October 21, 2005

Varying Wilma Models Confound Forecasters

MIAMI -- In the time Max Mayfield has been at the National Hurricane Center, the forecasting of killer storms has gone from flying kites to satellites and computer models to help pinpoint the ferocity and landfall of storms. But Wilma has confounded the experts.

Simply put, models take information from satellites, aircraft flights, ships, buoys, water temperatures, winds at different levels and other sources to try to determine where a hurricane will go and how strong it will get.

Using those models, forecasters predicted Wilma would meander a few days in the Gulf of Mexico and then race across southwest Florida or the Keys. Its slow speed has somewhat confounded them. While forecasters believe Wilma will be picked up by the jet stream and zoom across Florida, it hasn't happened as quickly as the models have predicted.

"It's going to take a little patience," Mayfield said.

Mayfield said one of the models used for Wilma has been a "windshield wiper," widely varying each time it was been used. In some computer runs, it showed Wilma off the coast of Maine after five days. In others, the same model showed the storm off the coast of Cuba.

When the models agree, it is simple for forecasters to determine the path and speed of a hurricane. When the models disagree, forecasters often use a consensus to determine what they believe is the correct path and speed of a hurricane, Mayfield said.

John Celenza, director of weather technology for The Weather Underground, a private weather company in San Francisco, said his company is having the same trouble predicting Wilma as the Hurricane Center.

"We're all the same boat," he said. "It's really up in the air right now."

Celenza noted that a small change in the data put into a model can make a huge change in the forecast.

"What the model is doing is responding to something in the atmosphere," he said. "The models are very important," he said.

The models look like an alphabet soup of letters, such as GFDL, NOGAPS, and UMET.

NOGAPS is the U.S. Navy's forecast model; UKMET is the global forecast model run by the UK Meteorological Office; the GFDL is a model developed specifically for hurricane prediction by a lab at Princeton, N.J.

But experience has proved hurricanes can be unpredictable.

Last year, Hurricane Charley's 145-mph force took forecasters by surprise and showed just how shaky a science it still is to predict a storm's intensity.

Charley quickly grew from a Category 2 to a Category 4 storm and its course took a sharp turn to the right, which put it some 70 miles south of the originally projected bull's-eye.

With so much media focus on Tampa and St. Petersburg, many residents in and around Punta Gorda were caught unprepared for the Aug. 13 hurricane. That hurricane left at least 13 people dead in its wake, which might not have been nearly as big if the storm had stuck to its original path and struck the big evacuated cities farther up the coast.


On the Web:

National Hurricane Center: http://www.nhc.noaa.gov/

The Weather Underground: http://www.weatherunderground.com