Staying a step ahead of hurricanes via home computer
San Francisco — After a relatively quiet two years, the Gulf Coast seems to be experiencing a more active hurricane season this year. While that is an unpleasant news for the region's inhabitants, William G. Lesso sees a silver lining: A series of hurricanes will allow him to test a system he has developed for predicting where these awesome storms will strike land.
The University of Texas professor's method is unusual in two respects. First, it consists of a program that will run on an Apple or Radio Shack home computer. Second, it makes its prediction several hours faster than the larger and more complex computer models used by the the National Oceanographic and Atmospheric Administration's (NOAA) National Hurricane Center.
More generally, Dr. Lesso's program may also prove a forerunner of a broad range of more sophisticated weather information that could become available to the public through a combination of state-of-the-art weather research and computer networking. Some of the possibilities are:
* The use of computers to get more accurate information about the weather along specific highway routes, for trip planning.
* Portable computers that would provide construction foremen with two- to four-hour forecasts of the weather at a building site to help make decisions as, for instance, when to pour concrete.
* Heating and air conditioning systems run by computers that automatically dial weather networks to get the latest projections of local temperatures, humidity, and winds. This would enable more efficient operation.
* Home computers that can act as storm alarms, monitoring weather data and alerting residents if a severe storm approaches.
All of the latter are based on work now being done at NOAA's Environmental Research Laboratory. A program is being conducted there to utilize the latest in computer technology to create the weather system of the future. After three years' effort, a prototype of such a system has been designed, built, and is now being tested.
''The marriage of meteorology and computer science holds revolutionary capabilities for the field of forecasting,'' says the project's director, Sandy MacDonald.
The focus of this federal effort is to provide a detailed picture of local weather patterns as they happen. Computer information services such as CompuServe and The Source are extremely interested in NOAA's work, because the type of weather information it is developing would lend itself to dissemination over computer networks.
Lesso's hurricane program, by contrast, relies on the current state of meteorological knowledge. In fact, it is a refinement on the oldest method of trying to predict hurricane paths: Looking at what similar storms have done in the past. He has taken a century's worth of records on hurricane movements and reduced them to a statistical model that predicts the direction a given storm is most likely to take. When fed the coordinates of a storm, the computer takes 15 minutes or so to come up with its prediction. This is displayed graphically, by displaying an outline of the Gulf Coast and moving a spiral hurricane symbol along the expected storm path.
For this kind of model, ''his program represents an improvement over what we have had,'' confirms Charles Newman, a computer modeler with the National Hurricane Center in Coral Gables, Fla. But the basic approach does have some weaknesses, the expert says. Such models do well on typical storms, but can fail badly on unusual ones. Thus, Lesso's model quite accurately forecast the landfall of the recent hurricane Alicia, which struck the Houston area, but failed in the case of the next storm, Barry, which acted anomalously, Dr. Newman explains. It takes much larger and more complex computer models, and the instincts of professional hurricane watchers, to predict the unusual storms, he says. One problem is that these large models take several hours to make predictions.
Lesso recently added an error-correcting routine that allows updating of the prediction as a storm moves. Whether this will improve its performance has yet to be verified.
Unlike Lesso's hurricane forecasting program, the kinds of sophisticated weather information system Dr. MacDonald talks about would cost somewhere between $500 million to $1 billion, according to NOAA estimates. But last spring a consortium of 50 universities involved in atmospheric research urged that such a capability should be developed. Such a program could ultimately save billions of dollars and hundreds of lives annually, it said.