Predicting in Real Time

For the real-time forecasting experiments during spring storm season, forecasters in Norman determine the basic structure of the atmosphere each morning. If there is any potential for storms, these data become the starting conditions for an ARPS run. A CAPS researcher at Norman feeds the data to Pittsburgh by noon (central daylight time). By 1 p.m., ARPS output is available to forecasters in Norman, in time for the daily afternoon forecast.

For the first experiment, in May 1993, the center scheduled ARPS to run on four of the C90's 16 processors each day for the entire month. The model usually took about 45 minutes to complete its daily run, which simulated four hours of storm evolution. (ARPS runs at a sustained rate of over 400 million calculations a second on a single C90 processor.) Along with other output, the model sent forecasters in Norman a sequence of images they could display on workstations as a moving picture of how the storm would evolve during late afternoon and early evening.

Because the experiment was directed primarily at testing operational procedures, the model left out information -- surface features of the land and Doppler radar data -- that were included in subsequent more full-scale ARPS evaluations. Nevertheless, experienced forecasters found that even this simplified version of ARPS was useful. "It gave us an integrated estimate of rain water at various levels," says Paul Janish, a forecaster who participated in the experiment, "and we were able to evaluate other parameters to give us an idea what type of storm might develop." Forecasters gained enough confidence in the model that on May 21 they referred to it in an official National Weather Service release indicating the potential for intense thunderstorms over Oklahoma that evening.

Stormy Mother's Day in Oklahoma

These images show snapshots from the ARPS prediction for rainfall over central Oklahoma on Mother's Day, Sunday, May 9, 1993. Each image represents a 100 x 100 kilometer square horizontal slice through the storm four kilometers above ground level, showing what the storm would look like on a radar scope, where reflectivity corresponds to how hard it's raining. Color indicates rainfall intensity, increasing from light blue to pink.

Severe storms on Saturday, May 8, and again on Sunday resulted in several tornadoes and heavy flooding in parts of Oklahoma, and ARPS successfully forecast this threat. It showed the development of "supercells" -- storms with potential for rotating updrafts and tornadoes. The rainfall images show multiple supercells forming into a line. "The storm tracks were right on top of one another," says Droegemeier, "so a given point in space would get almost continuous rain. This agreed well with what was observed."


Saving Money and Lives

By increasing the lead time for severe storm warnings, ARPS has the potential to reduce property damage and save lives. It should also make it possible to give more precise information about impending storms. "It's one thing to say there are going to be thunderstorms," says weather forecaster Paul Janish, "it would be another to say there's a likelihood of strong thunderstorms with potential tornadoes, or that tornadoes seem most likely in this part of the state and will form three hours from now." ARPS, says Janish, should give forecasters a better feel for the type of storm likely to form, where it will go and how long it will last.

A fully operational ARPS model could also reduce costs for the airline industry, which loses millions of dollars annually to weather-related flight re-routing and delays.

"Oddly enough," says Droegemeier, "there's no plan at the national level to run a model like this operationally. In some sense, we've been given a challenge to prove the usefulness of a forecast of this type. There are people who, for good reasons, question whether you can actually do storm-scale numerical predictions. The mission of CAPS is to demonstrate that capability."

go back to the main screen