Hurricane Forecasters
Try Model That Focuses
On Chances of Landfall
June 10, 2005; Page B1

Forecasting the future is tough enough. But when it comes to hurricanes, "predicting" the past is no cakewalk either.

To predict the coming hurricane season, scientists look at climate factors in late summer that are linked to hurricane activity. Then they see how well they can predict those factors -- ocean temperatures and currents, El Niño conditions, wind patterns -- and thus the number and intensity of coming storms. Next, they test this model, plugging in the numbers from a particular year in the past and seeing if the model correctly "predicted" that year's hurricanes. If not, they fine-tune equations, adjust the weight they give each factor ... and order in crystal balls and chicken entrails.

I exaggerate only slightly. But seasonal hurricane forecasts clearly need help. In May 2004, the National Oceanic and Atmospheric Administration, home of the nation's meteorologists, forecast a 50% chance of a higher-than-normal Atlantic hurricane season, with two to four major hurricanes. That August, NOAA revised the odds -- down: It pegged the chance of an unusually intense season at only 45%. The respected team at Colorado State University, Fort Collins, also lowered its forecast in August, just days before Charley formed.

But as NOAA wrote in a postmortem after the rampages of Charley, Frances, Ivan and Jeanne caused a record $22 billion in insured losses and killed at least 3,100 people, 2004 "had well-above-normal activity," with six major hurricanes.

With the 2005 hurricane season under way as of June 1, the unforeseen (by many) devastation of 2004 has led critics of the traditional methodology to argue that it is time to throw out the standard crystal ball, which relies heavily on the sea surface temperatures from which storms draw their fury. They are also calling for forecasters to focus not just on how many storms will form but on how many will make landfall. That's one of the toughest parts of forecasts, but it is also where scientists are making surprising progress. In a promising new model, the number of hurricanes making landfall in the U.S. depends on conditions you'd never suspect.

When atmospheric scientists Mark Saunders and Adam Lea, of the Tropical Storm Risk unit at University College London, scrutinized 54 years of data and looked for correlations between wind patterns and the hurricanes reaching U.S. shores, one set of measurements stood out: wind patterns 2,000 to 22,000 feet up, over six regions of North America and the eastern tropical Pacific and North Atlantic oceans during July. How the strength and direction of these winds deviate from the norm, says Prof. Saunders, "is strongly linked to upcoming hurricane activity."

The reason is that wind patterns either favor or block hurricanes from making landfall. For instance, when the usual high-pressure area around Bermuda is shifted north and is stronger than usual in July, it tends to stay that way. "Once these wind patterns are set up in July, they persist through October," says Prof. Saunders.

The Bermuda high is a crucial factor in determining if a hurricane will make landfall, agrees Steve Smith, an atmospheric physicist at Carvill America, a reinsurance intermediary in Chicago. "From 2000 to 2003, the Bermuda high was closer to Europe and steered hurricanes away from the U.S. coast. But last year it was more westerly," he says. Parked off the U.S. coast, it generated winds that blew hurricanes onto land.

Oddly, winds over the Rocky Mountains have been even better hurricane harbingers. Strong southerly winds over the Rockies in July set up a low-pressure zone over the western Gulf of Mexico. That produces steering winds that push hurricanes toward the Gulf Coast and Florida. Winds over the tropical east Pacific strengthen the low pressure over the Gulf, setting up a wind pattern that arcs north to drive storms onto land.

Measured by its ability to retrodict past hurricane seasons from the wind anomalies in July of those years, Tropical Storm Risk is twice as precise as the sea-temperature method. Its August 2004 forecast said the chance of an unusually intense hurricane season was 86%, compared with NOAA's 45%.

Its 2005 forecast, issued this week, says there is an 86% chance that landfalling hurricanes will put 2005 in the top one-third historically, with two to five intense hurricanes. The Colorado team agreed, upping an earlier forecast to eight hurricanes, half of which would be whoppers with sustained winds above 110 m.p.h.

Get used to it. Surface temperatures in the Atlantic have been elevated since 1995, relative to an historical average that goes back 150 years, notes NOAA's Stanley Goldenberg. From 1995 to 2000, the number of hurricanes almost doubled from the historical norm. Elevated sea temperatures might be part of a normal, 50-year cycle, "but you have to wonder if it is also linked to global warming," says Prof. Saunders.

Either way, "2004 was not unprecedented," says Dr. Smith. "Simple statistics say the return period for storm losses like those of 2004 is 50 to 70 years. But there is reason to believe it might be shorter."

And reason, too, to believe the season will be earlier. As of yesterday, this year's first named storm, Arlene, was swirling through the Caribbean, almost two months ahead of 2004's first.

• You can email me at sciencejournal@wsj.com.