Open Thread II – Severe Events

We’ll have another feature article up in a couple of days.

Meanwhile, any topic is fair game as commentary to this article (though as always we prefer comments that are interesting, thoughtful and kind over ones that are boring, trite or mean).

In the event that y’all are tongue-tied in coming up with a topic, I’d like to propose an increasingly common question and increasingly important question that is likely to get some special attention in the next few weeks, and which I expect to be at the center of a legitimate controversy for a long time to come.

(Wow. A legitimate controversy.)

The question is how we should be thinking about extreme events. Notably, extreme weather events, and extreme environmental events of sorts that are connected to weather and climate, such as wildfires and infestations.

There is a tendency for those of us who are alarmed to look at every such event, and say “See! We’ve gone too far already!” Meanwhile those who tend to be more sanguine will flip through the history books, find analogous events, and say “it was ever thus”. These claims seem like they ought to be reducible to objective questions. But there is considerable controversy within science as to how to think about them.

I hope my paraphrases are not too far off the mark, but as I understand it Kevin Trenberth suggests that the human influence on climate is already so profound that every event is a climate change driven event to some extent. On the other hand, Myles Allen suggests that no event has occurred to date that is demonstrably impossible in the preindustrial climate. Both of these claims are cogent and reasonable, indeed undeniable. Yet their cogency carries the same sense as it does in the following anecdote, commonly told of engineers, but here repurposed for atmospheric scientists.

Two men in a hot air balloon lose their bearings, and yell down to somebody on the ground “Where arrrreee weeee?”. To which the distant response is heard: “In an Balloooooon!” One of the two daredevils turns to the other and says “That must be a climate scientist. Every word he says is true and still we have no better idea what to do than before we asked.”

The simple answer seems to be that no individual event has any established connection to climate change, yet all events are causally related to it.

This one is not just the usual political back-and-forth between the groups that have associated their identity politics with one or another side of the question. This is important, because how hard we work to avoid climate change depends on the associated risks, among which severe events figure prominently.

Whil we have many many observations of common events and can determine with astonishing precision the measures of the mundane. Even with terribly calibrated instruments, after all, we have essential unanimity in the global mean temperature trajectory! But the statistics of the extraordinary are much harder, both in practical and in theoretical terms.

Right now there are astonishing floods in Thailand. How, for example, should we think of them? Are they once-in-a-lifetime floods? Or unheard-of  floods? If they are once-in-a-lifetime floods, are they becoming twice-in-a-lifetime floods? How much does that matter?

I have seen some images that make me wonder if Thailanders don’t take floods in stride more than most of us. One picture showed a family, perched up on tables and sofa backs, watching television in their knee-deep flooded house. People seem to be grinning and bearing it. Maybe I am misreading it, but perhaps the lack of presence of the Thai floods on the front pages is appropriate. This is a rare event, but not an unheard of one, and people are prepared for it, like a blizzard in Montreal or a hurricane in Texas. Yet if a comparable area were underwater in most countries it might be much more severe, because it would have no precedent.

This past summer, the ongoing drought in all of Texas, most of the neighboring states, and half of Mexico took hold. In Texas, it yielded a summer without a close precedent in the evaporation rate in Texas or any other state. (The high desert states tend to have cool nights.) How should we think about this?

Is it the combination of an unprecedented drought, unrelated to global warming, plus some additional warming from the background trend? Note that while most models dry the southwest in the upcoming century, they are somewhat equivocal on the fate of Texas. Should what the models say work into our degree of attribution of particular events? Are we already seeing the effects of climate change in the occurrence of severe events? If not, how will we know when that happens? And if it does occur, how should it figure into our risk calculations?

This looks likely to be an ongoing theme for a while. Comments welcome here. Longer articles are more than welcome here ( ) and links to such articles elsewhere would also be appreciated.



  1. 'Extreme' things ties more broadly to my own unfamiliarity with impacts. I'm not, at the moment, very good at arguing with `skeptics' who try to make a cost-benefit case. I don't have much to come back with, since the answers depend on what sort of economic framework you're using. It's clearly crazy to think we can carry on stuffing carbon into the atmosphere past 2100, but it isn't actually impossible - we might just achieve that. I wonder, what's the clearest way to analyse and communicate how the costs are clearly going to outweigh the benefits of e.g. not spending money on mitigation.

    It's very difficult: trying to work out what to say about how you cost extreme event damage (and as you mentioned, what sort of social resilience people already have.) I was thinking of a thought experiment example: a world with 4 squares, with one town that lives in one of the squares. With some probability, an extreme event hits their town that forces a percentage of them to move. It's terrible, but economically it's actually not that terrible because it forces building work to take place. There's enough space for that movement to keep on going, below a certain level of extreme events, but people are still regularly moved on.

    Obviously there must be a point where enforced nomadism ruins that little world's economy, but it's also true that the non-economic impact on people's lives simply isn't measured. That's a point made quite a lot in the last year: we can adapt, but adaptation is e.g. people leaving a pile of kindling behind in Joplin.

    I suppose the underlying point: are there any other ways - apart from economic cost and benefit - of quantifying the human impact of extreme events, or at least of framing it in a way that allows us to talk about it semi-objectively? Is it possible - as I was implying above - that you can actually have a `keynesianish' response to extreme events that actually looks economically productive, but that we'd clearly want to avoid?

    I'm reminded of some slightly worrying comments after the Manchester bombing in 1996, with people claiming it actually ended up being economically beneficial...

  2. Sorry, another possible discussion topic: this story on Nigerian growth and returning workers - oil has caused massive problems there, of course, but it's also a key part of Nigeria's development. At some point, we need to be keeping carbon in the ground. Yet I cannot think of any way one can demand Nigeria cease seeking oil revenu0e. What alternatives are there? How could those alternatives possibly be developed? How much more time might we have to use carbon / oil / gas as a source of investment? (Which ties to how we quantify the cost and benefits, of course.)

    Just re-reading Diane Coyle's Soulful Science: she does a very good job of painting a broad-sweep picture of development economics since 1945. The biggest lessons for me: decades where certain development theories appear to have led in completely the wrong direction, that we're still in a world of massively competing paradigms, and that climate change policy only makes that more complex. Where on earth do we start in unpicking all that?

    Anyone following Krugman's writing (recent example) has seen him repeatedly argue that `Serious People' are seriously misunderstanding the economics and thus effectively going in exactly the wrong direction. How are we meant, collectively, to find the right paths?

    MT was discussing the status quo recently. That's also another massive issue: two economists have a theory, they attempt to persuade current hegemonic powers. Are they more likely to adopt the theory that undermines their power? As Coyle points out - "when the evidence available can be used in support of two opposing views of economic development, each held by well-intentioned and intelligent people, we are clearly not in the realm of hard science." (soulful science p.65)

    The meta-question there: how are (and should be) economics, democracy and power connected?

  3. Pingback: Collide-a-scape » Blog Archive » Collide-a-scape >> Burning Down the House

  4. Climate change does not deterministically cause extreme weather events, but it modifies probability of extreme weather events.

    (Cf. Low dose radioactivity does not deterministically cause cancer, but it modifies probability of cancer.)

    Let me try "Bayesian" reasoning, though it is unfinished.

    Bayesian reasoning, Version 1.
    EWE = extreme weather event;
    AGW = anthropogenic global warming;
    P(AGW|EWE) = P(AGW) * P(EWE|AGW) / P(EWE) .
    As we get information that EWE really happens, our subjective probability of AGW is modified by the factor which is the ratio between the probability of EWE under the condition of AGW and the un-conditional probability of EWE.

    This reasoning does not address the question whether we can say something like "AGW caused EWE" (but not exactly that). Perhaps we should discuss the probability of "AGW being (one of) the essential cause(s) of an EWE". But I cannot elaborate it now.

  5. I'd start out by making a distinction I've tried to push before, but I'm not sure if I've made it coherently or ever convinced people: the difference in "extreme" between (a) "fundamental changes in statistics", and/or "anomalies not present in the foregoing statistics; and (b) "just moving the distribution".

    In case (b), for example, European summer temperatures that are now considered "extreme" will become commonplace in, say, 2100. But this doesn't represent anything fundamentally interesting happening; all that happens is that as a normal distribution shifts its mean up, what was once in the 5% tail becomes part of the more likely section.

    But in case (a) ridiculously extreme floods that were never part of the distribution start occuring, for example. Is Thailand a case (a)? Is there a better recent example?

  6. Pingback: The Blackboard » Bangkok floods: 1942

  7. I think that the burden of proof is a problem in AGW. The attitudes of Trenberth and Allen that you refer to above are not exclusive. An event does not need to be "demonstrably impossible in the preindustrial climate" in order to have a climate change component in it, but that is the standard we are held to by some to make our case.

    Unlike other more mundane aspects of AGW, like the temperature record, the nature of the statistics of extreme events will mean that it is probably the last thing for which a reliable signal will make itself visible above the noise.

    But knowing what we know about how the atmosphere has changed and how weather works, it is plausible to suggest that AGW is affecting some kinds of extreme events. Saying that doesn't mean that the SNR is strong enough to clearly see the AGW signal _now_. It just suggests that the mechanism by which that signal will eventually show up can be described.

  8. My opinion is that, regardless of whether Thais are content with watching television with their knees deep in water, there's one undeniable fact: floods are a problem. So perhaps, in a way, the question of whether such and such a flood is "once-in-a-lifetime" or not is a bit of a distraction. The real question is whether we're going to reduce the flood problem, or to exacerbate it.

    And while it's possible to do cost-benefit analyses, I'm not sure that ordinary people are actually thinking in those terms. Perhaps part of the appeal of inactivism is simply that people are fearful of change -- and switching the world over to sustainability will require massive change -- even if the changes will benefit them in the end.

    -- frank

  9. Stoat's distinction seems theoretically relevant, but difficult to be applied to actual cases.

    The size of tsunami in northeast Japan in 11 March 2011 was in hindsight considered to be within the range of natural variability in 1000-year time scale, but was outside of the range conceived by experts before the event. (This is a issue of geodynamics rather than climate, though.)

    I am almost sure that the present flood in Central Thailand is an event of the type "(b) just moving the distribution". The rainy season is May to September, so it is normal there that the water level is highest in October. The level is higher than normal this year, and I suspect that the "normal" has shifted somewhat. But we do not need to introduce a novel climate regime to explain the situation.

    But the level of flood is outside the design of modern constructions, in particular, of factories built with investments from Japan. Perhaps it would not be outside of the range which pre-modern Thai people traditionally adapted to. But, in the suburb of Bangkok, even though urbanization is a new phenomenon, modern civil engineering to control water for rice farming has changed water regime already one century ago. The seasonal change of river water levels has been minimal since any person living there was born (and the situation in Japan is similar). It is a tough issue for the people to adapt to a new climate regime.

    In the Mekong river basin in Northeast Thailand, Laos and Cambodia, the normal seasonal range of river water level is still around 10 meters. So, if people from these areas lead the design of construction, it would be more resilient. Unfortunately, people grown up in modernized areas tend to be influential.

  10. I appreciate being asked to participate in this much needed forum and, as a sustainable land developer, I especially respect Michael for his candor in admitting that climate scientists should keep the following anecdote in mind:

    "Two men in a hot air balloon lose their bearings, and yell down to somebody on the ground “Where arrrreee weeee?”. To which the distant response is heard: “In an Balloooooon!” One of the two daredevils turns to the other and says “That must be a climate scientist. Every word he says is true and still we have no better idea what to do than before we asked.”

    Well, my profession is in the balloon and we must make important decisions everyday based on current scientific consensus. We have a term for extreme events, environmental or otherwise - "Black Swans". While these events have occurred throughout history, the frequency (and intensity?) does seem to be increasing. I'll focus here on the environmental climate "Black Swan" which currently dominates your sustainability discussions and submit the following articles we have published for a triple-bottom-line audience both in and around our industry:

    Beyond the Bailout … a Bigger Problem… and a Solution

    Sustainable Land Development Goes Carbon Negative

    While I eagerly await further advances in scientific knowledge to aid our efforts, I am convinced that a balanced approach to sustainability will be called for, no matter what the specific best practices are.

    Your comments and participation are welcome.

    Sustainable Land Development Initiative

  11. If we're trying to answer the question "How much worse is this under current conditions than, say, 30 years ago" - floods can be quantified. (To an extent.)

    If increasing temperatures have increased water vapour by 4% (on average) then we can look at an event and see what flood level might have been reached if the 'same' weather event occurred with 4% less water involved. In many places that does mean that particular floods would not have resulted in anything like the extent of damage they did cause.

    And then we do the sensible thing and say if 4% increase is the **average**, obviously a weather event causing floods is probably in an area on the high side of the numbers contributing to a 4% average. So we re-do the calculations with 6, 8, 10% less water involved and see what damage might then have resulted.

    Not convinced that this sort of thing is any more than just playing with the numbers, but it can be useful as a thinking tool.

  12. “Prediction is difficult, especially about the future, as the wise Yogi Berra is often quoted as saying.”

    No, that was Niels Bohr.

    “We are considering adding a separate section to the site in which we look at the excesses of the denialati with some pleasure”

    Why don’t you add a section on reviews of cookbooks and perhaps another one on optimal sports betting strategies while you’re at it. I’m sure those must somehow also fall under you’re all encompassing umbrella of what “sustainability” involves, somehow or other.

  13. Jim,

    Re Bohr vs Berra, thanks.

    Re off topic sections - if I had credible information that these would be attractive to people likely to be participants (as opposed to readers) I think it might be a good idea. But more to the point, deciding the boundary between legitimate scientific disputation and illegitimate nonsense is difficult. It is exactly that difficulty which is among the key weapons of the obfuscators. If the community develops as I hope, we need some place to thrash this out in the open.

    Re "sustainabilty", I hereby invite an article from you defining the term as you understand it.

  14. This suggests that there are indeed several linked problems (not that we didn't know that). What we're experiencing in extremes may not be unprecedented at all on the scale of time frames that are only slightly longer than a century. But we have built a massive agricultural and industrial infrastructure within a very short period, and not only may it be brittle relative to small changes in the regime of extremes, it may be directly exacerbating them (as the increase in paving increases flooding). That suggests we might have problems even without an anthropogenic contribution to climate change. But it sure suggests we don't want to risk accelerating the changes.

  15. Well, I suppose we could include a word count with every excerpt.

    The great thing about the web is that there is no unit cost per word. It's no more effort or cost to repost 10000 words than 100. Maybe it's best if people don't know what's under the hood on each article?

  16. I see Pielke Jr's taking a rather different view of the extreme weather data. Actually making underlying points about the same as Paul is, but is concluding that people 'bigging up' extreme weather talk are ignoring the empirics. Hmm.

    On floods: how does the rainfall data look, rather than flood data? That'd help remove the human landscape change component.

  17. Can I ask, particularly of those who are professionally immersed in this, does the difficulty in talking about extreme events stem from a genuine lack of knowledge and understanding about the science or is it predominantly an issue of articulation, the problem of parcelling up the concepts into language, perhaps for ourselves as much as for others?

    I know this is to some extent a false dichotomy and maybe the better question to ask would be 'to what extent does the way in which we attempt to parcel up our understanding of the science into language impact on that understanding?'

  18. The problem is that extreme events, being extreme, don't fit statistical analysis well. So if you look at, say, simple global surface T, you can do statistical analysis to show that yes, it is increasing and yes, it is significant. Doing the same with extreme events is much harder.

    One argument (which I'd push) is that the correct response is not to use the extreme events. We don't need them to demonstrate GW. But people do like to use them for "oh look how bad it will be".

  19. "One argument (which I’d push) is that the correct response is not to use the extreme events. We don’t need them to demonstrate GW. But people do like to use them for “oh look how bad it will be”."

    There's a mixing up of two meanings of extreme event though, isn't there? Statistical extremes are very specific - but we're also talking about events that have an extreme impact on infrastructure. The latter is dependent on the statistical extremes, but also on the infrastructure itself and institutions in place to deal with those extremes.

    Both are vital to understand - understanding the likelihood of infrastructure damage and strategies for avoiding that is clearly important.

    p.s. I guess everyone's probably seen this recent article on extremes at realclimate.

  20. Pingback: Collide-a-scape » Blog Archive » Collide-a-scape >> Debating How to Debate Severe Weather

  21. Well, it's not just an exercise in worrying. It matters just how much we need to be worrying. If the intuitive sense that some meteorologists seem to be expressing is right, then things are much worse than we thought, whatever the actual temperature sensitivity might be.

    The fact is that we have very little basis for predicting what the sensitivity of cost-weighted severe events to a given global temperature change might be. Neither our statistical methods nor our models give us much of a handle on it.

    As an engineer, I would say that systems far from equilibrium demonstrate modes that can be neglected in systems near equilibrium. (Fenders do not crumple on cars under ordinary operating conditions, but they do when the car is subject to forces outside its ordinary operating range.) So sooner or later we can expect transient dynamics that don't appear in the near-equilibrium dynamics.

    Probably the best place to look for evidence of this in nature would be the Younger Dryas period. I would not be surprised if GCMs completely fail to capture such behavior even if it turns out to be real.

    My current belief is that it is real, that extraordinary extreme events have started quite a bit earlier than anyone might have expected. This accords with public perception but that perception may be skewed either by reporting or by confirmation bias.

    It is not obvious how to test this hypothesis, but science is obviously starting to try to grapple with it.

    If it is real, it just shortens the fuse on reacting, and I suppose that fuse is already too short anyway. So it may not affect the optimal policy all that much. But it may add to the urgency quite a bit.

  22. Good point Michael,

    As I said over at KK's, I would suggest that those of us that fall into the ‘alarmist’ camp are more concerned with the implications of these sorts of changes than we are with coarse global metrics like climate sensitivity or statistical trends in extreme weather events.

    It seems to me that people who take a more pollyanish view tend to think about climate change in linear terms and adjust their view of impacts and adaptation on that basis.

  23. Peter Stott just gave a talk this morning at the WCRP conference, on attributing extreme weather events to climate change this morning. His approach is to calculate the Fraction of Attributable Risk (FAR), which is the difference in likelihood of exceeding a particular threshold in a climate changed world versus a pre-industrial world, using ensemble model studies. So this allows him to calculate that, for example, climate change was responsible for doubling the likelihood of the UK floods in 2000.

    He calls for a properly resourced attribution service, which would do this in real time in response to public concerns about extreme events, and cut through some of the back and forth about whether or not climate change is responsible for stuff.

    See his white paper here:

    See also the twitter feed from the conference:!/search/%23WCRP11

  24. What basis do we have for trusting model efficacy here?

    There is a perverse underestimation that happens. Suppose neither the perturbed model nor the unperturbed model displays a certain type of behavior at all. Then the trend in that behavior is zero. So when a 10,000 year event shows up, the attributed FAR is zero. Fine, fair enough. Suppose another one shows up two years later, and then three more in the next decade. Still zero, right?

    This is precisely the error I see being pervasive. The more extreme the event, the more confident my internal Bayesian calculator is that it is a side effect of anthropogenic forcing. Most people will say the same. When climatologists plug in this particular formula, don't they get the wrong answer?

  25. MT: I don't think that's the right way to think about it. Sure, the models have various kinds of weaknesses. If some phenomena of interest doesn't show up in the model, then you can't compute a FAR.

    But the point made by both Peter Stott and Myles Allen this week is that the unusual heatwaves, floods, snowdumps, etc of the last decade or so all *do* show up in both forced and unforced models runs, if you use a big enough ensemble. Hence, you then have the basis for to quantify the effect, and the reasoning is defensible, to the extent that the models capture our current understanding of the climate system.

    The kerfuffle a few weeks ago over Myles Allen's piece in Guardian, where he argued the "painting extra spots on the die" metaphor is wrong (and got blasted by Joe Romm for it) is entirely down to the fact that in all the attribution studies they've done of recent extremes, all of them show up sooner or later in both forced and unforced ensembles. The different is that a 500 year return rate might get pushed down to a 100 year return rate; a 100 year return might get pushed down to a 10 year return. So the net impact of climate change *so far* is to make these extremes much more frequent, but there's no clear evidence yet for entirely new kinds of extremes that were not possible before.

    Where I part company with Myles is that he's only focussing on what's happened so far. Extrapolating his graphs, it won't be long before we do start to see extremes that only show up in the forced ensembles. At that point, we can say we've painted new spots on the die. The only question is how soon that point will come. Myles thinks we haven't reached it. Others (Trenberth for example) think we have.

  26. Steve, you've done a good job explaining FAR, so I will concede that under some circumstances it may be meaningful to compute the FAR.

    But it is not obvious a priori which circumstances those are. It is one thing to have a general confidence in the behavior of a model. It is another to have confidence that it does extreme events for the right reasons, and another yet to say that the sensitivity of those events to various forcings is captured.

    Again, I call your attention to the fact that if the sensitivity in some particular case is too small or trending the wrong way, this method will adamantly refuse to recognize an event as anthropogenic no matter how unlikely it was in the past, no matter how frequent or extreme the observations. To the contrary, people will be all too willing to say, look, see, no anthropogenic climate change, because these particular weird events were not predicted.

    So using this method indiscriminately will systematically understate costs due to anthropogenic interference. It does this by systematically attributing costs only to the predicted component of the change. Anything unpredicted, no matter how bizarre, is automatically attributed to "natural" variability by construction. This can confuse the public and can confuse the economic calculations, and both in the direction of even further insouciance, of which we have altogether too much at present.

  27. Recent Thailand floods are at the approximately 25 year recurance level. The extraordinary things this time are (i) in retrospect unfortunate decisions by the dam operators (ii) an extra 40 years of relatively unplanned so-called economic development since the last flood of this magnitude. The hardships primarily are caused by the lack of hydrologic planning in allowing those thousands of factories, houses and parking lots in a flood plain (which otherwise would partially absorb all that water and otherwise be largely harmless).

    Somewhat more shocking are the floods in Pakistan in 2010 (north) and 2011 (south). Each, taken independently is recurrent at around the 50 year recurrence interval. The two taken together are not the combination of two independent events, so have a recurrence time of much less than an imagined 2500 years; unfortunately there does not seem to be enough history kept to have an idea of the recurrence time for such a pair of floodings. Now wiat until next summer. Based on La Nina it seems there is a good chance for a third flood in a row in the Indus River valley. If that happens, the sequence of three in a row is certainly a rare event, possibly one never before experienced during the Holocene. Stay tuned.

Leave a Reply

Your email address will not be published.