Pittsburgh Supercomputing Center

FOR IMMEDIATE RELEASE                         CONTACT:
July 1, 1998                                     Michael Schneider
                                                 Pittsburgh Supercomputing Center

On the Horizon: Accurate Storm Forecasts

Timely, reliable forecasts of bad storms will be available within five years, says leader of nationwide numerical weather prediction experiment.

PITTSBURGH — Thunderstorms and tornados struck southwest Pennsylvania and other parts of the country with unusual ferocity this spring, taking lives and inflicting millions of dollars in damage. It won't be long — perhaps five years, say scientists — until new numerical weather prediction technologies tested during May at the Pittsburgh Supercomputing Center will avert some of the loss from similar storms in the future.

"There's a huge change on the horizon," says Kelvin Droegemeier, who directs the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma. Within the next five years, the National Weather Service and some private companies, such as airlines, will implement new computer-based models that can provide detailed information on severe local weather. "It's a major transition in technology and weather science," says Droegemeier. "The effect is analogous to introducing the X-ray and CAT scan in modern medicine, and it's going to take forecasters and all users awhile to learn how to use this information."

Giving forecasters an opportunity to work with the new high-resolution data was one objective of a major nationwide study called SAMEX (Storm and Meso-Scale Ensemble Experiment) conducted this spring. Along with CAPS and the National Centers for Environmental Prediction (NCEP), which provides numerical forecast guidance that serves as the starting point for nearly all U.S. weather forecasts, SAMEX involved the National Center for Atmospheric Research (NCAR), the Air Force Weather Agency (AFWA) and the National Severe Storms Laboratory (NSSL), and it included as participants more than 40 meteorologists and forecasters from around the country.

Another SAMEX goal was to test a concept called ensemble forecasting, and for this the project necessarily relied on the Pittsburgh Supercomputing Center. Ensemble forecasting involves using many individual forecasts to arrive at a mean forecast that, in theory, averages out much of the uncertainty inherent in weather prediction and gives a quantitative measure of forecast accuracy. "It may sound like a brute force approach," says Droegemeier, "but ensemble forecasting actually is a very sophisticated way to deal with a big problem of the atmosphere, which is that it's inherently chaotic and nonlinear."

This condition, often referred to in popular writing, including the movie "Jurassic Park," as "the butterfly effect," means that since it's impossible to precisely specify the state of the atmosphere (temperature, barometric pressure, wind speed, etc.), very slight differences in the starting conditions for a numerical forecast can lead to vastly different forecast outcomes. SAMEX relied on the diversity of numerical models from the five participant agencies, each of which generated several 30-kilometer-resolution forecasts over the continental United States (each day during May) to develop an ensemble forecast.

"This is the first time that multi-model ensembles at that resolution have been done," says Droegemeier, "and we're seeing all kinds of interesting things." Significantly, the early results — presented and discussed in mid-June at NCAR in Boulder, Colorado — show that the mean forecast is better than even the best individual forecast from any of the models.

Because ensemble forecasting by its nature requires generating many forecasts (25 were used in SAMEX), it multiplies the demand for computational resources. Other centers, notes Droegemeier, aren't yet set up to provide the needed computational horsepower on a dedicated basis.

Because their own computers were saturated with operational requirements, NCEP turned to Pittsburgh's CRAY C90 system to run its SAMEX ensembles, and as part of the project developed a regional ensemble system it will incorporate into operational forecasting. "This major research experiment wouldn't have been possible without the resources and cooperation of the Pittsburgh Supercomputing Center," says NCEP atmospheric scientist Jun Du.

Since 1993, CAPS has conducted spring storm forecasting experiments in collaboration with PSC. These experiments, which for the first time successfully predicted location and atmospheric structure of thunderstorms as much as six hours in advance, have proven the feasibility of storm-scale forecasting. For this work, CAPS and PSC won the 1997 ComputerWorld Smithsonian Award for Science, and CAPS also won a 1997 Discover Magazine Award for Technology Innovation.

Storm-scale forecasting involves a much tighter focus — down to the scale of counties — in contrast to current weather reports on TV, which derive from larger-scale models that predict atmospheric structure over the continental United States. "We want to get down to be able to say, for instance," says Droegemeier, "that over Pittsburgh this afternoon at 3:30 there'll be a thunderstorm with 30 mile-per-hour wind, golfball-sized hail, two-and-a-half inches of rain, and it will last 10 minutes, and to give you that forecast a few hours in advance."

More information about SAMEX, including graphical results from daily forecasts, is available at http://www.caps.ou.edu/CAPS/samex.html.

More information, including graphics and video animation, about CAPS storm-scale numerical weather prediction experiments at PSC is available at: http://www.psc.edu/science/droeg.html

The Pittsburgh Supercomputing Center is a joint effort of Carnegie Mellon University and the University of Pittsburgh together with Westinghouse Electric Corp. It was established in 1986 and is supported by several federal agencies, the Commonwealth of Pennsylvania and private industry.

# # #

go back to contents page
© Pittsburgh Supercomputing Center (PSC)
Revised: July 1, 1998

URL: http://www.psc.edu/publicinfo/storms_07-01-98.html