An argument preferring recurrence questions to attribution questions
We have had a very severe weather event in South Carolina. As I write it is still an extremely serious situation. Before taking off onto the science and implications, it is worth taking note of the human suffering and environmental damage. There have been 14 reported deaths and enormous property damage, much of it to residences without flood insurance, bringing catastrophic and unexpected expense to many households.
Even now, before the rivers have stopped rising, the usual Disaster Tango has ensued, with the dance partners partners each dancing to a different tune. The tunes were “climate change caused it” / “it has nothing to do with climate change”. This dance inevitably follows a severe weather event, especially in the USA. Little is achieved by it.
Both positions are very problematic in my opinion.
We have to look for better ways to talk about these sorts of things, and in my opinion the only way to do that is to stop pointing vaguely at science and instead develop the patience to actually think about the situation. Science is commonly used as a weapon in a polemical battle, but that’s abusing it.
There are reasons to argue from authority, but every severe event begins as an open question. Open questions give us the opportunity to think about science as a rational community of interest, rather than as a pair of competing debate teams.
We should be looking at what science says, and using it to bring the conversation closer to reality. Instead, each side picks their own evidence and uses their favorite points as a cudgel. That may be how politics is done. But it’s not how we attain to a world that is informed by reality.
If you need to quickly decline to dance the “did not/did too” dance, Kevin Trenberth et al have provided a sensible escape:
“The climate is changing: we have a new normal. The environment in which all weather events occur is not what it used to be. All storms, without exception, are different. Even if most of them look just like the ones we used to have, they are not the same. ”
To that, Andy Revkin raises this objection:
“In this piece, Kevin Trenberth’s insistence that greenhouse-forced climate change is in every extreme weather event has no meaning without some quantification. The same could be said for every sunny unremarkable day, as well.”
Of course it is true for every event, remarkable or otherwise, that it is a part of a changed climate! That is not a refutation of Trenberth because it is exactly what he claimed.
To be fair to Revkin, Trenberth is looking at the attribution question itself, while Revkin’s attentions are further down the policy chain. After all, if the world’s climate were to somehow become more benign (however you choose to define that) it would still be a climate change but wouldn’t imply any policy change.
Trenberth is however responding to an overvaluing of the formal attribution question that has plagued climate change conversation from the beginning. When we see something odd in the weather, it is natural to ask whether it is “because of” human interference. This is formalized into scientific questions of various sorts, and the result is often inconclusive or misleading. Climate change is perhaps the only scientific discipline where inconclusive results, rather than being buried, are trumpeted from the mountain-tops by those who would like to imply that nothing at all is known.
Here’s a classic example reported in mass media of work by Marty Hoerling, who is consistently equivocal about finding traces of climate change in actual events on the ground.
It’s to this pernicious pattern that Trenberth et al is responding.
We should simply bypass those questions in public discourse. I propose extending Trenberth et al’s assertion something like the following:
As Trenberth et al. assert, “The climate is changing: we have a new normal. The environment in which all weather events occur is not what it used to be. All storms, without exception, are different. Even if most of them look just like the ones we used to have, they are not the same.” This in itself does not inform us about the extent and immediacy of the climate risk we are taking. What we should care most about is the prognosis for the future climate that is dramatically much more altered than the one we face today. To inform that, we should not look at individual events, even the most extremely destructive ones, without taking a historical perspective and seeking comparable disasters in the past. ”
This is to say that formal attribution is the wrong question. (One might say it is applying frequentist reasoning to very small sample sets, and so necessarily yields inconclusive results.)
You might argue that we should take a Bayesian perspective, to see how events that we see line up with our priors, but that’s problematic as well. That is, we might ask whether this or that particular event tends to confirm or change our detailed expectations. Without, for the moment, getting into why this is a very difficult question to address, I’ll just say that it’s still the wrong question insofar as Revkin’s complaint, (assuming it makes any sense at all) is concerned.
What we care about collectively is whether things are already getting worse because of our actions, not whether our theories are correct in every detail. This, it seems to me, is the quantification Revkin is asking for.
(Regarding sea levels, which did play a role in the South Carolina disaster as well as the Sandy disaster, this is a slam dunk. Definitely, to the extent that sea level is part of the event, human action is partly responsible. Definitely, coastal flooding is worse on account of human interference in the climate system. But let’s leave that aside for now.In other disasters, drought for instance, clearly this is not so. There are very interesting questions at hand in cases where sea level is not involved.)
This brings us around to “recurrence rates”. To speak of recurrence rates seems to me a sounder approach to the climate/disaster connection than speaking of attribution.
To be sure, once one accepts Trenberth’s observation that we live in a changed climate, there is a separate question as to whether human activity dominates the changes. On this matter, the consensus is clear enough; the battles with those who try to squirm out of the concensus won’t go away because we wish they would.
But even so, recurrence conversation gives us something we can talk about together, in a non-oppositional way. It might help us improve the extent of our shared understanding.
If we do accept that climate change is and for the foreseeable future will be human-dominated, then we can make more fruitful use of the disaster record to inform the scope and urgency of the problems we face in coping with our new responsibilities.
The “recurrence rate”, useful though it is in thinking about our situation, is commonly misunderstood.
Of course, “a thousand year recurrence rate” or a “1000-year storm” means that the expectation of seeing such an event in a given year is 0.001, i.e., that the average 1000 years contains one such event.
Claims like ”a 1000-year recurrence rate” are being bandied about for the South Carolina flood, and being duly mangled by politicians and in turn duly mocked by comedians. The governor of South Carolina seems to have claimed that “South Carolina has seen nothing like this for a thousand years”, as if there were some sort of of millenial clock spurting out weird moisture jets in every year ending in ‘015.
Recurrence time does give us something we can effectively apply quantitative reasoning to. If we are seeing more disasters that one would expect on this quantitative scale, then we could unambiguously apply (non-sea-level-rise) costs to human activities. A recurrence argument is likely to be more fruitful than attribution arguments.
Still, Colbert has a point. We don’t really have a lot of information on the distribution of the rarest events, as typically their occurrence, if any, would be prehistorical. There is much speculation about the shape of the tails of these distributions. Unfortunately, the stranger an event is, the more important it is, but also the less confidence we have in its recurrence rate.
On the other hand, the most extreme events ARE the ones we should be paying the most attention to in evaluating whether climate change has really begun to bite. The current California drought, the Texas drought of 2011, recent heatwaves in Australia, the Pakistan floods and Russian fires of 2010, such events are arguably so far outside the norm that they arguably would simply not occur in a stable Holocene climate. If so, their recurrence times would literally be infinite. It looks that way, but it is intrinsically hard to know this, as any recurrence time longer than the duration of the instrumental record has to be somewhat speculative.
Unfortunately any honestly inconclusive result can easily be twisted to yield aid and comfort to those-who-must-not-be-named; let’s call them the ignorophiles. Of course I mean those who ask “you can’t *prove* anything so why should we change our policy?” We must patiently point out to those wielding this tiresome argument that they are analogous to a doctor, unable to come up with a firm diagnosis, asserting that a patient, however uncomfortable, cannot possibly be sick. Applied science is not afforded the luxury of certainty. The balance of evidence is tricky enough – it is more so when there is a dedicated group trying to dig up ambiguities and red herrings. Strange things are afoot, and we expect them to get stranger, even if we can’t exactly quantify that expectaion.
On the other side of the ledger, there’s something misleading about the way these recurrence times are bandied about. There are plenty of locations on earth, and various sorts of events (wind, rain, flood, drought, snow) to worry about. A 1,000 year event in one place is a big problem for that place, but it sounds scarier than it is. In an unchanging climate, if there are 1,000 places the size of South Carolina and 5 types of events, we’d expect an average five of these a year globally, one per type. A single thousand year event is big news for the people who experience it, and may be a big deal economically or even geopolitically, but it isn’t in itself climate news.
But we have to be willing to roll up our sleeves and dig into the dirt. The recurrence question can provide an honest motivation to engage with the details of the science. However tragic it was, the South Carolina firehose was also a fascinating bit of atmospheric dynamics. We should be able to take the occasion to delve into the details, rather than shrugging it off.
We should be willing and able to say “What’s the recurrence time of such an event? We don’t have a great answer, but it’s a great question!” It’s certainly a better question than “is this our fault? Hell yes!” (or Hell no!”) It opens the door to reasoning and to attention to evidence, rather than shutting it. It engages our minds rather than our guts.
Regarding the South Carolina floods, I suspect we’ll even find that 1,000 year recurrence is something of an understatement. But before we dig into it, we should think a bit about what a recurrence interval really means. Stu Ostro has some points to make about this question (toward the end of the linked article). I agree with him that it’s considerably more complicated than it sounds at first. But it is neither polarizing nor boring, so I think it’s worth thinking about.
UPDATE: From Tom Fuller’s blog:
ATTP has generously added me, Michael Tobis, Ph.D. (Atmospheric and Oceanic Sciences, University of Wisconsin – Madison, 1996), commonly referred to as “mt”, as a co-author of this site, so I guess this doesn’t count as a guest post.
I am a retired climate scientist and much influenced by computer science and information theory. I never aspired to a faculty position or published much in the formal literature, but I am familiar with the science and am personally acquainted with some of the most prominent and impressive climate scientists. I will continue to post smaller climate-related observations and possibly longer pieces of economics – skepticism on my own blog, Only In It For The Gold, but will endeavor to contribute here regularly.
I am working on a book on how to think about disasters in a changing climate.