Predicting Storms in Oklahoma

In Oklahoma, "where the wind comes sweepin' down the plain," April through June means thunderstorms, huge ones that dump inches of rain in a few hours, swelling streams and flooding crops, businesses and homes. Storm season in Oklahoma also brews the ominous dark funnels that can smash houses flat in seconds.

What would it be worth in reduced property damage and saved lives to have reliable warning four or five hours in advance rather than 30 minutes -- as current forecasts give, of flooding or tornadoes headed in your direction? Kelvin Droegemeier and his colleagues at the Center for Analysis and Prediction of Storms (CAPS) intend to make four hours the norm. An associate professor of meteorology at the University of Oklahoma and co-founder of CAPS, Droegemeier leads a research team that has developed a new weather-prediction model, the Advanced Regional Prediction System (ARPS), that can track evolution of severe storms over an area larger than a state or as small as a city or an airport.

Beginning in 1993, in a unique collaborative experiment, Droegemeier's team used the CRAY C90 and, this past year, the CRAY T3D at Pittsburgh Supercomputing Center to see how ARPS works under the constraints of daily weather forecasting. Each stormy day during Oklahoma's spring storm season, numerical predictions from Pittsburgh provided a real-time, direct feed of information to forecasters at the National Weather Service's Experimental Forecast Facility in Norman, Okla. These experiments mark the first time a fully-automated system of this type has been used in an operational environment.

During the 1995 storm season, ARPS ran on the scalable, parallel CRAY T3D system, allowing the model to run at full capacity and yield predictions in an hour or less, a shorter time than previously possible. Drawing on this computing capability, ARPS marked a milestone in meteorology by successfully predicting the structure and location of individual thunderstorms six hours in advance. This is the first time anywhere this has been accomplished.

Numerical Forecast versus Radar for June 8, 1995.

The ARPS forecast (left) created at 1 p.m. for conditions at 7 p.m. compares well with a radar image at 7 p.m. (right). Color indicates rainfall intensity, increasing from light blue to pink.


This animation, created with ARPS storm forecast data, shows development of a thunderstorm over Oklahoma. It covers a surface area 67 kilometers square extending 17 kilometers in altitude. The white isosurface shows high concentrations of combined cloud and rain water. The red isosurfaces represent columns of high vorticity, which have the potential to spawn tornadoes. The blue ribbons show wind velocity at two planes, about 600 and 4,800 meters above ground level.

Researcher: Kelvin K. Droegemeier, University of Oklahoma at Norman.
Hardware: CRAY C90
Software: Advanced Regional Prediction System (ARPS)
Keywords: weather, forecasting, storm, storms, storm prediction, storm warning, tornado, tornadoes, supercells, weather models, Advanced Regional Prediction System (ARPS)

Related Material on the Web:
More information about the Center for Analysis and Prediction of Storms
PSC News Release about this research.
Projects in Scientific Computing, PSC's annual research report.

References, Acknowledgements & Credits