Global Warming’s Pretend Pause

A new paper by Risbey et al. examines the so-called “pause” in global temperature, and demonstrates convincingly that it wasn’t a real phenomenon, it was just random fluctuation that can look like a pause all along. I’ve been saying this for some time now. I’m also a co-author on the paper.

Much of the heart of the matter is the statistics behind the issue, a topic which I (and my regular readers) know pretty well. You can find this discussed in the paper, in numerous other papers, and in quite a few posts on this blog. The bottom line is: when claims of a “pause” or “hiatus” or “slowdown” are actually put to the test, they fail. Sorry, Charlie … no pause.

But this paper also gathers together lots of information about the claims of the pause. I was surprised to learn that there are over 200 published papers discussing it as though it were a real phenomenon! Most take its existence as their starting point, as a given, and those few who do actually test the question, have tended to leave out some crucial statistical aspects (e.g. the “multiple testing problem/selection bias problem”).

There’s also a surprising amount of confusion about just what and when the pause was. Not only are the start and end times all over the place, things are vague enough that in a great many papers it’s just not possible to pin down when the pause took place.

And there’s some interesting discussion about the social aspect of the “pause” claim and its effect on the public discourse. The paper is open access, I recommend a good read.

A companion paper by Lewandowsky et al. goes a step further, testing claims about a supposed “pause” in comparison to model projections of temperature change. It too is open access. Spoiler alert: dead pause.

UPDATE: An interesting take on the issue in The Conversation by Stephan Lewandowsky and Kevin Cowtan, two other authors on this pair of papers.

This blog is made possible by readers like you; join others by donating at My Wee Dragon.


24 responses to “Global Warming’s Pretend Pause

  1. What is noise? I try a definition: Noise is information, that cannot (or not yet) be attributed to a known cause. The “pause” – and any succeeding “pauses” – is a noise fluctuation and no signal, because we cannot attribute it to any cause. Opposed to this, we can attribute the trend, and indeed some other fluctuations, to things like volcano eruptions, solar power fluctuations and ENSO. According to above definition they would be no noise, but signal.
    IMV it does make sense to take the “pause” (and any following one) seriously and inquire its causes, in order to let science progress and possibly transform some noise into signal.
    The problem in this case is the wording and its associations. “Pause” somehow insinuates, that not only the effect, but the whole thing including the causes did magically stop working. Instead, some of the dozens of more important interactions cancelled out, as to produce a temporary slowing of the mean temperature.

    • “Noise” is any unwanted signal. It may contain unwanted information , e.g., a competing nearby conversation, or random (i.e. unpredictable) fluctuations (no information at all), e.g., static during AM reception of a broadcast.

      A “pause” is a cessation or period of no change. A traffic light pauses between state changes. Nothing magical about it at all.

      The climate pause refers to a temperature change statistically indistinguishable from no change — a plateau. Its length is/was determined by extending a regression backward in time from the current date. Why it existed is anyone’s guess but, at least perhaps until recently, it was possible to regress a zero trend line backward from the present or then present for something like 15-18 years. That it happened at all — and was a surprise — is an indicator that there is much left to understand about the causes of recent climate change.

      • Here’s the thing: it was *not* a surprise to anybody who knew the record reasonably well–certainly it was not for Our Host, who frequently pointed out the lack of statistical significance in the purported ‘pause.’ Certainly it was not for Santer et al., 2011:

        Because of the pronounced effect of interannual noise on decadal trends, a multi‐model ensemble of anthropogenically‐forced simulations displays many 10‐year periods with little warming. A single decade of observational TLT data is therefore inadequate for identifying a slowly evolving anthropogenic warming signal. Our results show that temperature records of at least 17 years in length are required for identifying human effects on global‐mean tropospheric temperature.


        You can find such periods in the observed record, too,–for instance:

      • That it happened at all

        Here’s the thing: it didn’t happen.

      • @Doc Snow, December 30, 2018 at 3:23 pm

        There doesn’t seem to be a reply button on you post so I’ll put it here.

        Interestingly, the length was 28 yrs 7 months exceeding Santer’s 17 years. Maybe he meant 19 years?

        The last two sentences in your quote are non-sequiturs regarding the pause. The pause shows a disconnect in understanding of climate causes regardless of what they are believed to be.

        Which models reflected observations can only be known after the fact. You don’t know ahead of time which will do so. It’s a bit like predicting the winner or a horse race by saying one of them will win. With the climate models you can’t even be that certain. Thus the need to average all of them and the average doesn’t conform to observations.

      • Sorry. Meant 18 yrs 7 mo.

      • 17 years was regarding a Type II error, not a Type 1 error. An estimate of how many years–on average–before the null hypothesis is NOT being falsely rejected at an appropriate level of reliability. Or, how many years, on average, you expect to need to reliably identify a trend which really does exist in the data. A year more is simply not that unexpected.

        The pause was a “surprise” in the same way that undergrad stats students are “surprised” when they find that so-called “statistically unlikely” runs of 5 or more heads in a series of 100 coin flips are expected with a reasonably high probability, NOT “unexpected”. Any surprise here is more unawareness of how scanning a series for likely “significant” areas and then testing for it is simple statistical nonsense unless the biases introduced by the scanning procedure are explicitly accounted for.

        The simple fact is your procedure does not yield the correct probabilities because you are breaking the fundamental assumptions which underlie your (incorrect) statement.

        Tamino has been at great pains for many, many years now on this point. Why it never seems to penetrate is beyond me, personally. All it takes is a half hour with some pennies or 10 minutes with R or a spreadsheet to prove it to yourself.

      • Our results show that temperature records of at least 17 years in length are required…

        I trust I don’t need to expand on ‘at least’?

        I would also point out that in fact you can’t say that the ‘pause’ was this or that length, because there has never been any generally agreed definition of said ‘pause.’ If you search the literature, you’ll find quite a variety of candidate periods.

        Particularly amusing in this regard is our old ‘friend’, Lord Montwhatsisname of Brenchly, who periodically put forth graphs purporting to show “no warming for 18 years and some odd months.” What made them funny was that the *start date* of those graphs kept creeping forward. For years, warming was ‘invisible’ for a roughly 18-year retrospective period, rather as if it were chasing us forward through time. Then in 2014, it definitively caught up. The gods of variability giveth, and they taketh away.

        (I suppose in all fairness I should also mention that he also had to cherry pick the data set involved to get his results, by selecting the RSS satellite record. Then they had to go and spoil his fun by correcting for an orbital drift bias that had become obviously wrong.)

        I don’t agree that the last two sentences of the quote from Santer et al. are ‘non-sequiturs’. The main reason that people care about the alleged ‘pause’ is because they think it provides evidence that the mainstream science is wrong about anthropogenic warming. Of course there is some legitimate investigation of the specific factors driving the observed variability, just as, on the other hand, there is investigation into the factors driving periods of more rapid warming, such as the last four years. But it would be naive, even disingenuous, to think that that is what is driving the persistence of interest in–not to say, dogmatic defense of–a supposed ‘pause.’

  2. Given that there will probably be an el Niño this winter, 2018 and 2019 will have a temperature boost relative to the trend. The “pause” was derived by starting from the top of the warmest el Niño of the 20th century, 1997-1998.

    • “The “pause” was derived by starting from the top of the warmest el Niño of the 20th century, 1997-1998.”

      Indeed. And as tamino and Risbey, et. al. point out, this means that frequentist assumptions are completely violated if one does not correct for the selection procedure of “starting at the top” here.

    • Yes–though as I understand it, we’re *already* seeing some of that bump in warmth, since Pacific SSTs have largely been above normal since June. They’ve just been waiting for the atmosphere to come around to typical El Nino conditions to make the official call, which it seems they confidently expect to do soon:

      Click to access enso_evolution-status-fcsts-web.pdf

  3. “Noise is information, that cannot (or not yet) be attributed to a known cause.”

    Noise is a huge topic. But so much as the underlying view here is a 19th century view that so long as we know enough about the present state perfect future prediction is possible, I disagree. Some noise is inherently unknowable. Radioactive decay, for example. Or a really low friction compound pendulum’s location exactly one day from this instant. Some physics-steeped persons might disagree, I know. But I don’t think physics allows an out here either. I stand to be corrected.

    It appears to me, at least, that sometimes God does indeed play dice.

  4. Good to see Michael Mann being involved in that companion paper, since he was also involved in the Fyffe, et al, paper from a couple of years ago, which you showed was flawed but Mann didn’t seem to want to consider rebuttals at the time. Hopefully, he’s learned his lesson.

  5. David B. Benson

    Good for you, Tamino.

  6. “Noise is information, that cannot (or not yet) be attributed to a known cause.”

    Noise is any part of the signal that is not pertinent to the problem. In terms of global warming, minor changes in where heat is stored in the oceans or transported from the tropics to poles can add variability to the annual temperature change.

    • Good point. It leaves to define the problem properly, though.
      If it is to explain the past course of a curve as well as possible, the said fluctuations could be comprised into the signal.
      But if it is to predict the temperature curve in the future, this is at least partly undue, because all 3 causes (volcano eruptions, ENSO and solar irradiance changes) can only partly be predicted. So those effects do contribute to the prediction – but mostly to the expected error margin, not to the line of maximum probability.
      Then, if it is to predict 30-year-means of the temperature, AKA the climate, their effect is small and can be replaced by statistical means.

    • rhymeswithgoalie

      Aye, measuring *surface temperatures* is a problematic proxy for *accumulated heat*.

      [italics test]

  7. Were tropospheric temperatures featured too little in the literature to include in the paper?

    • Well the title is: “A fluctuation in surface temperature in historical context…” So that should give you one small hint as to why.

      What do you propose a study of tropospheric temps would add except more pages? Why don’t you suggest the paper examine ocean heat content uptake as well, I wonder? Surely OHC is more relevant than temps 8 km up where no one lives, or can live. Or, perhaps I don’t wonder.

    • From Lewandowsky 2016, cited in the paper:

      Contrarian discourse about a “pause” in global warming has found traction in climate science even though there is little evidence for anything but a fluctuation in the warming rate similar to earlier deviations from a longer-term trend.

      As tropospheric temps have featured prominently in the blogosphere regarding the so-called pause, I’m curious if that metric had been explored much in the wider ‘pause’ literature. In particular I was wondering how much the informal ‘literature’ had influenced the interest of researchers in this respect.

  8. off topic, but Regime Shifts?

    Cascading regime shifts within and across scales
    Study at Science is behind paywall.

    The Guardian has article about the study.

    As has been the case for as long with most of the revelations on global heating for the past 15 years, things are worse than we thought, in terms of this wicked problem.

    Happy holidays,