London

Future in their hands: children at Gosford Hill school in Oxfordshire examine a simulation. Credit: S. BEBB

A major source of uncertainty in models of climate change could be reduced by a project utilizing the spare power of desktop computers.

The backers of climateprediction.net, due to be launched on 12 September, are inviting the public to download software that will help to probe the assumptions used in sophisticated models of Earth's atmosphere. By running thousands of these models simultaneously on home and office machines, researchers hope to identify the ranges of assumptions that are realistic, and hence to generate better predictions of future climate.

Climate models use rules to describe how large-scale factors such as temperature and humidity determine smaller-scale properties such as cloud cover. A range of rules are compatible with observations, so researchers pick the ones that most accurately make the models follow the course of climate in recent decades.

According to Myles Allen, a climate researcher at the Rutherford Appleton Laboratory near Oxford and the originator of climateprediction.net, it is important to explore the full range of possible rules in order to understand the range of realistic future-climate scenarios.

With time on climate supercomputers scarce, this has not been done. But other big tasks, such as the analysis of extraterrestrial radio signals for signals from intelligent life forms, are already being performed by parcelling out the task to large numbers of personal computers. “Distributed computing projects have attracted hundreds of thousands of participants,” says Allen. “Such an ensemble size would form an extraordinary experiment.”

Slightly simplified versions of a model that has been developed at the Hadley Centre for Climate Prediction and Research in Bracknell, UK, will be available for downloading from the project website. Each model uses a unique set of rules and simulates a 15-year period, during which the stability of the rules is checked. It then simulates 15 years of current climate conditions, and finally a world with increased greenhouse-gas concentrations. This should produce better estimates for a range of parameters such as temperature increase, and allow scientists to single out the rules for determining the small-scale processes to which climate is most sensitive.

Allen aims to generate publicity at this week's launch, which takes place at the Science Museum in London and coincides with the British Association's annual science festival. The first phase of data collection should be completed by the end of the year, he says, and the first results could be available early in 2004.

→ http://www.climateprediction.net