Faster Supercomputers Aiding Weather Forecasts

Brian Handwerk
for National Geographic News
August 29, 2005
Increasingly powerful supercomputers are giving U.S. weather forecasters an edge on Mother Nature—though she still holds plenty of surprises in store.

"Thirty years ago people were delighted when we occasionally got a forecast correct," said James Hoke, who directs NOAA's Hydrometeorological Prediction Center in Camp Springs, Maryland. "Now their expectations have risen. Clearly society is becoming more weather sensitive."

Enter the new generation of weather supercomputers, designed by IBM to help forecasters crunch an ocean of weather-related data.

The computer firm houses Blue, NOAA's primary forecasting computer system, at an IBM facility in Gaithersburg, Maryland. A backup system, White, is located at a NASA site in Fairmont, West Virginia.

The two machines rank as the world's 69th and 70th most powerful computers, according to the TOP500 List of Supercomputers.

"Those two machines are pretty impressive," said Jack Dongarra, computer science professor at the University of Tennessee, who compiles the TOP500 list with colleagues in Germany.

Satellites to Ocean Buoys

Hoke, the NOAA scientist, is a 30-year veteran of the weather forecasting business. He said, "To make a forecast for tomorrow, you have to have a current snapshot of what the atmosphere is like."

"These supercomputers are so powerful that we're able to take advantage of the huge amount of data that streams [in] around the clock."

Weather data arrives from an array of sources, including mountaintop observation stations, ocean buoys, and some 700 global weather balloons. Most wide-body aircraft in the United States even carry sampling instruments.

But satellites increasingly provide extensive coverage that can penetrate and monitor different layers of the atmosphere. The number of daily weather observations crunched by NOAA's supercomputers will soon top 200 million. NOAA uses the data to churn out some 200,000 different meteorological products every day.

Some of these "might just be a map with temperature or wind patterns. Others are [more sophisticated] regional and global models," said Gary Wohl of the National Weather Service's National Centers for Environmental Prediction.

Todd Gross, chief meteorologist for WHDH-TV, an NBC affiliate in Boston, Massachusetts, said, "They are getting better and better and better."

"The four-day forecast is now as accurate as the old two-day forecast," he said. "The two-day forecast is as accurate as the old-one day—though after about 72 hours, it's still difficult to get an accurate forecast from these computer models."

Life, Death, and Dollars

When dangerous storms occur, accurate forecasting can have life-saving consequences, particularly for the growing population found in U.S. coastal regions.

"With respect to hurricane landfall … it costs about a million dollars a mile to evacuate [people from] the coast," Hoke, the NOAA scientist, said. "So every mile that you don't evacuate saves a considerable amount of money."

The global economy is also increasingly dependent on weather forecasts. As much as one-third of the U.S. gross domestic product—three trillion dollars' worth of goods and services—is at least partially dependent on weather, according to estimates.

Obvious examples include road, sea, and air transportation. Less obvious, perhaps, are power companies, which depend on temperature forecasts to anticipate consumer demand, and school systems.

"My wife directs a preschool, and the parents and staff are highly weather sensitive," Hoke added. "She pays more attention to the local weather in the Washington, D.C.-area than I do."

Powering Up

In 1965 Gordon Moore, the co-founder of computer chip maker Intel, posited that the number of transistors possible on a chip would double every two years. The prediction, which became known as Moore's Law, has proved a tad conservative.

Such progress has led to dramatic growth in supercomputing power. NOAA's machines pack a serious punch.

But their critical responsibilities keep them a bit off the cutting edge of computer capability, said Dion Rudnicki, an IBM vice president based in Bethesda, Maryland.

"The main characteristic for NOAA is that they live and die by a 24-by-7-by-365 clock—[and are] concerned about reliability and availability," Rudnicki said. "This is the system producing the output that they need to be able to run the nation's weather services."

The machines still feature staggering abilities.


The early computers of the 1950s worked at speeds of about a thousand calculations per second, a rate known as a kiloflop.

By 2000 supercomputers processed data at a rate of trillions of operations per second—a speed described in teraflops.

To put that in perspective, think of a kiloflop as a piece of paper with a thousand numbers recorded both sides in the span of a second.

By comparison, "a teraflop [is] a stack of paper of the altitude achieved by Spaceship One—[62 miles/100 kilometers], the edge of space—every second," explained Dongarra, the University of Tennessee computer expert.

NOAA's top machines currently run at more than four teraflops per second.

Dongarra reports that by 2010 computers will likely achieve speeds measured in petaflops, a rate of data processing represented by a stack of paper 62,140 miles (100,000 kilometers) in elevation—or one-fourth the distance to the moon.

As processing speeds continue to soar, weather forecasters and those who count on them stand to benefit.

"The progress in weather prediction is right up there with the progress in, say, medical science," Hoke said. "The improvement is just undreamed of."

Free E-Mail News Updates
Sign up for our Inside National Geographic newsletter. Every two weeks we'll send you our top stories and pictures (see sample).


© 1996-2008 National Geographic Society. All rights reserved.