National Geographic News: NATIONALGEOGRAPHIC.COM/NEWS
 

 

Spare PC Power Aids Climate Scientists

John Roach
for National Geographic News
January 26, 2005
 
Thousands of people around the world have loaned the unused power of their personal computers to help scientists model Earth's past and future climate.

The project aims to illuminate the range of possible climate changes in the 21st century. Results so far suggest Earth's climatologic future is even more uncertain than previously believed.

"We've just started down the road to understanding uncertainty in climate change. This has not really been dealt with before," said David Stainforth, a climate physicist at Oxford University in England and chief scientist for the climateprediction.net project.

The initiative is helping scientists understand the complexity and chaos of global climate in a way that supercomputers cannot. Supercomputers typically run single versions of a state-of-the-art, software climate model, known as a general circulation model, or GCM.

GCMs are designed to represent, as accurately as possible, every aspect of Earth's climate system. The programs can represent some aspects accurately. But others, such as how quickly cloud droplets convert to rain, can only be estimated.

Given the chaotic nature of the climate system, where such estimates fall could determine, for example, how long a drought may last in Africa.

There are approximately ten GCMs in the world, according to Stainforth. He said that since climate scientists tend to collaborate, their models tend to be similar. For example, the researchers' estimates of variables, such as the cloud-to-rain threshold, cluster around what seems most probable.

"That's not a surprise. It's probably a sensible thing to do. When you only have one model, you want to make it the best you can. And to some extent that means making it similar to other models," Stainforth said. "But we don't care about that. We just want to explore what the models can do. We'll rule out the [scenarios] that aren't realistic or sensible."

Distributed Computing

To explore the range of climate possibilities, the climateprediction.net project uses the excess capacity of thousands of personal computers around the world to run unique versions of a GCM. Each version contains slight tweaks to variables that scientists know have a range of possibilities.

This allows climate scientists with the project to explore how climate may change in the next century under thousands of potential climate scenarios.

The project is based on the same distributed-computing technology as the popular SETI@home project. In the SETI (Search for Extraterrestrial Intelligence) case, millions of people have downloaded, via the Internet, a software program to help space scientists listen for radio signals beamed at Earth from extraterrestrial beings, should such a thing occur.

Climateprediction.net launched in September 2003. Today there are more than 26,000 personal computers currently running unique versions of the project's GCM. The model runs in the background on PCs and does not affect ordinary computing tasks, such as surfing the Internet, writing a letter, or running a spreadsheet, Stainforth noted.

Ultimately the project aims to produce the world's most comprehensive probability-based forecast for Earth's 21st-century climate. In tomorrow's issue of the science journal Nature, Stainforth and his colleagues will report the results from a test drive of the models.

Test Drive

The first step in the climateprediction.net experiment is for researchers to get a feel for how the differing versions of their climate model work. Stainforth likens it to taking a new car out for a test drive. "When you push down the accelerator, sometimes it makes little difference, and sometimes [the car] shoots away very quickly."

To find out how their climate models "drive," Stainforth and his colleagues sent to project participants more than 2,000 GCM versions. The models realistically represented various past and present climates on Earth but were tweaked by a doubling of carbon dioxide, an atmospheric greenhouse gas.

A model's response to the doubling of carbon dioxide is a scientifically useful measure of the model's sensitivity.

The climateprediction.net models showed a sensitivity range between 2 and 11 degrees Celsius (3.6 and 19.8 degrees Fahrenheit). This was more than twice the sensitivity of the most sensitive GCMs currently used to predict future climate, Stainforth said. Those have a range of 2 to 5 degrees Celsius (3.6 to 9 degrees Fahrenheit).

Gavin Schmidt, a climate modeler with the NASA Goddard Institute for Space Studies in New York, said he is surprised by the sensitivity of the climateprediction.net GCM runs. However, he said the results need to be examined more closely before any implications from the study are drawn.

"The team involved here needs to be congratulated on getting this far," he said.

Stainforth cautioned that the fact that the climateprediction.net GCMs have a sensitivity range of 2 to 11 degrees Celsius does not necessarily mean that global temperatures will rise within that range.

For example, there's no guarantee that carbon dioxide levels will only double. Also, more model runs with more climate variables—such as an ocean that changes temperature throughout the seasons—are needed to accurately assess the range of future climate scenarios. These will be done in the next stages of the project.

Schmidt, the NASA climate modeler, said that during the last glacial period, about 20,000 years ago, changes in ice sheet cover, vegetation, greenhouse gas concentrations, and atmospheric dust most likely forced a 5-degree-Celsius (9-degree-Fahrenheit) cooling, suggesting a sensitivity of around 3 degrees Celsius (5.5 degrees Fahrenheit).

"If the sensitivity was really 11, the [forces] required to produce 5-degree cooling must have been proportionately less. This appears to be unsupportable, given what we know about glacial climate," he said.

Stainforth and his colleagues have only completed the test-drive phase of their project. They hope future runs with refined and more complex models will lead to an accurate, validated forecast for the 21st-century climate. To participate, log on to www.climateprediction.net.

Don't Miss a Discovery
Sign up our free newsletter. Every two weeks we'll send you our top news by e-mail (see sample).
 

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