CALVERTON, MD. - Even a century ago, scientists working out equations on paper understood that gases in the atmosphere absorbed and emitted energy, keeping Earth from being a ball of ice. Today they use supercomputers to make increasingly refined predictions about how the Earth's climate will change.
The new efforts take the question from global to local scale. Nations, states and communities have lots of climate-related questions: Should they divert water from one area to another? Build higher sea walls? Store and manage water the way Israel does today? Plan for many more 100-degree days in future summers?
"We can't answer those questions with the capabilities we have today. That's why we're using supercomputers to push the limits of what we understand and how well we can predict," said James Kinter, a professor at George Mason University in Virginia and director of the Center for Ocean-Land-Atmosphere Studies.
Better models
Scientists have used computer-generated models for decades to understand the past, present and future climate by studying the interaction of the oceans, atmosphere, land and ice. Kinter said climate models today show changes on a continental scale, but that faster computers will be able to make better predictions at regional and local scales.
Better computers should help with the difficult climate problem of clouds. Clouds interfere with the flow of energy between the Earth and the sun in two ways, Kinter said. They reflect some of the sun's energy back to space, a cooling effect. But they also absorb and send back some of the energy that the Earth emits, just as gases such as carbon dioxide in the atmosphere do. That's a warming effect.
Recent models looked at the Earth as if it were covered by a grid of 2 degrees by 2 degrees, or boxes that were more than 19,000 square miles each, which is roughly half the size of Kentucky. The computer model sees everything within the box as being the same, but of course no clouds are that big.
Today's models are better. And scientists hope to have a computer that's 1,000 times as powerful as those today by the end of this decade. That still won't be robust enough to deliver models as precise as desired, but they'll be closer than today's, Kinter said.
