Friday, February 18, 2011

Blame human emissions for British floods


Al Gore famously had his knuckles rapped for implying that human-induced climate change had caused hurricane Katrina. The scientific party line then was "No single weather event can be attributed to climate change". It's a line that has held strong but is beginning to fray.

This week, a study has shown that the devastating floods which damaged nearly 10,000 properties in England and Wales in 2000, and cost £1.3 billion in insurance losses, were made significantly more likely by climate change caused by humans.

It is the first study to quantitatively link a severe rainfall event and climate change. The team that carried out the work, led by Myles Allen of the University of Oxford, had earlier linked the 2003 European heatwave to climate change.

The team's method looks simple at first glance. They run climate models over and over again, thousands of times, to predict the likelihood of a specific event happening – in this case, the floods that affected the UK in 2000. This is done for two different scenarios: first, a "realistic" scenario in which the mix of gases in the atmosphere is a representation of the atmosphere in 2000; and second, a world in which humans do not and have never produced greenhouse gases.

By comparing the two, the team is able to say how greenhouse gases changed the likelihood of the floods happening as and when they did.

Soggy science

The study of the 2003 heatwave, which used similar methodology, found that greenhouse gases made the event two to four times more likely to happen.

Modelling flooding is quite a bit harder than modelling a heatwave, however. Predictions can't just take into account how much rain might fall: they also have to account for things like how waterlogged the soil was, and where the rainfall flowed.

To complicate matters further, the level of detail – in terms of both space and time – in global climate models is too coarse to allow the simulation of individual rainfall events. The models can make only broad statements about how rainfall is likely to change regionally over the course of decades or a century.

To overcome these problems, Allen and his team combined a seasonal weather-forecasting model with a rainfall run-off model, which predicts the flow of water over land.

Then they harnessed the number-crunching power of the thousands of personal computers that have signed up to Climateprediction.net. In this "citizen science" experiment, people offer up slack capacity on their computers to Allen's team, who use it to run their climate models in the background. Nearly 55,000 people are currently signed up.

Cause for confidence

The bottom line of all this? Allen and his team found that human greenhouse gas emissions "significantly increased" the likelihood of the 2000 floods. They can say, with a 66 per cent confidence level, that emissions nearly doubled the risk of the 2000 floods.

Conversely, says Allen, there is only a 10 per cent chance that the increase in flood risk rose by just 20 per cent as a result of human contributions to climate.

Allen's findings tally with global predictions, such as those offered by the Intergovernmental Panel on Climate Change, that climate change will raise the risk of extreme rainfall in regions that are already wet – such as England and Wales. Another study published this week shows that this is already happening globally (Nature, DOI: 10.1038/nature09763).

As for Katrina, Allen points out that he is not a hurricane modeller but is encouraging his colleagues who are to consider this approach. In the meantime, however, his team has launched an experiment called Weatherathome, which will use distributed computing to model all weather events, year after year.

"It's important not to look just at the weird events," says Allen.

A bigger collaboration between the British Met Office and National Oceanic and Atmospheric Administration in the US began last year, also with the aim of boosting our ability to link specific events to climate change.

Journal reference: Nature, DOI: 10.1038/nature09762

No comments: