John Drake
News-Sentinel photo by Michael Patrick

Computer scientist John Drake uses complex mathematical equations to help build models for studying the earth's climate.

from Knoxville News-Sentinel
October 21, 2002
original URL:,1406,KNS_4257_1484544,00.html

Climate calling

Powerful processors enable 100-year simulations of physical conditions from mountains to oceans

By Frank Munger, News-Sentinel senior writer
October 21, 2002

Using computers to simulate the world's climate is a novel and extremely complex undertaking, but John Drake offers a relatively simple description:

"You take the Earth and chop it up into little boxes and apply Newton's Laws to them and lick them all together and see how it works."

Somehow, in mathematical form, that covers everything from leaves falling from trees to winds sweeping up and over mountains to oceans interacting with the clouds above.

It's a big job, requiring some of the nation's brightest minds and fastest computers. Oak Ridge National Laboratory is one of the research institutions involved in the project.

The good news is that scientists have achieved a working computer model that simulates the long-term meteorological conditions and can be used to test a wide range of scenarios about climate in the latter half of the 21st century. The bad news is that it's far from perfect at this point.

"This is the best we have for now, but there are things that need to be added," said Trey White of ORNL's Center for Computational Sciences.

For instance, chemical and biogeochemical interactions in the natural systems are not well represented at this stage, he said.

Drake, whose academic training is in mathematics and fluid dynamics, was an early participant in the virtual experiment with climate. He started around 1990 with a pilot project to see if massively parallel computers could be used to simulate segments of the climate processes.

"Nobody really understood whether that was going to work," he said.

The latest computer model, completed this summer by researchers at ORNL and the National Center for Atmospheric Research, doubled the resolution of the previous version, providing greater regional climate detail.

Instead of looking at box-like segments 200 kilometers on a side (roughly the size of Colorado), the new model is able to focus on areas half that size. This helps scientists better understand local impacts on the global climate conditions.

The greater resolution also means additional details in topography and other climate factors.

"The state of Tennessee is rather long and narrow, and you'd like to say that the weather is not the same in Memphis as it is at the top of Clingmans Dome," Drake said.

Instead of the Rocky Mountains being a just a big hump, it's now a series of big humps, and the Appalachian Mountains now are starting to show up, with recognizable climate effects of downslopes and upslopes, he said.

Radiant energy from the sun essentially dictates the Earth's climate. Some of that energy is absorbed, warming the surfaces, but much of the heat radiates back into the atmosphere.

The Earth's annual revolution around the sun and its tilted position govern how sunlight falls on different parts of the planet during different parts of the year.

Charting the energy balance, input and output, is critical to the climate model.

"I think of it as an accounting nightmare," Drake said.

Buddy Bland, a group leader in ORNL's computational sciences center, said the project in essence requires computing the weather at 20-minute increments for 100 years for the entire world.

"On our big computers here, it takes weeks of time," Bland said. "But if you had one little desktop, you could never finish the problem. It would take too long."

The researchers are running the new climate model on ORNL's Cheetah, an IBM supercomputer capable of performing trillions of calculations per second. The problems are divvied out to the computer's many processors, solving mathematical equations that represent the circulation of the oceans and atmospheric movements and ice cap flows, as well as complex interactions with landmasses and vegetation.

Warren Washington, senior scientist at the National Center for Atmospheric Research in Boulder, Colo., said the computational power at Oak Ridge has helped researchers deal with natural variabilities in the climate model.

The computer programs must be run again and again and again to identify the role of such things as volcanic eruptions, greenhouse gases, solar changes and alterations in land mass over the past century and a half, Washington said.

"Cheetah allows us to do an ensemble of experiments," he said. "It's been a real important tool for us because we can run multiple experiments all at the same time. We can make much faster research progress. It's a very fast computer."

One reason to push for greater and greater resolution in the climate models is to answer questions from politicians and policy-makers, who want to know more specifics about their regions.

Climate information from the best computer models is likely to form the basis of future decisions on energy production, especially the burning of fossil fuels because of the known ties to global warming.

Drake said experiments on future climate run the gamut, ranging from simulating a complete makeover in energy production to no changes at all, the "business-as-usual" scenario.

"Almost any question you can ask is an experiment that is being run on the computer," he said. "What the future will be like is pretty unpredictable."

One of the standard ways to validate the new computer models is to run them against detailed climate records from historical periods and compare the results.

"It matches remarkably well," Drake said.

Good climate records are generally available from 1870 forward, and some researchers are starting to reconstruct comparative models for even earlier periods based on data acquired from tree rings, he said.

Factual data fed into the computer are also important, coming from many sources. Drake said a milestone was reached earlier this year when, for the first time, complete information on the ocean state - temperature, sea levels, salinity, etc. - became available from a group at Princeton University. That forms a starting point in computer simulation.

"We were not able to do that before and have it work," he said. "We were missing data."

ORNL performs multiple roles in the national program, particularly making sure the climate models run well on the latest computer architecture.

A consortium of six Department of Energy laboratories is working on the project as part of the Scientific Discovery Through Advanced Computer (SCIDAC) program. ORNL's work focuses on land and atmospheric processes, including atmospheric chemistry, one of the key areas of climate change.

White, a computer scientist who provides applications support, said conducting a virtual experiment is the only way to tackle such a massive problem.

"There is just no way to guess what the future of climate is going to be except to model it on supercomputers," he said. "We can't do any experiments because we don't have another Earth to play around with."

Copyright 2002, KnoxNews. All Rights Reserved.
Mirrored with permission.

This page is part of the CSMD web archive and is not maintained.
Please visit for the latest CSMD information.