The science behind tornado warnings has come a long way since the first one was issued 70 years ago. But typically about three-quarters of the time, when forecasters issue a warning, a tornado never actually happens. Some new research based on sound could help those warnings be more accurate.
After a tornado passes through a community, people talk about how the storm sounded, almost inevitably comparing the sound to that of a freight train.
But there’s another sound, one that you can't hear — that’s just below what the human ear can detect. It starts as much as two hours before the storm.
"The sound is specific tornado and it carries information about the size, the wind speed, the core pressure," says Brian Elbing, who teaches mechanical and aerospace engineering at Oklahoma State University.
Scientists have known for years that weather events including earthquakes and volcanoes make so-called inaudible sounds — but they haven’t figured out how to use that information to predict big weather events. So, last year, Elbing and a group of students put out some microphones on campus, and in May, they picked up a tornado.
"We could look at radar data of the storm system," says Elbing. "When the rotation there was really strong, it produced almost no infrasound. Then, 10 minutes before the tornado formed is when you start seeing this big increase in sound levels, and then that lasted the duration of the tornado."
If you boost what the microphones recorded into the range of what a person can hear, it sounds like wind, kind of like an actual tornado.
With more testing, this sound could help predict the formation of a tornado. Rick Smith is the warning coordination meteorologist at the National Weather Service office in Norman, Oklahoma. Right now, forecasters combine all sorts of data from radar, atmospheric conditions and human observation before they issue a tornado warning.
"It’s still a mystery though," says Smith. "We don’t know everything. The science is helping them issue better warnings, but the science is not the point where we can definitively say storm A will definitely have a tornado, and storm b definitely won’t."
With the little amount of data scientists gathered last year, Elbing says his team is refining its hypothesis. For example, it showed that it was the tornado making the sound, not the larger storm. And they suspect by studying the amplitude of the sound they might be able predict how strong the tornado is in real time. He believes, with just $50,000 of funding for equipment, the system could be operational across Oklahoma within five years.