INDUSTRY

In a multi-partner spring program led by NOAA, the Center for Analysis and Prediction of Storms and the Pittsburgh Supercomputing Center generated the highest-resolution numerical weather forecasts yet attempted, with results suggesting that storms may be more predictable than previously thought. As it has during many storm seasons over the past decade, the Pittsburgh Supercomputing Center (PSC) this spring collaborated with the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma, Norman, to produce real-time numerical forecasts of storms. This year, however, rather than forecasting for a small region over the Great Plains, CAPS harnessed PSC resources to generate forecasts over two-thirds of the continental United States. "This was an unprecedented experiment," said CAPS director Kelvin Droegemeier, "that meteorologists could only dream of several years ago." Relying on LeMieux, PSC's Terascale Computing System, the CAPS team produced the highest resolution storm forecasts that have yet been attempted. CAPS used the Weather Research and Forecasting Model (WRF), an advanced model designed for research as well as operational use, and with LeMieux -- running on 307 nodes (1,228 processors) -- successfully produced an on-time, daily forecast from mid-April through early June. A one-of-a-kind collaboration among weather researchers and forecasters, the spring program -- sponsored by NOAA (National Oceanic and Atmospheric Administration) as part of NOAA's Hazardous Weather Testbed -- is operated jointly by NOAA's Storm Prediction Center and the National Severe Storms Laboratory, both in Norman, in close collaboration with the NOAA National Weather Service office in Norman. Three external organizations -- the NOAA Environmental Modeling Center (EMC) in Camp Springs, Md., the National Center for Atmospheric Research (NCAR) in Boulder, Co. and CAPS -- partnered to generate daily forecasts. Using several different versions of WRF, the partners generated forecasts three times daily. EMC and NCAR provided high-resolution forecasts that predict thunderstorm systems in greater detail than current operational forecast models. Using LeMieux, however, CAPS went a step beyond. "Our daily WRF-model forecasts," said Droegemeier, "had twice the horizontal resolution and nearly 50 percent greater vertical resolution than the other two experimental products." The spacing between grid points was two kilometers (1.2 miles), more than five times finer than that used in the National Weather Service's most sophisticated operational model. As a result, the CAPS/PSC experiment required 300 times more raw computing power. This higher resolution meant that the experimental PSC forecasts were able to capture individual thunderstorms, including their rotation. This experiment offered an unprecedented chance for forecasters -- as well as researchers -- to work with advanced technology on a daily basis, which Droegemeier notes may be five years from being incorporated in daily forecast operations at the resolutions used. This year's spring program may well be a watershed event in the understanding of atmospheric predictability. "Results from the spring experiment," said Droegemeier, "suggest that the atmosphere may be fundamentally more predictable at the scale of individual storms and especially organized storm systems than previously thought." Such results could lead potentially, he added, to a revision of classical predictability theory put forth by Edward Lorenz, the now retired MIT professor, whose pioneering research led to chaos theory. In some cases, the WRF captured the initiation of storms to within 20 miles of their actual location, and within 30 minutes of their actual time of formation -- more than 24 hours in advance. "That type of result," said Droegemeier, "pretty much sets conventional thinking on its ear." "Real time daily forecasts over such a large area and with such high spatial resolution," said Droegemeier, "have never been attempted before. These model runs were conducted daily with virtually no problems. We had enormous help from the Pittsburgh Supercomputing Center. For nearly 20 years, notes Droegemeier, PSC has provided outstanding personalized support to large and challenging projects such as this spring program. For his part in this year's project, Droegemeier especially credits PSC scientist David O'Neal. In 1997, CAPS and PSC shared the Computerworld Magazine-Smithsonian Award in Science for their pioneering work in computer weather prediction.