As I type this sentence, my putty-colored Dell GX1 computer is helping scientists predict the future of Earth's atmosphere. Aside from a half-inch-diameter globe at the bottom of my screen, I don't notice my scientific contribution. The machine is running a climate prediction model that I downloaded from climateprediction.net, a project started by University of Oxford researchers whose goal is to find out what the weather will be like as greenhouse gases build up in our atmosphere.
"What are the chances that we'll cook in the twenty-first century or freeze in the twenty-first century?" asks Myles Allen, a physicist and the project's principal investigator. He's only half joking. While all but the most intractable scientists agree that global warming will alter the climate (and potentially whole ecosystems), they do not yet know how catastrophic the changes might be. Notoriously chaotic, the weather is shaped by millions of factors, including the way the ocean interacts with the atmosphere and especially the behavior of clouds. By changing the way factors like these are represented in their climate prediction models, researchers can make wildly different prognostications. The authors of a recent article in the journal Nature, for instance, found that extinction rates for a group of 1,103 animals and plants could range from 15 to 37 percent by 2050 -- depending on the climate-change scenarios employed. "We're trying to limit the range of uncertainty," Allen says. But that takes an enormous amount of computer power, and unfortunately there just aren't enough supercomputers to go around.
That's where my PC comes in. To find out which models are most accurate, the Oxford researchers need to run hundreds of thousands of theoretical climate scenarios. My machine is running just one of them. That may sound like a minor achievement, but when you consider that 45,000 other computer users are doing the same thing, and that no two of us are running the same simulation, it adds up to real results. The software program that manages our models kicks on when we boot up our machines and runs continuously in the background of the computer. Depending on the speed of the processor (Allen recommends a Pentium 4 or the equivalent), each simulation takes between four and eight weeks to complete. Combined, according to Allen, we have the power of Japan's Earth Simulator, the fastest climate prediction supercomputer in the world. When my model wraps up, sometime next week, the results will be sent back automatically via the Internet to Oxford for analysis. Meanwhile, I can check my temperature map and cloud cover anytime (right now it is April 1817, and clouds have moved over Africa). "We're screening millions of different models to give us a reasonable spread of reasonably possible futures," Allen says. While this is less than absolute certainty, it at least brings us closer to the truth.
-- Jill Davis
