Droughts, wildfires, heat waves, intense rainstorms—these are all extreme weather phenomena that occur naturally. But climate change is now increasing the frequency and magnitude of many of these events. Flooding in Paris and the Arctic heat wave are just two instances where climate change contributed to extreme weather in 2016—and there are many more examples.
Yet how do scientists know that global warming influenced a specific event? Until recently, they couldn’t answer this question, but the field of “attribution science” has made immense progress in the last five years. Researchers can now tell people how climate change impacts them, and not 50 or 100 years from now—today.
Scientific American spoke with Friederike Otto, deputy director of the Environmental Change Institute at the University of Oxford, about how attribution science works and why it’s a critical part of helping communities prepare for and adapt to climate change.
[An edited transcript of the interview follows.]
What exactly is attribution science?
Whenever an extreme event happens, usually people ask, "Did climate change play a role?" We aim to provide scientific evidence to answer that question within the news time frame—so within two weeks of the event occurring.
Are researchers trying to connect specific extreme events or patterns of extreme events to climate change?
They’re doing both. My work focuses on specific extreme events—for example, the last attribution study we did was on the heat wave in the Arctic that’s currently happening.
All extreme events have different forcings [factors that influence Earth’s climate], and one of the forcings can be climate change. With this research, we can now say an event of this magnitude has been made more or less likely due to climate change, or we can say what was once a one-in-150-year event in the past is now a one-in-50-year event.
How does attribution science work on a technical level? What kind of methods do researchers use, and what kind of evidence do they look for?
It's kind of like when you roll a die and you roll 10 sixes in a row. You become suspicious that something's wrong with that die, but from these 10 rolls, you can’t determine that it’s a loaded die. To be able to do that, you have to roll over and over again, do statistics on the numbers you roll, and only then will you be able to get an answer.
That’s the same way we study an extreme weather event. We simulate what is possible given the current forcings, and we figure out the likelihood of this event occurring in today's climate. Of course we don't have observations of the world that might have been without climate change—what would be the normal die. So we have to simulate that world by removing the anthropogenic warming from the climate models, or by doing statistical modeling on observations of the late 19th and early 20th century. Then we determine what would have been possible weather in a world without climate change.
If the likelihoods of the extreme weather event we're interested in are different [between these two worlds], then we can say climate change caused this difference. It can be a difference in intensity or frequency.
You said that your team can provide an answer within a news cycle. How is it possible to be so fast?
We use peer-reviewed methods, so we basically just apply the same methods used in previous studies to a new event. It's a little bit like doing a seasonal forecast. We also have all the model simulations done in advance, so that if an extreme event happens, we just need to do the analysis.
How do you separate global warming's effect from other factors, like El Niño?
In our experimental set-up, we simulate the event in today’s world, and then we remove anthropogenic emissions from the climate model's atmosphere, and do the same experiment again. So the only thing we have changed is the anthropogenic forcing. In the simulations, the world we live in and the world that might have been have the same large-scale patterns, like El Niño. So we’re asking, "Assuming everything else being equal, what is the influence of greenhouse gas emissions?"
How do scientists interpret extreme weather events where they find no climate change signal?
This result is not unexpected and not inconceivable. For example, we know that increased global mean temperature raises the risk of heat waves, just because the baseline is warmer. But it's not just the temperature that is influenced by climate change. For instance, in the recent two-year São Paulo drought, there was more precipitation and more evaporation. In the end, these two effects canceled each other out, and the drought risk itself did not change. Climate change had no impact on the overall risk of drought in this case—but it had an impact on the individual components.
We also look at what's the influence of climate change today. It may well be that for some events we don't see an influence today in a world that is one degree warmer. But just because we don't see a signal now, doesn't mean that this is an event that will never be affected by climate change.
Are scientists more confident of climate change's contribution to certain types of extreme weather events versus others?
There are events where we expect to see an increase, like heat waves and extreme rainfall. In particular, the signal is already quite large with heat waves. Other events are much more complicated. With droughts, for example, the feedback with the land surface plays a huge role, and the atmospheric circulation plays a much more important role. There are also events like hurricanes, where you need very high-resolution models to be able to say something about it—that’s a situation where the technology is just not there yet.
I’ve heard scientists say that 20 years ago they couldn’t answer the attribution question. What has allowed the field to advance?
The science really only came into existence within the last five years. We first had a technical breakthrough—you need to be able to simulate weather over and over again, and that was technically impossible even in the 1990s. Only in the 2000s did it become an option because of greater computing power.
Then in 2003 the methodology to do this kind of research was suggested—the idea that we could use advanced computing power to look at extreme events in this way. But it still needed some conceptual work. In the last five years, we really had a conceptual breakthrough.
What’s next for attribution science?
The next big challenge is to work on disaster-risk reduction, and on the impacts of extreme events. Because the question people ask is not, “What is the risk of three-day rainfall in Paris?” The question they ask is, "What's the risk of flooding in Paris?" And that depends not only on the meteorological event, but also on other factors, like the size of the river catchment, the management of the river, and all these aspects of vulnerability and exposure.
This may seem obvious, but why is attribution research important?
Three things: one is that we don't currently know very well what the actual impacts of climate change today are. We can predict the large-scale changes, but global average temperature increase does not kill people. What kills people are extreme weather events. This research allows us to get a more comprehensive picture of what climate change actually means.
Second, it provides scientific evidence to the public discourse. When extreme events happen, people ask if climate change played a role. Quite often in the past it has been a politician who has answered that question, and it was completely independent of any scientific evidence.
This research also allows us to make better planning decisions. When we know a drought is becoming more likely by a factor of 10 because of climate change, then we know we need to focus our adaptation efforts on that.
What extreme event in 2016 had the clearest connection to climate change?
The heat wave in the Arctic that’s ongoing. It has been made orders of magnitude more likely due to climate change.