During the 2004 hurricane season, six major storms formed in the North Atlantic basin and four of them made landfall in the U.S. to devastating effect. Now researchers writing in the journal Nature report that they have developed a new method for forecasting hurricane activity, which could help hurricane-prone states better prepare.

Mark A. Saunders and Adam S. Lea of University College London based their new system on measurements of wind taken at heights between 750 and 7,500 meters above sea level. By analyzing wind patterns recorded in July from six regions in North America and over the oceans, the model predicts the amount of wind energy in hurricanes expected to hit land between August and October. When the team used the new model for retrospective forecasts of hurricane seasons between 1950 and 2003, it correctly classified the extent of damage in 74 percent of the cases. Under "real-time" conditions in 2004, the model performed well, the scientists say, and insurers and others could have reduced their losses by acting on the forecast.

Because of the costly nature of hurricanes--the average annual damage tops $5 billion--any steps toward alleviating their destruction would be welcome. "For over two decades scientists have been attempting--with limited success--to deliver seasonal predictions of hurricane activity reaching the coast of the U.S.," Saunders says. "This study is the first to offer forecast prediction which is high enough to be practically useful." The researchers plan to release their predictions for the 2005 season on August 4.