A British team claims a breakthrough in predicting hurricane strikes
Hurricanes are the most expensive natural disasters to hit the United States every year, yet there is no reliable way to forecast how many will strike land each season. The best predictive techniques tested against the historical record from 1950 to 2000 have turned out to be less than 40 percent accurate. But a team of British scientists, citing those estimates in a recent Nature article, claims to have come up with a method that almost doubles the forecast accuracy, to a reliability level of 65 to 70 percent.
The team, led by climate physicist Mark Saunders of University College, London, has made projections for the current hurricane season, which began in June and stretches through November. "Based on current and projected climate signals, Atlantic Basin and U.S. land-falling tropical cyclone activity is forecast to be about 200 percent above average in 2005," says Saunders. (A hurricane is a tropical cyclone with winds of at least 119 kilometers per hour.)
On average, two major hurricanes hit the United States every three years. Last year, in an exceptionally fierce season, four major hurricanes struck Florida coasts, an unprecedented number. More warning of such above-normal hurricane activity, even just a few months in advance, could help save lives, cut property losses, and enable people to buy insurance as the season begins.
But even with its highly skilled scientists, supercomputers, aircraft that gather storm measurements, and weather stations that feed it hourly data, the U.S. National Weather Service does not have a good track record in predicting hurricanes. Because of the huge number of variables influencing weather, models inevitably are complex and tax the abilities of even the fastest supercomputers to yield timely, useful predictions [see photo, " Worrying"]. To improve performance, modelers resort to parameterization--a process in which mathematical tricks are introduced to make computations more manageable--but the reliability of the forecasts can then suffer.
The method developed by Saunders and his University College colleague Adam S. Lea relies instead on hurricane correlations with July wind anomalies in regions over North America and the eastern Pacific and North Atlantic oceans to predict hurricane frequency during the main season, which occurs from August through October. Because wind anomalies in those areas have been found to either favor or hinder the development of big hurricanes that strike land in the United States, Saunders and his colleagues believe they do a better job of predicting worrisome storms than the more complex models.
The British pair initially noted links between U.S. hurricane activity and anomalies in sea-level atmospheric pressures over the northeastern United States and the Gulf of Mexico, and they observed that the telltale anomalies were already present in July. Those associations proved difficult to incorporate into models, but then the scientists realized the anomalous pressure patterns were also linked to unusual patterns in the overlying winds.
Assessing the British team's work, Gerald Bell, a meteorologist with the Climate Prediction Center, calls it "a good first step." But he also says that the center, a part of the weather service and under the umbrella of the National Oceanic and Atmospheric Administration (NOAA), in Washington, D.C., has no plans to use the technique. (The weather service is based in Silver Spring, Md., and the Climate Prediction Center in Camp Springs, Md.)
According to Bell, NOAA's hurricane forecasts are based on various trends known to influence hurricane formation. He agrees with the British scientists that some of the key ingredients are already in place for an above-normal Atlantic hurricane season this year, though the ingredients NOAA looks at are different. The National Weather Service predicts 11 to 14 tropical storms this season, with 7 to 9 becoming hurricanes and 3 to 5 of these becoming major (or intense) hurricanes.
The forecast of the British team is similar but more specific. Saunders predicts "15 tropical storms for the Atlantic basin as a whole, with 9 of these being hurricanes and 4 intense hurricanes." The decisive difference, as Saunders sees it, is the greater confidence he attaches to his forecast. NOAA said on 16 May that the probability of severe Atlantic storm activity's being above normal this season was 70 percent; Saunders, on 7 July, estimated the probability of above-normal activity at 97 percent. His revised estimate for the August to October season, based on complete data for July wind anomalies, was to be issued on 5 August.
NOAA's forecast is posted at http://www.cpc.ncep.noaa.gov/products/outlooks/hurricane.html.
The British team's forecast is available at http://www.tropicalstormrisk.com.