UW News

January 24, 1997

A supercomputer gives the Pacific Northwest a super-detailed view of its often-capricious weather

It was not entirely a victory for hostile elements when severe winter storms devasted the Pacific Northwest in late December. It was also a victory for the National Weather Service, the University of Washington and six local, state and federal agencies, whose weather-forecasting supercomputer is providing local details of the Northwest weather with an accuracy never before possible.

The forecasting system–among the nation’s most advanced–had been running only a matter of weeks when the storms hit. The detailed forecasts on temperature and winds, says Brad Colman, scientific officer at the National Weather Service in Seattle, “helped us to decide that the snow would turn to rain over central and southern Puget Sound.”

The forecasts are emerging from a powerful Sun Microsystems computer on which Clifford Mass, a UW professor of atmospheric sciences, and Mark Albright, a UW research meteorologist, are running a simulation, or model, of the region’s atmosphere. The computer forecasts many aspects of the complex meteorology of an area stretching from northern California to southern British Columbia. The model, as well as new systems such as Doppler radar, “are producing rapid and dramatic improvements in our ability to understand the details of local weather and to predict them,” says Mass.

Twice a day, data from the National Weather Service are fed into a computer program that solves a series of complex equations every 12 kilometers (7.5 miles) over the region. The software simulates the evolution of the atmosphere over the Northwest for the next 36 hours, detailing such weather patterns as the dry zone to the northeast of the Olympic Mountains, the dispersion of smoke from fires, or precipitation from weather features such as a band of clouds and showers called the Puget Sound convergence zone.

The new local predictions already are being used by the sponsoring group in a number of ways. As one dramatic example, the Port of Seattle, which operates Seattle- Tacoma International Airport, says it was given such advanced and accurate warning of the December storms that it had snow-removal teams in place when the storm moved in. “Without the warning, we would not have been prepared,” says a spokesman.

The Puget Sound Air Pollution Control Agency, uses the forecasts to call air- quality alerts when there is a danger of pollution build-up in “stagnant” air. And the National Weather Service is using the data to forecast local weather, from wind and snowstorms to heavy precipitation in the mountains.

Two other members of the funding group, the Washington State Department of Natural Resources and the U.S. Forest Service, will use information on winds to plan controlled burns, so that smoke will not drift over urban areas or wilderness areas and national parks. The Washington State Department of Ecology needs to know wind directions before it can issue permits for ocean oil-spill burn-offs or agricultural slash burning. The ecology department and the federal Environmental Protection Agency plan to feed the data into air-quality computer models, which until now have lacked detailed descriptions of the area’s meteorological conditions.

The processing of the vast number of calculations needed to produce these detailed forecasts has been made possible by the advent of high-speed computers, such as the Ultra Enterprise 4000 with its bank of 14 processors, which Sun Microsystems partially donated to the group. The funding effort is the result of a recognition by the sponsors “that we have a common need for more detailed weather information,” says the group’s coordinator, Naydene Maykut of the Puget Sound Air Pollution Control Agency.

A computer simulation pictures the atmosphere in terms of a three-dimensional grid, with each grid point representing a location where a weather calculation is made. The closer the grid points, the more detailed the weather forecast. But reducing the distance between the grid points requires increased computer power: Just to halve the distance requires a computer eight times as fast.

The daily forecasts at the UW are prepared on computer grids for two regions. One encompasses a vast area stretching from California to Alaska and from the Rockies thousands of miles into the Pacific Ocean. Within this area the grid points are separated by 36 kilometers (22.4 miles). But getting the finer points of local weather requires the more-detailed 12-kilometer grid.

“You have to get under 15 kilometers grid spacing (9.3 miles) to even start to get the fine detail of local weather right,” Mass notes. With the 12-kilometer grid, “we get the major regional weather features, whereas before we didn’t get them at all.”

As one example, he cites the Puget Sound convergence zone, a band of clouds and often-heavy showers that frequently covers central Puget Sound. Eighteen months ago the UW researchers began experimenting with a 29-kilometer (18 miles) grid. But even that wasn’t fine enough, says Mass. But the new system clearly details the storms.

Now Mass and Albright are preparing to make another advance. Sun Microsystems has agreed to donate its even-faster UltraSparc processors. These will give the researchers the computer power to set up a four-kilometer (2.5 miles) grid over the region, which will mean forecasting on a yet smaller scale. “This will provide even more accurate prediction of local weather conditions associated with air pollution,” says Albright, “as well as windstorms along the western slopes of the Cascades, and mountain rain and snow.”

###

Excerpts (soundbites) from Mass are available at (206) 543-4984 after10:30 a.m. on 1/24/97 (times and outcues are on the tape).

Narrated videotapes of weather graphics are available on request.

The forecasts are available on the World Wide Web, providing weather maps updated throughout the day, at http://www.atmos.washington.edu.

For more information contact Mass at (206) 685-0910
or cliff@atmos.washington.edu

<!—at end of each paragraph insert

—>