Uncertainty monsters

I was trying to get a copy of Stephan Lewandowsky’s latest papers on Scientific uncertainty and climate change: Part I. Uncertainty and unabated emissions and Scientific uncertainty and climate change: Part II. Uncertainty and mitigation. Unfortunately, my university library doesn’t seem to have a subscription to the Journal, so I haven’t been able to access them. They are discussed briefly in this article. There is a related article by Kerry Emmanuel on Tail Risk vs Alarmism.

The issue of uncertainty is, I think, a very interesting topic and is not particularly well understood. It may be partly semantic. People perceive the term uncertainty as implying that we aren’t sure, when – in some sense – it’s the other way around. The uncertainties define our level of certainty and should maybe – more correctly – be referred to as confidence intervals. I sometimes come across people who argue that because we’re still uncertain, we shouldn’t do anything yet. One thing that I find ironic is how certain some can be about how we should proceed, given these uncertainties.

Another aspect of this argument that I don’t fully understand, is that it is often accompanied by a suggestion that we’re innovative and therefore will come up with a solution. What confuses me then, is what such people mean by do nothing? I don’t know how we can solve anything, if we don’t do something. To me, it comes across as a “closing down the debate” type of suggestion. Somehow, magically, our future selves will solve whatever problems our future selves will face, despite our current selves deciding that it’s not worth discussing, or considering, in any detail. It may be true that waiting will allow us to constrain the uncertainties, but it also means that we might discover that climate sensitivity is indeed high and that we’ve now got to the point where acting is much more difficult than it would have been had we acted sooner.

Additionally there are people who seem to think that because a low climate sensitivity is about as likely as a high climate sensitivity (and, I know that this isn’t strictly true) that we should just bank on it being low. They seem to be arguing that there is a risk associated with doing something, and hence – given this risk – we should gamble on climate sensitivity being low. The problem with this, in my view, is that it’s the wrong comparison. What we should really be doing is comparing the risks of continuing as we are if climate sensitivity is high, with the risks of making significant changes to our economies if climate sensitivity turns out to be low. Those, as far as I’m aware, are the two extremes. Of course, it’s not a simple issue and, of course, changes will likely have to ultimately be global, not local. However, given that some of the extreme risks we face include significant damage to ecosystems and extinction of a significant fraction of the species on the planet, the risks associated with changing our economies would seem to have to be quite severe if we were to conclude that doing nothing yet, is the right approach. Rachel has a recent post discussing some of the risks we face.

There’s also the possibility of positive benefits from the technology development that will almost certainly be needed. It seems ironic that there are those who argue that we are sufficiently innovative that we will find ways to adapt, but don’t seem to accept that the same applies to our ability to find ways to mitigate.

To be clear, I’m not trying to make any kind of policy specific argument here. I’m simply suggesting that how we consider the uncertainties and risks is non-trivial and we should not only be willing to discuss and consider the various options, but we should also be doing the risk analysis properly. Just because we aren’t certain of what risks we face, doesn’t mean that we are incapable of making sensible decisions that minimise the future risks. Of course, it’s not going to be easy and not every policy decision is going to be optimal or sensible. On that note, I also think it’s time that we stop blaming silly policy decisions that have happened in the past on climate alarmism. I might think that our policy makers should be starting to act, but that doesn’t mean that I’m wanting them to make decisions that aren’t sensible and that aren’t going to be effective.

This entry was posted in Climate change, Climate sensitivity, Global warming, IPCC, Science and tagged , , , , , , . Bookmark the permalink.

17 Responses to Uncertainty monsters

  1. Think about speeding on an unfamiliar road when fog rolls in. Do you react to the added uncertainty of the fog by driving faster or slower? Note that “doing nothing” here actually means speeding through fog, in much the same sense that “doing nothing” about climate change actually means releasing CO2 ten times faster than before the end-Permian extinction.

  2. OPatrick says:

    What we should really be doing is comparing the risks of continuing as we are if climate sensitivity is high, with the risks of making significant changes to our economies if climate sensitivity turns out to be low.

    Can I recommend a Burundian flag metaphor. White represents neutral outcomes, red negative ones and green positive ones. If we do nothing but adapt gradually, represented by the centre of the flag, we may be able to maintain a comfortable level of development but there might be unexpected surprises, which could be catastrophic but may also have positive fringe benefits, from an evolutionary perspective. The vertical axis of the flag represents different courses of action, ranging from rapid and intense mitigation to wholesale commitment to exploiting fossil fuels. The horizontal axis of the flag represents different responses of the climate system, ranging from sensitivity to CO2 being zero and ecosystem responses to CO2 as positive, to high sensitivity and negative ecosystem responses. The combination of uncertainties clearly produce regions of net benefits and net costs and the interactions in these uncertainties can be better understood by applying this metaphor to our thinking. It is clear that in any combination of actions and responses there are neutral paths which justify inaction and these form the major diagonals of the uncertainty space, which therefore represent paths that cut across, and so can unite, all participants in the debate.

  3. Vinny Burgoo says:

    Wotts, Lewandowsky’s new papers are probably based on four blog posts he did in 2012:


    IIRC, he confused his means with his modes, but that’s probably been corrected in the published papers.


    From the press release:

    ‘They show that as uncertainty in the temperature increase expected with a doubling of CO2 from pre-industrial levels rises, so do the economic damages of increased climate change.’

    That should surely be the *possible* economic damages. If maths can prove that the more imprecise your estimate, the more it’s going to cost you, then there’s something wrong with the maths.


    Wotts: “What we should really be doing is comparing the risks of continuing as we are if climate sensitivity is high, with the risks of making significant changes to our economies if climate sensitivity turns out to be low.”

    A good formulation as far as it goes but surely you have to exclude the near-impossibly high costs of near-impossibly unlikely events and where do you put the cut-off? (Cf discussions of Weitzman’s Dismal Theorem a few years ago.)

    You also have to consider the harm caused by overly rapid mitigation. Lewandowsky touches on this in his blog posts.

    And I think OPatrick is on the same track. Care to elaborate on your Burundian vexillology, OPatrick?

  4. Vinny,
    I’ll grant you the “possible”.

    surely you have to exclude the near-impossibly high costs of near-impossibly unlikely events and where do you put the cut-off?

    You’ll have to explain this a little more, as I’m not sure what you’re getting at here.

    You also have to consider the harm caused by overly rapid mitigation. Lewandowsky touches on this in his blog posts.

    Agreed, but when I said “minimise the risks”, I didn’t mean only with respect to climate change, I was meaning all the risks, both those associated with climate change and those associated with possibly extreme policy options.

  5. The abstracts of the two papers tell:

    Paper 1:

    . We show that increasing uncertainty is necessarily associated with greater expected damages from warming, provided the function relating warming to damages is convex.

    That’s certainly true, and no new analysis is needed to state that. The rest of the abstract does not add anything more novel. I have the full paper, but haven’t really checked, whether it has something more substantive. Based on rapid skimming everything they discuss has been known for long.

    Paper 2.

    This article examines the role of uncertainty about future climate change in determining the likely success or failure of mitigative action. We show by Monte Carlo simulation that greater uncertainty translates into a greater likelihood that mitigation efforts will fail to limit global warming to a target (e.g., 2◦C). The effect of uncertainty can be reduced by limiting greenhouse gas emissions. Taken together with the fact that greater uncertainty also increases the potential damages arising from unabated emissions (Lewandowsky et al. 2014), any appeal to uncertainty implies a stronger, rather than weaker, need to cut greenhouse gas emissions than in the absence of uncertainty.

    When uncertainties related to the actions rather than to the warming are included, the variety of uncertainties that affect the conclusions widens. It’s certainly possible to introduce assumptions in the model that guarantee an outcome like that presented in the abstract, but it’s also possible to make assumptions where added uncertainty weakens the argument in favor of some mitigative action. Nothing generic can be concluded without restrictive assumptions.

    One very common case is that uncertainties support postponing the decision until better knowledge is available. That’s typical assuming that the development of knowledge and available options is likely to proceed rapidly. Then the same amount of resources is likely to lead to better outcome, when used later in spite of the loss from delayed action.

  6. “If maths can prove that the more imprecise your estimate, the more it’s going to cost you, then there’s something wrong with the maths.”

    That is quite easy. Imagine the the damages rise fast with the amount of climate change (let’s call it temperature increase) and that they rise faster than a linear relationship, which is, as far as I know, what climate impact researchers expect. Let’s for example assume that the damages double for every degree temperature increase.

    Now if there is no uncertainty, the temperature in 2100 will increase by 4 degrees and the damages with be D dollar.

    In the case with uncertainty, the temperature in 2100 will increase on average by 4 degrees, but it could also be 3 or 5 degrees. The damages for 4 degrees are still D dollar, but because the damages rise so fast, the average damage for 3 and 5 degrees is higher than D dollar. For example at 3 degrees the damage could be 0.5*D and at 5 degrees 2*D dollar., which is on average 1.25*D.

    That is the funny thing about the blog of Judith Curry. She thinks that by talking about uncertainty she can make the problem smaller, or at least she thinks her audience is sufficiently uneducated to believe this. Like Wotts writes above, she probably assumes that by talking about uncertainty the non-scientists will read: we are not sure. That is not what the word uncertainty means in science, it means we expect that the value lies in the uncertainty range, which is indeed better called confidence interval or range of possibilities.

    We are taking the climate system outside the known range at an unprecedented speed, that makes accurate predictions difficult. It is thus no wonder that the range of possibilities is large.

    In reality uncertainty is not the friend of people arguing against mitigation. The same thinking error is also made by climate change activists that sometimes ask scientists to emphasis uncertainty less.

  7. AnOilMan says:

    Victor… Thanks that was insightful.

    For the denial community it is most definitely desirable to emphasis certainty in any way other than what it means. They need to sell the image that the science is a Wild Ass Stab in the Dark (WASD), or that it shows with Certainty that it is a Low Risk. Indeed this is the emphasis with all their articles.

    From now on I will measure them on which extreme they choose… WASD, or CLR.

  8. OPatrick says:

    Care to elaborate on your Burundian vexillology, OPatrick?

    Oh, I could witter on about it endlessly. But I can guarantee you wouldn’t make head nor tail of it.

  9. Vinny Burgoo says:

    Wotts, re the cut-off thing, this discussion at Stoat’s might be helpful:


    I’d put it in my own words but it makes my brain hurt.

    Peka, the second abstract suggests that that paper is based on the blog where Lewandowsky plugged in illustrative numbers and distributions and proved that the probability that we have already exceeded a safe carbon emissions budget is already 1000 times greater than an Australian’s annual probability of being killed on the roads. Fun with maths!

    Victor, you have described how maths can prove that greater imprecision increases possible damage, not actual damage. Some might see this as pointless pedantry. I don’t. It’s pedantry, but it’s not entirely pointless. Imprecise language causes a lot of pointless alarm. (Maths can prove this.)

    And yes, I do get the point that, given this, that and the other, a larger feasible range for climate sensitivity means a larger range of feasible levels of damage. It’s just that there are people out there who are eager to spin this as ‘uncertainty increases harm’ rather than ‘uncertainty increases the risk of harm’ and things shouldn’t be made too easy for them.

    An example of these eager deceivers? Er…

    Ah ha! That was easy.

    ‘We can understand the implications of uncertainty, and in the case of the climate system, it is very clear that greater uncertainty will make things even worse.’

    That’s Lewandowsky himself, in the press release.

    That quote is probably what encouraged this bloke …


    …to write this: ‘The core correlation here is not one of increased temperatures causing economic harm, they might well do, but the researchers advance the conclusion that increased uncertainty equals increased damage.’

    I wonder what it’ll have morphed into by next week.

    OPatrick, good. My brain hurts enough already.

  10. Vinny Burgoo, you are right, one should formulate very carefully, because climate dissenters are determined not to understand what is written.

    I would have formulated it differently myself, but I see no problem calling a higher expected value for the damages a worse situation. I was able to understand what Lewandowsky intended to express. Maybe that is because I am sceptical in the old sense and not in the modern interpretation of the word.

  11. OPatrick says:

    ‘We can understand the implications of uncertainty, and in the case of the climate system, it is very clear that greater uncertainty will make things even worse.’

    I’m not sure of the exact context of this quote, you haven’t linked to it, but this seems a reasonable statement if you are considering the costs of adaption. If we are confident what the response, regionally, to a given amount of warming was going to be then it would be easier to adapt to it. The greater the uncertainty the harder adaption would be and the greater the range of possible outcomes we would need to prepare for. This also seems to be the interpretation in the allvoices article you linked to, where the Lewandowsky quote you gave is preceded by this:

    Taking the case of future sea-levels as an example, the authors highlight that the larger the degree of uncertainty over sea level rise, the greater the precautionary actions required in managing flood risk.

  12. AnOilMan, you are welcome. :0) This is a special case of my main theme, the importance of getting the distribution right if you want to study nonlinear processes. Many people have been indoctrinated by the beauty of linear methods where only the mean is important. That is often a good approximation and the power of the methods can make it worthwhile, but sometimes you need to consider the distribution to get accurate results. That means lots of low hanging fruit for me.

  13. Here’s “eager deceiver” Lewandowsky (my emphasis):

    “increasing uncertainty is necessarily associated with greater expected damages from warming, provided the function relating warming to damages is convex. … greater uncertainty also increases the potential damages…”

    Ignore the bolded words. Fantasize that Lewandowsky is an eager deceiver who claims to have measured the actual damages in 2100. He’s obviously using black magic to time travel. Burn the witch!

  14. Vinny,
    I’ve read Stoat’s post and James Annan. Haven’t quite digested it fully, but I think the point is slightly different to what I was getting at here. As I understand it, Weitzman has suggested that because of the long CS tail and because the damage increases faster than the increase in CS, then we are doomed. The point I think James Annan is trying to make (and which I agree with) is that the CS is essentially a number and not some possible distribution of numbers. The distribution simply represents our uncertainty about that number. Therefore, even though there is a long tail, it may never be realised and so to imply that because the damage likely increases faster than the CS we are doomed is missing the point that such high CS may never be realised (i.e., the actual CS may – probably isn’t – as high as the values in the tail).

    My point was slightly different and much simpler (I think). I was simply trying to suggest that the correct risk analysis is to consider the risk of doing nothing and CS being high, versus the risk of doing something – which may itself be economically damaging – and CS turning out to be low.

  15. “The dismal theorem” suggested that the combination of distribution of likelihoods and related damages leads to an infinite expected value. That would mean that we are doomed if all alternatives would be true in turn, i.e. if the distribution would be a frequency distribution, not a likelihood distribution. Annan’s point is that it’s a likelihood distribution.

    Other scientists, including Nordhaus, have noticed that the other part of the assumptions fails as well. I.e. the result is not infinite either.

    All this tells only that the outcome of that approach is not infinite, but it does not contradict the observation that most severe cases are very important for the conclusions even if their likelihood is rather small.

  16. dana1981 says:

    I can email you the papers if you still want them. I may still have copies, provided by Lewandowsky when I was drafting my post on the papers.

    You hit on a key point that ‘doing nothing’ isn’t really doing nothing, it’s continuing to pump greenhouse gases into the atmosphere and hoping that this doesn’t result in very damaging consequences, even though the body of scientific research indicates it most likely will. When people argue for doing nothing, they’re just arguing for maintaining the status quo, because that’s the easiest thing to do. But from a risk management standpoint, it’s also one of the dumbest things we could do. As Dumb Scientist notes, like continuing to drive at a high speed through the fog, while we’re pretty sure there are roadblocks not too far ahead.

  17. Dana,
    Thanks, but I managed to get copies of the papers.

    Yes, the “doing nothing” issue is interesting. It’s almost as if people think that continuing as we are is somehow natural while not continuing as we are is somehow unnatural.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.