Weather Forecasting

An Improving Forecast

Cliff Mass enjoys cycling to work. In the past, this sometimes meant braving unanticipated rain showers. But now Mass avoids such surprises by checking the frequently updated radar imagery and local forecasts for his commuting route.

Such specific forecasts are a recent development in forecasting. They are particularly satisfying for Mass, UW professor of atmospheric sciences, because he is among those who have been
working to improve regional and local forecasts.

From the Dark Ages to Sophisticated Models

To understand how far forecasting has come, consider where it started. Until the mid-1960s, forecasting was in “the dark ages,” says Mass. Forecasters had very primitive numerical weather forecasting models and radar, designed to track planes but later used to track precipitation as
well. Weather satellite imagery, now an essential tool for forecasting, was not available. “Without satellites, we couldn’t tell what was happening offshore,” says Mass. “Most major storms were not forecast properly.”


Cliff Mass checks recent weather maps in the Department of Atmospheric Sciences' map room. Photo by Mary Levin.

Things began to improve in the late 1960s, as weather satellites became available, and technology has continued to advance. Radar now provides valuable information about wind speed as well as precipitation. Many commercial planes carry weather equipment, and a device known as “the profiler” collects temperature and wind data several times an hour from some 100 locations around the U.S.

Yet the biggest change in forecasting is how scientists use the data collected from these sources. Scientists feed the data into powerful computers, creating a three-dimensional description of the current atmosphere. Then they use equations—computer models—that describe the physics of the atmosphere. Based on these equations, they can predict how the weather is likely to evolve over the course of hours or days. “We’re continuing to develop equations to simulate the physics of how radiation works, how clouds work, and how precipitation works,” says Mass.

Computer models require computer power—the more, the better. So as computers have become more powerful, scientists have been able to refine their forecasts accordingly. “In the 1990s, computers were getting better at forecasting large storms,” says Mass, “but not local conditions. Now we’re able to run computer models at much higher resolution so we can input local weather features.”

The Department of Atmospheric Sciences’ computers—among the most powerful at the UW—are able to run computer models at four kilometer (2.4 mile) resolution. The Northwest Weather Service’s current forecast model, in comparison, is limited to a 12 kilometer resolution.

Taking Flight to Gather Data

How does higher resolution translate to better forecasts? It allows forecasters to take into account local weather features such as mountains and bodies of water. “Very different things can be happening 20 miles away due to mountains and water,” says Mass. “That’s one of the
biggest challenges of forecasting around here. It makes it interesting.”

With higher resolution come new challenges, however. The weather models that work for large-scale forecasts don’t translate effectively when forecasting local weather patterns. For these local forecasts to be accurate, scientists need to improve the physics in the computer models. This requires more detailed Northwest data than has been available. Recognizing this, the Department of Atmospheric Sciences recently completed an ambitious field project, called IMPROVE, to collect such data in two Northwest regions: the Washington coast and the Oregon Cascades.
The major focus of IMPROVE was two familiar Northwest weather features: clouds and precipitation. “We need to understand what’s going on in clouds that causes some to produce precipitation or thunderstorms while others do not,” says Peter Hobbs, professor of atmospheric sciences.


	Through the years, Peter Hobbs has flown this Convair 580 aircraft through smoke, clouds, and ash to gather research data. Photo courtesy of NASA.

There’s no better way to study clouds than to be inside them. So Hobbs spent more than 130 hours flying a research aircraft through the grey stuff to study individual particles.

Taking flight to collect research data is nothing new for Hobbs. He has flown research aircraft through the acrid smoke from burning oil wells in Kuwait, burning jungle vegetation in the Amazon River basin and Africa, and through ash spewing from many volcanoes, including Mount St. Helens. The clouds of Washington and Oregon are generally a more pleasant destination than others Hobbs has experienced, but the questions raised about cloud physics are no less intriguing.

Clouds are made up of water and ice particles that can grow by collisions as they fall through a cloud. At the freezing level the ice particles melt and produce rain. That’s the basic idea, but the details get more complicated. For example, Hobbs is studying whether different mechanisms operate in orographic clouds, which are created over mountains by air being lifted.
In the IMPROVE project, the aircraft measured 60 different parameters every second—the size of droplets, the types of ice particles, the temperature, the humidity—as tiny ice crystals zipped past instruments under the wing of the plane at 100 miles per hour.

“Each particle was recorded digitally,” says Hobbs, who recalls early data collection methods that involved putting a stick with special adhesive out the window of the plane and then retrieving
it with crystals attached. “Now we can collect thousands of images in a few seconds, and see them in great detail.”

Back on the ground, Hobbs and his students are now working with the collected data. “We have so much data, it could keep us busy for years,” he says.

The Next Step

With all the cutting-edge technology being directed toward forecasting the weather, there are still unexpected weather events. Forecasters, the brunt of endless jokes about the accuracy of their weather predictions, are the first to admit that 100 percent accuracy is unrealistic. But Mass also wants people to appreciate how far we’ve come.

“Twenty years ago, we never predicted the big storms,” says Mass. “Now we predict at least 90 percent of them.”

And the other 10 percent? Some of those forecast errors, he says, can be attributed to insufficient understanding of the physical processes at work in the atmosphere. Others are due to poor observation data—particularly in areas downstream of oceans, like Seattle.

To address the latter, Mass has been urging the federal government to fund a weather radar on the Washington coast, which he says has the “worst coastal radar coverage” in the country. “An additional radar would produce significant improvements in weather forecasting here,” says Mass. “It would mean that those forecasts that go wrong in 6 to 12 hours would become far less frequent.”

Even when forecasts are correct, there can be a problem sharing the information with the public. Television forecasts, which must address the needs of a large region in a few short minutes, cannot always provide details about regional variations. Radio forecasts face a similar challenge, as Mass—who presents a weekend forecast on KUOW every Friday—knows all too well.
The best place to turn for weather information, according to the experts? The Internet.

At www.atmos.washington.edu/data/, forecast information from various sources is updated hourly. The site, part of the Department of Atmospheric Sciences’ website, gets hundreds of thousands of hits on an average day, with the numbers rising dramatically when the weather is stormy.

“Having the ability to use the Internet for weather information has been a great boon,” says Mass, whose interest is both professional and personal. “I used to get wet all the time bicycling to work. I don’t get wet anymore. I look at the web and see what’s out there.”

[Winter/Spring 2003 - Table of Contents]