Getting Weather Forecasts Right

It was a weather misforecast that Peter Sousounis would rather not see again.

''Last April we had an amazing situation,'' says the assistant professor of atmospheric science at the University of Michigan, Ann Arbor. ''The forecast called for rain in the morning, then clearing, with temperatures about 40 degrees. It rained in the morning. But within 10 minutes, the rain turned to heavy sleet. Within another 10 minutes, the sleet turned to heavy snow with thunder. We got 2 to 3 inches of snow in an hour, and it continued to snow through 2 p.m. The temperatures never got above 31.''

Dr. Sousounis is obviously not the only one who would would like to avoid a repeat performance.

The National Weather Service is in the midst of a $4.5 billion modernization program to enable forecasters to deliver increasingly precise warnings and updates when the weather turns nasty.

Forecast offices in Pittsburgh, McLean, Va., and Taunton, Mass., recently began shakedown tests of the final element of the technological upgrade - an advanced network that pulls together data from satellites, radar, and automated observing stations, which lie at the heart of the modernization effort.

''Our existing equipment was becoming obsolete. Some of our radar was using 1950-style vacuum tubes,'' says Robert Thompson, manager of the Southern New England Forecasting Office in Taunton. ''At the same time, advances in the science of meteorology were not getting into our operations fast enough.''

To deal with the equipment problem, the NWS, along with the Federal Aviation Administration (FAA) and the Defense Department, settled on a series of upgrades, including:

r Doppler radar. Where its predecessor could only indicate the presence of rain or snow and its intensity, Doppler radar tracks the speed and direction not only of precipitation, but of air masses of different densities. ''We're even able to watch the edge of a sea-breeze front coming in,'' says James Lee, science operations officer at the Taunton forecasting office. The NWS began installing Doppler radar in 1992 and hopes to have 121 of its own units in place by the end of the decade. Forecasters also will have access to data from 29 additional units being installed by the FAA and the Department of Defense, which also are participating in the modernization program.

r Five new satellites. Placed in geostationary orbit, the units are designed to return highly detailed images of weather systems moving across the globe. GOES 8, the first in the series, is operational, while GOES 9 is undergoing engineering tests.

r A network of 850 automated observing units. Known as the automated surface observing system (ASOS), each unit collects, archives, and transmits weather data that once required a human observer. Prior to ASOS, Mr. Lee says, only four or five observing stations in his region were staffed 24 hours a day, and observers were required to report information once an hour unless conditions warranted more-frequent observations. ''Now I've got 10 ASOS stations reporting 24 hours a day and updating observations o nce a minute,'' he adds.

r An interactive data network to tie all the information together. Because the other elements of the system can operate on similar time scales, the network processing system can correlate the information in time as well as in location, allowing forecasters to see conditions and stages of storm formation and progress they've never seen before.

One byproduct of the upgrade program has been increased interest in the detailed weather data the NWS is accumulating on the part of university researchers.

''There's a natural pull,'' says Tim Spangler, director of the Cooperative Program for Operational Meteorology, Education, and Training, a collaborative program that grew out of the modernization effort. ''One group has the data and wants to know more about what to do with it. And the university people are salivating to get at it.''

The collaborative program pairs NWS forecasters with academic researchers to develop better computer models for use in forecasting.

In one instance, the program linked a professor at the University of Michigan with a forecaster in Anchorage, Alaska to develop an automatic tracking system for volcanic ash, which can pose a hazard to airborne planes.

For Dr. Sousounis, a similar collaboration has helped refine his model for a subtle 300-mile wide circulation pattern that develops in the Great Lakes region during the winter and affects cloud cover and precipitation.

The program also focuses on education for forecasters. Mr. Lee points out that training on the old radar took at most a couple of days. ''Forecasters were basically learning how to push buttons,'' he says. By contrast, training on Doppler radar now takes four weeks.

''Before the modernization program, forecasters were in a dimly lit room,'' Dr. Spangler says. ''Now they see things they've never seen before. Everybody has to learn about their area all over again.''