Speedup Skeptic

I enjoy reading the Climate Progress blog, often getting good information and ideas. But when it comes to one recent post, I have doubts about both its correctness and its wisdom.

The theme is encompassed in the title: “Record-smashing August means long-awaited ‘jump’ in global warming is here.” I’m skeptical.


The post says this:


As I reported last year, climatologists have been expecting a “jump” in global temperatures. There is “a vast and growing body of research,” as Climate Central explained in February 2015 that “humanity is about to experience a historically unprecedented spike in temperatures.”

It goes on to say:


A March 2015 study, “Near-term acceleration in the rate of temperature change,” makes clear that an actual acceleration in the rate of global warming is imminent — with Arctic warming rising a stunning 1°F per decade by the 2020s.

First let me point out that I dislike referring to a “March 2015 study” by linking to a Climate Progress blog post. It should link to the study itself.

Second, let me point out that “a stunning 1°F per decade by the 2020s” is neither an imminent consequence of accelerated warming, or something to expect “by the 2020s.” C’mon, guys, look at the data: Arctic warming has been happening that fast already.

arctic

The linear trend since 1990 puts the warming rate in the Arctic at 1.1°F per decade.

arctic90

I mainly object to the level of exaggeration which I regard as misleading. For instance, the March 2015 study does not make it clear that “an actual acceleration in the rate of global warming is imminent.” It presents some evidence (computer model simulations) and makes a viable argument … but “makes it clear”? No.

As for “a vast and growing body of research,” that “humanity is about to experience a historically unprecedented spike in temperatures,” as far as I can tell that just ain’t so. There’s research to that effect and research contrary to it. In my opinion, the claim is just exaggeration.

But probably the most clearly wrong statement in the whole post is in its title: that the record-smashing August means the jump is here, now. No, it doesn’t, a single hot month or string of hot months or years doesn’t do that, any more than a cold month or string of cold months or years would disprove global warming.

It’s the “hiautus/pause” mistake, in reverse. I’ve spent years knocking down the “pause” nonsense we had to put up with from deniers. As for “acceleration,” let’s freely discuss the possibility but let’s not declare it a fact, until the evidence is really there. It isn’t, yet.

Here’s the annual average global temperature since 1970, including this year in spite of the fact that 2016 isn’t over yet:

nasa

The most recent value is farther above the trend line (shown in in red) than any other. Does this “make it clear” that acceleration is “imminent?” No. It’s possible, but we don’t know yet. This year, as high above the trend line as it is, isn’t enough to justify claims of a trend change statistically. If we look at the residuals from the linear trend it’s very suggestive:

resid

But “very suggestive” is not “makes it clear.” Furthermore, the extremity isn’t out of line with what one could expect if the residuals follow the normal distribution. This is well visualized with something called a “quantile-quantile plot”:

qqnorm

It’s well quantified by testing the normality of the residuals with the Shapiro-Wilk test. With a p-value of 0.43, it’s not even close to statistically significant evidence of significant departure.

As for other tests for departure from the linear trend, of which there are many, none of them gives the evidence required.

It’s a pity, because we don’t need to exaggerate the warming rate … the present rate is plenty scary enough. It’s more the pity because if the coming years show claims of acceleration to be as mistaken as claims of the “pause” were, it’s an embarrassment, and gives plenty of ammunition to deniers to crow about how climate bloggers overstate the changes in their unbridled alarmism.

Please, please, we don’t need to exaggerate. We definitely don’t need to claim as “clear” and “imminent” things for which the evidence isn’t yet there. What we already have the evidence for, is scary enough.


This blog is made possible by readers like you; join others by donating at Peaseblossom’s Closet.

41 responses to “Speedup Skeptic

  1. Concur. The Q-Q plot is an especially powerful argument.

  2. Christoph Rose

    I’ve read this blog for a long time, have learned quite a lot, and am able to follow it most of the time.

    I’ve never heard of or seen a quantile-quantile plot, and to be honest, I don’t really understand it.

    I would really appreciate a short explanation of how to read it.

    Thanks!c

    [Response: It’s a plot of what the values are *expected* to be if they follow the normal distribution, vs what they actually are. If they do follow the normal distribution, the points should fall approximately along a straight line. If they depart greatly from a straight line, it’s probably not normal-distribution.

    The Shapiro-Wilk test quantifies that, by fitting a straight line to the quantile-quantile plot and computing the departure from linearity (allowing for how many data points there are). The Q-Q plot is a good visual guide, but the Shapiro-Wilk test is the real test.]

    • Uh …., unlike Tamino, who I respect a great deal, I’d go with the Q-Q plot as being definitive (refs: Harrell, Regression Modeling Strategies, in the vicinity of Section 15.6 in the first edition and 7.6 in the second; Williams, Grajales, Kurkiewicz in Assumptions of Multiple Regression: Correcting Two Misconceptions, section “Normality assumption simulation”). While the Shapiro-Wilks test is powerful, the present case is in the wrong direction. Sure, the residuals aren’t blowing away the tentative hypothesis of Normality, but y’can’t conclude Normality either. The Q-Q is much more comforting, and with justification.

      However, even the Q-Q can mislead, and resampling tests have been argued as being better all around (paired two sample t-tests, replicated, or BEST), especially for small sample sizes, something which does not apply here.

    • I, too, was confused by that Q-Q plot, especially after reading https://en.wikipedia.org/wiki/Quantile . I think the use of the word “quantile” here is confusing as the data isn’t divided by quantiles; AIUI now each dot on the Q-Q plot corresponds to an individual year.

      Here’s how I think the plot was made:

      1) Take the residual anomalies for each of the years (47 numbers, if I counted right) and sort them into ascending order of residual (i.,e., ignore the year order).

      2) Lay them out in that order along the x-axis as they would be distributed if they followed a normal distribution. This spacing only depends on the fact that there are 47 numbers, there actual values are irrelevant for this.

      3) Find the mean (zero presumably because they’re residuals?) and standard deviation of the numbers.

      4) Lay them out along the y-axis by their values scaled to that mean and standard deviation. Because the numbers are sorted the resulting plot will be non-decreasing.

      Is that roughly right? If so, why is the scale on the y-axis one tenth of that on the x-axis (y-axis values seem to only go from roughly -0.2 to +0.2)?

      [Response: Basically yes, but the y-axis values aren’t scaled they’re raw. If the data follow the normal distribution, then the points will lie close to a straight line whose mean and slope reveal the mean and standard deviation of the underlying distribution.

      And perhaps some of the confusion comes from the similarity of spelling and pronunciation of the words “quantile” and “quartile.”]

      • My spelling of “their” (as “there”) in step 2 is a bit embarrassing.

      • Basically yes, but the y-axis values aren’t scaled they’re raw.

        Ah, yes. I had to think for a while why they’d come out linearly if you do that but, of course, that’s what you’re testing for.

  3. One minor contention: It’s not entirely clear, but I think that when the thinkprogress blog post says, “with Arctic warming rising a stunning 1°F per decade by the 2020s”, they might mean the warming rate increasing by an additional 1°F/decade (thus why they say “rising” in reference to the warming, rather than temperature), which from your figure 2 would mean warming of ~2°F/decade. I might be wrong—the sentence is a little ambiguous in that regard—but it’s a possibility to consider.

    As for the rest of this post, great work. Before I read your post, I checked out the Climate Progress blog via the first link you gave, and happened to notice and quickly skim a portion of the very post that you discuss here, without knowing you were going to talk about it. I actually had some of the same initial thoughts you expressed here, though I didn’t invest much time or thought into it initially, particularly a bit of incredulity at the claim that this August is somehow Exhibit A for a jump having occurred.

    Once I realized that was the post you were going to talk about I gave it a more thorough reading, and I completely agree with everything you said. Thanks for the insightful commentary, as usual!

    • I agree Romm’s wording leaves that possibility open. But if you click through to his post about the 2015 study (https://thinkprogress.org/rate-of-climate-change-to-soar-by-2020s-with-arctic-warming-1-f-per-decade-85db70fb9d1#.6wty4wx0o), it’s clear he meant an acceleration to just north of 1 degree F.:

      “Because of Arctic amplification, the most northern latitudes warm two times faster (or more) than the globe as a whole does. As this figure from the study shows, the rate of warming for the Arctic is projected to quickly exceed 1.0°F (0.55°C) per decade.”

      The study is actually about the trailing 40-year multidecadal warming trend. Eyeballing the graph Romm reproduces from the study (based on an RCP 4.5 scenario projection), it shows a 40-year Arctic warming trend of about 0.55 C (1 F) now, peaking at ~0.63 around 2030.

  4. I really wish Romm, with a physics Ph.D. from MIT, wouldn’t go overboard with a mix of weak analysis (or rather lack of analysis), advocacy and (partly) overwrought tone like this. Sure, he didn’t have to flood a post aimed at a general readership with too much detail and rigor, but this is going too far the other way. AGW ‘skeptics’ are rightly mocked for their motivated statistical innumeracy and cherry-picking; these don’t look any better coming from the majority side of the consensus.

  5. I agree. Exaggeration just leads to more grief down the road, because you end up trying to defend statements that really aren’t defensible. Eg., “Arctic sea ice could be gone by 2013”. Yes, deniers distorted what was actually said, but why make it easy for them?

    The Trenberth quote toward the end of the article was interesting:

    “The increase in carbon dioxide and other heat trapping gases from human activities is relentless. The effects on global mean surface temperatures can be masked by natural variability for a decade or a bit more, but as the natural variability goes in the other direction, suddenly it is quite a different story and record after record gets broken.”

    Sounds an awful like someone else on trends and fluctuations, if you ask me. ;-)

    Presumably, the slightly over-the-top spin is Romm’s take.

    • Except, I believe Trenberth has written that the negative phase of the PDO is perhaps a better explanation for mid-century cooling than aerosols, and that a positive phase of the PDO will bring… not knowing what else to call it. If I can’t call it accelerated warming, how about a whole bunch of extra warming?

  6. This kind of post is what keeps me coming back to Open Mind for reliable information on climate trends. As someone who has no background in climate science, I need to rely on honest brokers who do have that kind of expertise, but who also aim to educate without ideological bias coloring what they have to say. Thank you!

  7. Non-linear feedback systems tend not to follow normal distributions. If the system is in “control” then then the values tend to be relatively stable. If the system goes out of control / crosses a tipping point, the the system tends to move rapidly until new feedback loops act to impose new controls on the system.

    One of my favorite examples is a pipe of a plutonium solution. It can sit there for a long time. At tipping point 1) it drains into a bucket (change on a timeframe of hours), then it goes critical at tipping point 2, (change on a timeframe of hours) starts to heat and the water boils off, point 3) the bucket melts releasing the Pu. Fortunately, a smart guy walked by and tipped the bucket of solution out on the floor, avoiding full criticality.

    None of the tipping points can be predicted by statistical trends in the system prior to the tipping point, but all of the tipping points in the example can be predicted from other knowledge of the world. JR knows of the bucket incident, and the nature of non-linear systems that go out of control.

    I learned this stuff from Jay Forrester and Ed Deming, but there are many good texts on it. When we were building chip factories for IBM and Motorola, both of them made me take Ed Deming’s course, and they both wanted original transcripts, so I had to take the course twice to get 2 original transcripts.

    It saved my life when we were designing a hazardous waste treatment facility.

    I do not expect that you will again ever see a time sequence of weather data that fits a normal distribution. Nor, do I expect that you will ever again be able to statistically predict the future trend of a time sequence of weather data. Future trends will have to be derived from other knowledge of the world.

    • “Future trends will have to be derived from other knowledge of the world.”

      Which is pretty much the opposite of fossil at discourse, much of the time. Eg., linear extrapolation of sea level rise to date to ‘prove’ there’s no cause for alarm.

  8. While I agree that we should not exaggerate the data, we should act as if the worse case will happen. You would not put your child on an airplane if the plane’s mechanic is uncertain if the engine’s oil seals are installed correctly. Uncertainty is a reason for action, not inaction, when the possible results of inaction are catastrophic.

  9. I know this is simplistic, but by my reckoning the average 30-year rate of global surface temperature rise since the mid 20th century to the present is 0.14 C/dec, with a standard deviation of 0.04 C/dec. Since 1976 the average 30-year rate of global surface temperature rise is 0.17 C/dec with a standard deviation of 0.00 C/dec (I used the WMO metric).

    The more recent period obviously has a much shorter sample size. Even so, it’s not pushing it to tentatively infer from this evidence that there has been been an acceleration in the long term rate of global surface temperature rise since the mid 20th century. There has also been less internal fluctuation in this rate since the mid 1970s, compared to the period as a whole.

    There’s no evidence to suggest a recent ‘step change’ style acceleration in warming. There is some evidence to suggest a slow acceleration in the rate over the longer term. That’s probably about as far as we can confidently go.

  10. How would it look after removing the ENSO influence (using MEI) to reduce the noise?

  11. @Tamino

    “Please, please, we don’t need to exaggerate”

    Got that. But many people say, climate change/global warming is “very, very gradually”. Is that correct or is that an understatement?

    [Response: It’s relative. Compared to the lifetime of a tse-tse fly, it’s achingly slow. Compared to the lifetime of a civilization, it’s frighteningly fast.]

  12. I suspect that one of the reasons that people get out ahead of the data is that AGW deniers use the uncertainty to make the case of “Why should we spend trillions of dollars to address a problem that we can’t even predict with certainty?”. Climate realists realize that that uncertainty makes the problem even more urgent, even based on the current trends, because those are bad enough. But a wild card tipping point could bring us rapidly to the point of disaster that will have us scrambling to deal with it while society begins to crumble due to extreme pressures.
    That idea of utter chaos as a consequence of inaction can create the urge to try to find the right button to push to wake up a sleeping and disinterested population. I kind of get that. We can’t afford to look like the ‘boy who cried wolf’ though. What to do?

    • @skepticmac57

      “What do do?”

      That’s the final question. To me, it’s a psychological problem in the first place:

      How to get rid of greed and ignorance?

      In the economic system, greed is a good thing, it is the motor of the economic system. That’s a devastating paradigm, because it causes ignorance regarded to the ecological consequences of greed. The ecological system is treated as “externalities” by economy, that’s a huge problem. The economic paradigm of greed is a grinder:

      You put natural ressources into it and money comes out.

      We need a total paradigm shift (see Fritjof Capra eg), especially in the mass media. The message of the mass media is consumerism non stop, almost all fields of human life are commercialized. Greed, consumption is like a drug, a vicious circle: The more you use, the more you need. So, the modern paradigm needs to change and the need of that paradigm shift must be communicated in the mass media, in school, in economy, everywhere. If that paradigm does not change, we are doomed sooner or later, rapidly or, quote: “very, very gradially”.

      But I am very much afraid, that this paradigm shift will not come. The lobbyists are too strong, the “elite”, the system, the military-industrial complex (see Eisenhower’s farewell speech/warning) completely depends on the paradigm of greed, of consumerism (and on ff too). As I said:

      Greed is the motor of the economic system and we must turn around the system aroind by 180°. I doubt, that this will happen. So, it seems, the system must brake down first. It will brake down, if it goes on with BAU, no doubt.

  13. “First let me point out that I dislike referring to a “March 2015 study” by linking to a Climate Progress blog post.”

    One of the reasons why I read there less often now than I used to. I once went through link after link after link after link after link trying to get to a source document and every single link was to yet another Climate Progress blog entry. I’m sure it’s a site policy. Most people/sites with an item on a topic will provide links to _both_ their previous item and the source or other relevant docs or posts.

    CP seem never to do that. Maybe it’s extremely seldom. Either way, that makes them unusable when providing information to people who ask for background or further evidence. Romm writes intelligently and often has good insights. But his writings are of no value other than the reading of them. I never bookmark them for later use because they are of no further use to anyone else.

  14. Dr Gavin Schmidt just weighed in with a detailed projection for 2017 entirely consistent with what Tamino is saying. He also published another version at fivethirtyeight.com. Essentially, Dr Schmidt is betting temperatures will regress to the mean trend.

    • It would not be surprising if 2016 ends up closer the the purple, actual data through July, than the blue, the prediction. As for 2017, I don’t think it will drop that far. Time will tell.

  15. One of these days – unless we soon manage to put a serious cork on our emissions – there will be a sustained, significant acceleration in the rate of warming, not explainable by natural variations, but attributable to positive feedbacks or tipping points. Per Tamino’s post, there’s no evidence in the ebb and flow of the surface temperature record that we’re there yet any more than we were during similar warming sprints in the 70s-80sand the 90s: http://woodfortrees.org/plot/gistemp/from:1945/mean:12/plot/gistemp/from:1975/to:1982/trend/plot/gistemp/from:1981/to:1994/trend/plot/gistemp/from:1992/to:1998.5/trend/plot/gistemp/from:1998.5/to:2012/trend/plot/gistemp/from:2009/to/trend/plot/gistemp/from:2011/to/trend/plot/gistemp/from:1993.5/to:1998.5/trend/plot/gistemp/from:1964/to:1974/trend/plot/gistemp/from:1964/to:1982/trend