Spencer’s Paper Reviewed; Remote Sensing Editor Wolfgang Wagner Resigns


In one of the most asinine, self-promoting, sniveling, absurd, nakedly political moves Wolfgang Wagner has resigned, with trumpets blazing, his editorship of Remote Sensing.

Why? Because the journal under his command dared follow its editorial guidelines, and follow them properly.

Because while adhering to procedure he allowed the Spencer and Braswell paper “On the Misdiagnosis of Surface Temperature Feedbacks from Variations in Earth’s Radiant Energy Balance” to be published, as it should have been published.

What happened was this: Spencer and Braswell submitted their paper, and Wagner’s staff gave it to three reviewers. Wagner said, in his public apology, the reviewers were

three senior scientists from renowned US universities, each of them having an impressive publication record. Their reviews had an apparently good technical standard and suggested one “major revision”, one “minor revision” and one “accept as is”. The authors revised their paper according to the comments made by the reviewers and, consequently, the editorial board member who handled this paper accepted the paper (and could in fact not have done otherwise). Therefore, from a purely formal point of view, there were no errors with the review process.

Wagner felt so badly about this excellent process that he resigned. He claims that he did so because the authors “and like-minded climate sceptics have much exaggerated the paper’s conclusions in public statements.” This is irrelevant, so he also said other papers by other authors “refuted”—misusing this strong word—Spencer and Braswell’s central claim.

But this is far from the first time that rival groups have published papers that come to opposite conclusions. It happens so often as to be utterly unremarkable. So why did Wagner remark on it?

Spencer and Braswell’s paper is not that exciting (see my review below; very few papers are exciting), and neither Wagner nor any of the reviewers or other staff thought that it would be that big a deal. And so it passed normal peer review into print.

But the paper turned into a spectacle. And then Wagner probably caught hell from “the” consensus. “How could you!” must have been the subject line of all the emails in his inbox. Wagner then said he became concerned because scientists and non-scientist “engaged observers of the climate change debate pointed out in various internet discussion fora” that Spencer and Braswell were all wet.

In other words, the blogosphere erupted and Wagner took it seriously. But Wagner also dismissed those on the “internet discussion fora” who supported Spencer and Braswell—those “public statements” he mentioned. So much for consistency.

It takes a man to stand up to this kind of barrage. Wagner was in danger of not being invited to meetings, of being a pariah, of even—it hurts to say this—losing future grants. He had to do something to distance himself from his own journal, a journal he helped create, using rules he helped devise.

Wagner took the path of self-aggrandizing cowards.

Consider: Old Wagner was the first editor of Remote Sensing, and had served nearly two years. My guess is that his term of office was up this January. So by resigning, he was only leaving office a couple of months early. I might be wrong about this, of course, but I’d like to hear the denial.

Richard Black of the BBC and others in the media immediately began promoting Wagner as a hero. Gushing, in effect, “That man loves science so much!”

What rot. If the paper is flawed and its conclusions are genuinely refuted by other papers that Spencer and Braswell, with malice aforethought, ignored, then Wagner would have retracted the paper. That would have been the “honourable” thing to do.

Since Wagner did not do this—he could not, because there is nothing in it to retract—he instead put on his hair shirt and pretended to be affronted. A brazen move transparent to anybody except the ideologically addled. Tellingly, Wagner’s wailing convinced the BBC.

Incidentally, the BBC, in an idiotic but typical move, captioned Roy Spencer’s picture with the words, “Dr Spencer is a committed Christian as well as a professional scientist.” What in the holy hell is the purpose of this except to disparage Spencer and imply that his religious beliefs corrupted his findings.

Are Wagner or Blacks “committed atheists,” or devotees of yoga, or who knows what else, as well as being a professional scientist and reporter? It’s irrelevant and only the devious would use such a cheap trick.

Paper Review

I emailed the same criticisms noted below to Roy Spencer and he answered all of them. He was of course under no obligation to do so, particularly as my questions were sharply worded and more than a little combative.

I assume the reader is familiar with the paper, so I will not summarize it here. Here is a portion of the conclusion of Spencer and Braswell (another is below):

[The] atmospheric feedback diagnosis of the climate system remains an unsolved problem, due primarily to the inability to distinguish between radiative forcing and radiative feedback in satellite radiative budget observations.

I concur with this conclusion and agree that the pair have presented evidence to support it.

The evidence, however, is likely overstated. For those who would copy and paste that statement, you would be negligent in your duties if you did not also copy this: this overstatement is likely minor and does not imperil the conclusion. My detailed criticisms follow.

Major: In Fig. 2, the pair ran a smoother over the data to “reduce sampling noise.” Of course, smoothers can’t do this. If there is measurement error in the data it has to be measured separately or estimated by some external means to account for it. Smoothers, unless the serendipitously mimic the measurement error process (extraordinarily unlikely here), merely make the data look more pleasing to the eye.

The smoothed data was then used in the regressions (Fig. 3). The problem is that when two times series are smoothed, it always increases the correlation between them, even if the series have no relation to one another. So when the smoothed data is used as input to the regressions, the effect is to inflate (in absolute value) the size of the regression coefficients.

Since a central finding is the difference between the regression coefficients of the observations against the same from the GCMs, the effect is to claim a larger difference than exists. This is not to say that no difference exists, just that the actual difference is likely to be less than Spencer and Braswell indicate.

Roy told me that he tried his models “Without smoothing, with smoothing” and that “The same conclusions” were found. I see no reason to call his honesty into question.

Minor: I also wondered why the full hundred-plus years from the GCMs were used instead of the same eleven years as the observations, as that seemed a fairer comparison. The answer is that more data gives a better estimate of the GCMs’ behavior. This is true, but it would have been interesting to see what limiting the GCM data did. Best guess is not much.

Minor: The authors had to “detrend” the GCMs’ output. This is always a tricky thing to do, and if done incorrectly it can add spurious correlations or remove existing ones. The authors did not use the most sophisticated methods to model the time series. But, then, neither do most authors in this field.

This is a niggling criticism, since my guess is that even a more sophisticated model won’t change the conclusion much. But skeptics have to be like Caesar’s wife, where even the suspicion of infidelity to best practices is a kind of lapse.

Minor: I would have preferred that Figs. 3 and 4 share the same axes limits. And even that the GCM results (from Fig. 3) would be overlaid on Fig. 4. This would make comparisons easier. With the limits as they are now, the GCM results appear smaller (ni Fig. 3) in comparison to Fig. 4 than they actually are.

Here is another portion of the conclusion:

While the satellite-based metrics for the period 2000–2010 depart substantially in the direction of lower climate sensitivity from those similarly computed from coupled climate models, we find that, with traditional methods, it is not possible to accurately quantify this discrepancy in terms of the feedbacks which determine climate sensitivity.

This also is consonant with their findings, even taking my criticisms into account.

It should be understood by other critics that Spencer and Braswell are not claiming much. They say only that we are not as sure of ourselves as we thought we were. They do not claim the AGM thesis is certainly wrong. They do not claim to have produced the final word on the subject of feedback versus forcing. They do not, really, claim very much.

This is a minor paper—I mean no insult by this; nearly all papers are minor—in a minor journal, which adds to our store of knowledge only a minor nugget. That this nugget appears shiny and of great worth to some and fraudulent to others, that it caused so many people to bleat and moan, that it resulted in the shedding of crocodile tears from a petty scientific official, and that his wailing garnered widespread public notice, only confirms the ridiculous politicization that has befallen climatology.

Update 7 Sep.

I have since calmed down since the day I first learned of Wagner’s action. I was furious that such an obvious political trick could have taken in journalists like the BBC’s Richard Black and others. Too, the pressure put on Spencer and Braswell by the scientific community is astonishing and appalling. If the folks in opposition research put in the same efforts on their own papers, we wouldn’t be in the sad situation we now face.

But Jerry Pournelle, a man whose opinion I respect deeply, who read this piece said:

Briggs is a competent statistician, and his analysis, once he cools the opening rhetoric down, is both comprehensive and competent. (Note that I tend to agree with his opening rhetoric, but I might have preferred it if he had reserved it for his conclusions.)

I wrote the “Politics” in the heat of the moment, bubbling mad. The review of the paper, I had started weeks before. My criticisms of Wagner, Black, etc. would have had more effect if I had written them coolly.

I have also since heard gossip that Kevin “Travesty Doesn’t Mean What You Think It Means” Trenberth put the screws on Wagner. Trenberth sits on a committee which controls funding of Wagner’s position. Wagner might have felt that if he hadn’t done something, life would not go well for him.

I still wonder if he was ending his regular tenure as editor, but in any case, I can now understand, if this rumor is true, why Wagner did what he did. I say “understand” not condone. If the rumors are true—and I emphasize if—Wagner threw Spencer and Braswell under the bus (as the saying goes) to save himself. He might have reasoned that these two were already persona non grata, so he was doing little harm.

But this, if true, is wrong. He should have, as editor, either retracted the paper if he really did have information S&B cheated (you know what I mean), or if he felt the paper merely weak, he could have organized a rebuttal. Those were the only “honourable” actions. Resignation surely was not.

Regular readers of this blog know that peer-review is only the weakest filter of quality. Papers which are not even fit for the bottom of the birdcage are published regularly. Bandwagon papers, especially those purporting to show the doom that awaits us once global warming hits, are especially atrocious. The statistical methods used in these works would have to improve fourfold to approach rotten. (Click on the left “statistics” or “start here” at the top of the page for examples.)

Most climate “skeptics” are like Socrates in that all they claim is that they are not as certain as everybody else. To say that we can not only measure, but predict and even possibly control (the statistical artifact) “global average temperature” to the tenth of a degree fifty years hence strikes us as surpassing hubris.

And let’s not forget that the Greeks thought that there was no greater crime.


  1. I think the last line you wrote is the root of the problem. I’m not a scientist but I know what science should look like and I follow politics enough I know a politician when I see one. And when I look at the AGW debate I don’t see much science on the pro side but I do see a lot of politicians.

    If I speak of “Rev Al throwing the race card” you don’t know if I’m talking about Al Gore, the High Priest of the Warmers or the Rev. Al Sharpton, race hustler and general aparatchik of the Democrat Machine. And that ain’t science, can we at least get a consensus on that?

    If a career can be ended not for disagreeing with the ‘consensus, but for only questioning whether the science really is settled beyond all doubt we aren’t talking about science now are we? No, we have a political litmus test, like pro-life democrats being barred from higher office for disagreeing with a core group that holds veto power in the Party. In science Newtons Laws got tossed overboard, Einstein and Heisenberg’s replacement theories are constantly being questioned, refined and tinkered with by the string theorists. It is what science is supposed to do. Progressive political dogma on the other hand is immutable law and immune from facts and measurable results. AGW looks a lot more like dogma than science.

  2. You claim that Spencer and Braswell’s paper is a minor one, so are other papers you’ve criticized in your blog. My guess is that we tend to be less critical of people we know or on your side.

    As far as I am concerned, all the discussions or speculations as to whether Wagner should resign and why he has resigned can’t change the fact that the paper contains shoddy statistical analysis and shouldn’t be published. A solution is for the journal to accept another paper that corrects the problem. Which is not unusual at all.

  3. JH,

    With respect to “shoddy statistical analysis,” I agree to a point. It obviously doesn’t rise to the level that our host would use, and so could definitely be better. My guess (from what I’ve seen and read about) is that the stats certainly weren’t shoddy with respect to standard practice outside of statistics specialists, which is probably a part of why the reviewers said to publish it.

    I think it just underlines that all these climate guys really need someone who understands the mechanics of analyzing data.

    The really amusing thing here, given the local theme of, “We are too certain of our conclusions,” is that (if I read the review correctly) Briggs is now saying that someone is possibly exaggerating (or at least erring in that way) the uncertainty of conclusions!

  4. JH

    Your claim that S&B contains “shoddy statistical analysis” is not borne out by the review above by our host. Other than the issues he raises with smoothing, and a couple of graphs, can you specify any “shoddy statistical analysis” that Mr Briggs has missed.

    It is trivial to name large numbers of papers published despite such minor issues, and with much more serious shoddy statistics which invalidate the conclusions – Mann Bradley Hughes 98 (hockey stick) and Stieg 09 (Antarctic warming) are two that spring to mind. No resignations there. The proposition that minor issues like this should prevent publication, let alone lead to the editors resignation, is ridiculous.

    If, despite the endorsement of our host (and Pielke Snr, and Lindzen) the paper is genuinely flawed, the normal rebuttal process is quite adequate to correct it, with all the arguments for and against in the open- why the theatrics?

  5. Dear 49er,

    I ask the same question. One obvious answer is that Mr. Wagner is not JH. ^_^

    I guess that sometimes things can be easily resolved but get out of hand for no good reasons. I can’t tell you whether Wagner’s decision to resign is rational since I am not in his shoes. It can be seen by others as the right or coward thing to do. And maybe our beloved DAV is correct.

    Mr. Peter Wilson,

    I agree. If things went wrong, why not just correct them? Why the theatric and the resignation? Beats me. But it’s Wagner’s decision.

    Not trying to be arrogant, but I read the paper myself and rely upon my own statistical knowledge. Please see my comments on the paper in a previous post in this blog.

  6. Mr. Peter William,

    I have hinted in my previous comment that I don’t understand why Mr. Briggs thinks those issues are minor. Also just because other flawed papers are published doesn’t mean it’s OK to publish another one!

  7. This may be a silly question, but since papers such as MBH, Steig et al., S&B [to name a few of the more prominent examples] all seem to be sub-optimal in their handling of statistics, why doesn’t every paper which involves such data analysis have a statistician as a co-author? Or reviewer? That’s a real question, I’m not intending it to be rhetorical — how difficult would it be to have a good statistician included in the preparation or review process?

    Based on our host’s analysis, relatively minor changes in data processing would be required to put the conclusions on a more solid footing. Of course, until we see the results, it remains possible that implementing those changes might yield a different conclusion…

  8. I question the hypothesis of positive feedback. I used to design control systems and we were very careful to avoid positive feedback. Systems with positive feedback do two things, they latch up, full on or off, or they go into limit cycles, on-off-0n-off, AKA oscillate. A system with positive feedback is uncontrollable. When the climatologists claim the climate has positive feedback, they are implicitly saying it is uncontrollable and then they claim you can control the climate by adjusting the carbon dioxide.

    Dr. Wegman, professor emeritus of statistics, in his critique of the hockey stick pointed out that climatologists need help from statisticians.

  9. I just read Bernie’s link to Peter Gleick’s article. Gleick has an amazingly childlike faith in the models.

    If it holds up to scrutiny, it become part of the scientific literature and knowledge, safe until someone can put forward a more compelling theory that satisfies all of the observations, agrees with physical theory, and fits the models.

    Evidently fitting the models is now necessary for a hypothesis to be accepted.

  10. >The problem is that when two times series are smoothed, it always increases the correlation between them, even if the series have no relation to one another.

    Why don’t they get it? Maybe on day one of Stat 001, the prof should say “Don’t ever use smoothed data for anything but Power Point graphs.”

    Bill Drissel
    Grand Prairie, TX, USA

  11. Nick Stokes says,

    “No, Wagner’s boss (DPMI honcho Mr Elvis Wang) gave it to three reviewers that he chose.”

    So, even Wagner’s boss was tired of his Gatekeeping?? 8>)

  12. Just what do you think the probability is that Wagner is a “progressive?” Ironically (and happily) he has REALLY hurt the very commie supporters he is trying to impress/lead/whatever (all IMHO, of course!). LOL! The FORCE is not with the warmers, that’s for sure 🙂

  13. While I agree with your post in general, I do have a few issues with your major problem. I do agree, however, that “smoothing” data will necessarily increase correlation between different data sets, though it is not always a bad thing, and I will explain that below. First…

    Major: In Fig. 2, the pair ran a smoother over the data to “reduce sampling noise.” Of course, smoothers can’t do this. If there is measurement error in the data it has to be measured separately or estimated by some external means to account for it.

    Actually, they can do this, but to what extent depends upon what they mean by “sampling noise.” Smoothing functions tend to be lowpass filters. If the “signal” bandwidth is small relative to the bandwidth of the smoothing filter, then what the filters remove are effectively “noise” (rather, artifacts and/or noise) that are not part of the correlation being sought.

    As I note, the problem depends upon what they mean by “sampling noise.” If they are literally referring to the sensor sampling systems – some form of analog to digital converter, then this noise is (most likely) uniformly distributed quantization niose along with integral and differential non-linearities. These noises are almost always broadband (they cover the entire Nyquist sampling range, though INL and DNL can also result in spurs) and a smooting filter will definitely remove this noise. If the smoothing filter is sufficiently wide (can’t say what sufficiently is without more information,) then any noticeable increase in correlation is actually something that should have been there.

    Smoothers, unless the serendipitously mimic the measurement error process (extraordinarily unlikely here), merely make the data look more pleasing to the eye.

    I think I’ve sufficiently stated why this may not be true. If the sampling noise is indeed what I pointed out, then it really is noise that they are removing and the increase in correlation is not only expected, but real (not an artifact of the smoothing itself.)

    The smoothed data was then used in the regressions (Fig. 3). The problem is that when two times series are smoothed, it always increases the correlation between them, even if the series have no relation to one another. So when the smoothed data is used as input to the regressions, the effect is to inflate (in absolute value) the size of the regression coefficients.

    If the smoothing filters are not actually dipping into the bandwidth of the “signal” itself, then the necessary increase in correlation is not a bad thing per se. In fact, it can be seen as an obvious result simply from the standpoint of removing excess signal energy that was added as a process of the samplers themselves. In other words, the correlations were likely higher before sampling which ultimately degraded the signal.

    Since a central finding is the difference between the regression coefficients of the observations against the same from the GCMs, the effect is to claim a larger difference than exists. This is not to say that no difference exists, just that the actual difference is likely to be less than Spencer and Braswell indicate.

    Maybe, maybe not.

    Roy told me that he tried his models “Without smoothing, with smoothing” and that “The same conclusions” were found. I see no reason to call his honesty into question.

    This result would tend to indicate that my assessment is correct. However, I do think that Spencer should actually publish these “smoother free” results to remove all doubt.

    I can go into more detail if you want. I’ve spent more than my fair share of time analyzing data from sampled systems (from A/D converters) and can easily point to references regarding the points I made. It is extremely common for signal processing systems to filter time series before doing further processing, though always with foresight into what the filters are actually doing to the signals themselves.


  14. The issue is neither technical nor a matter of the reviewers; those are just distractions from the essential fact that the journal followed its procedures; a certain faction did not like the result; and it acted to cleanse the matter. Figure what will run through the mind of the next editor faced with publishing a paper of a skeptical nature, and you will understand the game being played out.

    So here we are, following the path of the Soviet Union in the Lysenko era, complete with public recantings and retractions. It was another terrible day for science; the public will hear about it, further eroding trust in science and scientists, and deservedly so.

  15. Mark T,

    I’m no expert, but it sounds like you’re talking about some sort of digital signal processing. The paper was smoothing a time series of global temperature anomalies, which is already a few steps away from anything I’d be willing to call a raw measurement.

    FYI, I couldn’t find a link in Briggs’ post, but Dr Spencer hosts his paper here (pdf).

  16. RE: John Morris– EXCELLENT commentary sir. Not only were honorable truth seekers such as Einstein and Newton dismissed at times by their contemporaries, but others were JAILED for it– Copernicus and Galileo come to mind– and by those duly authorized to shut them up, by the theocratic oligarchy of their time– which for all practical purposes is NO different than the “oligarchy” of today, in positions of institutional power and influence, and who pursue a political agenda, at the expense of TRUTH.

  17. There may be a political “consensus”, but there is no scientific “consensus” on climate change/global warming. I support any research that questions the dogma of hysterical carbon dioxide vilification.

  18. All climate data is smoothed to one degree or another.

    Monthly data, hourly data is already smoothed. Energy is moving in the atmosphere at the nano to picosecond level.

    Having said that, one should use the highest resolution data available that produces a result which can be understood and/or matches the timeline that the climate effect you are measuring operates on.

    Sometimes even monthly data is too noisy to produce usable results but the ENSO operates with 3 month lags so any smoothing has to be done on a shorter timescale than that.

  19. To spepper,

    Copernicus was never jailed, for his beliefs or otherwise. Of course, the fact that he published his sun centered system posthumously allowed him to avoid controversy. Galileo was placed under house arrest but never jailed. I also think that it is too strong to say that Newton and Einstein were dismissed at times by their contemporaries. The contributions of both were quickly recognized and although there were the usual scientific disputes and clarifications they were never dismissed. Not everything that Newton or Einstein said was correct. The Nazis objected to Einstein’s Jewish heritage and dismissed his work for that reason, but this is a separate issue.

  20. The actions by the progressive warmist consensus crowd are getting more and more desparate. Like Senator McCarthy they are bound to collapse in on themselves when the lies start to build a house of cards.

    People need to see what Burt Rutan is saying about the AGW movement. All you climate scientists have lost credibility. I don’t care if you are a scaremongering AGW proponent or a skeptic. No one believes you anymore. Looks at the latest Rasmussen poll numbers.

    Spencer’s and the skeptics problem is that they keep thinking that playing in the game that is rigged, somehow they can win anyway. Well you know the old argument…when arguing with a fool you also look like a fool.

    Spencer and all other skeptics need to realize the game is up; it is rigged. Eventually all journals will be rigged. This is when the downfall of the AGW movement will happen. Remember what the climategate emails called Lindzen…their token skeptics.

    Stop being fodder for these AGW idiots. Start working together with people like Rutan that have real credibility.

  21. Anyone that wants to understand how this AGW scaremongering will end should read Walter Russell Mead. Especially the three part series on “the failure of Al Gore.” He will help you skeptical scientists understand why going after the science isn’t a good strategy. The real strategy to bring out the rediculousness of the warmists is public policy. Whether the AGW theory (actually never saw a theory, just saw the results of what will happen if we don’t believe this non-theory) is correct or not, the public policy options that are being presented by the politicians and UN are ruinous of the economy. Not now, nor at any time in the future will these policies be enacted, except by stupid countries and states (UK and California). The places that do enact these policies will suffer greatly and be made to look fools in the end.

    You should also read Walter Russell Mead on the Green jobs myth. You can search for it on Google by entering “feeding the masses on Unicorn ribs.” You can say what you want about the weather in 100 years, but if you pour money into green jobs and the jobs don’t materialize, people notice. Now the green economy is in shambles. Solar companies are going bankrupt every day. People are starting to notice.

    You skeptical climate scientists keep beating your heads against the wall. It is people like Rutan, Morano, Mead that will turn public opinion. Stop playing in a game that is rigged.

  22. None of the controversy about CAGW (the basis of all controversy is the “catastrophe” approaching and the need for pre-emptive action) would exist if the science were not said to be settled and the outcome of man’s addition of CO2 to the atmosphere, certain. Spencer and Braswell’s paper is a threat to Trenberth et al – and, hence, Wagner – not because it contradicts major tenets of orthodox, IPCC climate views, but because it introduces reasonable unsettledness and uncertainty into discussion. If those like S & B are recognized as having a defensible position, even if somewhat minor, then those of the illuminati cannot claim to have a universality of knowledge and insight.

    The entire skeptical position is anathema to the warmist creed. As they might say, there are no valid, multiple positions on some things, as we would state about whether gravity causes objects to move toward or away from other objects. The warmist position is not even that the Precautionary Principle says we should support their position in the interim while the world shows us what is happening. The warmist’s claim is they are the holders of truth in all its significant elements; all others who disagree, even in a minor way, are by definition holding falsehoods.

    This is the dilemma of what I call the Unique Solution Syndrome. If you hold to the USS – no pun intended – then you merely need to find A solution to know you have found THE solution. It is an engineer’s failure, a black-or-white view that is based on a belief in having total knowledge or total control. It results in economic and military disasters when its apparent solidness causes no other options to be considered or prepared for.

    The warmists are well aware the weakness of the CAGW story lies in their profession of “settled” and “certain”. They explicitly express outrage at those who cast doubt. A scientific investigation is open to doubt because scientists truly exploring what was considered unknown, recognize limits to their knowledge and understanding. That is why they have others replicate their results, and why graduate students work the same field as their professors. There is always something more to learn and room to modify, perhaps even to tear down.

    If the MSM ever understand and publicize that the science is not settled, and the outcome of more CO2 in the atmosphere, not certain, the hold the Trenberth-Gore-IPCC gang have on local, national and international affairs will wither away. The public will want to have hard, personally recognizable evidence that CAGW is a reality. That is unlikely to happen in the next five years as it hasn’t happened in the last 23.

    We – the world – is at a critical point in the CAGW hysteria (if true, hysteria is justified, by the way). The ravings of messiahs or madmen are about to be put to the test by both powerholders and the planet. The planet can wait for the outcome either way. But the powerholders cannot. If CAGW is not embedded in social structures very quickly, the chances of them being embedded shrink with each day that the world does not spin into a biocidal rage.

    If the world were to start killing off its planetary life, all these desired provisions would be implemented immediately, so you could wonder why the alarmists are concerned about deep-sixing “deniers”. The introduction of doubt is immaterial if that tornado you forecast has just spun your neighbour’s house into the sky. The problem is that the science is not just unsettled and the future, uncertain. The problem is that the science is likely seriously flawed and the future no where near as bad as predicted, and certainly at a level amenable to our ability to change it.

  23. MarkT,

    Agreed. Filters simply attenuate a part of the spectrum. If the noise is stronger than the signal at those frequencies, then the accuracy is improved. Smoothing will only work if the signal is concentrated at low frequencies (likely), or the noise weighted more towards high frequencies (unlikely) or some combination. The odd situation can arise where the signal is entirely at high frequencies which are lost by smoothing, but buried in noise of even stronger amplitude, when smoothing will entirely remove all trace of the signal and yet still gives a more accurate result. (Simply zeroing the data would make it more accurate still.)

    The situation is similar with correlation – which is the covariance of two normalised series. The lagged covariance is the convolution of one (real) signal with the time-reverse of the other, and hence its spectrum is the product of one spectrum with the complex conjugate of the other. You can add up the covariance at low frequencies and high frequencies to get the total covariance, and smoothing both series first just drops the high frequency component of it. If the correlation between low frequencies is the same or stronger than the correlation between high frequencies, the magnitude of the smoothed correlation will increase because of the reduced variance of each smoothed series.

    Both statements are true for a particular (arguably common) class of signal and noise models. But they’re not always true. However, it is definitely true that smoothing does potentially have an effect on the size and significance of the result. So in any case, some thought ought to be given to what effect it has.

  24. Matt – signal processing theory applies irrespective of how the data are generated. However, Spencer is also using data acquired from a satellite, likely from something akin to a CCD, which gets converted to digital data before being sent back to the earth. That is specifically the sort of process i was referring to. I would expect that data to be filtered before processing. The earth station stuff probably is already heavily processed…

    Nullus – indeed, which is why i took issue with a few points. Spencer shows the raw work and he removes all doubt.


  25. Mark T,

    Yes, I would totally agree with you with respect to the raw data from the satellite instruments. But again, we’re already talking about averages of averages when we look at “global temperature data” as used in climate analysis.

    Distinguishing between noise and signal in temperature data is really just begging the question, IMHO.

  26. @ Mark T.

    I couldn’t agree more – your point about people who do DSP generally understand something about the nature of the noise, its relationship to the measuring system and the properties of the underlying signal. A personal climatalogical bugbear of mine is why do climatologists use a rectangular filter, which has poor properties, when with a little effort they could design a filter with characteristics appropriate to the question?

    I quite agree – they should. In my field, which is invetigative medicine, I have always done my own statistics, guided by a statistician, which is a pleasure because I also have a numerate background.

    However, I suspect that many climate scientists have as much insight into statistics as medics and only want a “significant” result. Most statisticians I know find dealing with this approach tedious, at best because those they are helping are generally not interested in formulating the question as a statistical question. Many take the line of least resistance annd point them to a package and tell them to get on with it!

  27. Which is why i predicated my counterpoint on Spencer’s definition of sampling noise. If the satellite samples the image array at some higher rate than than the array bandwidth, then the “signal” is guaranteed to be lowpass w.r.t. the entire data set even if there is no clear distinction between signal and noise. The same cannot be said for the ground data, however.


  28. jae says:
    4 September 2011 at 9:12 pm
    Just what do you think the probability is that Wagner is a “progressive?”

    Wagner is an Austrian professor at the Tech Uni, Vienna.

    Of course he is a “Progressive” (your limited US view of politics shows). Austria’s ruling party is the Socialist Party, the right wing in Austria’s politics is well to the left of Pelosi.

    The whole EU is dominated by the Greens these days, witness Germany’s abandonment of nukes (Austria shuttered it’s only A-power plant decades ago, just weeks before it was to be turned on).

    As it is, Wagner was just trying to keep his livelihood going. Without the AGW scam he probably wouldn’t have had his main job:

    “The Institute of Photogrammetry and Remote Sensing (I.P.F.) conducts research and education in the fields of photogrammetry and remote sensing. It has research groups dealing with geometric modeling (Norbert Pfeifer), physical modeling (Wolfgang Wagner), and image processing (Josef Jansa).

    The groups cooperate closely to fully utilize technological capabilities offered by photogrammetry and remote sensing to provide essential geo-spatial data for environmental and societal applications (flood warning, urban developments, climate change impacts etc.).”

    A tempest in a teapot.

    This quote from Delingpole on this issue sums up Wagner’s importance:

    Incredibly Obscure Editor Of Magazine You’ve Never Heard Of Resigns Over Immeasurably Trivial Issue For No Apparent Reason.

  29. JH writes: “As far as I am concerned, all the discussions or speculations as to whether Wagner should resign and why he has resigned can’t change the fact that the paper contains shoddy statistical analysis and shouldn’t be published.”

    This statement is fundamentally incorrect and represents a misunderstanding of how science and peer review works. In peer review the reviewers generally have three options: reject, revise and resubmit (with major or minor revisions), and accept. Most reviewers, particularly in highly quantitative fields go straight to the conclusions and then validate the numbers. If the analysis is indeed “shoddy” enough to warrant rejection (JT’s assertion) then it’s a pretty good bet the reviewers would catch it (given their credentials as described by the editor). That didn’t happen which means that either (a) the reviewers though the problems identified in the above analysis were trivial or (b) that they accepted the authors’ responses to their feedback regarding this matter. Without being privy to the reviews or the associate editor’s decision-making process, I expect (b) is what actually happened.

    But even if we assume that a third alternative happened – none of the reviewers noticed the statistical flaws and a fundamentally flawed paper was accepted, something which the above analysis does not suggest – then the appropriate mechanism for refuting the claims of the paper is to submit a paper with a contrary analysis using the same data that explains why the conclusions drawn in the original submission are erroneous. The response is not to spike a publication or to browbeat an editor because he adhered to the established peer review process.

    Science is a dialogue and no scientific fact is fixed for eternity. The response to what is rightly described as a “minor paper” reflects an awareness that such a paper introduces a crack in the dominant paradigm. If this paper gets accepted, then more papers like it get accepted, and before you know it the very fundamentals of the paradigm are challenged, the old ways of thinking about problem are no longer relevant, and science moves on. That’s the way science should work, and will work, even if those invested in the AGW == catastrophic environmental change paradigm actively try to resist.

  30. Dr. T. ,

    Ok. I take back the word “shoddy.” Instead, I’ll assign a grade of C to the paper. ^_^

    Let’s forget about all the solvable issues raised here. Making conclusions based on a plot of the resulting estimates without any measure of uncertainty?? Is this flaw major or minor? What is the essential concept of Statistics?

    Yes, my experiences tell me that an associate editor or referees were probably satisfied with the authors’ responses to the referees’ reports and hence accepted the revision. How did you come to the conclusion that I misunderstand how science and peer review works though? Fundamentally incorrect?

    I don’t know exactly what the reviewers think about the statistical analysis. However, I’d like to take a guess that they are probably not statisticians.

    Check out some papers discussed by Mr. Briggs in this blog, the statistical analyses were simply wrong, much worse than shoddy, and they were referred and published!

    I’d like to say that the peer-review process in general works very well for the top journals in the area of Statistics. And I am positive that you know better than I do why this minor paper can cause all the noises in the blogospheres.

    And I usually don’t believe everything, statistics or not, Mr. Briggs writes.

    Mr. Briggs,

    An honorable editor would retract a paper with the agreement of the authors only, I think. Wagner may request the authors to retract the paper, but it really is up to the authors

  31. Mr. Briggs,

    You really seem intent on misrepresenting science on your blog. First it was false “translating” Kirkby’s et al. CLOUD results into “cosmic rays cause clouds”. Now you’re creating a “minor nugget” out of a paper that’s at best deeply flawed and at worst simply fraudulent. Please read this blog entry — read it carefully —


    and then please answer one small question: on what basis did Spencer and Braswell select the 6 models — out of 14 analyzed — that were compared to the observational data?

    I find it hard to believe that Spencer and Braswell chose models that poorly simulated ENSO variability by pure accident. I am also a bit astonished how readily you pass judgements about other people actions and attitudes when it’s quite clear now that you don’t understand the physical reality of the science you choose to comment.

  32. Grzegorz, you and Bickmore (who clearly has a personal grudge against Spencer) are completely wrong. See Spencer’s site. If you had bothered to read anything you would know that the basis for the model selection was the 3 most sensitive and the 3 least sensitive. If you want to see the results of all the models you can see them on Spencer’s website.

    I am astonished that you pass judgements (and accuse of fraud) people whose research you clearly have no understanding of.

  33. Jumbo,

    The point is that Spencer was trying to make the data say something about climate sensitivity, so he selected the three least and three most sensitive models, but he says he did analyze the other 8, too. Several of the models with “medium” climate sensitivity do much, much better than the others at mimicking the data pattern. Therefore, how well the models do at that obviously isn’t strongly correlated to climate sensitivity.

    And just because I think Spencer needs a good debunking doesn’t mean I have a “personal grudge” against him. I don’t even know him.

  34. @Jumbo

    I’m sorry, it’s really hard to see the choice as anything else than cherry-picking those models whose output doesn’t agree most with observations. I don’t think Spencer and Braswell are so ignorant that they simply didn’t understand the general picture when they got much better fits using some of the models that they chose to ignore.

  35. “To say that we can not only measure, but predict and even possibly control (the statistical artifact) “global average temperature” to the tenth of a degree fifty years hence strikes us as surpassing hubris.

    And let’s not forget that the Greeks thought that there was no greater crime.”


    This is exactly what I thought the moment I heard on the news, and later read:

    “We must be honest about what we have got” said UNFCCC Executive Secretary Yvo de Boer. “The world walks away from Copenhagen with a deal. But clearly ambitions to reduce emissions must be raised significantly if we are to hold the world to 2 degrees” he added.

    Amazing to me.

  36. Grzegorz Staniak says:

    7 September 2011 at 5:58 pm

    I’m sorry, it’s really hard to see the choice as anything else than cherry-picking those models whose output doesn’t agree most with observations.

    Except that the others don’t agree any better, either, as Spencer has shown. He chose to represent the range using least/most sensitive. I suppose he could have chosen two from each end and two from the middle, that point is certainly arguable, but it would not have changed the conclusions. Hd he done so you still would have been arguing “cherry picking” simply because he did not show the hundreds of combinations that would have been required for all 14. At best this is nothing more than the same old red herring that people like you would prefer to use as refutation instead of the truth of the results themselves.

    I don’t think Spencer and Braswell are so ignorant that they simply didn’t understand the general picture when they got much better fits using some of the models that they chose to ignore.

    They neither got better fits nor ignored other models.

    I understand you seem to have a fascination with Bickmore, but his fundamental argument seems to be a personal one, not a scientific one. That’s equivalent to argumentum ad hominem.


Leave a Comment

Your email address will not be published. Required fields are marked *