Statistical Analyses of Surface Temperatures in the IPCC Fifth Assessment Report

Guest essay by Douglas J. Keenan

Temperatures on Earth’s surface—i.e. where people live—are widely believed to provide evidence for global warming.  Demonstrating that those temperatures actually provide evidence, though, requires doing statistical analysis.

All such statistical analyses of the temperatures that have been done so far are fatally flawed.  Astoundingly, those flaws are effectively acknowledged in the IPCC Fifth Assessment Report (AR5).

The flaws imply that there is no demonstrated observational evidence that global temperatures have significantly increased (i.e. increased more than would be expected from natural climatic variation alone).

Despite this, one of the main conclusions of AR5 is that global temperatures have increased very significantly.  That conclusion is based on analysis that AR5 itself acknowledges is fatally flawed.  The correct conclusion is that there is no demonstrated observational evidence for global warming.

Full critique at  http://www.informath.org/AR5stat.pdf

===========================================================

He closes with these quotes:

The reason for so much bad science is not that talent is rare, not at all; what is rare is character. People are not honest, they don’t admit their ignorance, and that is why they write such nonsense. — Sigmund Freud

Half the harm that is done in this world is due to people who want to feel important. They don’t mean to do harm — but the harm does not interest them. Or they do not see it, or they justify it because they are absorbed in the endless struggle to think well of themselves. — T.S. Eliot

You should, in science, believe logic and arguments,carefully drawn, and not authorities.—Richard Feynman

0 0 votes
Article Rating

Discover more from Watts Up With That?

Subscribe to get the latest posts sent to your email.

84 Comments
Inline Feedbacks
View all comments
BioBob
October 30, 2013 12:33 pm

I see two major issues with climate analysis that are basic to all further considerations.
1) Temperature data is the result of chaotic processes and can not be accurately predicted. This results in the prohibition of parametric analysis. Assuming observations drawn from today’s “population of temperatures” will be drawn from the same population tomorrow are false. Temperature has guaranteed black swan events – essentially 1% chance for a new record each year for any 100 year period. There are no many sided dice that encompass all possible temperatures.
2) ALL of the currently available data stinks ! Any rational analysis of current data sets will conclude that the errors are larger than any “average” trend. ALL current analysis have laughable estimates of ALL error types. There are no random samples. There are generally no replicates or an inadequate number of replicates. Data discontinuities are glossed over. Bullsh*t adjustments are de rigeur.

Janice Moore
October 30, 2013 2:15 pm

SUGGESTIONS (from a non-scientist)
Dear Dr. Keenan,
Given that the above excellent analysis was largely written to educate non-scientists, I suggest:
1. Page 9 – re: “… the 90%-confidence interval for the global temperature in that year: [−0.10 °C, −0.36 °C]… ” — a brief explanation of anomalies would be helpful (otherwise, the figures are undecipherable and meaningless to a non-scientist). Chapter 2.1 of Bob Tisdale’s new book Climate Models Fail is one source to link for an explanation of anomalies.
2. To prevent misunderstanding, I suggest reconciling the following two clauses:
“The flaws imply that there is no demonstrated observational evidence that global temperatures have significantly increased (i.e. increased more than would be expected from natural climatic variation alone).” (Executive Summary, p. 1)
With:
“Simply put, no one has yet presented valid statistical analysis of any observational data to show global warming is real. Moreover, that applies to any warming—whether attributable to humans or to nature.” (Sec. 7, p. 11)
3. “Having the confidence interval so far away from including 0… .” (p. 11) — Need to re-word to prevent non-scientist confusion.
4. “… SPM ignores what is said in Chapter 10. It does that even though responsibility for the statistical analysis lies with Chapter 10.”
Suggest: “… SPM ignores Chapter 10 even though per Chapter 2 the only valid statistical analysis is done in Chapter 10.”
5. p. 12 — correct for capitals in title: “paper is entitled “Statistical Models and the Global Temperature Record”.
6. minor vocabulary clarification: “tabled” – Does that mean to pose a question? In the U.S., it would mean to set aside and, possibly forever, ignore it. (Can’t Lord Donohue simply issue a subpoena and force the witness to appear and testify? — never mind my queries here if this paper is completely aimed at a U.K. audience — not important, in any event, to the fine discussion you present in your paper.)
7. While your bending over backward (on p. 16) to portray the climastrologists as innocently mistaken does credit to your good nature, it is so implausible (not a justified assumption, ahem!) that I would suggest you delete it. It is obvious to a layperson that these Ph.D.’s know good and well that what they are doing is wrong. You need not say so explicitly, but to toss a high-necked, full-length, flannel nightgown on a prostitute in an attempt to portray her as naive is a little ridiculous and, thus, detracts a little bit from an otherwise well-written expose.
8. p. 17 — capitals in title: ““Global Temperature Trends”
9. Fascinating reporting of the religious-like attachment of those who believe in Neo-Darwinism to a mistaken radiocarbon dating regime. You were treading on hallowed ground, there!
EXCELLENT REPORT.
Thank you for sharing it with us for free.
Your grateful student (and humble proof reader),
Janice
Note: I apologize if I have duplicated suggestions above — I did not take the time to review the comments written while I was writing this.

Janice Moore
October 30, 2013 2:20 pm

Hi, Geran. Re: my redeeming myself — huh?

1sky1
October 30, 2013 2:26 pm

Keenan is entirely correct in pointing out that the critical question of assumed time-series model strongly prejudices all statistical conclusions about “trend” drawn from short measurement records Those assumptions are typically made ex cathedra–on both sides of the AGW debate. Based upon model-free spectral analyses of actual meaurements as well as proxy data, I have long argued that linear-trend plus AR(1) models are simply inappropriate. What is fitted to the data by linear regression or by nonlinear methods cannot be taken as a SECULAR trend, persisting far beyond the time-frame of the records.
I must quibble, however, with the claim that Koutsoyiannis, who resorts to Hurst-Kolmogoroff persistence models married to the unproven notion of self-similarity at all time scales, advances our insight very far. The power density spectra almost invariably show the presence of strong oscillations over a very wide range of frequencies that are incompatible with simplistic H-K “red noise” spectra. Sadly, virtually none of the statisticians whose voices are repeatedly heard in “climate science” are truly adept at geophysical time-series analysis and modeling.

Editor
October 30, 2013 2:30 pm

Mosher
Instead of carping from the sidelines, perhaps you can answer Doug’s central point, and tell us what evidence you have that any warming recorded is “statistically significant”.
We know you are much cleverer than the rest of us, so that should not be terribly difficult.

geran
October 30, 2013 2:40 pm

Janice Moore says:
October 30, 2013 at 2:20 pm
Hi, Geran. Re: my redeeming myself — huh?
>>>>>>
You’re on the “come back trail” from calling me a screaming monkey….

geran
October 30, 2013 3:10 pm

Paul Homewood says:
October 30, 2013 at 2:30 pm
Mosher
Instead of carping from the sidelines, perhaps you can answer Doug’s central point, and tell us what evidence you have that any warming recorded is “statistically significant”.
We know you are much cleverer than the rest of us, so that should not be terribly difficult.
>>>>>>>>>>
Paul, Mosh will not be returning any calls tonight. He has no evidence, only his personal wishes. He wishes so much that mankind was ruining the planet. He is angry because CO2 has gone up but we are not yet “boiled”. To him, his computer models cannot be wrong….

Janice Moore
October 30, 2013 3:20 pm

My dear Geran,
I believe I said that you were acting “like” a screaming chimpanzee.
#(:))
Did I just make it three feet farther along the “come back trail?”
I hope so.
Your truth in science ally,
Janice

geran
October 30, 2013 3:31 pm

Janice Moore says:
October 30, 2013 at 3:20 pm
….
Did I just make it three feet farther along the “come back trail?”
>>>>>>>
We’re talking centimeters here, but stay with the program. Rehab will set you free….

1sky1
October 30, 2013 4:02 pm

As an example of the long-term natural variability manifested by paleo data, see the power density of GISP2 del18O-isotope data at:
http://i1188.photobucket.com/albums/z410/skygram/graph1.jpg
P.S. Some day, i’ll have time to organize my photobucket account.

October 30, 2013 4:11 pm

In his guest essay Douglas J. Keenan said,
In AR5, Volume I, Chapter 2 is devoted to observations of the atmosphere: observations of temperature, humidity, wind, etc. The statistical method used to evaluate trends in those observations is described in Box 2.2, which is subtitled “Trend Models and Estimation”.

[The first part of Box 2.2 from AR5, Volume I, Chapter 2]
“. . .
[paragraph #3] The quantification and visualisation of temporal changes are assessed in this chapter using a linear trend model that allows for first order autocorrelation in the residuals (Santer et al., 2008; Supplementary Material 2.SM.3). Trend slopes in such a model are the same as ordinary least squares trends; uncertainties are computed using an approximate method. The 90% confidence interval quoted is solely that arising from sampling uncertainty in estimating the trend.
. . .
[paragraph #4]. There is no a priori physical reason why the long-term trend in climate should be linear in time. Climatic time series often have trends for which a straight line is not a good approximation (e.g., Seidel and Lanzante, 2004). The residuals from a linear fit in time often do not follow a simple autoregressive or moving average process, and linear trend estimates can easily change when estimates are recalculated using data covering shorter or longer time periods or when new data are added. When linear trends for two parts of a longer time series are calculated separately, the trends calculated for two shorter periods may be very different (even in sign) from the trend in the full period, if the time series exhibits significant nonlinear behavior in time (Box 2.2, Table 1).
. . .”

The third paragraph states that the IPCC has chosen a statistical model that comprises a straight line with first-order autocorrelated noise. If you are unfamiliar with autocorrelated noise, that does not matter here. What is important here is that a model has been chosen, yet there no justification given for the choice. The failure to present any evidence or logic to support the assumptions of the model is a serious violation of basic scientific principles.
The fourth paragraph acknowledges that “residuals from a linear fit in time often do not follow a simple autoregressive … process” (indeed, that is well known). This means that the chosen model does not fit the data; i.e. the model is ackowledged to be statistically inappropriate.
Box 2.2 concludes with this statement: “The linear trend fit is used in this chapter because it can be applied consistently to all the datasets, is relatively simple, transparent and easily comprehended, and is frequently used in the published research assessed here.”
Box 2.2 can be summarized as follows. A statistical model was chosen, without any statistical justification. Moreover, the chosen model is believed to be statistically inappropriate for climatic data. The model was chosen anyway for two reasons: first, choosing a more appropriate model would require more effort; second, almost everyone else has been using the same model—though also without statistical justification.

– – – – – – –
Douglas J. Keenan,
Thank you for making it clear; inappropriate statistical models were chosen in AR5 and the authors of AR5 admit it.
The time is fast approaching for a formal independent statistical expert panel to do a comprehensive audit in full public view. An audit with full access to all internal docs, communications and archives of their non-open forums. N’est ce pas?
John

Dr Burns
October 30, 2013 4:18 pm

I do wonder about IPCC’s claimed accuracy of global average temperatures, especially as listed by CRU to 3 decimal places. They rely on Jones et al and the √n rule. The surface stations project shows that at least 6% of these measurements have errors greater than 5 degrees C. Now this isn’t much better than putting a finger in the air. If we could get all 2,400,000,000 internet users to put their fingers in the air to estimate their local temperature, we should be able to calculate an average global temperature and get an error in the average of 0.001 degrees C. I’m not a statistician – what’s wrong here ?

Janice Moore
October 30, 2013 4:25 pm

Dear Geran,
I did not realize how deeply I had offended you by my exaggerated description of your comments on that thread. I have learned a lesson (heh, well, I hope I have) about hyperbolic and carelessly (and poorly) executed repartee. I am sorry for the hurt I have caused you. I was wrong.
Please, forgive me.
Your wiser, humbler (we can only hope), friend,
Janice

geran
October 30, 2013 4:44 pm

Janice, no harm done here.
“Words” don’t hurt me. I am not harmed, and you owned up to your loss of control. So, all is level. We skeptics fight amongst ourselves, but it only makes us stronger to fight the REAL enemies.
Now go get the Warmists, and take no prisoners! (They FEAR the TRUTH.)

Janice Moore
October 30, 2013 4:51 pm

Thanks, Geran.

Paul Penrose
October 30, 2013 5:05 pm

Douglas, you have already been labelled as a “denier”, therefore your entire analysis is invalid; at least to the warmists and the great unwashed masses. Too bad too, because everybody needs to understand this issue.

Manfred
October 30, 2013 7:11 pm

Is anyone able to explain why Mosh isn’t happy that we’re not boiling ourselves to death?

October 30, 2013 8:32 pm

I garnered from Met Office’s replies to Parliament that they chose the AR(1) model because everyone else was using it. In other words, they had no theoretical basis to their choice. Maybe I shouldn’t be, but I am somewhat surprised that the IPCC is clinging to first order autoregression for determining significance, especially after Met’s admitting it was unsuitable. Don’t they not learn from those behind the woodshed experiences?

October 30, 2013 8:34 pm

Ugh… I meant “Don’t they learn from those behind the woodshed experiences?”

Matthew R Marler
October 30, 2013 11:20 pm

Steven Mosher: “The correct conclusion is that there is no demonstrated observational evidence for global warming.”
Yes, that was a mistake. Touche’
His other punchline is correct: The flaws imply that there is no demonstrated observational evidence that global temperatures have significantly increased (i.e. increased more than would be expected from natural climatic variation alone).

October 31, 2013 5:09 am

Janice Moore, October 30 at 2:15 pm
Thank you much and very kindly. About your first three points….
1. Good point—will amend.
2. The Executive Summary is intended to be a true executive summary; in particular, it is not an abstract. I am unskilled at writing something like this; so the Summary will be written jointly with someone else (Lord Donoughue). For now, I decided to include a rough version. That was perhaps a mistake, though, because you, and others, have found valid problems with it.
3. Thanks—will do.
1sky1, October 30 at 2:26 pm
About the work of Koutsoyiannis, I agree that the model that he has found does not fit the data well; indeed, I had a brief conversation with Koutsoyiannis about that last year. The critique, though, does state that he has not found a viable model.
What makes the work of Koutsoyiannis especially important is that it provides the only example of attempting to use physics to constrain the set of candidate statistical models. Using physics to constrain the candidate models is an obvious strategy, but implementing that strategy is difficult. Virtually all other researchers who attempt to statistically model climatic time series constrain the candidate models by assuming that the models are analytically simple, e.g. by assuming linearity. That is wrong, but it allows those researchers to claim “results”. Koutsoyiannis has taken the correct, but much more difficult, path.

Janice Moore
October 31, 2013 11:03 am

Dear Mr. Keenan,
You are so very welcome. And, thanks, for letting me know that I could be of some little help. Re: Koutsoyiannis, his work IS a good example for the point you were making; I guess you just need to make the purpose of citing K. a bit more clear for readers like 1sky. Of course, you can’t write a paper carefully enough for those who read it in haste. Just have to ignore the criticisms of such readers.
Best wishes for success with this, your latest endeavor on behalf of truth.
Yours,
Janice

October 31, 2013 12:08 pm

Doug, what you write for me, both in your essay and in comments, is extremely flattering. Thank you very much.
Also, thank you Janice for your kind comment.
1sky1, I appreciate your criticism. Indeed, self-similarity at all time scales (using just one parameter) may be too simplifying. However, I find it extremely useful to derive simple and parsimonious models by using the simplest possible constraints. Then by adding more constraints, one may approximate reality better. Of course, there is much more to do than what is reported in a single paper to which Doug mostly refers. We are continuing our studies in this direction (in very tough conditions though—note, our university has been closed for the last couple of months due to substantial cuttings and firing of half of the personnel, and its future is uncertain). Anyhow, you may find additional stuff and replies to many comments in a recent discussion in http://www.climatedialogue.org/long-term-persistence-and-trend-significance/

Janice Moore
October 31, 2013 2:09 pm

Dear Professor Koutsoyannis,
I’m so sorry to hear that you are working under such tough conditions. A world-class scientist like you should have the finest that academia has to offer. You come from a land of heroes, who gave to the world great science, art, and literature…….. who fought for democracy………. who ran 14 kilometers in bare feet and cried,
“We have won!”
Just a little encouragement from an American admirer. An American man singing in a rather ordinary way, but with that optimism which has been and always will be the gift of America to the world. You — can — do — it!

Truth will win.
Take care,
Janice

October 31, 2013 3:10 pm

Dear Janice, it’s a great feeling to hear from so far away that “You’ll never walk alone”; thank you indeed for this friendly encouragement.
Demetris