Bishop Hill Keenan does AR5
Oct 30, 2013 Climate: Statistics Climate: Surface Doug Keenan has written a critique of the IPCC's handling of statistics in AR5. Suffice it to say he is not impressed.
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 be en 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 in creased 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.
You can read it here.
Reader Comments (93)
Roger Longstaff, Martin A
"could you give the provenance for that formula?"
My pleasure.
http://www.globalwarmingequation.info/eqn%20derivation.pdf
To those of you who jumped on my "250 year" figure upthread. The point I was making was that if the LIA is regarded as a minimum then there seems to be an upward trend in temperatures since then. I don't think I'm being controversial in this and Mosher's comments at WUWT rather reinforce my comment upthread, that Doug's article appears to imply that the LIA was not real.
I'm with steveta_uk, and without a re-write Doug will be dismissed as a crank.
For those interested, this is a fairly complete history of the concept of CO2 as a greenhouse gas.
There is also a good reference list, though many of the older papers are not available online.
http://www.aip.org/history/climate/co2.htm#M_24_
Was the LIA real? That is not the question or the implication of Doug's work. Was the LIA significant is more of a question. Or was it merely the lowest recent point in a 'random' walk. Not that temperature movements are random per se, but that each annual temperature is the effect of that year's conditions on the temps at the start of it. With all your talk of cycles and forcing, maybe it is just that. After all, the temperature right here in Oxfordshire varies more in an hour than all the 'climate change' in all the decades and centuries in the record. Point 8 of a degree isn't a whole lot to worry about.
So, was the LIA real is a question which must be resolved by postulating a mechanism by which it might be real rather than a random artifact, hypothesising a model to describe the effect of that mechanism and testing the data for significance vs that model.
OR, I don't understand any of this. That's the null in this case.
I take Doug Keenan as meaning what he says: that there’s no evidence for global warming, man-made or otherwise. In previous articles he’s compared it to a random walk, which I take it is what his first graphs show. Perhaps it would be better to talk about “a causal process of global warming”, or some other formulation that brings out his basic point: that without a valid statistical analysis there really is no evidence for any kind of non-random global warming.
We all understand that saying “it’s warmer today than it was yesterday” is no evidence for anything. My understanding of Keenan’s article is that he’s simply extended this logic over two centuries. Have I got it wrong?
I think the point Doug Keenan is making is that his analysis of the temperature history demonstrates that the observed ups and downs are indistinguishable from natural variation. Of course global temperatures have varied and the Little Ice Age was colder than present times but there is no systematic process of warming or cooling that can be dignified by a law of Global Warming (or Cooling) that can be used to make predictions of future behaviour.
Or I too am covered by Rhoda's null case.
Okay, I have changed “there is no demonstrated observational evidence for global warming” to “there is no demonstrated observational evidence for significant global warming”.
I had considered doing that originally, but thought that when people generally use the term “global warming”, they actually mean significant global warming, and so I had omitted “significant”. Obviously, the omission was muddying the waters. Thanks much for the feedback.
http://www.globalwarmingequation.info/eqn%20derivation.pdf
Oct 31, 2013 at 8:31 PM Entropic man
Thanks EM. I looked at the paper. I rest my case.
I recall Fred Singer describing the response of greenhouse warming to logarithm of CO2 concentration. And Fred Singer really is a scientist, and an atmospheric scientist to boot. I tend to take what he says seriously.
Rhoda: "Or was it merely the lowest recent point in a 'random' walk"
That's exactly the point. If you don't know the long term processes, and the periods, that are at work, you can erroneously consider a process to be non-stationary when in fact it is a stationary process with a long period of some sort. Like a 70 year cycle, or a 220 year cycle, or a 1500 year cycle, or a superposition of multiples of such cycles.
Therein lies the problem of statistical inference.
Thinking Scientist / Rhoda - I recall this was all discussed a few years ago on BH or WU or maybe Tallblokes. The key point being if there are multi-decadal cycles e.g 80 years, then it is pointless trying to identify trends until you have at least 2 or more wavelengths of data. As it seems likely that the Holocene climate is being influenced or determined by 60-80 year Ocean cycles, and a much longer 1000 year cycle, Doug's point that there has been no warming in the last 100 years has some validity. But, it may be best for him to add 'significant' for reasons of expediency.
I can't find/remember the original thread now but this more recent thread is relevant - http://www.bishop-hill.net/blog/2013/8/5/akasofus-model.html
As usual others can express what I'm trying to say better than I do (Rhoda, geoffchambers and Ronaldo). I think the redrafting of the sentence by Doug will help prevent semantic discussions about what is being said which would detract from the intent of the arguments advanced.
I've been reading the CA post about the majority of SH composite temperature proxies amusingly stuffed with NH data. That and Doug's article pointing out the absurdities in the AR5 statements about models and trends and how they're "useless but easy-to-understand" does make me wonder if I've entered some parallel existence replete with bizarre quasi-para-meta-cargo-cult science. Surely there must be many climatologists who know all is not right in the discipline?
Another point many miss is that natural variability could well be much greater than we expect on long time scales - to the point that natural variability could easily swamp any putative CO2 driven warming, even with the greenhouse gas equations as they are.
It could well be that in the context of climate variability, natural climate change could be a far bigger risk than AGW.
All you "natural variation" enthusiasts, remember that changes in climate are not magic. A cooling trend is due to reduced energy input ;a warming trend requires increased energy.
It's all very well to waffle vaguely about such things, but nobody outside your own church is going to take it seriously until you can show mechanisms and energy budgets.
Martin A complains about my use of the CO2 forcing equation, a quantitative description of atmospheric behaviour based on physical theory and with observational backing from changes in temperature, OLR and DWIR.
Why can none of you provide something quantitative for your preferred natural variations?
Regarding cycles, the AMO is reputed to have a roughly 60 year period. Unfortunately anyone trying to use it to explain the variations in the temperature record will find it does not fit the record very well.
Look at the GISS graph.
That shows peaks at 1880,1940 and 2000. It also shows minima at 1910 and 1970 which would match a 60 year cycle. Take away the long term trend and the difference between each peak and trough is about 0.3C.
The problem is that this alone cannot explain the shape of the graph. In the two cycles between 1880 and 2000 the temperature rose 0.8C. If the cycle were the only driver 1880, 1940 and 2000 should have had the same temperature.
In practice the long term trend has produced a much greater variation than the cycle.
EM "It's all very well to waffle vaguely about such things, but nobody outside your own church is going to take it seriously until you can show mechanisms and energy budgets."
So please provide the explanation by "your church" of the mechanism for the LIA and the energy budget for that period. A demonstration of an error in a theory does not require a competing theory to be offered. That is an un-scientific logical fallacy presented by those supporting (C)AGW to deflect criticism.
You are putting the cart before the horse, as are the CAGW supporters who assign all atmospheric changes to world domination by CO2. If you don't know the periods or nature of processes such as PDO or others then you cannot make statistical inference. This is the whole point of Doug Keenan's essay, just to bring it back to the subject of the thread. There is no justification for the statistical models that lead to the simplistic view that recent temperature changes are a trend plus noise. So to your subsequent comment of "In practice the long term trend has produced a much greater variation than the cycle." the proper answer is: what long term trend? What statistical model are you using as the basis of that claim, given that climate is thought to be a coupled chaotic system and therefore might be expected to exhibit some long term random walk behaviour?
Just because your eye tells you that you can draw a straight line through the data, so using Excel you do that and then, using statistics based on independent observations with random errors such as a Student T test, you say "look here's a statistically significant trend" does not mean your conclusion has any validity. Doug's point is that climate is so complex, we don't really know even what statistical model to choose, to test for trends or significance. Any random walk will have apparent trends, depending on the parameters and the length of the series, as Doug shows in the essay. I could easily generate stochastic simulations, with complex properties, that you could mistake for having trends, even though they were generated by a stationary (ie trend-less) model. That is exactly what I show students when I teach them this stuff.
And the further issue is that the moment you assume some particular statistical model, not only are you defining the part of the model you think is predictable (the trend), your uncertainty model is then formed from the magnitude of the residuals. At that point, if your model assumption is wrong you are making a Type 1 statistical error ie an inaccurate prediction with great confidence. Exactly what I think is happening with CAGW, and the reason why the models have failed to predict a 17 year plateau in temperature. It looks to me like a classic type 1 prediction error. We are not so many years away now from the period of no trend being longer than the period over which the CAGW got so excited and hysterical in the 1980's and 1990's
You have just jumped to the wrong conclusion again and failed to take on board any of the valid points made by Doug in his essay. Presumably because you want to believe its CO2 wot dunnit. After all Entropic Man, why let data get in the way of a good theory?
EM - as Edim pointed out on here a few months ago, it can be argued that the AMO is a misnomer - and it is effectively a global multidecadal oscillation (GMO). The correlation does seem rather good:
http://www.woodfortrees.org/plot/hadcrut4gl/from:1880/detrend:0.762/plot/hadcrut4gl/trend/detrend:0.762/plot/esrl-amo/plot/esrl-amo/trend.
As for your point about warming period needing an extra energy input and the converse for cooling, a 1 or 2% decrease in cloud cover results in more energy input from reduced albedo than the alleged extra 'forcing' from man-made CO2.
Just look at the cloud cover data and Hadcrut4 for the last 30 years, i.e. compare these graphs: Global cloud anomaly (%), 1983-2012) and Hadcrut4, 1982-2013. (if you have an image editor, invert the Hadcrut4 graph and resize it by a ratio of about 1.13 to get the scales to match).
CO2's warming effect is negligible once it is above 300ppmv. It may be Halloween, but don't you get bored trying to scare the children (and gullible adults) with this cargo-cult science all year round?
EM - You state these things with total apparent confidence based on your belief that things are really pretty simple and can be reduced to formulas with coefficients known to two decimal places.
Yet the system and its dynamics are essentially simply not understood. Why should a cooling trend not be due to other causes? [increased energy output, transfers of energy within a system having complex and essentially unknown dynamics, change in atmospheric water content,...] Likewise for warming trends.
"Martin A complains about my use of the CO2 forcing equation, a quantitative description of atmospheric behaviour based on physical theory and with observational backing from changes in temperature, OLR and DWIR."
You make it sound impressive but it's essentially rubbish all the same.
If someone is talking evident rubbish, the fact that someone else has not provided a correct and comprehensive explanation does not stop what they are saying being rubbish.
So, Steven Mosher doesn't think much of Keenan's argument. Perhaps more significantly lucia does't think much of it either ("I’m very reluctant to rely on Doug Keenan for anything factual."). Is there anyone with acknowledged statistical expertise who has spoken out in favour of it? If not, why is it still getting prominent and implicitly favourable coverage on BH?
Nov 1, 2013 at 9:09 AM anonym
Anon - you seem to be one of those who holds the notion that something is only valid if a consensus of experts has endorsed it. In any case, a good number of regular BH posters have statistical expertise - but so what?
The beauty of of Keenan's paper lies in its clarity and simplicity - the validity of what he says is directly apparent from its straightforward logic. Surely you can read it and evaluate it for yourself, rather than having to quote 'lucia' - whoever they are.
em, you need to learn some basic physics. There is no 'magic' required to generate long-term fluctuations in temperature in complex systems, it happens entirely naturally. People with a solid background in applied mathematics, differential equations, physics and modelling, such as Spence and I, understand this well. The view that you seem to have acquired, that every little wiggle in past temperatures has to be 'explained' as a result of some kind of 'forcing', is a very naive one. You yourself have mentioned the AMO. What do you think causes that?
Of course it goes up and down. It is chasing equilibrium on multiple timescales. Every instant that artifact we call the global average temperature is subject to a range of influences which cause it to move in accordance with the resultant. It is quite possible that the range of influences will never be repeated. It may be that we see cycles because we are designed to see cycles. Some of them are real, perhaps though some are not. We know that the solar cycle does not have a reliable period or amplitude. That may apply to all the observed cycles. So the conditions of today or 2013 or 21st century never come round again.
Now, I may be wrong. Your pet theory be it CO2, ocean cycles, the sun, orbital cycles, whatever, may be correct. But you can't show it from a few decades of global average temps unless you have a valid statistical model and a good fit.
..which I believe is a paraphrase of Doug's contention. It makes no assertion for any theory but does assert the inadequacy of that particular time series to prove anything much is happening beyond the null hypothesis which I choose to state as 'nothing much is happening' but others might say 'it's all within the range of natural variation'.
anonym
With all due respect to Lucia, she’s a statistician. She would say that, wouldn’t she?
If I understand the paper correctly, Keenan’s argument strikes at the heart of time series analysis in general. An awful lot of people make a good living from interpreting time series graphs. They’re the chicken’s entrails of our times.
Lucia is somewhere between a physicist and an engineer. I would take what she says very seriously: her knowledge is wide, her honesty is obvious, and her judgement is good. But note that one of anonym's links takes you to a comment by Nick Stokes, a far more slippery figure.
Is Keenan right? On the narrow technical points he makes, yes. Does this matter? Now that's a more subtle question. The more intelligent critics of Keenan (such as Lucia) start by pointing out that this time series is physically generated and so a priori we can rule out certain mathematically elegant generating processes; this is pretty much the line Mosher takes too. I don't find their arguments quite as convincing as they seem to, but they are not bad arguments.
Anonym,
The comment on Lucia's blog is made by Nick Stokes and not Lucia, so should be taken with a large pinch of salt. It was in relation to Doug Keenan's account of parliamentary questions asked by Lord Donohue. Since these are a matter of public record, it is difficult to see why Nick Stokes would make such a complete foll of himself. Again.
@ Jonathan Jones, 10:19 AM
That the time series is generated by a physical process is indeed important. My critique discusses this, especially in §10.
Oct 31, 2013 at 5:36 PM | Registered CommenterMartin A From whence the logarithm?
I seem to recall that the logarithm was mentioned by Arrhenius, 1896, first at p. 244 of this English translation -
http://www.rsc.org/images/Arrhenius1896_tcm18-173546.pdf
@Jonathan Jones, @andymc
Gah, yes. That "I’m very reluctant to rely on Doug Keenan for anything factual." comment was by the egregious Stokes, not lucia: what a clanger, sorry about that. Overall it seems I represented lucia's opinion fairly correctly though.
Nov 1, 2013 at 9:41 AM | Registered CommenterPaul Matthews
While on this topic, I often see the explanation "Of course the globe is warming. It's recovering from the cold period of the Little Ice Age."
This worries me unless it is qualified by a physical mechanism - often, it is not.
Do we have any preferred mechanism, such as energy stored in oceans, that can back up this claim? With numbers?
Interesting thread but not for me helped by the intrusions of EM and the various responses. I wouldn't say that about all threads where EM has challenged something or other. In other words it's both context and reader dependent - making the job of host devilish difficult. Meanwhile anonym did something extremely sly with that quote from Nick Stokes. Doug I'm sure will have already risen about it but the attempt to drive a bigger wedge than necessary between known people, with informed but different perspectives, is something we should recognise as significant and indeed encouraging.
Douglas Keenan 10:53
Indeed you do discuss this point. My understanding of Lucia's position is that she thinks you don't take this point seriously enough, much as your position is that other people don't take your points seriously enough. In particular she argues that temperature cannot arise from a genuine Hurst-Kolmogorov process. At which point things get very technical very rapidly.
@Richard Drake
No, I did something extremely stupid. Please take that back right away.
@Martin A
This a travesty of what I said. The fact that it's hard to find even one person of acknowledged statistical expertise who finds Keenan's argument convincing - even among the statisticians active in the climate debate on "his side", loosely speaking - is a rather bigger problem than the lack of a consensus behind it. I am not convinced by the argument that Keenan's case is so self-evident that I should disregard this.
I won't hear any criticisms of Lucia. She is intellectually honest and a very talented scientist. She runs one of the best science blogs on the net with very light-handed moderation.
Interestingly, though, I did disagree with her position on this particular issue. It says much about her integrity that she raised no objection at all to my posting an article on her site (subsequent to the one referenced above) which largely supported Doug Keenan's viewpoint on the statistical analysis. The article is here - for nerds and statisticians only, I'm afraid:-
http://rankexploits.com/musings/2013/the-occams-razor-oscillatory-model/
My apologies for the shameful self-plug. For those of you who cannot follow the science, I would just ask you to imagine Real Climate or Skeptical Science allowing someone to post an article which questioned something it had published. Go Lucia!
"There is no 'magic' required to generate long-term fluctuations in temperature in complex systems, it happens entirely naturally."
"The view that you seem to have acquired, that every little wiggle in past temperatures has to be 'explained' as a result of some kind of 'forcing', is a very naive one. You yourself have mentioned the AMO. What do you think causes that?
Paul Matthews
I hope climate scepticism has not abandoned cause-and-effect, or abrogated the Laws of Thermodynamics!
@ Jonathan Jones, 12:08 PM
It is true that temperature cannot arise from a pure Hurst-Kolmogorov process. My critique effectively says that in §10—see paragraph 4.
Nov 1, 2013 at 12:25 PM anonym
Anonym - You originally said "Is there anyone with acknowledged statistical expertise who has spoken out in favour of it? If not, why is it still getting prominent and implicitly favourable coverage on BH?" . To me that seemed pretty clearly to indicate that the absence of such endorsement seemed to you that it was not worthy of discussion here.
I don't see that my saying "you seem to be one of those who holds the notion that something is only valid if a consensus of experts has endorsed it" was a gross distortion/travesty of what you said. If you think it does misrepresent what you said, that was not my intention.
Doug Keenan's point is that you need to have a valid statistical model if you are going to do meaningful statistical analysis. Is that something you disagree with?
Statistical analysis inherently requires a model. In some situations, you can obtain a model by prolonged observation of of signals - undersea noise, for example, in applying statistical analysis to target detection in passive sonar systems. Then if something changes (eg a noise-generating target appears), statistical analysis can indicate it. In other cases, where the system is well understood physically, you can obtain a model by direct analysis - from the differential equations for the system for example.
In the case of the climate system, neither is possible. In the absence of validated statistical models for the climate system, it has always seemed absolutely obvious to me that you can't draw conclusions from statistical analysis.
When Professor Phil Jones and the Met Office talked about climatic changes being statistically significant it seemed to me such obvious bullshit I was astounded. The only possible explanation I could think of was that they simply did not have a clue what they were talking about and were just making it up, yet unaware they were talking nonesense.
Paul_K: Thanks for that story. Lucia has done remarkable things. I like Doug very much as well.
anonym: No harm done.
Douglas, I think it was a good idea to change some of your wording. But I think I'm on the side of those who think that the short-hand you use in writing makes it too easy for people to dismiss you as a crank. E.g. the very first sentence of your Executive Summary says:
This is really hard for some people to parse, because if global average surface temperature has risen (for the present purposes, lets leave out of the discussion issues such as the role of increasing UHI effects, accuracy of measurements, and the meaningfulness of a mean global temperature anomaly) in the last 100 or so years, then surely the "Temperatures on Earth's surface" provide a tautologous proof for global warming. It is warmer, on average, now than 100 years ago. Now, I know that that is not what you mean by "global warming" - you mean "significant global warming that could not - to some level of probability - arise due to natural variation alone". I think you may lose quite a few people who would be sympathetic to your argument by not making this more clear. Even adding the word 'significant' won't really help here, as it can be understood in many ways - e.g. if you assume that the error on determination of the global temperature anomaly at a particular point in time is 0.1 degrees, then a rise in mean global temperature anomaly by 0.6 degrees is 'significant' with respect to measurement error - though not in other ways.
For what it is worth, I think that your claim that it is impossible to prove anything about the contribution of CO2 emissions to the last 100 years of warming based on the temperature time series alone is a good one. I also think that it is not that unorthodox - the IPCC-consensus-upholders would, I guess, tend to agree, and say that the shape of the time series provides evidence only in the context of fairly strong prior belief in the fact that we have a reasonably good understanding of what the causal factors affecting temperature are.
It is important to point out that the anomaly time series is NOT physically generated, seeing that Lucia mentioned forebodes flashinglights
there is scant evidence that the cherrypicked 1000-odd surface thermometers indicate anything other than er 1000-odd surface thermometers.
To have a physically generated timeseries one should consider the averaging that is done with respect to the physics, and the location of the thermometers. For example: how can in a calorific system earth the land surface be a measure of overall calorific content variation?Never mind imbecillic surface proportionally averaging.
I think this is a very interesting summary on the obfuscations published by IPCC+MET, just a note of warning on playing around with random bernouilli walks: WFeller-I ch 3.6, drawing-you-know-which-one(fig4)..
having a near match of 1/4 says a lot about the random number generator in your SW library..You do not have an ideal coin, you have a coin with a magnet..it would be interesting to see the plot.
"I hope climate scepticism has not abandoned cause-and-effect, or abrogated the Laws of Thermodynamics!"
No, we leave that to climate science.
@ Jeremy Harvey, 1:39 PM
I agree that the first sentence is really problematic. 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—trying to simplify but still be accurate; 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.
"That was perhaps a mistake, though, because you, and others, have found valid problems with it"
I'd say it was a good step forward and identified some things that will now be taken care of and which otherwise might have been overlooked.
ptw:
Too right. Thanks for making this crucial point.
Doug: That is a very encouraging answer. We await with real interest.