Support Our Work
Blogs Section
Thoughts on everything from climate modeling to energy policy.

Pushing the Envelope of Climate Science ‘Attribution Studies’

Claudia Tebaldi

By Claudia Tebaldi

On August 17th and 18th, in the midst of a series of extreme summer weather events around the world, a meeting took place in Broomfield, Colorado, convening atmospheric scientists interested in the problem of event attribution, along with potential users of the attribution research information and experts in the communication of science to the larger public. I was fortunate to be among them.

Event attribution seeks to explain the complex causes behind a given weather event, be it an especially cold winter, an intense heat wave or a devastating flood, with the particular aim of detecting a possible departure from “normal” conditions, and the role – if any – that human activities played.  

The timing of the meeting, which was planned several months ago, could not have been better. In fact, I wondered if its main organizers, Martin Hoerling, a research meteorologist at the U.S. National Oceanic and Atmospheric Administration (NOAA), and Peter Stott, a researcher at the U.K. Met Office, had secretly discovered a new weather forecast method that enabled them to anticipate the climatic mayhem that supplied us with our compelling background.

The ability to confidently attribute specific extreme weather events to manmade or natural causes would constitute a major step forward in our ability to understand the effects of human-caused climatic changes, particularly in their regional manifestations, and would supply us with the knowledge that so many in the media world wish we had (and I use “we” here as in “we climate scientists”) when they ask us those dreaded questions, along the lines of “Is the Russian heat wave due to climate change?” or “Are the Pakistan floods due to our flooding the atmosphere with greenhouse gases?” You may want to read our recent post on this topic.

Image from the Landsat 5 satellite of flooding along the Indus River in Pakistan, near the town of Kashmor on August 12. Credit: NASA

So far, attribution research has resulted in studies of the causes of observed trends in global temperatures over recent decades, such as continental scale temperatures, large area-averaged precipitation trends, ocean temperature trends, long-term changes in atmospheric humidity and more. Using sophisticated computer modeling and high quality observations, we are able to say with great confidence that in these  changing aspects of our climate system, the fingerprint of human causes is already evident, as stated in the U.N. Intergovernmental Panel on Climate Change's Fourth Assessment Report.

You’ll notice, though, that I’m talking about broad trends, large area averages, and long-term changes. Such studies are very different from trying to attribute a measure of human causes to a one-off weather event, like the heat wave in Russia. The presence of high natural variability in the day-to-day weather makes it difficult to isolate a signal of systematic change, especially when we target extreme events with our forensic studies. Plus, the behavior of local climate is influenced by many causes that are difficult to measure, take apart, and simulate.

But science is always progressing and tackling the next challenge, and there are ideas, methods, models, and approaches out there that we are exploring. Studies are bringing us closer to being able to say something about the human factor in the complex mix of forces affecting weather events. You can read about one of the first studies, which represents just one of several approaches, that tackled the problem of attributing the increasing risk of an event like the 2003 European heat wave to the human-caused long-term warming of European summer temperatures.

The meeting in Broomfield this week was an occasion for scientists to compare notes on this nascent area of research, and allow us to forge alliances among a small community active in this area, particularly in the United States and the United Kingdom. It also gave us the chance to sit together with potential users of this information, and its mediators, to learn more about how such information can be used and communicated.

Composite image of vegetation stress from August 4-11, taken by the NOAA AVHRR satellite sensor. Note the high drought risk in western Russia, where wildfires were raging at the time. Credit: NOAA Visualization Studio.

Why is the notion of event attribution so compelling? It is not only to appease the media, of course.

We can already talk about how the latest extreme events fall in line with trends all over the world – those long-term trends of increasing temperature and increasing heavy precipitation that have been attributed to our impact on the climate system – and we can of course remind the various audiences that in the future, it is very likely that these events are going to be more frequent and intense.  

These are general statements that talk about the world as a whole, or at best large regions of the world, and do not pretend to say something specific about Moscow’s and Pakistan’s future heatwaves and floods, their intensity or frequency. An analysis that is able to drill into the specific events and their causes could not only make us more aware of the effects that our actions have on our local climate, but be the strong basis for saying something about what people on the scorched ground in Moscow, or feet deep in the water in Pakistan, should be prepared to face in a world where greenhouse gases are not kept in check (or not to face, if the result of event attribution concluded that the component from global warming was not significant).

The results from event attribution could thus speak much more cogently to needs for climate change adaptation and mitigation.

I heard some interesting thoughts at the meeting that I would like to close with. Much talk pertained to the educational value that a sustained activity of event attribution would have, in teaching us about our climate system and the factors affecting its behavior – not only human emissions but urbanization, changes in land use like deforestation and agricultural practices, and natural sources of climate variability like El Nino/La Nina (thanks to Eileen Shea of NOAA for speaking eloquently about this aspect). Related to this educational value was the observation that transforming information into clear and meaningful messages is not a straightforward task, and that is where expert communicators could help (thanks to Ed Maibach for making this point very effectively).

Another interesting thought was about our inclination as scientists to structure our studies so that we minimize the chances of saying that something “unusual” is going on if in fact nothing besides natural variation is going on. (This is known in statistical analysis as controlling the chances of a type 1 error, also known as a false positive). This cautionary choice though does not protect us from the opposite kind of error, and in fact makes its chances higher: we are more prone to conclude that nothing is going on, even when in fact the opposite is true (which is to say, our chances of a type 2 error, or false negative, may be high). But we know, at this point in time, with high confidence, that we are changing the behavior of the climate system. We also know that in many cases the error of not detecting that something has changed could have very dire consequences for populations or systems that would be better off adapting to the changing conditions. When – if ever – should we switch our scientific methods to minimize the false negative rather than the false positive? (Thanks to Kevin Trenberth of the National Center for Atmospheric Research for his thought-provoking words on this issue)

I don’t see us changing our methodology any time soon, and on the one hand my prudent side as a scientist – and in particular as a statistician-- is comfortable with that thought, but the same thought provokes me to seriously consider the dangers in this attitude, especially when one realizes that as we alter the earth more and more through our collective activities, it becomes more difficult to talk about what is natural climate variability (i.e. if modern human activities were not taking place on this planet) and what is “forced” behavior, in a system where the two are already inextricably mixed.


2013 Record Highs vs. Record Lows For the first time in 20 years, the U.S. is on track to set more record lows than record highs in 2013.

View Gallery