Epidemic of Denial

Wouldn’t you know it, I just posted about how dumb it is to talk about “trends” without enough data to talk about trends intelligently, when it’s brought to my attention that there’s yet another example of doing just that.


This time it’s the truly bizarre claim that Arctic ice is “recovering.” In fact GWPF (the Global Warming Policy Foundation) says that data from MASIE (Multisensor Analyzed Sea Ice Extent) “prove” it. They’re really just parroting a blog post by Ron Clutz.

Once again the trick is to use data that only covers a brief span of time — in this case less than 10 years — while ignoring data that cover a longer span of time. Then use the too-brief data to talk about how the trend is going your way. Add a bogus rookie mistake for good measure.

Here’s the data used, from MASIE:

masie

Yeah, it doesn’t start until 2006. Which makes it downright bizarre to use this for studying climate, for two reasons. First, we have data (passive microwave from satellites) covering quite a bit more time — starting in late 1978. Second, the MASIE people themselves tell you that their product isn’t the best for climate studies, instead you should use that passive microwave data from satellites:


“If one is interested in long-term trends in sea ice or how it responds to changing climate forcing, generally, it is best not to use an operational product, but rather one that is consistently produced and retroactively quality controlled. The NSIDC Sea Ice Index monthly ice extent, and the satellite passive microwave data sets upon which it is based, is one example.”

So why did Ron Clutz decide to use MASIE data instead? You make the call…

For the time span in question, it doesn’t make a whole lot of difference. Here’s the MASIE data again, but this time I’ll add the passive microwave satellite data from NSIDC as a red dashed line:

masie_nsidc

Once again, the two don’t agree perfectly but do agree extremely well within their period of overlap. The problem is, their period of overlap is a smidgen less than 10 years. For sea ice extent data, that’s just too little for an honest appraisal of the present trend.

Here’s a comparison of those two data sets over a longer span of time:

ice_all

Kinda looks like the ice has been decreasing, doesn’t it?

Clutz decides that rather than look at the yearly minimum or yearly maximum, he’ll look at the yearly average, presenting this graph:

masie-annuallarge

He then goes on to announce that there’s an “increasing trend.” I don’t think he knows much about trends.

I can compute yearly averages too. I got this:

annave1

Do you notice a difference? Yes, the latest value (for 2015 so far) is lower than Clutz shows. Y’know why? Because his blog post was two days ago. In the meantime, MASIE has added a couple more days’ data. Why, you may be wondering, would a mere couple of days make such a difference? Because this is the time of year that sea ice extent is low.

When Clutz averaged 2015 “so far,” of course he left out data from now (from a couple of days ago, actually) to the end of the year. The funny thing is, the way the seasonal cycle goes, during the time he left out ice extent tends to be below average — on average, that is. During the time he did includ it tends to be above average — on average, that is. It’s kinda like computing the year’s average temperature, but leaving out winter and only including summer. Not exactly like that, mind you, but the essential error is the same.

And that, on the part of Ron Clutz, is a rookie mistake.

If you really want to compute a yearly average including a year that isn’t complete, for data that show a seasonal cycle, you need to transform the raw data to anomaly data. Let me do that for Ron Clutz:

masie_anom

Now we can get some honest yearly averages. For the last one (2015) the year is incomplete, so it won’t be as precise as preceding years, but it least it won’t be biased high by a rookie mistake:

annave2

But the most important bogus mistake in Ron Clutz’s post is his talk about “recovery” and “trend.” One wonders, did he bother to compute what the trend is and what its uncertainty is? Because this case is yet another example for which, when you do that, you find out that the estimated trend from such a short time span is meaningless.

What if we used more data? The data recommended by the MASIE people? Why, we’d see this (using anomalies for annual averages):

nsidc_ann

Does it look like it’s in recovery to you?

Claiming recovery based on a meaningless “trend” and a rookie mistake, using an all-too-brief data set whose creators advise you to use something different for climate studies, is rather ironic. But perhaps the most ironic part of Ron Clutz’s blog post is the name of his blog: “Science Matters

42 responses to “Epidemic of Denial

  1. uknowispeaksense

    The unironic thing about this is his name. Where I’m from a clutz is someone whp is uncoordinated and clumsy.

      • Ron Clutz,I would charactertise it as more of a “pulled punch” than a “cheap shot”.
        How should the world characterise someone who is cretinous enough to note that because in the summer of 2014 SIE managed to drop to ~5M km while SIE in 2015 was less than 5M km for “exactly 28 days” and from there proclaim

        “Why is this fact not mentioned in articles talking about the 4th lowest extent recorded? How can a 28-day event (produced by a major storm) be called “climate change” when it is so temporary and natural?”

        I would suggest it takes a very deeply deluded soul to find the return of the northern winter a reassurance that AGW is not reality.
        Such a perverted view is perhaps the worst of your mucky little blog although the assertion preceding it (“As for minimum extents, 2015 September average will likely be the fifth lowest in the last ten years, so ranked in the middle of the years in this period. For it to be the fourth lowest ever would require ignoring earlier history, especially the 1930’s and the age of the Vikings.”), that runs a very close second.
        And note that these crazy things you tell the world are so dreadful that even the Gentlemen Who Prefer Fantasy, the GWPF no less, refuse to find a place for them even in their noxious Forum.

      • uknowispeaksense

        You’re right, it was a cheap shot. I don’t need to argue with you because you are starting from such a low base that your position is absurd beyond belief. This blog has gone to so much trouble to expose just how absurd your graph is in detail, it removes any need to engage you. All that’s left is ridicule, not undeserved.

      • Was that an apology?

  2. Perhaps it should read “klutz?”

  3. A funny thing is that Clutz’s article includes a table which shows the extents for March and September each year, and for 2015 these are both among the smaller ones. Given that, how can they think that the 2015 average extent could be the largest one?

    • It’s not about one year (this one) it is about a decade.

      • When I read that, I read a no excuse for this mistake, and worse, a failure to acknowledge it. That’s dishonest. Acknowledge the mistake, and then pivot to what you feel is your broader point. It’s a two step process.

      • If you have just ten years, then one year can make a considerable difference.

      • Indeed, that is the volatility of a short dataset, and wny I made no claim about the trend continuing.

      • “Indeed, that is the volatility of a short dataset, and wny I made no claim about the trend continuing.”

        So it’s not a trend then. And still you didn’t think to look at the longer data set? Don’t you think that would have been wise?

      • It is a trend in the observations. Whether it predicts the future is uncertain since only decade was provided. But it is an observed and recent trend.

        [Response: It’s abundantly clear that you don’t know what a trend is.]

      • Allow me (well, google really) to help: https://stats.oecd.org/glossary/detail.asp?ID=2769

        Again, you seem concerned that too short a period of data was available. Don’t you think that it would have been wise to look at the longer data set?

        “Here is what they are not telling you:” ~Clutz

        Only I’m not sure what you think ‘they’ aren’t telling people.

        If it’s that when you collect data from the wild, as opposed to controlled laboratory conditions, you get wiggly lines owing to external factors and natural variability and such, then… really?

        If you think some greater context about the state of the planet is being hidden, is restricting your own search to a data set that’s too short to say anything useful, err, useful?

      • For starters, a trend is not the coefficient you get from fitting a linear model Ron. Trends are the long term changes in expected values caused by the data generating processes themselves. We estimate trends with models but the model can only tell us what the “trend” would be given an error-minimizing calculation to the terms we let it use.

        To say a trend has changed is to say that the data generating process has changed. Figuring this out is more complex than merely graphing windows of the data and finding the coefficient, especially more complicated than graphing and saying “the line goes up”. Tamino has already written several posts on methods we have that do indeed show that the data generating process has probably changed—change point analysis shows a slightly reduced rate of ice melt, though still negative, and not particularly predictive since that model does show that these sorts of change points can occur (question is, where?).

        There are no statistics, however, that make a sufficient case for arguing that the data generating process has changed to cause long term expected ice extents to *increase*. No “recovery” (as poorly defined a term that is), no positive trend. I would propose an ordered list of explanations that people should consider for short term data anomalies, to be reorganized given higher anomaly values: (1) random variation itself, (2) exogenous factors, and only then (3) changes in the trend. With such a weakly positive coefficient for the annual data points from MAISE (~ 1 s.e. from zero), and especially with the larger data context, inferring a chance in trend should be the very last thing people do.

  4. They meant “See Ice” or how much they think they saw.

    But this does lead to the question of defining “Sea Ice” is that extent of coverage? Thickness? Multi year ice coverage? Snow on top? Salinity? Consistency?

    https://nsidc.org/cryosphere/seaice/data/terminology.html

    https://en.wikipedia.org/wiki/Measurement_of_sea_ice

    It is nice of you to post this article, but this not just denial, this is a contortion of stupidity designed to distract and deceive.

  5. One wonders, did he bother to compute what the trend is and what its uncertainty is?

    Based on his blog it doesn’t look like he did so I gave it a shot based on the data in his table. I got the following:

    The predictor is years the response average ice in millions of square km

    slope = .0631
    SE = .0233
    pvalue = .0267
    R2 = .478

    Hey his model didn’t do all that bad! What a shame the model’s premise is bogus and the data’s no good.

    Maybe there’s an inadvertent upside to all this phony GWPF BS. Shell announced today that they’re pulling out of the Arctic:

    http://www.reuters.com/article/2015/09/28/us-shell-alaska-idUSKCN0RS0EX20150928

    • The data is what it is: Navy people working each day to describe the ice conditions for the safety of ships operating in the area. They chose to release 10 years of data and the trend is what it is, Now some people think only data between 1978 and 1998 is real, because it was cool with lots of ice in the beginning, and warm with much less ice at the end. So that 20 years is meaningful, but the last 10 years is not.

      • Mr. Clutz, who thinks that only data before 1998 is real? “Real” science doesn’t cherry start and end dates? Why not use all years available to you, especially when recommended to do so by the people whose data you’re using? These are Navy people working each day to describe ice conditions who are telling you to use NSIDC Sea Ice Index. You used their data against their recommendation!

      • I withdraw my last comment. I would need to read the source material.

      • Which part of the phrase “statistically significant” do you not understand?

        Which part of the law of LARGE numbers do you not understand?

        Which part of “twice as much data is going to be much better” do you not understand?

        Why bother to do analysis on a cherry picked data set that doesn’t have enough data to support said analysis?

        What is wrong with you?

      • The data is what it is .

        The issue Ron is what the data isn’t. If you read the critique of your analysis in the blog post you will see it is not the correct data to use. See the quotes in italics above in the post. Just above that quote is a link: “MASIE people themselves tell you” and if you click to it then scroll down to 2. Background Info you will find the quote and why they recommend against using MASIE data. It states:

        While operational analyses are usually the most accurate and timely representation of sea ice, they have errors and biases that change over time .

        So when you did your analysis how did you handle the time varying biases and errors referred to?

        They recommend using the NSIDC Sea Ice Index monthly ice extent for one, which would probably preclude the compensations required by MASIE data. They give a link to it but you don’t have to bother, its simpler just to use the link given at this site which you can find under Climate Links.

        Just a couple of final points. The blog post identifies a number of other issues that you should address for this forum, like the criticism of the questionable averaging you used for 2015 mean ice and the implications regarding the trend. Even though you may think the “the trend is what it is” you will find that it isn’t what you say it is. And also the question that references the last plot in the post deserves an answer:

        Does it look like it’s in recovery to you?

        Hopefully at some point in this discussion you will read this blog post. I think you, like the rest of us will find it very informative.

      • What elspi said.

        In frequentist statistics there’s no such thing as a statistically insignificant ‘trend’, unless that ‘trend’ is zero, flat line, nada.

        Go back to school Clutz.

      • Chris O'Neill

        The data is what it is

        Or more accurately, the data you pick out is what it is and the data you choose to ignore is not what it is.

  6. Bull. Straw manning on your part. No one,least of all Tamino is saying any such thing. Time and again he has clearly said one has to take all the data together not cherry pick it solely to one’s advantage. YOU obviously are the one trying to say one portion of the data is not important. Have the good grace to admit your silly ploys were caught out by a much,much better mind that you will ever be. I caught the name irony right away and think your comments have done nothing but show it was no cheap shot at all,but a pretty accurate characterization of your intellectual status.

  7. No, Ron. You use ALL the data unless you have a good reason for doing so. And you don’t pick the dataset that whispers reassuringly to you.

    Don’t feel bad. Not everyone has the courage to face reality.

  8. Too many people have been following Arctic sea ice too long in too much detail for you to get away with that sort of uninformed, half-assed analysis, Mr. Clutz. Next time do yourself a favor and do your homework first.

  9. Ron,

    Why not admit to making an error or two, even give thanks for the correction. Nothing bad about being wrong. Not being able to admit it is telling.

  10. A point of correction.
    The GWPF are of course a bunch of AGW-denying morons but with some UK political muscle having been set up by an ex-Chancellor of the Exchequer. They formed themselves up to have charity status so they could lie to the world using UK tax-payer’s money. And would you credit it, they formed as an educational charity! Their egregious lies have thus caught them with their trousers round their ankles so they have had to re-arrange themselves. The gobshite discussed here is the work not published by the Global Warming Policy Foundation, the part of the GWPF that remains an educational charity, the part that only lies with moderation. Instead it is part of the gobshite that doesn’t pass muster and is published by the Global Warming Policy Forum which is allegedly a think-tank. It prides itself on being wholly-owned by the GWPF charity.
    Presumably it is on the GWPF’s Academic Advisory Council that we find all the experts in gobshite who decide which lies it is charitable to accept by the Foundation and which are too repugnant and thus jettisoned out into the Forum. The continued charitable status of part of the GWPF suggests they do an excellent job categorising all this gobshite and I feel their effort should not go without due acknowledgement.
    The present set of experts in gobshite are:-
    Professor Ross McKitrick (Chairman), Adrian Berry, Sir Samuel Brittan, Sir Ian Byatt, Professor Robert Carter, Professor Vincent Courtillot, Professor Freeman Dyson, Professor Christopher Essex, Christian Gerondeau, Dr Indur Goklany, Professor William Happer, Professor David Henderson, Professor Terence Kealey, Professor Deepak Lal, Professor Richard Lindzen, Professor Robert Mendelsohn, Professor Ian Plimer, Professor Paul Reiter, Dr Matt Ridley, Sir Alan Rudge, Professor Nir Shaviv, Professor Philip Stott, Professor Henrik Svensmark, Professor Richard Tol, Professor Fritz Vahrenholt, Dr David Whitehouse.

  11. The graphs and the data are the graphs and the data. They don’t hide anything, and the data is freely available. Sound unfamiliar?

    As for the last graph of anomalies, that’s just laughable.That will never show a recovery until the sea ice is well above what it was originally. Yet the sea ice will have been increasing in extent for years before that.

    So do explain why that’s not utterly misleading?

    [Response: Declaring a “trend” without any real evidence, is declaring a “trend” without any real evidence. Including a bias in your yearly averages, is including a bias in your one-year averages.

    You being so deeply in denial, is you being so deeply in denial.]

    • Oh dear, Tim. You are so lost. Why don’t you think what an an anomaly is. It will exhibit the same trends as the raw data, but with less noise. All you do with an anomaly is subtract the average for that period–a constant for each date. That can’t obscure an increasing or a decreasing trend.

    • “That will never show a recovery until the sea ice is well above what it was originally. ”

      “Recovery” is not a meaningful word. As in it has no mathematical meaning. If you want to show a trend change you can try fitting a polynomial model, try a change point model. You actually do not need to have the data “recover” to 1979 values in order to demonstrate a trend change. But unless you (to steal what might be this blog’s catchphrase) *do the math*, you should stop complaining blindly about people who do.

  12. “From a climate change perspective, a better metric is the average ice extent over the entire year.” ~ Clutz

    You know, I can’t help but think…. from a climate perspective, half the year is irrelevant because there’s no sun and therefore no change to albedo.

    • At the very least Clutz should provide some backup for this highly specific, categorical claim.

    • Half the year is not irrelevant. There is still increasing carbon dioxide up there, and it should be warming the earth below. There are many factors that mask that in the short term, and a decadal trend can easily be misleading.

      Of course, we see a much larger trend at the minimum since there is so much land in the way at the maximum.

      • Arctic sea ice extent has very little to say about, or influence on a warming world for half the year.

        The big climate change issue there is albedo (reflectivity) – the greater the area of sea ice the more sun light is reflected. Whereas less ice means light is absorbed by darker water – leading to additional warming.

        However during the polar night, there is literally no sun for months on end. Albedo (reflectivity) is effectively zero and sea ice extent makes no difference to it. As far as Arctic sea ice extent goes, It is the polar day and the summer melt season that is of particular interest from a climate change perspective.

        So the question ‘is it a better metric?’ is begged: why does Clutz believe that the average extent over the entire year is a better metric for measuring climate change?

      • “Of course, we see a much larger trend at the minimum since there is so much land in the way at the maximum.”

        We do, but I’m not sure that that’s why; does the shore really so greatly constrain ice expansion relative to an unconstrained hypothetical case that the onset of melt still shows little change in the constrained sectors?

        Or does another factor explain the differential trends for different times of the year? One might also think that the sudden near-absence of post-Solstice insolation in the far Northern sector, which clearly drives a very sharp change in radiative balance, might provide a strong ‘anchor’ to freeze rates even in the face of an increasing greenhouse effect.

        Frankly, I dunno. But I do wonder.

        We sure do see smaller trends at the max, though; using monthly February NSIDC data as a proxy for the actual minimum (which ought to be good enough for the purpose of a casual blog comment), we see a decrease of 2.9% per decade:

        http://nsidc.org/arcticseaicenews/2015/03/

        Still pretty considerable, if less drastic than the case for the minimum.

        The estimable Jim Pettit has an attractive graph–unfortunately only updated to 2012 at this point–which compares minimum and maximum volume (as modeled by PIOMAS), deriving annual ice loss by volume:

        https://sites.google.com/site/pettitclimategraphs/sea-ice-volume#asivamlir

        Had it been updated, we’d be seeing a modest upward bump in volumes over the last couple of years. (I almost called it a ‘modest recovery’, but there’s no reason to think it’s anything but variability–which is not what the ‘r-word’ suggests to some folks at least.)

        The PIOMAS data summary page is here, with lots of food for thought:

        http://psc.apl.washington.edu/research/projects/arctic-sea-ice-volume-anomaly/

        Last semi-random musing: I suspect that Mr. Clutz’s number for 2015 is going to look even worse when the year is over than one might think based on the anomalies to date. Already, the maximum was the lowest on record; the minimum was the 3rd- or 4th-lowest (depending on whose numbers you use, JAXA or NSIDC) and most of the peripheral Arctic waters are still showing considerable positive SST anomalies (and particularly in the Northern Pacific, where that El Nino is in progress.) That last hints at a pretty slow refreeze, and possibly even another low record at the 2016 maximum–though Arctic weather has a long record of dashing efforts at prediction over seasonal time scales.

        You can see in Tamino’s anomaly graph that 2015 is already 3rd-lowest in the (truncated) record. Given that the Arctic navigation season actually extends into November, we have a couple more months which typically have small extents relative to the yearly average. Put it all together, and I would not be surprised to see a record-low average extent for the current year. It still wouldn’t make a meaningful trend over the span of MASIE data, but the optics wouldn’t be nearly so attractive for our ‘friends’ over at GWPF.

        Oh, well, ‘a graph’s a graph,’ right?

  13. Interesting that Mr. Clutz showed up here to defend the indefensible. I assumed he deliberately chose MAISE because only with that short database could he get the result that he wanted to get. His comments here suggest he really doesn’t understand his mistake.

    I don’t know if that means there’s hope for him should he come to understand his mistake – his original bias will still be there. Maybe. This really isn’t hard stuff to understand, though.

  14. I take issue with your assumption that these incessant “recovery” or “cooling” graphs are the result of ignorance & rookie errors.
    Perhaps that was true years ago but I think that some deniers & confusionists have gotten very adept at making data show the results they want.

  15. Mr. Clutz is “trying but failing” to show something that can’t be shown from data that shouldn’t be used. That’s a pretty expectable result from the confusionists (good word!) . We have to credit that he honestly believed his analysis sufficiently to come here and respond. Doesn’t make him right about it, but we have to allow that he “thinks” he is and argue on that basis.

    Basic sciences and statistics are not taught before economics in our schools.

    Explains a lot when you think about it ;-)

  16. Kanji Sketch Pad Arcuate

    Over at Ron Clutz’s blog he is trying to portray this as a case of competing opinions that each have their own preferred graphing method, and he is doubling down on his original analysis in a way that reveals he is engaged, not in blind idiocy, but active deception. Also, he is making much of the fact that he came here to discuss things man to man and was insulted and then censored. The fanboys are claiming that Tamino’s blog is an echo chamber where we reinforce our misguided belief systems and howl down dissent.

    I posted the text below at Clutz’s site. Unsurprisingly, the comment did not survive moderation, allowing him to persist with his patently false image of a reasonable man whose ideas were stifled, but nonetheless revealing to me that he is a hypocrite.

    Ron,

    You made no attempt to engage with the substance of Tamino’s arguments, there or here. Apart from your rookie mistake producing an inappropriate value for 2015, there is the issue of focussing on short-term noise and missing the big picture. Your blog consists of propaganda, not science, and the scorn expressed on Tamino’s blog is well deserved. Your readers here appear happy with any sciency-looking graph that tells them what they want to hear and, like you, seem to know very little about actual data analysis, but your arguments are unconvincing when taken on their merit. It is you who live in an echo chamber, or you would realise this.

    Show us the comments Tamino deleted, and we can judge for ourselves whether it was appropriate. Moderation of a site like his is quite appropriate or it gets swamped in mindless denialism. He gave you a couple of chances to discuss the issues and you seemed to have nothing worthwhile to say. I suspect he is simply trying to preserve the signal-noise ratio of his blog.

    Leto.