How much is climate change to blame for extreme weather?

The science of risk

by Maggie Koerth-Baker

A lot of journalists think the public can’t possibly comprehend the complicated probability and risk assessments associated with climate change. Myles Allen thinks they’re wrong. There’s not much difference between explaining how climate change could contribute to a specific weather event and explaining how smoking for 50 years could contribute to developing lung cancer, he says. People understand what you mean when you tell them that smoking doesn’t always cause cancer and isn’t likely to be the only reason a cancer happens. Likewise, they can understand you when you tell them that climate change isn’t the cause of every weather disaster but is a contributing factor to many of them.

“The bigger problem is that the public has lots of things to think about and limited bandwidth, particularly for events far in the future or far away in the world,” Allen says. “But when an event affects them, personally, people are more than willing to focus on the role of human influence.”

Allen should know. It’s his job both to explain the role of human influence on weather events and to ascertain exactly what that role is. A researcher at the University of Oxford in the U.K., Allen is the creator of a branch of climate science—event attribution—that turns much of the discipline on its head. Instead of using patterns from the past and computer modeling to predict how climate might change over the next 50 or 100 years, Allen takes current weather events—a heat wave in France, a flood in Wales—and tries to determine whether those disasters had anything to do with anthropogenic climate change. If so, he figures out how much blame climate change should shoulder, compared to the other factors at play.

Allen isn’t simply hoping to change the way people talk about climate change. His work could also have a direct impact on how climate change affects us today. That’s because understanding the risks of weather events is an integral part of how insurance companies evaluate risk in the near future and thus is an important factor in deciding what you and I pay for our insurance premiums.

Allen’s research wouldn’t have been possible as recently as ten years ago. It requires too much computing power. To do single-event attribution, Allen and his team create virtual worlds over and over again. In a world where humans haven’t pumped greenhouses gases into the atmosphere for 100 years—where the only forces affecting climate and weather are natural ones—what happens? In a world much like our own, where anthropogenic climate change is a factor, what happens? How likely is a specific weather event?

The first time Allen tried do event attribution, by analyzing the causes behind a particularly bad spate of fall floods in the U.K. in 2000, it took him six years and 50,000 simulations with the help of a distributed computing project. That’s 50,000 alternative histories, some with floods and some without. Even then, he wasn’t able to say anything nearly as definitive as “These floods were caused by climate change.” Instead, Allen was still in the world of probability; he was able to say that, nine times out of ten, a world where anthropogenic climate change exists is 20 percent more likely to experience floods than a world with no climate change.

That might seem like a complex and not particularly useful answer. But without single-event attribution, Allen points out, scientists are left with nothing at all to say about a weather event beyond hand-waving statements such as “We might expect this sort of event to be more common under climate change.” However difficult it might be for someone to understand what “20 percent more likely in nine out of ten scenarios” means, at least it has meaning. That was the main reason why Allen set out in 2003 to develop a method for doing single-event attribution. “I think we should be able to do better than [vague statements], and we should be able to be more specific,” he says.

The insurance industry has always calculated risk by looking at the past. That’s why you pay more for flood insurance if your house is located in a place where flooding has happened historically; how much more you pay is connected to the historical frequency of the flooding. The trouble with climate change, though, is that trends of the past no longer predict what might happen in the future.

But that’s a problem traditional climate science can’t really solve, says Robert Muir-Wood, the chief research officer at Risk Management Solutions, a company that helps the insurance industry quantify risk. To do their jobs, insurers need to know which specific events are likely to happen—and what’s likely not to happen—on a five- or ten-year horizon. The kind of climate science that focuses on change decades into the future isn’t much help. Instead, Muir-Wood and his colleagues have begun to adapt the models they use to predict risk so as to include information about what’s happening now. So far, Myles Allen’s research hasn’t yet made its way into these calculations. But he and Risk Management Solutions are collaborating closely, banking on their conviction that single-event attribution will get faster and better.

And it is. This year, Allen’s team was able to tackle the sort of question that had once taken them a decade to answer, coming up with a meaningful risk assessment in just one month. They were able to say that climate change increased the risk of a recent flood by 25 percent. That’s not quite real-time evaluation of the role of climate change in natural disasters, but it’s getting damned close. In the meantime, Allen’s work can help improve the conversation about climate change by making clearer exactly how climate change affects us—and how it doesn’t. “It’s important to manage expectations,” he said. “One of the more interesting consequences of getting a bit more scientific is that we find attribution is more or less evenly split between positive and negative results. Some activists would like to blame any instance of bad weather on climate change, but we’re finding that’s also wrong.”

Maggie Koerth-Baker is the science editor at BoingBoing.net. She writes a monthly column for the New York Times Magazine and is the author of Before the Lights Go Out.