour, for example, a 10% or 50% chance, and then update users every 15 minutes. At home, if there’s a one-in-ten chance for a tornado in the next hour, you might wait for more information. For a hospital, a one-in-ten chance might prompt management to call in extra staff, just in case.

What are the current limitations?
We have a pretty good understanding of what environments produce dangerous storms, so warnings are a combination of the forecaster’s knowledge of the environment in which the storm is forming and the storm’s behavior on the Doppler radar. The forecaster predicts from those inputs whether a tornado will develop. This system has worked well for almost 30 years, but it’s now clear that forecasters are often limited to generating warnings with around 15 minutes of lead time.

How will Warn-on-Forecast improve that?
To extend warning lead times, a new paradigm is needed. This has happened before. The introduction of Doppler radar in the 1980s, combined with advancements in storm science, is how National Weather Service forecasters have extended tornado warning lead times from five minutes to 15 minutes over the last 30 years. Adding in high-resolution, rapidly updating, “storm-scale” weather prediction will help provide hour-long predictions to the public at an unprecedented level of detail. Since weather models—and forecasters, for that matter—are not perfect, a Warn-on-Forecast system is designed to provide probabilistic forecasts of severe weather. A future tornado warning might say there is a 10% chance of a weak tornado passing within a one-mile radius of your location over the next hour. That’s a very different type of warning than is currently issued, and we think this type of information can be used more effectively than the current “yes/no” warnings.

We think it will take another five to 10 years to finish this system and make it operational. Eventually, this type of weather prediction will be used for a broad range of hazardous weather prediction, from flash floods to hail to ice storms.