Did X Cause Y? A New Look at Attributing Weather Extremes to Climate Change

In a world filled with high-impact weather events, it’s only natural to wonder exactly why your town was beset with a heat wave, a destructive flood, or a deadly tornado. Today, such events occur in a different global atmosphere--one with more greenhouse gases than at any time in human history, thanks to human activity. A growing branch of atmospheric research is working to quantify the influence of human-induced climate change on various types of extreme weather, and there is real progress being made. “It is now possible to estimate the influence of climate change on some types of specific weather events,” said Rear Admiral David Titley (Pennsylvania State University) at a press briefing in Washington, D.C., last Friday. Titley chaired a U.S. National Academies committee that has just produced an important report, released on Friday. Attribution of Extreme Weather Events in the Context of Climate Change serves as a very useful guide to how this work is carried out, what it can and can’t do, and where the science is heading.

The idea behind attribution research is to provide reasonably satisfying answers to the query so often raised by policy makers and the public: did climate change have anything to do with this event? For years, scientists rightly pointed out that a changing climate doesn’t “cause” any particular event. Often, they added that it was impossible to know exactly what role climate change might have played in a particular weather happening, apart from basic conclusions about how the physics of a warming atmosphere should make certain events increasingly more or less likely.

Things are different now, as pointed out by Titley, the founding director of PSU’s Center for Solutions to Weather and Climate Risk. Attribution research, said Titley, “makes the future of climate real. It brings the future into the present.”


Figure 1. A schematic depiction of the National Academies assessment of the state of attribution science for different event types. The horizontal position of each event type reflects the level of understanding of the effect of climate change on that event type, and corresponds to the rightmost column in Figure 2. The vertical position of each event corresponds to scientific confidence in current capabilities for attributing specific events of that type to anthropogenic climate change; this vertical position draws on all three columns in Figure 2. In all cases, there is potential to increase event-attribution confidence by overcoming remaining challenges that limit the current level of understanding. Image credit: National Academies.


Figure 2. An overall assessment of the state of event attribution science for various event types. In each category, the committee has provided an estimate of confidence, based on the available scientific literature and the results of committee deliberation and judgment. Image credit: National Academies.


A spectrum of understanding
While this report has some fairly technical content, including heavy-duty statistical concepts, most of the terminology is well explained up front. This follows in the tradition of reports from the National Academies that speak to a broad, policy-interested audience on emerging science topics of keen public interest. It was high time for this particular report, said committee member Marshall Shepherd (University of Georgia), a WU contributing blogger and host of the Weather Channel’s popular “WX Geeks” series. In the headline of a Forbes commentary published on Friday, Shepherd heralds “The Death to One of the Most Abused Questions Ever: Was That Caused by Climate Change?” He cites this question as “so abused by ALL sides of the climate discussion.”

The ideas underlying the new report are summarized nicely in Figure 1, which places a variety of extreme weather events on a twofold spectrum: how well the influence of climate change is understood, and how confidently we can attribute specific instances of that event to climate change. Figure 2 elaborates on Figure 1 for each event category. As one might expect, our confidence in attribution goes up as our understanding increases--but not all events line up on this straight line. Tropical cyclones are a good example. There is a growing body of research concluding that we can expect the frequency of the most intense storms to rise in many parts of the world, a finding reiterated by the IPCC in 2013. Several studies in the last decade have suggested this trend is already under way, especially in the Atlantic. Yet tropical cyclones remain rare enough that it is tough to disentangle natural variability and observational uncertainty from the actual impact of climate change. Top global models are just now gaining the skill and resolution to simulate trends in Category 4 and 5 hurricanes over centuries, or to embed individual storms such as Hurricane Sandy into global-scale models that replicate the climate of the distant past and our greenhouse-warmed future.


Figure 3. Three potential ways in which a warming climate could affect heat and cold extremes: (top) shifting the entire distribution of observed temperatures toward a warmer average; (middle) increasing the variability of the distribution; and (bottom) skewing the curve in one direction or the other. Image credit: National Academies.


Heat waves and cold waves provide the best-understood and most-straightforward links: as climate change proceeds, we can expect more extreme heat and less extreme cold, with occasional exceptions here and there. In many areas, we have a century or more of good-quality temperature data, and this helps smooth the way toward making solid attributions. “Heat waves and cold waves may be the best candidates for assessing the reliability and robustness of attribution methods,” says the report.

Severe convective storms--intense thunderstorms and the hail, wind, and tornadoes they spawn--rank lowest on the spectrum in Figure 1. That’s not to imply that these storms are necessarily immune to the influence of greenhouse gases. The point is that (a) such influences are not yet crystal clear, in part because of sampling and observing issues, and (b) the huge natural variability in severe weather makes it hard to pluck out a climate-change signal for a particular storm. There is intriguing progress being made in these areas, though, some of which I’ll be covering in a future blog based on a meeting I attended last week at Columbia University (the 2nd Workshop on Severe Convection and Climate).


Figure 4. A man walks along the heavily damaged beach on November 2, 2012, in the Rockaway neighborhood of Queens, New York City, following the devastating arrival of Hurricane Sandy. Image credit: Spencer Platt/Getty Images.

Asking the right question
In order to study an attribution problem in a useful way, one needs to know exactly what the goal is. “Statements about attribution are sensitive to the way the questions are posed and the context within which they are posed,” the report notes. “For example, a scientific researcher might re-pose the question ‘was Hurricane Sandy caused by climate change?’ as ‘by how much did human influence on climate increase the odds of a tropical or post-tropical storm with winds greater than 65 knots making landfall in northern New Jersey?’”

Once a well-posed question is in hand, there’s a growing toolbox of techniques for researchers to draw on, amply discussed in the report. Many attribution studies include a mix of observations and model-based simulations, with the goal of determining how much more likely a certain type of event has become (such as a heat wave of a particular strength over the Midwest) due to the presence of human-produced greenhouse gases.

The most familiar source of attribution research for many of us is the dozens of studies compiled each year since 2012 in special issues of the Bulletin of the American Meteorological Society. These studies are useful in many ways, the report notes, but they shouldn’t be taken as a collective assessment of how climate change affects the whole gamut of extreme weather. For example, there is a selection bias at work. These reports focus on the highest-impact weather of a given year, so if a particular type of event has become increasingly rare (say, mammoth citrus-killing freezes in Florida), it’s much less likely to get examined.

Why the cookie crumbles
I helped take a very informal stab at an attribution spectrum in 2012 while part of the communications group at the University Corporation for Atmospheric Research. For this special project, we found ourselves touching on some of the questions of framing, phrasing, and methodology that are dissected much more thoroughly in this report. It’s encouraging to see how much the discipline has been growing, and how much vital knowledge it will be able to generate with the appropriate level of support and focus. The National Academies report should go a long way toward the “mainstreaming” of attribution research. It will also help journalists, educators, and others who need to put this often-esoteric science into terms that people can easily grasp. The links between greenhouse gases and extreme weather are far too important and intricate to be dismissed or broad-brushed.

Consider this tasty analogy provided by David Titley: we can think of a freshly-baked cookie as a weather event, with the ingredients being the factors that aligned to cause the event (i.e., large-scale atmospheric features), and the baking surface and the oven temperature representing conditions in which the event occurred (i.e., increasing greenhouse gases, decreasing sea ice). “When you bite into a cookie and it doesn’t taste right,” said Titley, “it can be hard to determine what went wrong.”

You can download a PDF of the full report at the National Academies website. Jeff Masters will be back on Thursday with coverage of NOAA’s highly anticipated report on global climate for February. See our report from this past weekend on NASA’s blockbuster data.

Bob Henson