NOVEMBER 1999

Krishnamurti

Accurate forecasts can save lives and fortunes

By Amy Welch
FSU Communications Group

The day before hurricane Floyd blasted North Carolina, Susan Schiller, a producer from CBS Evening News, was on the phone to ask FSU meteorologists where Floyd was going to make landfall.
The FSU researchers mentioned Wilmington, N.C., and the next day, Dan Rather was on a plane to Wilmington, just a few miles north of where Floyd hit land. That night, people across the country saw his broadcast from that city.

The team at FSU's Real Time Hurricane Forecast Center had predicted, five days before Floyd landed, that it would miss Florida and hit the southeastern part of North Carolina, while other forecasters predicted the storm would hit the east coast of Florida.

The researchers give all the credit to a new forecasting method, the Superensemble technique, which works exactly as it sounds. FSU researchers take 11 forecasts from weather centers all over the world, including three of their own, and punch the coordinates into a supercomputer. The supercomputer, similar to IBM's master chess player, Deep Blue, compares those forecasts with past ones and removes errors among them. What's left is a super accurate forecast.

FSU professor T.N. Krishnamurti, an internationally recognized tropical meteorologist and a pioneer in numerical weather predicting, developed the idea for the Superensemble in January. By June, he and his research team had the National Science Foundation, NASA and Science magazine eager for more.

"If you just look at one or two days, the differences may not be that large but if you extend this to four or five, six days then you can see a huge difference," Krishnamurti said. "The Superensemble seems to outperform most models for longer ranges. Under shorter range, the differences are smaller, but still it's probably one of the best models that exist today."
But the researchers don't just forecast hurricanes; they can predict how cold this winter will be or how much rain farmers can expect, and they make one-to-six-day regional or global forecasts. In fact, we may stop blaming the weatherman for giving us the wrong daily forecast if the Superensemble has anything to do with it.

"We believe that substantial improvement to these areas can be made using the Superensemble technique," said Eric Williford, an FSU research assistant.

The FSU team is currently performing forecasts only for research, but there's no doubt that predicting a hurricane landfall five days before it hits could be useful information, and not just to tell CBS where to send its evening anchors. It could save millions of dollars in evacuation expenses, and those who will be hit could have more time to evacuate.
In the case of Floyd, the Superensemble method predicted the storm's path accurately five days before it hit.

Krishnamurti said he hopes his idea will help move forecasting to a higher technological level.
"There's a lot of scope for improving the research so we can keep improving what we are doing," Krishnamurti said. "We need to do this on a much finer grain so we can compute these forecasts more accurately and tell where there will be heavy rains and landfall during hurricane events."

Send a letter to the Editor:fstimes@unicomm.fsu.edu

Copyright ©1999 Florida State Times