Predicting Ivan's course remains inexact science
September 10, 2004
Where will Hurricane Ivan come ashore?
Supercomputers in facilities from Florida to Maryland to California have been trying to answer that question for days. Programmed with decades of hurricane history, updated with wind speed and barometric pressure and humidity from dozens of satellites, buoys and airplane-dropped sensors, they whir and process for an hour or more to produce -- on Thursday, at least -- this answer:
Anywhere from Louisiana on the west to Savannah, Ga., on the east. And maybe not there.
So when the humans at the National Hurricane Center produced a forecast track early Thursday showing Ivan coming through Orlando early next week, they also acknowledged that no one can say for sure when or where the storm will hit.
Orlando, the center said, at that time represented the middle course of a dozen or more computer models that show Ivan following paths west and east of Florida, as well as right up the spine of the state. But trying to predict the course of any hurricane five days out is beyond a supercomputer's abilities; even three-day forecasts can be hugely wrong, as Hurricane Charley showed last month.
Said hurricane-center meteorologist Gene Hafele, "The further you get out into time [before landfall], the more the uncertainty." Two of the biggest factors influencing the possible path of Ivan are a storm heading east through the Rocky Mountains and a high-pressure system in the Atlantic known as the Bermuda high, which typically is in place during the summer and early fall months.
If the storm moves the Bermuda high out into the Atlantic or weakens it, Ivan could head right into Florida and possibly to Orlando. If the Bermuda high strengthens or rebounds from the storm, Ivan could go west into the Gulf of Mexico. There is no agreement yet on what will happen, though the computers are "trending" toward the Bermuda high's staving off the storm.
"This is the stuff people spend decades and decades trying to understand," said Josh Darr, a meteorologist with Risk Management Solutions in California, a company that helps insurance companies forecast potential losses in a natural disaster such as a hurricane.
With tropical-storm-force winds extending 175 miles from its eye, Ivan was headed north-northwest Thursday night. Earlier, its top winds were 160 mph, though it was expected to weaken after crossing Jamaica and, quite possibly, Cuba.
After that, the computers disagree on what happens next.
The reason: Forecasting is an inexact science that -- while growing in sophistication -- cannot take into account the myriad atmospheric factors that eventually determine the course of a storm.
Anything from the temperature of the water to a shifting high- or low-pressure system to wind shears to a hurricane's forward speed can alter the track and intensity of a storm.
The computer models, most fed with data from hurricanes as far back as 1900, try to consider countless nuances in producing a potential course. That's why they can change several times a day, whenever a new set of programs is run.
Although hundreds of hurricane-tracking programs are available or in development, the National Hurricane Center typically pays most attention to five models. They are called NOGAPS, AVN, BAM, UKMET and GFDL.
On Thursday, their plots showed Ivan could head as far west as Louisiana -- or as far east as the open Atlantic. These, and a few others, were averaged to produce a "consensus" track that showed Ivan passing near Orlando but later revised to show it passing west of Tampa.
"It's always better to have several opinions rather than just one," said Pete Dailey, a meteorologist and manager of atmospheric science for AIR Worldwide Corp., a risk-modeling agency in Boston.
Two of the most respected models are NOGAPS, produced by the U.S. Navy, and GFDL, created by a government-funded laboratory at Princeton University and run on National Weather Service supercomputers in Maryland.
Morris Bender, a research meteorologist at Princeton and a developer of the GFDL, said the model solves complex mathematical equations involving fluid dynamics to predict Earth's ever-changing environment.
"It's called a model because it's an approximation of the behavior of the atmosphere," Bender said of the GFDL, which stands for Geophysical Fluid Dynamics Laboratory and was developed in the 1970s. "You're trying to reproduce what the storm is doing, but much faster. The atmosphere is complicated, and nobody's model is perfect."
The GFDL prediction Thursday: Ivan would make landfall in Florida sometime Tuesday near Tampa.
NOGAPS, which stands for Navy Operational Global Atmospheric Prediction System, uses wind, temperature, humidity, barometric pressure and other atmospheric-profile data gathered worldwide by satellites, ships at sea, weather buoys, weather balloons, military and commercial aircraft. It's all crunched by a supercomputer in Monterey, Calif.
The NOGAPS prediction: Ivan would make landfall Tuesday somewhere west of Tallahassee.
"Different models make different assumptions, on how they form clouds, the convective energy in thunderstorms, how radiation is absorbed into the atmosphere. So the models will behave differently," explained Mike Clancy, chief scientist for the Navy's Fleet Numerical Meteorology and Oceanography Center.
"We don't understand the exact physics of the atmosphere," he added.That's no surprise to John Williams, a veteran Florida meteorologist and co-author of a book about Florida hurricanes and tropical storms.
"Babies like these," Williams said of Ivan, "have a mind of their own. They do what they want."