Skip to content

“There is no way to prove that [an extreme weather event] either was, or was not, affected by global warming.”

This post is by Phil, not Andrew.

It’s hurricane season, which means it’s time to see the routine disclaimer that no single weather event can be attributed to global warming. There’s a sense in which that is true, and a sense in which it is very wrong.

I’ll start by going way back to 2005. Remember Hurricane Katrina? A month afterwards some prominent climatologists (Rahmstorf, Mann, Benestad, Schmidt, and Connolley) wrote “Could New Orleans be the first major U.S. city ravaged by human-caused climate change? The correct answer–the one we have indeed provided in previous posts (Storms & Global Warming II, Some recent updates and Storms and Climate Change) –is that there is no way to prove that Katrina either was, or was not, affected by global warming. For a single event, regardless of how extreme, such attribution is fundamentally impossible.”

Well, that’s just nonsense. How on earth could Katrina not have been affected by global warming? There’s no way. You can argue that a major hurricane might have struck New Orleans on August 29, 2005 with or without global warming — sure, could be. Or maybe it would have happened a day earlier or a week earlier or a year earlier or a decade earlier. But sure, OK, maybe it would have happened on August 29, 2005. It’s extremely unlikely but not impossible. But there’s no way, literally no way, that it could have been the same storm. Katrina was definitely affected by global warming.

Does it matter? “We all know what they meant”? Well, I don’t know what they meant! And I’ve seen similar statements hundreds of times.

The weather is different than it would have been without global warming, every day and in every location. In some places and at some times the differences are large and in some places they are small. On some days there are fewer tropical cyclones in the Atlantic than there would have been, and on some days there are more; on other days there are exactly the same number of tropical cyclones but they are not in exactly the same places with exactly the same winds.

To say we don’t know whether a given city would have been destroyed by a hurricane on such-and-such a date in the absence of global warming, OK, fine, coincidences happen. But to say that we can’t say whether the storm was affected by global warming, that’s just wrong.  That goes for Hurricane Dorian, too.

I’ve been waiting 14 years to get this off my chest. I feel better.

This post is by Phil Price


  1. Carlos Ungil says:

    We could reuse the discussion about the butterfly effect…

  2. John Hall says:

    It seems reasonable to do some kind of regression for category 4 or 5 hurricane frequency on average weather in order to get a sense of how extreme hurricanes are driven by weather. I imagine such analysis has been done…

  3. Dave says:

    I guess the distinction is:

    1. All hurricanes are a product of their climates.
    2. No single hurricane can serve as proof that a changed climate (versus some counterfactual climate) makes hurricanes generally more dangerous or frequent.

    People generally mean the second when they suggest we can’t tell whether climate change affected a specific hurricane, but they are misstating it due to the first (assuming they believe in the counterfactual climate).

    • Phil says:

      I think you’re right, this must be what is going on. But it seems extreme, unnecessary, and simply wrong to say that we don’t know if Katrina (or any other storm) was _affected_ by climate change.

      Hurricane seasons are variable and I think have autocorrelation on the timescale of many years, so of course from a statistical standpoint it’s not like you can look at the time series of hurricane number and intensity and say “this portion of the series _could not_ have happened under the previous climate.” At least I don’t think we have reached that point yet. But then why not say that, or some layman-friendly version of it?

      • Anoneuoid says:

        But it seems extreme, unnecessary, and simply wrong to say that we don’t know if Katrina (or any other storm) was _affected_ by climate change.

        The resolution to this “paradox” is really simple:

        1) Every event is caused by everything in its past timecone (ie, everything that happened earlier) including the vaguely and somewhat arbitrarily defined aggregate phenomenon we refer to as “climate”.
        2) The climate is always changing, has always been changing, and is always about to change.
        3) Every change in climate affects every future hurricane.

        This question is on the level of “how many angels can dance on a pin”, and should be of no interest to anyone.

        The question to be asked is: “what practical use can we get out of the predictions about hurricanes derived from climate models”?

        • Phil says:

          Anon, maybe “what practical use can we get out of the predictions about hurricanes” is the only question that matters to you, but it is obviously not the only question that matters to everybody. One of the questions that matters to other people is, evidently, “was this specific storm affected by climate change.” The answer to that question is ‘yes’, and I don’t understand why people deny that.

          Not to put words into Dave’s mouth, but I think he would agree that people only think they are interested in “was this specific storm affected by climate change”, really they are interested in broader questions about whether the evidence for climate change is clear in the hurricane record, and whether it’s clear that hurricanes are getting stronger and/or more frequent because of climate change. Or…well, whether Dave would agree with that or not, I do think it’s possible that that’s what people are really getting at. And I think these are interesting questions — in this way, Anon, I disagree with you — but if this is really what people are getting at when they talk about whether a single storm was ‘affected’ by climate change then I think they have a very confused way of thinking about this stuff, or at least talking about it.

          • Anoneuoid says:

            One of the questions that matters to other people is, evidently, “was this specific storm affected by climate change.” The answer to that question is ‘yes’, and I don’t understand why people deny that.

            It is because they have been trained to think correlations and causes are rare and interesting. This is wrong and leads people to waste insane amounts of time on pointless questions like that.

          • Greg says:

            I don’t think “was this specific storm affected by climate change” is the actual question interesting most people actually. I think the real question for the general public is “Can we attribute Katrina’s devastation to climate change ?” and its corollary “If we do something about climate change, can we avoid more Katrinas ?”.
            If the effect of climate change on any given phenomenon was really the question that matters for most people, it would be asked about every weather forecast. And yet it’s only asked when something we’d prefer to avoid happens.

      • Dave says:

        Yes, I agree with your points. I wish people were more precise with their language (or understood statistics a bit better). I’ve learned to translate incorrect statements about statistical concepts to what I think they intend, and hopefully I’m not too far off in most cases.

        There may be cases in which imprecise language leads to misunderstandings or misinformation though, so I suppose corrections are necessary sometimes.

        • Paul Hayes says:

          There may be cases in which imprecise language leads to misunderstandings or misinformation though, so I suppose corrections are necessary sometimes.

          There certainly are such cases and they can be very hard to correct, especially when probability theory – both classical and quantum – is involved.

        • Bob says:

          This is exactly what the blog poster did, switched from the term ‘attributed to’, to the term ‘affected by’, pure sleight of hand.

          We all know what people mean, the claim is that ‘extreme’ weather events are ‘more frequent’ due to climate change. This is an empirical claim, which can be tested for a given level of ‘extreme’.

          Funnily enough, when it is tested, it doesn’t stand up. Hence the need to oscillate between ‘this weather event is terrible, these events will be more common as a result of climate change’, to the opposite ‘climate isn’t weather stupid’, when events don’t support the narrative.

  4. Yeah, I don’t know what they mean either. It sounds like political mumbo jumbo.

  5. jd says:

    This does sound like the butterfly effect discussion again, which, I still find a bit hard to grasp. Certainly, it would seem that overall changes in global temperatures would have some non-zero effect on hurricanes as pointed out above. Global temperatures would seem like an important initial condition for predicting weather.

    However, Radford Neal’s comment and the subsequent discussion, and the ‘butterfly effect’, I don’t quite understand.

    I can see that systems are sensitive to initial conditions. But what constitutes initial conditions? Every tiny thing in a system always? Or everything sometimes? Is there a limit to this? (these aren’t rhetorical questions)

    I would have guessed that, in Radford Neal’s example, given millions of simulated worlds and counterfactual worlds, instead of one, there would be no difference in storm locations in most real-world/counterfactual pairs and maybe a difference in some pairs. But there wouldn’t always be a difference of some kind.

    In a counterfactual world if I remove 1 atom of Nitrogen from the air, are the storms in different spots?

    If I drop a single atom of carbon onto the ground, in a counterfactual world, would earthquakes occur at different frequencies after many years?

    Does every small piece in a system always constitute initial conditions? What constitutes the system? All matter?

    These aren’t rhetorical questions, I was just wondering how far this can be taken…it doesn’t quite sound right.

    • I can see that systems are sensitive to initial conditions. But what constitutes initial conditions?

      What constitutes initial conditions is whatever is in the “state variables” that are manipulated by the equations of motion of the system.

      In a counterfactual world if I remove 1 atom of Nitrogen from the air, are the storms in different spots?

      In theory, the (classical, non quantum) equations of motion of the earth system are defined in terms of the position and mass and type of every atom on the planet. Those equations will diverge exponentially quickly for an infinitesimal perturbation. Given enough time, the answer is *yes*

    • Just to give some idea of what we’re talking about here, the equations of motion of the atoms of the earth … it’s about 6×10^24 kg of stuff… There’s of the order of 6×10^23 atoms per kilogram, so around 3.6×10^48 atoms. Pretend we need say 1000 bits of precision in position and velocity in 3 dimensions to get accurate predictions for 10 years, so 6000 bits of precision x 3.6×10^48 objects… that’s 2.7×10^39 terabytes.

      So if you had around 10^39 terabyte hard drives you could store *the initial conditions* for the earth… With an “infinitely fast computer” and this knowledge at a point in 2010 I could tell you not only what the weather was yesterday but which movies you watched over the last year and the precise nanoscale length of the hairs growing out of your forearms.

      Removing a single atom from the 2010 data might well have changed the entire hurricane season this year… certainly if not this year, then eventually some year in the future… the scale of all this is kind of unimaginable, and the sensitivity to any given perturbation is not necessarily uniform. There are probably particular atoms you could get a dramatic effect from removing… like maybe particular atoms that control aneurisms in the brains of particular political figures or something.

      • Anoneuoid says:

        I don’t think a chaotic system works like you seem to imply. There are still “attractor” states that it will tend to be found in. It isn’t just “anything goes”.

        • Sure, I mean, taking an atom out of the system isn’t going to cause the ocean to boil, or the whole earth to freeze over like a snowball, it isn’t going to cause 40 foot long giant crabs to take over the earth and eat all the people… it’ll still be recognizable at a gross statistical level, like there are oceans and hurricanes and fish and etc, but predicting precise statistics like the mean surface temperature of each 1km patch of ground or the aggregate wind velocity over each 1km cubic air cell… that could all be drastically different 10 years in the future from very small perturbations in the 2010 dataset.

          • And some statistics that are particularly of interest to certain people could be drastically changed. Like if there’s a particular storm that causes a particular tree to fall on a particular person and kill them whereas without the atom the tree doesn’t fall down or if it does they aren’t in that particular spot…

            I don’t know about whether a particular atom’s perturbation would grow fast enough over 10 years, that’s a quantitative question that would require ridiculous computing power to address, but I guarantee that there are what most people would consider ridiculously small perturbations, like the butterfly flapping or a particular person sneezing or whatever that would result after 10 years in a difference big enough that some people would die and some people would live.

            With the most detailed weather models we have today, roundoff error level perturbations in initial conditions can cause divergence in the tracks of hurricanes by hundreds of miles over a few hours of simulation time as I understand it. This is why we can’t really predict the weather with much accuracy past a few tens of hours.

      • Anonymous says:

        > There’s of the order of 6×10^23 atoms per kilogram, so around 3.6×10^48 atoms.

        The first number is off by a couple of orders of magnitude, you’re making me want to join Andrew’s crusade against over-precision… Say 10^50 atoms and call it a day.

    • Phil says:

      jd, I don’t think this has anything to do with the butterfly effect, really. Global mean temperature has increased something like 1.5 F; in some parts of the world the increase is even larger (and of course in others it’s smaller!) Water temperatures and water levels have also increased. I don’t think this is at all analogous to a single butterfly in Brazil flapping its wings: unlike the butterfly, this is a worldwide phenomenon that has measurably changed a major feature of the climate.

      The ‘butterfly effect’ is all about a tiny perturbation in initial conditions, at a specific place and time, which _eventually_ leads to substantial worldwide changes in behavior. One can imagine that it would take thousands or even millions of years for that large a change to take place. Only a pedant would say that a butterfly flapping its wings in 1960 ‘affected’ Hurricane Katrina. Of course it’s literally true — not every air molecule would be in exactly the same place — but that is not the kind of trivial effect I am talking about. Compared to a counterfactual ‘no-global-warming’ climate, there have been major changes, certainly enough to substantially affect hurricane tracks and wind speeds.

      • Anoneuoid says:

        Only a pedant would say that a butterfly flapping its wings in 1960 ‘affected’ Hurricane Katrina

        Mathematics/statistics is very pedantic. Best not use those tools then if you don’t want to end up confused and arguing about nothing.

        • Dan F. says:

          “Mathematics/statistics is very pedantic. Best not use those tools then if you don’t want to end up confused and arguing about nothing.”

          Best to use physics when talking about global atmospheric dynamics. Almost nothing that pedantic mathematics or statistics can say is remotely relevant to the discussion here.

          • Anoneuoid says:

            Almost nothing that pedantic mathematics or statistics can say is remotely relevant to the discussion here.

            This is bizarre to me. First of all, physics without math? Second, Climate (long term averages) without statistics?

            I have no idea what point you were trying to make.

            • Dan F. says:

              The “butterfly effect” is pseudo-scientific nonsense. In any real physical system on the scale of an atmospheric system, intrinsic noise and measurement error (and a host of other issues) swamp any effect at the scale of a butterfly. In any case the models are not accurate to such a scale.

              • You seem to miss the point of the “butterfly effect”. The part about the butterfly was just an analogy made up by Edward Lorenz. The actual effect he originally observed (if I remember correctly) was that he typed in the initial conditions of his simulation to somewhat less than full machine precision so he could repeat his calculation and found that it diverged wildly from the original calculation. See the quote here:

                There’s nothing pseudo-scientific or nonsense about this: real world dynamical systems diverge exponentially quickly from their original trajectory after even infinitesimal perturbations.

                That mathematical fact gives a fundamental limit on how long any simulation of a dynamical system can successfully predict real world outcomes, and it’s not just about “if we had better measurements” or “if we could refine the scale of the simulation” because fundamentally you need to refine the scale down to *a 1-1 mapping onto the actual system*. You need a “clone” of the system to “do the calculations”

                The fact that in real world models of real world systems there are already effects that totally dominate butterflies isn’t so important, what’s important is that even if you could somehow build a model of the earth that had a resolution down to the scale of a butterfly (say 1cm cubes) you still couldn’t really get much out of it in terms of accurate prediction… it’d diverge exponentially from reality just due to the missing information from an unobserved butterfly flap.

              • Matt Skaggs says:

                “There’s nothing pseudo-scientific or nonsense about this: real world dynamical systems diverge exponentially quickly from their original trajectory after even infinitesimal perturbations.”

                No they don’t. When we think of “systems” we are usually talking about systems with some stability. Stability is created by damping, so humans design systems that are strongly damped. Any natural system is also likely to be strongly damped.

                The butterfly analogy was wrong. Lorenz made a guess at what was happening and he guessed wrong, but he created a social meme that lives to this day. It is a confusing topic.

                What does happen is that systems are subject to resonance points. It is at this resonance point where a small perturbation can create a large effect. Resonance points exist because of feedback. Only feedback can cause a system to fly out of control due to a small perturbation. There is no feedback to amplify the flap of a butterfly, so it is damped out to nothing.

              • Anoneuoid says:

                DanF. and Matt Skaggs:

                Neither of you seem to be familiar with what “butterfly effect” refers to. It is simply that it is impossible to tell the future state of a deterministic system unless you know the exact initial conditions. This happens when measurement error and “noise” are present as well. Here is the classic lorentz attractor with noise added:


              • Matt Skaggs:

                This is an instance of the mind projection fallacy that Jaynes discusses.

                Because of the actually very chaotic dynamics of the weather system, even detailed knowledge today of the actual weather conditions doesn’t help us predict the weather *at all* 120 days from now. But because of the attractor states of the weather system, our knowledge about what’s going to happen 120 days from now isn’t zero, it’s that the weather will be sort of like any of the previous say 20 years worth of weather data for days during that same week each year…

                Given our knowledge of the weather system today, our distribution over the weather tomorrow is *not* like the distribution of previously observed weather over the last 20 years for Sept 6-13 in the years 1999-2018 or whatever, it’s more specific because for short term predictions weather models give useful and accurate information.

                As prediction time goes out farther, our distribution over weather converges to a constant distribution informed only by history.

                The incorrect “mind projection” assumption is that knowledge today doesn’t help us predict the weather 120 days out, and so whatever happens today, the physical weather will “be the same” 120 days out. But of course it will be vastly different, just as if we choose different random number seeds, the 120th random number we draw will be very different between two sequences of RNG outputs. It’s only our knowledge of the weather which will be the same independent of knowing all sorts of detailed sensor and satellite data.

      • jd says:

        Actually, that was the main point of my comment, which I perhaps should have worded better. Global temperature changes seem like an initial condition that is of import to the location and severity of storms, where as removing a single atom of N in a counterfactual world does not. Which led to my questions about what really constitutes initial conditions, to which it appears to be all matter, according to Daniel. I understand that in a giant deterministic math model, that the math would work out to show this, but I guess I’m still a bit skeptical if that is actually how things work in the real world.

        But yes, I agree with your point, global temperature changes are not a tiny perturbation. I was asking more if there was a limit to how small a perturbation could be and still make an effect, and if an effect always exists in every real-world/counterfactual pair.

        • When Anoneuoid says that everything causes everything else, he really means it, and it’s true.

          Taken long enough through time and single atom perturbations will cause a system like the earth to diverge wildly from alternatives. But there is no computable mapping from perturbation to outcome for most kinds of perturbations, so this component is called “random” in most models.

          The point is still important however because if you ask “is there an effect of x on y” for any two things, the answer is logically yes for basically any two things. The better question is “is there a sizeable and somewhat consistent and manipulable effect of doing x on outcome y, and how big is it?” but this isn’t answered by the logic of NHST.

        • My mention of NHST is just because that’s the context in which Anoneuoid usually brings up the everything causes everything idea…

          Suppose that a small child drags his feet while his mother tries to get him to cross the street. This delay causes hormones to be released in a driver of a car that increases his anger level to the point that later in the evening he finally decides to kill his wife. If the child had not dragged its feet the abusive man would have waited a few days, and the wife who had been making plans to run away to a shelter the next day would have survived. Clearly, the child (and many other things) caused the murder.

          But because detailed knowledge of these connections is required to manipulate the system through the child to prevent the murder, we can’t reliably make a connection between events so that we can act to prevent the murder by strategically giving certain children toys to distract them, or convincing people not to wash their cars at strategic times to avoid glare under certain circumstances that lead to particular traffic accidents or whatever….

          What we can do is things that reliably and predictably have a good effect, like reporting abusive behavior to the police and filing for restraining orders and funding charities that help abused people, or putting up traffic warning signs or designing better automotive safety equipment or soforth.

          Once you understand that everything affects everything it becomes much easier to ask a more useful question, for example “what is the range of sizes of the effect we can expect from this particular manipulation” and “how much do we care about that effect size” and “what do we have to give up (the cost) in order to implement that manipulation” and “which of these different manipulations produces the best overall outcome including costs and benefits”

          I think if we could get scientists to ask the right questions in the first place, we could eliminate a lot of bad science. For example to return to the climate issue that started this, instead of “is there anthropogenic global warming” which is a stupid question which is still asked in all seriousness in public discourse, we should ask “how much has human release of carbon dioxide altered the balance of energy flux, and how would that change over the next few years if we imposed certain policy changes?”

          Other useful questions are ones like what James mentions below:

          How has human released CO2 affected the risk of storms in a statistical sense? This may be relatively accurately predicted from model simulations even if any one run of the simulation doesn’t correspond to anything that actually happens on the earth. It is often much easier to correctly predict statistical properties of aggregates than any specific thing that might happen, this is because aggregates are relatively insensitive to the specifics, as they can occur in many different ways.

          • jd says:

            I can see your point, and as I mentioned, I can see how this would all work in a giant math model on a theoretical mega computer. Also, the point is a good one about NHST, and is along the lines of the Meehl 1967 paper that Anon or Anonymous linked to on some post not long ago. For sure these lead to better questions and better thinking as you, Phil, and other comments here point out.

            However, I still remain a bit skeptical that the world works this way for the examples I gave, with the tiny perturbations.
            As for the examples above, a child dragging its feet doesn’t always cause a murder, and a washed car doesn’t always cause a wreck. It can. But it doesn’t have to. You say “strategically”, and that seems critical. It still seems that for tiny perturbations in many simulations of real-world/counterfactual pairs that the majority of them would be little to no effect over the long run with some of them having some detectable effect and maybe few a large effect.

            • MP says:

              My understanding (very much non-expert) is that even these very tiny, atom-scale perturbations can have very large effects in a surprisingly short time. So, just making up numbers here, but I’d guess that after a few years (rather than thousands of years), even the tiniest of changes would mean weather so different that it wouldn’t even make sense to talk about how a particular storm was different because there wouldn’t even be an identifiable counterpart in the counterfactual scenario.

              At the same time, the idea of attractors is that you’d have completely different details with broadly similar characteristics — average number of storms, broad locations, rough time of year, distribution of intensities, etc.

              But, as many people have already said, there’s good reason to think that climate change is leading to changes in those broad characteristics.

  6. RIchard McElreath says:

    There is a literature on fraction of attributable risk for climate change. I’ve collected some of it, but never gotten around to reading it. Relevant Bayesian example from 2008: DOI 10.1111/j.1539-6924.2008.01070.x


    The recent decision of the U.S. Supreme Court on the regulation of CO2 emissions from new motor vehicles shows the need for a robust methodology to evaluate the fraction of attributable risk from such emissions. The methodology must enable decisionmakers to reach practically relevant conclusions on the basis of expert assessments the decisionmakers see as an expression of research in progress, rather than as knowledge consolidated beyond any reasonable doubt. This article presents such a methodology and demonstrates its use for the Alpine heat wave of 2003. In a Bayesian setting, different expert assessments on temperature trends and volatility can be formalized as probability distributions, with initial weights (priors) attached to them. By Bayesian learning, these weights can be adjusted in the light of data. The fraction of heat wave risk attributable to anthropogenic climate change can then be computed from the posterior distribution. We show that very different priors consistently lead to the result that anthropogenic climate change has contributed more than 90% to the probability of the Alpine summer heat wave in 2003. The present method can be extended to a wide range of applications where conclusions must be drawn from divergent assessments under uncertainty.

  7. Dave Smith says:

    My city was not destroyed today by a major hurricane. Should I thank global warming?

  8. KC says:

    I like to distinguish between “necessary cause” vis “sufficient cause”. Global warming is not a “sufficient cause” of any particular hurricane. But it is a “necessary cause” of unusually strong hurricanes. Would you agree? Is it helpful?

  9. ariel says:

    May be the influence of climate change on that specific weather event could have been insignificant.

    • Phil says:

      Ariel, this is what I disagree with. We wouldn’t even have had that ‘specific weather event’ in a no-global-warming world. It’s like…eh, OK, this isn’t the best analogy, but suppose a football team scored a touchdown at 3:53 PM on September 23, 2018. It would make no sense to ask “would they still have scored that touchdown if they had the same team they had in 2015.” Sure, maybe they would have scored a touchdown at the same time on the same date — very unlikely, but possible — but there’s no way it could have been exactly the same touchdown.

      The climate has changed a lot in the past 50 years. The weather today is different everywhere in the world from what it would have been if the climate hadn’t changed.

      • jim says:

        “but there’s no way it could have been exactly the same touchdown.”

        But it’s so similar no one can tell the difference.

        A week ago I walked on the beach at Mukilteo. It was a great walk. The other day I did it again. The beach – in PhilTalk – was utterly and totally and irrevocably and unchangeably different. Relentless tides, the wakes of boats, the hundreds of beach walkers, seagulls dropping on the sand, crabs washing up in the waves…they all must have had an impact on the beach! How could I ever enjoy another walk on the beach exactly the same as I did the previous week? The beach *was* *extremely* different! Perhaps, at some nannoscopic level, my enjoyment was different – maybe even less. And I had absolutely no clue.

        I put my fist in a bucket of water and pulled it out again. Everything changed. But nothing happened. The world returned to it’s matrix of randomness and in the end the event was meaningless.

  10. David Paterno says:

    All sound points above.

    But isn’t all this a distraction from the real cause of hurricane impact – storm names?

  11. jim says:

    Hi Phil,

    I’m so glad you got that off your chest! :)

    But it doesn’t matter if a storm is “affected” by climate change or not. No single storm says anything about climate or climate change or the lack thereof. For all you – or anyone – knows, storms could be *negatively* “affected” – that is, of *lower* intensity and frequency – because of climate change. No one knows one way or the other. So in the end, since you can’t tell *how* a storm is “affected” by climate change, it hardly matters whether it is or not.

    It doesn’t make any sense to talk about the relationship between storms and climate change unless you’re referring to periods of 20 years or more to account for variations in all the major oscillations like ENSO, PDO, NAO, IOD etc. Really, even if a category 8 storm whopped Goose Bay Labrador tomorrow, it wouldn’t have any meaning for climate. Climate is a long-term average. A single storm is the position of an electron at time t. Climate is the ten bazillion different positions of the electron over time over dt.

    • Phil says:

      I agree with your second paragraph, and I agree with your first paragraph inasmuch as I don’t think ‘was a storm affected by climate change’ is a question that it even makes sense to think about. But — and this is what drives me nuts! — people do think about it. They think about it so much that a large fraction of reporting on climate change includes some sort of statement like the one that prompted this post…and that statement was written by climate scientists!

  12. James says:

    There is a long and strong literature on climate attribution. It’s imperfect in that it relies heavily on climate models, but it’s still interesting. The basic premise is that you run a really long simulation at each of several levels of carbon dioxide in the atmosphere and you compare some frequency and intensity statistics for the event you’re caring about. So for example climate change shifts the distribution of rainfall for events like hurricane Harvey to the right, which one can interpret either as making Harvey more intense or as making the frequency of Harvey-like events higher.

    Again this is always model-based, hence imperfect, but overall seems quite good. The huge amount of noise in hurricane paths and lots of external factors (both of which others have alluded to above) makes trying to do this with a regression analysis quite difficult as we just don’t have enough good data. But bottom line is that the story on hurricane intensity seems to be fairly clear (warmer = stronger) while the story on trajectories is much less clear.

    None of which contradicts Phil’s excellent point that it’s not reasonable to say whether a particular storm was *affected* by climate change any more than to say it was *caused* by climate change. But we can talk about shifting probabilities, and we can estimate these shifts [moderately well] using climate models.

    • Dalton says:

      +1 and then some.

      Except I wouldn’t caveat (deterministic) model-based inference so much. With weather and climate models are the only way we can get true replications on which we can do statistics. There is no planet N with all factors identical except for our variable of interest.

  13. Emmanuel Charpentier says:

    For the francophones:

    “Un barbu, c’est un barbu. Trois barbus, c’est des barbouzes” (Michel Audiard).

    [ Very roughly : “One beard: that’s a beard. Three beards : that’s spooks”, but it’s funnier in French ]

    I think that this is relevant…

  14. matt says:

    Yes.. global warming affects specific hurricanes the same way literally anything effects everything else. Is this really that profound of a point? It seems like you are making a fuss out of nothing. Obviously the question people are asking is “has climate change made it more likely for strong hurricanes to form?”

    Do you really think they are trying to claim that “in a world where global warming existed on a lesser scale than it does today, hurricane Katrina would have occurred exactly as it did in our lived experience”. What are you, an idiot?

    • Terry says:

      “Yes.. global warming affects specific hurricanes the same way literally anything effects everything else. Is this really that profound of a point?”

      This was my reaction.

      If you interpret the statement literally, it is a completely empty statement. So, being a normal human being, I assume the writer was saying something non-trivial and just shortened the statement the way ordinary humans do in ordinary human speech.

      As best I can see, there is an implicit “significantly” in the statement, so the writer was saying that “there is no way to prove that “Katrina … was … *significantly* affected by global warming.”

      • But that’s VERY WRONG. If back in 1800 we had somehow invented a CO2 free way of generating similar energy and done broadly all the same human scale things we’ve been doing without CO2 emissions there is *absolutely no doubt* that Katrina or even a very similar intensity storm to Katrina which made landfall within a week of the same date and a few miles of the same eye track… it simply would NOT have happened….

        now, whether some other storm on some other day with some similar level of intensity would have made landfall in the vicinity of New Orleans within say a decade of the actual date… who knows. At this scale it’s a largely statistical question of how often big storms hit regions in the gulf… those statistics over decade long scales might still be similar enough so that our assessment of the risk of a Katrina like event during the year say 2005 as made on Jan 1 2005 would remain basically the same… but this is just saying basically that with or without global warming, our knowledge of events even weeks in the future much less months or years is very limited.

        • Carlos Ungil says:

          I think when Terry says “significantly” means “in a meaningful way”. At least that’s the qualifier that is the implicit qualifier in the original formulation so it’s not just an empty, pedantic, trivial statement.

          If Mount Pinatubo had not erupted in the Philippines in June 1991 there is no doubt that Katrina or even a very similar intensity storm to Katrina which made landfall within a week of the same date and a few miles of the same eye track would not have happened…. but understandably nobody writes newspaper articles or blog posts about that.

          • If humans had emitted only say 10% of the CO2 that they’ve actually emitted, a hurricane like Katrina would never have happened… how is that “not affecting Katrina in a meaningful way”?

            The only way that I can understand statements connecting global warming and Katrina is something like “if humans had emitted CO2 at vastly reduced rates, then the risk in any given 10 year period of a Category 5 Hurricane making landfall anywhere in the US would have been reduced from X to Y” but as you yourself pointed out Carlos there’s no such obvious relationship in the data.

            • Carlos Ungil says:

              You’re giving yourself your own answer to how statements about global warming and Katrina can be understood in a meaningful way. I don’t think the difference between having Katrina in New Orleans in late August or having an unnamed similar hurricane in Cuba in early November is “meaningful” in the context of this discussion but obviously affected people may disagree and they may want to blame global warming or Mount Pinatubo.

              • Sure talking about risk rates is meaningful.

                But I don’t think a statement

                “no way to prove that Katrina either was, or was not, affected by global warming. For a single event, regardless of how extreme, such attribution is fundamentally impossible”

                can be interpreted as a statement about the risk rate of large hurricanes somewhere in the western hemisphere.

                And there *is* a way to at least begin to figure out what risk rates should be like. For example you can run two different batches of GCMs, one batch involves releasing CO2 throughout the period 1800-now and one doesn’t, and you can look at the risk rate for large hurricanes in the two batches. This can be argued against, but it is a valid way to at least study risk rates…

                if that’s what they meant, then what’s all that stuff about “single events” and etc? It’s mumbo jumbo.

  15. Steve C says:

    The author of this article might be assuming that there will be more tropical cyclones, or more extreme tropical cyclones, as a result of global warming. I recommend reading chapter 3 of the Special Report on Extremes from the IPCC ( Nothing is clear.

    For example, Hurricanes and climate: the US CLIVAR working group on hurricanes, American Meteorological Society, Kevin Walsh et al (2015):

    “At present, there is no climate theory that can predict the formation rate of tropical cyclones from the mean climate state.”

    Another example, Modeled Impact of Anthropogenic Warming on the Frequency of Intense Atlantic Hurricanes, Morris A Bender et al, Science (2010):

    “Some statistical analyses suggest a link between warmer Atlantic SSTs and increased hurricane activity, although other studies contend that the spatial structure of the SST change may be a more important control on tropical cyclone frequency and intensity. A few studies suggest that greenhouse warming has already produced a substantial rise in Atlantic tropical cyclone activity, but others question that conclusion.”

    Here is IPCC AR5, Chapter 2, p. 216:

    “AR4 concluded that it was likely that an increasing trend had occurred in intense tropical cyclone activity since 1970 in some regions but that there was no clear trend in the annual numbers of tropical cyclones. Subsequent assessments, including SREX and more recent literature indicate that it is difficult to draw firm conclusions with respect to the confidence levels associated with observed trends prior to the satellite era and in ocean basins outside of the North Atlantic.”

    And here is the lengthy SREX conclusion, p. 159:

    “Detection of trends in tropical cyclone metrics such as frequency, intensity, and duration remains a significant challenge..

    ..Natural variability combined with uncertainties in the historical data makes it difficult to detect trends in tropical cyclone activity. There have been no significant trends observed in global tropical cyclone frequency records, including over the present 40-year period of satellite observations (e.g., Webster et al., 2005). Regional trends in tropical cyclone frequency have been identified in the North Atlantic, but the fidelity of these trends is debated (Holland and Webster, 2007; Landsea, 2007; Mann et al., 2007a). Different methods for estimating undercounts in the earlier part of the North Atlantic tropical cyclone record provide mixed conclusions (Chang and Guo, 2007; Mann et al., 2007b; Kunkel et al., 2008; Vecchi and Knutson, 2008).

    Regional trends have not been detected in other oceans (Chan and Xu, 2009; Kubota and Chan, 2009; Callaghan and Power, 2011). It thus remains uncertain whether any observed increases in tropical cyclone frequency on time scales longer than about 40 years are robust, after accounting for past changes in observing capabilities (Knutson et al., 2010)..

    ..Time series of power dissipation, an aggregate compound of tropical cyclone frequency, duration, and intensity that measures total energy consumption by tropical cyclones, show upward trends in the North Atlantic and weaker upward trends in the western North Pacific over the past 25 years (Emanuel, 2007), but interpretation of longer-term trends in this quantity is again constrained by data quality concerns.

    The variability and trend of power dissipation can be related to SST and other local factors such as tropopause temperature and vertical wind shear (Emanuel, 2007), but it is a current topic of debate whether local SST or the difference between local SST and mean tropical SST is the more physically relevant metric (Swanson, 2008).

    The distinction is an important one when making projections of changes in power dissipation based on projections of SST changes, particularly in the tropical Atlantic where SST has been increasing more rapidly than in the tropics as a whole (Vecchi et al., 2008). Accumulated cyclone energy, which is an integrated metric analogous to power dissipation, has been declining globally since reaching a high point in 2005, and is presently at a 40- year low point (Maue, 2009). The present period of quiescence, as well as the period of heightened activity leading up to the high point in 2005, does not clearly represent substantial departures from past variability (Maue, 2009)..

    ..The present assessment regarding observed trends in tropical cyclone activity is essentially identical to the WMO assessment (Knutson et al., 2010): there is low confidence that any observed long-term (i.e., 40 years or more) increases in tropical cyclone activity are robust, after accounting for past changes in observing capabilities.”

    Hopefully this clears up the subject. As in, it’s not clear and there is no consensus. And yet, all good people believe..

    • Bob says:

      It’s a religious position. They’re not interested in reality, weirdly the pedantic, hole-picking statisticians look at their shoes..

      “This is what climate change looks like: it hits vulnerable communities first.
      I can already hear climate deniers screeching: “It’s always been like this! You’re dim,” etc.
      No. This is about science & leadership. We either decarbonize & cut emissions, or we don’t & let people die.”

    • Phi says:

      Steve, when you say “the author of this article” I’m not sure if you’re talking about this post, or about the article that prompted this post (which had multiple authors).

      Assuming it’s the former: I am not assuming anything about whether there will be more or fewer hurricanes, or whether they will be weaker or stronger. I am aware that there are lots of uncertainties in the models.

      What I’m saying is that the hurricanes we experience are not the hurricanes we would have experienced in the absence of global warming. And I’m not talking about tiny little differences in where the air molecules are. The hurricanes occur in different places, at different times, and are of different intensities, than they would be without global warming. I make no claims about whether they are weaker or stronger, earlier or later, farther east or farther west, etc. etc., I only say they are different. I do not think a statement like “we don’t know whether Katrina was affected by global warming” makes any sense at all; it’s like saying “We don’t know if the touchdown the Cowboys made yesterday was affected by the fact that they are not the Dolphins.”

      • Or, like “we don’t know if the dinner you had last night was affected by the fact that your embryo was successfully fertilized 9 months before you were born”

      • Steve C says:


        Thanks for replying. I did mean your article.

        It was unclear what point you were trying to make, and I have seen numerous statements similar to yours which had the main point “things are getting worse because of ‘climate change’ (global warming is a better term), we can’t tie the cause of a specific event to global warming, but that doesn’t mean the frequency / magnitude of these events aren’t getting worse”. My short set of quotes from climate papers and IPCC reports was addressing that idea.

        If your point was “we don’t know much about anything regarding changes in tropical cyclones” then I’m a bit more aligned to your point.

        Perhaps you can clarify your point?

        • Phil says:

          Steve C,
          My point is that the weather everywhere at every moment is different from what it would have been in the absence of global warming. To ask whether any given weather event was affected by global warming doesn’t really make sense. If you must have a yes/no answer then the answer is Yes, it was affected by global warming, but that’s misleading because in general the ‘weather event’ would not have even happened in the counterfactual world. It is fairly unlikely that there would have even been a hurricane in the Gulf of Mexico during the week that Katrina hit New Orleans, so what does it mean to ask whether Katrina was affected by global warming? This is not a statement about whether hurricanes are getting more or less likely, or more or less severe.

  16. Luke says:

    If more people thought like Bayesians then Phil’s point would be obvious beyond statement. People generally seem to think like frequentists, either there is an effect or there is no effect, and if you can’t measure the effect then we say (wrongly) there is no effect. On a continuous distribution, the probability of 0 effect is 0, and so in practice all failures to reject the null are type 2 errors.

  17. MP says:

    It’s funny what people get worked up about. And I say that as someone who gets worked up about some very particular things.

    I mean, sure, any hurricane is “affected” by climate change, in the sense that everything is affected by everything. True, but trivial. Is that your objection? If it is, fair enough. But it’s like getting worked up over people using “literally” to intensify something that is clearly figurative. Which, to be clear, I do get worked up about.

    But do you really not know what they mean? Well, OK, I don’t know what they mean either, but I can take a pretty good guess that they’re countering (albeit sloppily) regular claims that this or that storm was CAUSED by climate change, claims that a whole different set of people get worked up about.

    • Phil says:

      Without climate change there would not have been a hurricane at that location on that day…or at least it’s quite unlikely. Perhaps an even stronger hurricane would have hit New Orleans earlier in the year, or perhaps Galveston would have been devastated instead of New Orleans. I am not making any claims about climate change making hurricanes worse or better or stronger or weaker or whatever. What I am saying is it is all different. Very different. So there’s a technical sense in which every hurricane we experience was ’caused by climate change’ but I don’t like that terminology either, it’s very misleading: it’s not like we wouldn’t have any hurricanes if there were no global warming!

      • Blah says:

        But Phil, can you respond to people here who have said your post / complaint is trivial? Climate change affected Katrina just like my mother turning left out of her house on June 25 1976 instead of right affected everything that came after it in our worlds history. This is simply uninteresting; everything affects everything. I can’t believe you are being this obtuse, reminds me of when you posted about NIMBYs way back. It’s a bit ironic really, how unwilling Andrew, and you, are to admitting mistakes on this blog. There is so much goalpost shifting that occurs so as to avoid admitting ones incorrectness.

        • If you think Phil is being obtuse, can you please say what you think was really meant by “there is no way to prove that Katrina either was, or was not, affected by global warming. For a single event, regardless of how extreme, such attribution is fundamentally impossible”

          because as far as I can tell it’s a meaningless statement and Phil is right to be annoyed by it.

          • blah says:

            I can tell you what it didn’t mean: “I think that in a world with less (or, just different) levels of global warming, I think that hurricane Katrina would have occurred EXACTLY as it did in our real experience”.

            They clearly just meant that it’s hard to say whether global warming has contributed to greater frequency or greater strength of these types of hurricanes. Any normal person should be able to see this.

            • If they want to talk about frequency and severity why say “For a single event, regardless of how extreme, such attribution is fundamentally impossible”

              It’s a clear statement that they are discussing “a single event”

              The point is they are muddling the waters by not saying what they mean. If what they mean is “as global warming progresses more intense hurricanes will become more common” then just say that, but also btw as linked above there is no clear data on such a relationship…


              this kind of frequency and severity statement is an almost purely model based prediction, and the models are far from unambiguous.

              the problem with the statement is it’s weasel words that sound meaningful to uninformed people but have no real meaning. instead it invites people to supply whatever substitute meaning they like best which you did.

              IMHO they don’t mean what you said at all. IMHO they mean that you CAN see greater frequency and severity but you have to look at trends from multiple hurricanes. so you see how weasel words lead to everyone supplying their own meaning.

              one reasons to not come out and say what you actually mean is that if you do, someone like Carlos will link you to some data that shows that your confident assertion about frequency and severity is not at all supported by data… so we get weasel words.

          • Bob says:

            It’s a meaningless statement that he’s created for himself to be annoyed by. We all know what the claim is, because every time an event like Katrina occurs it’s wheeled out as ‘evidence’ for a particular political position. It’s invoked all the time, so to be ‘confused’ by it and claiming ‘everything causes everything’ is deliberate obtuseness.

            • > We all know what the claim is

              I don’t… here’s how it parses out to me:

              “there is no way to prove that” = “we can’t be 100% or even say 90% sure”

              “Katrina either was, or was not, affected” = “that this storm would or would not have happened in substantially the same way”

              “by global warming” = “if CO2 had or hadn’t been emitted in the way it actually has been”

              but quite literally we CAN be nearly 100% sure that there wouldn’t have even been a storm in New Orleans that week if a couple hundred years of CO2 hadn’t been emitted… Because the weather is so chaotic that no doubt something else entirely would have been going on that week. If you choose a random week in August over the last 100 years then almost never was New Orleans being destroyed by a hurricane… so we can be pretty sure that if we totally altered the global weather by major changes to initial conditions… we’d just get something “like a typical August” where no major storm was occurring.

              This is so obvious that instead we each supply some alternative meaning “what they really meant was” … but the range of supplied alternatives is very broad, and incites flame wars.

              If they simply said

              “Global Warming is expected to increase the frequency and severity of storms and so we should expect more destruction from Katrina like events over the next few decades than we would otherwise have had if so much CO2 hadn’t been emitted” then they’d have a meaningful statement. Of course, the data linked above don’t really show either way whether that’s true.

              If they said “if we reduce our CO2 emissions by X amount over the next few decades, we expect to reduce the destruction to humanity from hurricanes by Y over the next 100 years” that’d be meaningful… but of course they don’t have anything like that specific of a model.

              But they didn’t say those kind of things, because they can’t really back those up with anything solid, it’s really just theory with a lot of uncertainty… instead they used some weasel words that don’t really parse out to anything specific. It’s all about negatives “we can’t be sure” and “there’s no way to know” and soforth.

              • Bob says:

                You’re looking for a precise statement of an hypothesis, there isn’t one, because it can’t be justified. That’s why it quickly becomes a moral claim, the more hysterical the better.

                The claim is this, just what I posted above:
                “This is what climate change looks like: it hits vulnerable communities first.
                I can already hear climate deniers screeching: “It’s always been like this! You’re dim,” etc.
                No. This is about science & leadership. We either decarbonize & cut emissions, or we don’t & let people die.”

                The storms are more frequent, climate change is racist, blah blah. It’s all variations on the same unfounded claim. All dishonest. The other side say you can’t prove it was a result of climate change, which is true.

                The poster is latching on to the use of the word ‘affected’, then goes down some rabbit hole of ‘everything causes everything’, but that’s an empty statement and he’s only making it because the wrong people are right and the right people are wrong.

                But the central claim is that of the environmentalists, they are making the claim of a link, and a claim of increased severity and frequency. That’s what needs to be scrutinized, but isn’t.

            • Phil says:

              I genuinely do not know what people mean when they say something like “we can’t say whether [such-and-such a weather event] was affected by global warming.” The claim seems nonsensical to me. And I am _not_ talking about the trivial fact that ‘everything affects everything else.’ As I’ve said several times in this comment string, I’m not talking about where the air molecules are in the hurricane, I’m talking about the fact that the hurricane wouldn’t even have happened. If, by coincidence, in the absence of global warming there would still have been a hurricane at that approximate location and at that approximate time, it would be just that — a coincidence — and could not be said to be the same hurricane but happening in a slightly different way.

              I don’t think I’m being ‘obtuse’ here, I am calling out people for speaking nonsense.

Leave a Reply