Few natural phenomena evoke a greater sense of awe and mystery than hurricanes, the powerful and destructive tropical storms that often menace coastal regions. Scientists at GFDL, led by Yoshio Kurihara, have worked steadily for two decades to develop physically based models that can be used to understand the basic mechanisms producing hurricanes and to predict the movement and intensity of individual storms.

A PREDICTION BREAKTHROUGH

Improvements in hurricane prediction have been made possible by improved observations, models, and methods for incorporating observations into the forecast models. Kurihara and collaborators Bob Tuleya and Morris Bender realized that current weather analyses were inherently inadequate as "initial conditions" for such models. They developed a method to circumvent much of the initial data problem and achieved a breakthrough in hurricane prediction.

The new forecast system achieved a series of remarkably successful forecasts, compared to other available methods for Hurricanes Andrew, Iniki, and Emily. For example, while conventional models predicted that Hurricane Emily (August 1993) would make landfall in the vicinity of Georgia, the GFDL forecast system correctly predicted that the hurricane would deliver a glancing blow to North Carolina's Outer Banks before veering back out to sea. Based on its impressive performance during the 1993 and 1994 hurricane seasons, the model has been adopted as the operational hurricane prediction model at the National Meteorological Center.

Three-day forecast of Hurricane Emily's track (red line) and intensity produced by the GFDL hurricane model. Thin contours show the maximum predicted wind speed (in knots) during the forecast period. The forecast was available at 8 a.m., Sunday, Aug. 29, 1993. The hurricane caused extensive damage along North Carolina's Outer Banks beginning late Tuesday afternoon (Aug. 31) before recurving to the northeast and out to sea, as predicted by the GFDL model.

Three-dimensional view of Hurricane Emily near the North Carolina coast as modeled by the dynamical hurricane prediction system developed at GFDL. The view is a 48-hour forecast (valid 8 p.m., August 31, 1993) obtained from the GFDL model. This forecast system provides improved hurricane track predictions roughly two days in advance of hurricane conditions and has the potential to provide useful forecasts of storm intensity and precipitation as well. In the case of Hurricane Emily, the model successfully predicted the occurrence of hurricane-force winds in the Cape Hatteras, N.C. vicinity. In the figure, the three-dimensional "cloud-like" volume (which is sliced vertically through the storm center) depicts the 91% relative humidity surface; the colored shading at the earth's surface shows precipitation (higher values are red); red and white arrows show vertical motions and surface winds, respectively; white contours depict sea-level pressure; the dark blue surface and solid green background represent the Atlantic Ocean and the U.S. East Coast, respectively. Note the "eye-wall" in the relative humidity surface at the storm center and the heavy precipitation north of the storm.

Given the expense and inconvenience of hurricane evacuations and emergency preparations, the public benefits greatly if unnecessary hurricane warnings can be avoided. Furthermore, residents are likely to take hurricane warnings more seriously if the forecasts are shown to be more reliable. With potential damage from a single storm exceeding $25 billion (e.g., Hurricane Andrew in 1992), reliable short-term preparations, while expensive to undertake, can result in large net savings by the public.

A NEW COASTAL FORECAST SYSTEM

In a complementary effort, GFDL is cooperating with scientists at Princeton University and NOAA's National Meteorological Center (NMC) and National Ocean Service (NOS) to develop a coastal forecast system for the East Coast. The system is designed to provide nowcasts and forecasts of currents, storm surges, and other information to coastal residents, the fishing and marine transport industries, and government users. The forecast system uses a high-resolution ocean model--developed by Princeton's George Mellor in collaboration with GFDL--and an atmospheric model from NMC.