Some people made last-minute calls. Others grabbed pets and searched for safe interior rooms. By noon on Aug. 13, 2004, Charlotte County residents understood for the first time: Hurricane Charley was their storm – not Tampa's.
Within four hours, Charley's 145-mph winds were pounding the banks of Charlotte Harbor, ripping off rooftops, mowing down trees, obliterating homes and businesses, and tossing the county into a power blackout that lasted longer than a month in some places.
Yet not everyone was caught by surprise.
A handful of professors and students at Florida State University in Tallahassee had correctly predicted the storm's path a full three days before touchdown. At the school's meteorology department, a hurricane team lead by Professor T.N. Krishnamurti had seen consistent results. Their "Superensemble" model repeatedly showed Charley sweeping farther south than the National Hurricane Center in Miami believed possible.
In the days leading up to Charley, Krishnamurti e-mailed his data to the NHC – as did other hurricane forecasters at the National Weather Service, the Navy, Princeton University and in Britain. Yet the NHC sided with the other models, which plotted the storm's course in the population-dense areas of Hillsborough and Pinellas counties.
Local emergency management crews tried to make sense of the unknown, and relayed all the NHC hurricane forecasts to their local governments. In Charlotte County, the models were projected onto a large screen as a series of lines on a map. The entire Suncoast was shown. All the lines but one crisscrossed Tampa and St. Petersburg in a tangled web of possible storm paths.
Charley took the isolated route.
"I think we were lucky," says Krishnamurti in a phone interview, referring to his team's success. But if luck was a factor, science was even more important; the storms may be the powerful creations of an unpredictable environment, but understanding them is the work of a scientific process he mastered in 2004.
Not only did his Superensemble rightly predict Charley's path, it also pegged Hurricane Ivan (the third of the four storms to make Florida landfall last year) within 10 kilometers, or about six miles. His models were so successful that by February, the NHC recognized FSU for having the lowest errors of all the 2004 hurricane models.
"Last year, he did very well," says NHC director Max Mayfield, who roughly two decades ago was a student of Krishnamurti's.
Essentially, the Superensemble is a hurricane forecast that crunches together all the different storm models received by the NHC. But it's more than a mere calculator. The Super tries to remove the models of past errors and, in the process, create a single, smart forecast with few mistakes.
"Adding info, that's the name of the game," Krishnamurti says.
This year, he notes, the competition has revamped its game. Both the National Weather Service and a NASA system have improved their data and mathematical figures. Now, Krishnamurti – best known around the globe for his studies on tropical cyclones and monsoons – must get in front of those models to truly understand how they work and where they go wrong. Then he has to improve upon them.
"Our job has become much harder this year," he says. "But we think the Superensemble will correct the errors and we are confident that we will still be able to do quite well this year."
Things didn't always run so smoothly.
Between 2000-2003, FSU's models were "extremely mediocre," Mayfield says. Many of the other forecasters changed their configurations, but those adjustments weren't accounted for by the Superensemble. Prior to 2004, the Super's best results dated back to 1999, the year it was created.
"It's like the stock market," Mayfield says. "Past performance isn't a guarantee of future results."
This year, Mayfield's got his eyes on newer projects. Years of hurricane research and projects from universities and National Oceanic and Atmospheric Administration labs will finally be tested during live operations at the NHC in Miami. Also, recent funds from Congress will, for the first time, allow NOAA planes to study hurricane winds felt at the surface/ground level. New data-collecting buoys will be added to the Gulf, which may be able to better tell the reach and span of hurricane-force winds.
"The improvements are always incremental," Mayfield says. "There's no silver bullet."
But the process of refining forecasts is helped by improvements in technology and data. Better numbers allow forecasters to become more precise, and as the competition hones their predictions, the Superensemble is right on their heels, says Krishnamurti.
"We're ready for them," he says.
allyson.gonzalez@weeklyplanet.com
This article appears in Jun 8-14, 2005.
