In-depth: the scientific challenge of extreme weather attribution

This is a re-post from Carbon Brief by Roz Pidcock

Working out whether human activity is supercharging extreme events, such as floods, storms, droughts and heatwaves, is one of the youngest branches of climate science. But it’s moving at breakneck pace.

So much so, that the US National Academy of Sciences has fast-tracked a report, taking stock of the science and where it’s heading.

Event attribution is the field of science that asks if extreme weather around the world would look any different if we could replay the last 200 years or so, without human-caused greenhouse gases.

Today’s report is an overview, rather than a showcase for new results. And at about 150 pages long, it’s not a light read. But its weightiness is apt for a topic that has come to underpin climate conversations everywhere from flooding in the UK to climate change adaptation.

Carbon Brief has been speaking to key scientists in the world of attribution about how far the science has come, experimenting with communicating the nuances, and the thorny issue of making results public at lightning speed, often before peer review.

One thing is for sure, Dr Heidi Cullen, chief scientist at Climate Central and a contributor to today’s report, tells Carbon Brief:

The days of saying no single weather event can be linked to climate change are over. For many extreme weather events, the link is now strong.

‘A universal talking point’

Storms, droughts, heavy rain, heatwaves and other extreme weather events are of huge interest to society because of their often disastrous consequences for people and property.

As Prof Ted Shepherd, professor of climate science at the University of Reading in the UK, explains in a recent commentary in the journal Current Climate Change Reports:

Just as weather is a topic of daily conversation, extreme weather events…provide a universal talking point.
A woman and a dog are rescued from flood waters in Chennai, India, 04 December 2015.

A woman and a dog are rescued from flood waters in Chennai, India, 04 December 2015. © BABU/epa/Corbis

Shepherd is one of the nine authors behind today’s report. The authors are scientists who are knowledgeable about attribution, but aren’t involved in the research they’re assessing.

When a flood, drought or heatwave hits, it has become normal to ask, what role did climate change play? And this is a question science is increasingly able to answer, through a branch of science known as “attribution”.

The concept of attribution has been around for a while. The famous statement in the 1995 Intergovernmental Panel on Climate Change (IPCC) report that “the balance of evidence suggests that there is a discernible human influence on global climate”, was based on the early crop of attribution studies published in the 1990s.

A huge body of literature in the intervening years has seen that statement strengthen enormously. The latest IPCC report said it is “extremely likely that human influence has been the dominant cause of the observed warming since the mid-20th century”.

But what’s causing the new flurry of excitement in the attribution community is grounded much more on people’s everyday experiences of climate change. It’s far more visceral, more immediate.

It’s the growing ability of scientists to work out how our own actions might be influencing the size, severity or frequency of extreme weather – often while an event is still going on.

Prof Myles Allen, head of the climate research programme at Oxford University’s Environmental Change Institute, first proposed the idea of extreme weather attribution in a 2003 paper. His basic premise was that scientists could use climate models to work out what proportion, if any, of the risk of a given extreme weather event could be pinned on human activity.

This is not the same as asking whether climate change “caused” an event to occur. Today’s report calls this framing a “poorly formed (or ill-posed) question that rarely has a scientifically satisfactory answer”. And as Shepherd explains in his review paper:

If a weather or climate event is truly extreme in the present climate…it requires unusual meteorological conditions, which means that climate change is at most only a contributing factor.

The point of extreme event attribution is to work out how big that contribution is and how it compares to other factors, such as natural fluctuations in the atmosphere and oceans.

Number crunching

So, how do scientists actually do attribution?

Typically, scientists use climate models to create a fictional world in which the industrial revolution never happened. In this world, the climate is influenced solely by changes in the Sun’s activity and volcanic activity. Then, they compare the chances of a particular event occurring in the fictional model compared to the version that best represents the real world.

Scientists tend to combine model studies with other methods, to improve their confidence in their findings. Statistical analysis of the historical record gives an idea of how unusual an event is, but this requires extremely good data going back a long way, which isn’t always available.

Just as the question of what role climate change played in an event can be asked in different ways, all equally valid, the results can be expressed differently, too.

Climate change made the exceptional amount of rainfall unleashed on the UK by Storm Desmond in December 2015 40% more likely, according to analysis by Allen’s group at the University of Oxford, together with the Royal Netherlands Meteorological Institute and Climate Central in the US. A follow-up study showed climate change had increased the odds of such a wet December in the UK by 50-75%, while separate research found climate change had made heavy rainfall in the winter of 2013/4 – another particularly wet period – 43% more likely.

The different percentages attached to these studies reflect the fact that attribution studies are done on a case-by-case basis, with no two events being identical.

But sometimes you’ll see the results presented another way. The very first study to apply extreme weather attribution in 2004 expressed the effect of climate change as having at least doubled the risk” of a heatwave as severe as the one Europe experienced in 2003. This could equally have been expressed as a 100% rise in the probability – they amount to the same thing.

study published this week expressed its results slightly differently, again. The authors found that climate change had raised the odds of extremely high temperatures in each of the last 16 record-breaking warm years, leading them to conclude that “without human-induced climate change, recent hot summers and years would be very unlikely to have occurred”.

Yet another way that attribution studies can express their findings is through a change in the frequency of a particular event occurring, rather than the magnitude it reaches.

For example, a 2015 study which found that extremely warm years, such as 2014, are 13 times more likely to occur in the Central England Temperature (CET) record under climate change also concluded that such events could be expected about every five years in 2020 compared to every 120 years in 1914.

Different ways to express the human influence on July heatwaves in Russia

Different ways to express the human influence on July heatwaves in Russia, in terms of the return time/frequency (top) or increased probability (bottom). Each figure compares the model result for the 1960s (green) and 2000s (blue). Source: Adapted for NAS Attribution of Extreme Weather Events in the Context of Climate Change report from Otto et al., (2012).

Need for speed

How authors of attribution studies choose to express their results comes down, at least in part, to a choice about what’s most effective, says Dr Peter Stott, head of detection and attribution at the UK’s Met Office and author of the 2004 study on the European heatwave.

Attribution has grown out of a demand for answers to specific questions, he says, and expressing results in different ways is part of an evolving understanding of what people want to know. The attribution community recognises how effective attribution can be as a communication tool, to connect climate change with people’s everyday experiences. Stott adds:

There is a power to be able to make statements while they’re still fresh in people’s minds.

In the case of the heavy rain during Storm Desmond and the heatwave across Europe last summer, scientists were able to turn the handle on the attribution studies while they were still going on. Cullen, whose research group was part of the team behind both studies, tells Carbon Brief:

Despite profound changes to our climate system, global warming and its associated impacts – sea level rise, ocean acidification, species loss – still feel distant in both time and space to many people. The most immediate and costly threat…results from the changing risks of extreme weather.

Cullen has an Op-Ed in the New York Times today, in which she compares today’s report linking climate change to a higher risk of some types of extreme weather to the surgeon general’s 1964 report connecting smoking to lung cancer. She writes:

Climate change can no longer be viewed as a distant threat that may disrupt the lives of our grandchildren, but one that may be singled out as a factor, possibly a critical factor, in the storm that flooded your house last week.

One important forward-facing use for attribution research is in prediction. Today’s report explains how it can be used to help vulnerable communities cope with climate impacts, present and future:

In the wake of a devastating event, communities may need to make a decision about whether to rebuild or relocate. Such a decision could hinge on whether the occurrence of an event is expected to become more likely or severe in the future—and if so, by how much.

Allen says scientists could use attribution predictions in the same way that meteorologists use weather forecasts. He told Carbon Brief recently:

My personal view is that, in the long term, attribution should be a routine part of any package of climate services, so a quantitative assessment of how various external drivers may be making weather events more or less likely to occur should become just part of the job of the world’s meteorological services.

For extreme event attribution to be useful in any of these contexts, time is of the essence. So, does the need for speed mean a compromise on quality?

Low waters of Lake Mead seen from above in this aerial view via helicopter

Oct 23, 2015 – Boulder City, Nevada, U.S. – Low waters of Lake Mead seen from above in this aerial view via helicopter. Years of unrelenting drought are straining a large reservoir of water between Nevada and Arizona. Lake Mead’s water level has dropped by about 120 feet (37 meters) from where the water reached 15 years ago, on July 6, 2000. © Ruaridh Stewart/ZUMA Press/Corbis

Quality control

The methods that underpin extreme event attribution are thoroughly peer reviewed, such that the scientists are just turning the same handle each time. Allen says:

Individual seasonal forecasts aren’t subject to peer-review before they are issued, but seasonal forecasting methods certainly are…I think we should see attribution results in much the same way.

“Effective, rigorous, and scientifically defensible” analysis of how climate change is implicated in extreme weather not only offers valuable information about future risks, but also satisfies the public’s desire to know, says Rear Admiral Dr David Titley in his introduction. Titley is a professor at Penn State university and chair of the author committee for today’s report.

And as long as those boxes are ticked, it’s usually better to say something than nothing at all, says Dr Friederike Otto, a senior researcher at the Environmental Change Institute who has been behind a lot of important recent attribution research. She told Carbon Brief recently:

As long as we make clear what we do, what our assumptions are, how we define the event, etc, it would probably not be bad for the public debate around the event if someone like us – in the sense of the attribution community – who can provide scientific evidence gives these answers, even if preliminary.

Someone will step in to answer these questions and it’s better that it’s the scientists actually doing the work, says Allen:

We currently get a lot of qualitative hand waving about how various drivers may have contributed, leaving the public pretty much in the dark (or worse, guided by which papers or websites they read rather than the evidence) about which drivers are most important.
Storm Desmond flooding newspaper coverage

Left: The Times, 8 December 2015; middle: Daily Mail, 10 December 2015; right: The Daily Express, 11 December 2015.

Quickfire peer review

After an initial outing in the media or policy circles, attribution results will often go on to be published in peer reviewed journals. Studies are emerging at such high speed that for the past five years, the Bulletin of the American Meteorological Society (BAMS) has published an annual special issue examining the human contribution to extreme weather events in the previous year.

The fast turn-around between an event happening and a study being is another area where event attribution is, arguably, revolutionising climate science. Stott, one of the co-editors of the BAMS report, tells Carbon Brief:

We’ve shown that you can do it. You can do rigorous, peer-reviewed science to this pre-set, short timetable.

But with so much at stake with extreme weather, the attribution community acknowledges it could up its game on joining the dots between different studies. Stott adds:

It seems to me that what hasn’t happened in the past is if you get an individual study that communicates its results in a particular way and another study about the same event communicates it in a different way, they can seem contradictory…What’s lacking there is a synthesis. There’s the potential to do a lot better.

Model limits

An important point today’s report makes about attribution studies is that the results are only as reliable as the models that underpin them.

If a model inadequately represents an aspect of climate system, or a process is not well understood, working out whether the probability of an event has changed over time won’t be grounded in reality either.

This means attribution studies tend to be, at least for the time being, reliable for some types of event, but not others. Heatwaves and, to some extent, extreme rainfall events, are well-understood because they relate directly to how much extra heat the atmosphere has accumulated.

Things get more complicated for types of extreme weather that involve the chaotic and complex dynamics of the atmosphere or oceans. This includes hurricanes (also called cyclones or typhoons, depending on which part of the world you’re in), floods and often droughts.

confidence in capabilities for attribution

Scientists have more confidence in attributing some events over others. Confidence is high for events in the top right corner and low for those in the bottom left corner. Source: National Academy of Sciences Attribution of Extreme Weather Events in the Context of Climate Change, March 2016.

The passage of a storm or the onset of a drought have a large element of chance, which can make it tricky to identify where climate change fits in.

Hurricanes pose a particular challenge, says Prof Gabriel Vecchi, head of the climate variations and predictability group at the National Oceanic and Atmospheric Administration (NOAA). There can be a palpable sense of frustration when scientists can’t answer those questions straight after an event, he tells Carbon Brief:

To really come up with a good answer, we do need to take our time…These questions are difficult…It may take more than a news cycle to come up with a good answer.

But it’s important that science pursues those hard-to-get answers, as vague statements aren’t enough. He says:

It’s true – and will always be true – that the world is now fundamentally different because of global warming, so everything we see is impacted by global warming…[But that] has very limited utility in meaning. It doesn’t help predict anything…We need to be much more specific and much more quantitative.

The good news is that the momentum behind event attribution research is heading in that direction, says Prof Gabi Hegerl, professor of climate system science at the University of Edinburgh and another of the authors of today’s report. She tells Carbon Brief:

We have fairly good capabilities on hot and cold events now…Moving forward, the more complex events are a more interesting and a more important frontier.
People ride their motorcycles over an empty irrigation canal in Lopburi province. Normally the canal is completely full at this time of year.

July 1, 2015 – Nong Phak Nak, Suphan Buri, Thailand – People ride their motorcycles over an empty irrigation canal in Lopburi province. Normally the canal is completely full at this time of year. © Jack Kurtz/ZUMA Wire

Proceed with caution

Scientists select events to study based on where the interesting questions are, not whether they think climate change is most likely to have had an effect.

For example, some studies have found no discernible impact from climate change on the odds of a particular extreme event happening. Rising temperatures are also making some types of events, such as extreme cold snaps, less likely.

But while the body of literature of extreme weather attribution is growing rapidly, there are still big gaps where data and resources just don’t exist. Allen says:

Clearly, we don’t have equal capabilities everywhere, although we are working with partners in vulnerable regions of the world to improve that, so right now there is obviously a bias towards our own backyards: northwest Europe, Australia and New Zealand.

This is partly why the back catalogue of attribution research shouldn’t be used as a metric to gauge the impact of climate change on extreme weather per se. Hegerl tells Carbon Brief:

Going to, for example, the BAMS report and counting how many hot and cold events are attributed to human influence, or how many wet or dry events, would give you a biased picture…There is a bias in what we can do in terms of region and event type.

Today’s report issues a few other cautions for interpreting attribution studies, too.

Determining how unusual an event is will depend a lot on how you define it – over what period of time, or how large an area, for example. And if a study is based on observational data, the historical period you choose to represent a time before climate change will make a difference too.

Direction of travel

In terms of direction of travel, another hot topic is how attribution could play a role in assisting nations that have suffered or are suffering irreversible loss and damage from climate change. Dr Dáithi Stone, a research scientist in attribution of extreme weather at Berkeley Lab in the US, tells Carbon Brief:

While the agreement coming out of the meeting in Paris last December was very explicit that any assistance should not be considered compensation or reflect any liability, one might expect that nations that can demonstrate solid evidence of impacts may be able to submit a stronger claim for assistance.

Bringing a similar theme down to a person rather than country level, there could be a role for extreme weather attribution in climate litigation – that’s to say, seeking damages from a company or government whose actions can be shown to be contributing to climate change-induced harm. Indeed, this is the picture Allen’s original 2003 paper paints, carrying the title: “Liability for climate change: Will it ever be possible to sue anyone for damaging the climate?”

In terms of where today’s report fits into all of these conversations, it’s essentially a stocktake of where we are in the fast-moving field of attributing extreme weather to climate change. Alongside the progress to date, the need for a strong evidence base, and the myriad ways attribution could be put to use, perhaps the point the report stresses most heavily is the need to be clear about the question being asked. As Shepherd tells Carbon Brief:

Being clear about the question being asked, and asking a well-posed question, is absolutely crucial in extreme event attribution…The question should be “What role did climate change play in the event?”. But even that question can be asked in different ways.

Or as Rear Admiral Titley says in his opening gambit to today’s report:

Like all areas of study, terminology matters.

National Academies of Sciences, Engineering, and Medicine. Attribution of Extreme Weather Events in the Context of Climate Change. Washington, DC: The National Academies Press, 2016. doi:10.17226/21852.

Posted by Guest Author on Tuesday, 22 March, 2016

Creative Commons License The Skeptical Science website by Skeptical Science is licensed under a Creative Commons Attribution 3.0 Unported License.