<mediar rid='m1'/>The greenhouse effect could be far more severe than experts had previously predicted, according to results from the world's biggest climate-modelling study. In the worst-case scenario, doubling carbon-dioxide levels compared with pre-industrial times increases global temperatures by an average of more than 11 ºC.

But as well as a predicting a bigger maximum rise, the project has also increased the range of possible temperature changes.

The results are the first from climateprediction.net, a project that harnesses the world's desktop computers to predict climate change. More than 90,000 people have downloaded software that uses the spare capacity of their computers to run global climate simulations.

A doubling of carbon-dioxide levels could eventually lead to an increase in worldwide temperature of anything between 1.9 ºC and 11.5 ºC, the project's researchers report in this week's Nature¹. That is a far greater level of uncertainty than the 2-5 ºC rise predicted by the Intergovernmental Panel on Climate Change.

The uncertainty is greater because climateprediction.net looks at more possibilities than previous models, explains the project's leader, David Stainforth of the University of Oxford, UK. Previous predictions of global warming have been based on just a few dozen simulations; Stainforth's team analysed more than 2,000.

The researchers cannot yet put a timescale on the temperature increases, although they suggest that extreme warming could take decades or centuries. Atmospheric carbon-dioxide levels, currently standing at 379 parts per million, are predicted to hit double their pre-industrial level of 280 parts per million midway through this century.

Policies aimed at keeping greenhouse-gas levels below a safe threshold may miss the point, says team member Myles Allen, a physicist at the University of Oxford. Uncertainty over global warming may mean that no such threshold can be determined; rather, we may need to keep cutting greenhouse gases for many years to come. "The danger zone is not something in the future," he says. "We're in it now."

Grand plan

Each simulation is a different version of a programme called a general circulation model. This model divides the globe into thousands of sectors, and estimates the future temperature based on certain assumptions such as cloud coverage, the rate of heat movement and rainfall rates.

Previous studies have included only the most probable values for these factors, whereas climateprediction.net's power has allowed the researchers to investigate two or three settings for each parameter.

The project's final predictions are based on the 2,017 simulations that were able to mimic the current climate. All predicted temperature rises. Most were about 3.4 ºC, the average value predicted by the Intergovernmental Panel on Climate Change; many were far more severe.

The researchers plan to improve their models, including a more sophisticated picture of how heat travels through the oceans, regional data and a more accurate picture of how temperatures will change during this century. "There's a huge database of which we've hardly scratched the surface," comments team member Mat Collins of Britain's Met Office in Exeter.

Meanwhile, they hope that more users will volunteer their spare computing power through climateprediction.net "There's lots and lots more to do," says Stainforth. 

University of Oxford

• References

  1. Stainforth, D. A. et al Nature 433, 403−406 (2005). | Article | PubMed |