 Forecast The Future From Your PC
|
Sept. 12, 2003 — The power of millions of personal home computers around the world is being harnessed to help forecast the climate for the 21st century and improve models of global climate change.
A new experiment in distributed computing called Climateprediction.net will be launched this week in London after being unveiled at the British Association for the Advancement of Science conference in Manchester.
In the experiment, people at home download a unique climate simulation, based on a model developed by the U.K. Met Office, and run it in the background on their personal computer. Each Windows-based version of the model, distributed to home computers, has a different set of parameters and runs for several months before returning data to a central server.
|
 
 |
|||||||
Climate change predictions have to date been plagued by inadequate estimates of their uncertainty, partly because limited computer power has meant that only one or two parameters could be varied at one time. This new system hopes to quantify more exactly the uncertainty of different climate scenarios, by running many more simulations at once than could be done on any single supercomputer.
"This is a way of accessing computing power which is absolutely impossible in the normal context," said co-developer Myles Allen from Oxford University.
Allen, a member of the Intergovernmental Panel on Climate Change (IPCC), said the panel had in the past been criticized for failing to assign a probability to its predictions about temperature rise. It had predicted that increased greenhouse gases would cause a rise in global temperature of between 1.4 and 5.8 degrees Celsius in the 21st century.
"Next time around, in 2007, we want the next IPCC assessment to be a systematic assessment of uncertainty in our long range forecasts," he said.
Up to 100,000,000 different variables (just one being, for example, how clouds reflect light) are now included in climate models. Each needs to be assigned with a range and a level of uncertainty, requiring vast processing power.
"You need multiple runs of similar models to get a handle on the possible range of results," Allen said. "We had one ... tester whose model went into a super greenhouse-ice age effect which I'd never see before — and a very rapid one."
To avoid being accused of adding to greenhouse gas emissions themselves, the project managers are restricting the number of users who can participate in the project. "We would stop at two million because then the actual greenhouse gases emitted by the participants would start to contribute to the problem, so we feel that's the limit!"
The climateprediction.net website is a collaboration between the University of Oxford, the University of Reading, the Met Office, the Open University, the CCLRC Rutherford Appleton Laboratory and Tessella Support Services.
Another high profile distributed computing project was the SETI@home project in which people help search the Universe for signs of extraterrestrial intelligence.
< news main
