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)
One of the things that amazes me in the number of people who believe in global warming based on the temperatures/weather being exactly the same for the near 40 years I can remember.
I liked this part: "Moreover, both consultants said that the analysis was so easy, there would be no charge."
Just a note to others here. Most of Mr.K's paper went over my head. I am not much of a mathematician...but everyone can learn from the matters set out from page 12 onwards. I suggest you all persevere! Disturbing stuff.
A superb paper. Thank you Doug Keenan. Written in language that even a LibDem MP could understand before completely rejecting. I will copy to a few MP's that I think will receive it well. And I gather questions in the House are still pending.
The paper is clear and convincing. But I fear it comes too late. It is interesting (indeed, it is astounding) that in all the recent talk about enery costs and green levies politicians have simply stopped talking about global warming. I have followed the debate closely and the term is virtually never used. "Climate change" sometimes gets a hurried mention in passing, but even this is rare. The discussion quickly moves to issues like energy security. The debate is now being framed exclusively in terms of "green" as a general (social) good. Obviously this is nonsense, but that is where we seem to have ended up. It is as if the existential threat of a warming globe underlying all this stuff is now something too embarrassing to mention. Keenan illustrates why.
Thank you Doug - What an interesting and informative read that was. He says dredging up his "A" level Statistics from half-a century ago. Very accessible for most I would think.
Who would have guessed that some people involved in the AGW team might select a "model" that shows significant warming???!!!! and ignore any that didn't give the right answer.
I would think a fertile area for research would be the "statistical" justifications for adjusting temperature records. How can thermometers always read too cold in recent times, but always read too high before WW2. NASA altering Icelandic temperature records in the GHCN from the 1930s which were embarrassingly warm for the AGW story is a case in point.
Paul Homewood covered this a year ago at WUWT.
http://wattsupwiththat.com/2012/10/15/ghcns-dodgy-adjustments-in-iceland/
There can be reasons to adjust temperature records, but how many times does the die have to show "6" before suspicion sets in??
http://wattsupwiththat.com/2012/08/19/july-was-also-the-329th-consecutive-month-of-positive-upwards-adjustment-to-the-u-s-temperature-record-by-noaancdc/
I guess The Met Office now realises that Doug Keenan is on their case - might just bring a little more honesty? No don't be silly David, as my Mother, always the parent even after I retired, would have said.
Keenan says there is no observational evidence for global warming,
Here is a quote from my guest post at WUWT see
http://wattsupwiththat.com/2013/10/29/commonsense-climate-science-and-forecasting-after-ar5/
"The key factor in making CO2 emission control policy is the climate sensitivity to CO2 . By AR5 – WG1 the IPCC is saying: (Section 9.7.3.3)
“The assessed literature suggests that the range of climate sensitivities and transient responses covered by CMIP3/5 cannot be narrowed significantly by constraining the models with observations of the mean climate and variability, consistent with the difficulty of constraining the cloud feedbacks from observations “
In plain English this means that they have no idea what the climate sensitivity is and that therefore that the politicians have no empirical scientific basis for their economically destructive climate and energy policies."
Taken these two statements together there seems to be little reason for anyone to pay any attention or even discuss anything the IPCC puts out. How can we get the politicians to realize this fact?
Be ready for: a lot of silence, or, a lot of bull***t with the word 'physics' mixed with it.
Thanks, Doug. Very interesting and well presented for the non-statistician. IMO though the first part discussing AR5 should stand on its own (as the title suggests) an the latter part discussing House of Lords, Phil Jones, Met Office, the Australian "experience" and radiocarbon dating etc. should be a separate commentary as it reads as significantly personalised attacks/ad hominems (I make no judgement as to whether the personal criticisms are warranted but they are directed at persons whom one would presume are "world authorities" in their respective fields).
One final point. While statistically there may be no way to conclude that significant GW exists, there does seem to be plausible reason to believe that global temperatures have risen in the last 250 years. I think the (first part of) the article could be improved by at least acknowledging this as it currently reads like you're claiming that GW is non-existent. I understand your position to be that GW cannot be shown to be occurring (or indeed dangerous) by use of linear trends on the output of current chosen models or the global temperature record. I'm sure the CAGW crowd will jump on your words to make you out to be a crank.
But then - what do I know? I work in the pharma industry and we torture logic to infer processes are robust!
Oct 30, 2013 at 7:57 PM | timheyes
You don't need to go back 250 years as you know that any changes that far back aren't anthropogenic CO2 related.
I have just read the paper which is brilliant. But there's a problem; how can it be made compulsory reading for decision makers in government? Can a copy be sent to Sir Mark Walport, so he cannot at the least plead ignorance at any future public inquiry?
I am also intrigued by Doug's example of the statistical error in both the 'industry standard' carbon dating software packages, and would like to know more, i.e. the typical percentage magnitude of this error and which way?
Surely it must be possible to present this report to the relevant parliamentary committee? There are now several supportive MPs such as Graham Stringer and Peter Lilley who would be able to get the committee to look at this subject.
I am thinking of writing a paper to support my oft-mentioned climate hypothesis, that nothing much is happening. It will be entitled 'Nothing much is happening' and the entirety of the text will be 'Well, nothing much is happening, is it?'
(When I mentioned my theory that the temp/time plot was little more than a random walk, Jonathan Jones smiled indulgently. He often does at my theories. I wonder why.)
People who have been following Doug Keenan for some time will not find anything fundamentally new in this document, but it is the clearest statement of his position he has so far produced: firstly because he gathers together several topics in a single coherent document, and secondly because have has given himself enough length to write a clear discussion. Well worth reading.
I have sent the link to the paper to Peter Lilley MP
If climate 'scientists' have shown on thing is that they will never be short of arrogance and unearned high opinion of their own work . And that is one of Doug's problems , and its the same issues with these people having little understanding of tree growth but decided that those that do are not the 'right type of experts ' so can be ignored .
Although ahead of his field by a long way , Mann is far from the only climate 'scientists' who cannot see the reality of their data becasue their to busy looking down their noses at others as well as keeping the gravy flowing .
When 'the cause ' falls climate 'science' will be lucky if it goes back to the uncared for unknown and poorly funded cousin of the physical science it was before the need for AGW 'made it ' and it will only have itself to blame .
ok another brilliant statistical tour de force. But so what? How does this relate to the underlying physical mechanisms of climate? It is great to be able to claim that the consensus position is wrong.....but it is so much better to be able to give an alternative point of view, if you want to appear on Women's Hour and get them thinking about your floppy hair.
Steve Mosher is unimpressed in the comments at WUWT. Can anyone shed light on why he thinks Newman's paper is fundamentally flawed?
Doug, outstanding essay. I would very much like to attend a statistics course given by yourself.
In the meantime, I shall be writing to my MP. Given the problems of energy prices, cold winters and the real possibility of people dying as a consequence of ill-founded government policies, perhaps he will listen.
Kudos to Koutsoyiannis too, I have read quite a few of his papers and have a lot of respect for his work too.
TimHeyes: "While statistically there may be no way to conclude that significant GW exists, there does seem to be plausible reason to believe that global temperatures have risen in the last 250 years. "
This cannot be demonstrated without understanding the physical processes. Fro example, my area of exertise is geostatistics. If I stochastically simulate a time series, with a temporal correlation based on a variogram function with a range in time of say 70+ years, you would not be able to tell the difference between a stationary and a non-stationary time series generated in this way, over this time period, without knowing the fundamentals of the random function model that generated it. I have made this point many times over the years. Think of a sine wave with a long period, but which you are unaware of. How can you tell the difference between a sine wave and a trend, if the sine wave you choose for the simulation has a period similar to or longer than the length of your observation window? And that window, for global temperature thermometer data is only about 130 years or so.
Oct 30, 2013 at 11:02 PM | Arthur Dent
Who's Steve Mosher?
Doug,
Thank you for your specific mention of Australia's Garnaut report (2008). The following URL, my personal submission to that report-in-preparation, shows some of the same criticisms that you make.
http://www.garnautreview.org.au/ca25734e0016a131/WebObj/D0840427GeneralSubmission-GeoffreySherrington/$File/D08%2040427%20General%20Submission%20-%20Geoffrey%20Sherrington.pdf
Example "The past global temperature record is flawed in many ways and should NOT be used
unreservedly to claim that global warming has a certain magnitude. The true error envelope of
100-year temperature estimates is hugely greater than the much-claimed 0.8 deg C rise in
that time".
Note: Errors in some fields like Analytical Chemistry are commonly treated as the combination of accuracy (bias) and precision (a statistical concept of scatter about a value). It is common, in climate work, to treat mainly the latter. For temperature work, the precision of reading a thermometer can be evaluated today. However, the accuracy of a given thermometer depends on testing it against a standard; since many thermometers are now lost, a retrospective accuracy test cannot be made and such accuracy error cannot be applied. Similar arguments hold for thermometer housings, thermometer locations, etc.
Oct 30, 2013 at 11:12 PM | Registered Commenter thinkingscientist re Geostatistics
Thank you for the comment, which I have to label as perceptive because I agree so much with it.
Geostatistics has much to contribute to the analysis of climate data. I do hope that you can stay around to pass on comments and information, because I've got too rusty on applied geostatistics via a long retirement. My colleagues were world leaders in the late 1970-end 1980 period, visiting Fontainebleu, bringing luminaries to Australia, applying the methods to ore resource calculations etc.
Even the simple semivariogram, with range calculated properly, can give an indication about whether one observation can help predict another taken at a different time.
Hi Geoff, thanks for your kind comments, I am not just passing through here...
Interested to read about your experiences with geostatistics in the 1970's and 80's. Bit before my time - I graduated in 1984 and did not get trained in geostatistics until 1992/1993, by some folks from Fontainebleau. So I have around 20 years in applied geostatistics. Its used widely in the oil industry for modelling and other 3D applications, which is what I specialise in and provide training courses in for others. I use software from Fontainebleau (and elsewhere) and have designed and sell quite a few specialist software packages incorporating geostatistical methods. I also rock climb a bit, so Fontainebleau is interesting for other reasons.
The decision of the stationairty assumption of the model is at the core of a variogram based approach, but with long range dependency in the data, as long as the observation window length, it is not possible to distinguish between long range depedency and non-stationary behaviour. I actually use a series of slides depicting climate temperature or proxies over different time scales, starting with the world temperature, going through the IPCC 1995 picture with a LIA and MWP, then Keigwin (3,000 year proxy) followed by Vostok with ice ages. It makes an excellent and lively class topic for discussion, attempting to assess by eye what stationarity assumption the class might reasonably make in each example!
The Met Office seems to have gone quiet recently.
Keenan should write "no demonstrated observational evidence for 'unusual' global warming." or 'unnatural' or 'manmade' warming. We skeptics know what he means but even independent thinkers would dismiss anyone who seems to be saying that global warming hasn't happened when there is overwhelming evidence that it has - and they immediately point to the land-sea record and call you a denier or conspiracy theorist. The way we present our case is very important.
a variogram function
Oct 30, 2013 at 11:12 PM | thinkingscientist
Scuse my ignorance. What's a variogram function - an empirical autocorrelation function maybe? Spatial or temporal?
Martin A,
I suspect the Met Office will ignore the paper until they have got their story straight with the rest of the team first.
Chur
Mailman
Notice that the numerical amount of claimed past warming has been creeping up in each successive IPCC report. At first it was reported as 0.6K in the last 100 years, then they increased it to 0.7K in the last 100 years by using a lower starting point 10 years later than the original. Now in AR5 it is a trend of 0.85K in the period 1880–2012, which is really just 0.6K per 100 years again, ie no actual change since they first predicted imminent thermageddon but they just didn't want to report it that way. Subtle.
Oct 31, 2013 at 8:34 AM | Unregistered CommenterJamesG Data creeps with age
The ramifications of adjustments done poorly cut deeper than just giving a wrong change over the last 100 years or whatever.
Remember that proxy calibration is mostly done using historic temperature data. If the last 100 years has twice the trend it should have, to a first approximation, proxy estimations will have half the trend they should have and a result would be a Mann-style hockey stick with features like the postulated Middle Warming Period depressed.
But Doug Keenan's larger thesis is that the uncertainty of measurement, previously improperly computed, is so large that one ought not make statements like "the 100 year trend is twice what it should be."
On that we are entirely in agreement.
Martin A What is a variogram function?
Here's a suggestion. Pull up a chair, give yourself 2 hours, and think deeply about this rather enjoyable topic. Your question is finally answered at 20 minutes in.
http://www.youtube.com/watch?v=SJLDlasDLEU
It's a patient presentation of a rather different way to think about earth science data including time series of temperatures. Eventually in real life mineral and oil work it gets used for creation of 2D or 3D maps of gridded or blocked data, so it's practical rather than theoretical.
But let's not get into a discussion where Doug's essay is the topic.
Martin A What is a variogram function?
Here's a suggestion. Pull up a chair, give yourself 2 hours, and think deeply about this rather enjoyable topic. Your question is finally answered at 20 minutes in.
http://www.youtube.com/watch?v=SJLDlasDLEU
It's a patient presentation of a rather different way to think about earth science data including time series of temperatures. Eventually in real life mineral and oil work it gets used for creation of 2D or 3D maps of gridded or blocked data, so it's practical rather than theoretical.
But let's not get into a discussion where Doug's essay is the topic.
One of the most important parts of the paper can be understood by even the most innumerate people; that is the section near the end where Keenan describes how people have reacted to statistical problems in other sciences. In most cases the practitioners didn't want to know. The glorious exception is medical research where, because it involves matters of life and death, great attention has been paid to the quality of statistical analysis (and I hope to the quality of the data as well!).
The findings of climate science also affect people's lives. If you are going to spend billions of pounds tackling a (possibly poorly identified) problem it could hardly be otherwise. I wonder if Sir Paul Nurse and the Royal Society would consider urging climate scientists to learn from the more professional (and honest) approach of medical researchers to statistical analysis.
An interesting essay. I have a question - have any blind tests been carried out with radiocabon dating? For example - ancient tombs that have actual dates carved on them. Perhaps such tests could indicate which methodology is correct.
Concerning the warming value of 0.85C, I am afraid that I am now so cynical that I think this is the answer because it had to be the answer - according to the "magic formula" 1.5*log2(400/270) = 0.85C. This in turn gives the climate sensitivity that is applied to Trenberth's fantasy physics of "radiative forcing". So now it seems they got the statistics wrong as well as the physics....
I have read the whole document, and understood most of it, but I do think that this is overstating the case to such an extent that it will be simply dismissed.
As Mosher (to the amusement of many on WUWT) has pointed out, this is in effect a denial of the LIA. Surely something as simple as the CET graph from the Met Office is observational evidence for global warming.
Yeah, I get that this does not demonstrate that there is evidence of warming that is outside the bounds of random or natural variation, and it is just a local record, and it says nothing about the causes or any anthropogenic contribution, but it is evidence for global warming.
The thing to remember is the head of the MET spends far more time fighting for funding and empire building than they ever do with any hands on science. Its an oddity that to get to that level you have to happy to do this and all the other boring 101 tasks that senior management require. Which is why those who are much happier as political operatives or administrators end up at this senior level.While those whose strength and interest is the actual science are unlikely to want these roles or last long should they get them .
So the MET heads 'science credentials' may simple not be that good for that is not their role, their task to make the end justify the means and its that 'end' which as lead the MET down its current path .
I have posted a link to the paper on John Redwood`s blog. My post is currently in moderation and is slightly off topic. But I hope the reference to the questions asked by Lord Donoughue (with the ultimate reply) and whether PPE graduates are expected to have some grasp of statistics should get his attention.
I thought that the paper was very clear (I am not a statistician).
Well done Douglas Keenan!
Your paper identifies the precise reason why the IPCC, Consensus "Science" and the literally hundreds of $billions wasted on climate "psience", have failed to improve on the figure of 1.5 °C-to-4.5 °C for a doubling of pre-industrial atmospheric carbon dioxide (Climate sensitivity), first quoted in 1979.
Were any of these climate models correct, one might reasonably expect (I'm sure there is a statistical test!) some refinement in this figure.
steveta_uk
The Central England Temperature record is observational evidence for Central England warming, nothing more.
Certainly you can try to extrapolate from this and where you can find other temperature records that appear to move in some sort of sync you can build up a pattern which may lead to the conclusion that overall it is likely that the globe has warmed by a certain amount over the period you are examining.
But as Phil Jones' own figures showed (replicated by BEST as I understand it) about one-third of recording stations were showing a declining trend which suggests that climatic variation may be regional rather than global, an argument that warmists themselves tried to use when attempting to downplay the MWP.
The result of all this is, as Geoff Sherrington points out above, that the error bars are so wide that the figure is meaningless.
I must admit I nearly gave up after the coin-tossing intro - but fortunately I persevered and found the paper really interesting and enlightening.
What surprised me were the obvious flaws in carbon dating - I suppose like a lot of people, I assumed that this was a case where the 'science is settled'. Far from it, it would seem - so doesn't this throw a lot of doubt onto various historic and prehistoric finds..?
Furthermore on the matter of medical research - you do have to wonder about the various bits if 'advice' that regularly find their way onto our tv screens and into our newspapers - eat this; avoid that - epecially when (for instance) we are now told that eating saturated fats is not all bad..! (oh, grill me that rib-eye steak this minute...)
Seriously, though - the sooner that Doug's article is read and understood by a wider audience the better - but I fear a serious outbreak of 'lalala I can't hear you' amongst the political classes and vested interests....
On the general matter of surface temperatures - I notice that tv weather reporters are at pains these days to say: 'That'll be the temperature in towns and cities; in rural areas it wil be two or three degrees cooler..'
So - the 'urban heat effect' really does exist - and as the house-building juggernaut blunders on, creating more and more urbanisation, surely surface temperatures are bound to be perceived to rise..?
Roger longstaff
If you are going to debunk 20th century radiative forcing you should at least use the correct formula.
deltaF = 5.35ln(400/270)
To convert this to a temperature change divide the forcing by 3.7.
The calculation gives 2.1 W and 0.57C directly due to the effect of increasing CO2.
The actual change is 0.8C. Recognising lag effects, the CO2 induced warming is increased by a minimum of 40% due to secondary forcing.
Martin A:
"Scuse my ignorance. What's a variogram function - an empirical autocorrelation function maybe?"
Effectively, yes, a measure of the power spectrum of the data. I should properly have used the term semivariogram (its based on an uncentred variance calculation...ironically!)
"Spatial or temporal?"
Defined in whatever coordinate domain you are using, so either.
Geoff Sherrington mentions 2D or 3D maps of gridded or blocked data, which is the application of kriging. It is the spatially dependent equivalent of the mean or average in classical statistics, can be stationary or non-stationary and like classical statsitics the kriging estimator returns both the estimate (or expected value) and the uncertainty on the estimate (as the error variance). For a variable following a Gaussian PDF and in the stationary case, the kriging estimator returns the mean and variance of a Gaussian PDF at each grid block or node.
Actually, in the oil industry estimation/kriging is not its most useful application, instead we use conditional simulation to generate non-unique realisations of the modelled random function process as we are most interested in computing non-linear dependent properties based on the uncertainty of the actual property, eg connectivity for fluid flow or volume (which is an integration of depth but non-linearly related to the depth uncertainty, which is what is being modelled).
Geoff thanks for posting that link to a video by Edward Isaaks. He is one of the authors of the most widely used undergraduate textbook "An Introduction to Applied Geostatistics" by Isaaks and Srivastava. He is a world authority. I use several examples from their book when teaching: elegantly and clearly explained, as in the video.
deltaF = 5.35ln(400/270)
Blimey, I haven't seen that one before. EM please could you give the provenance for that formula? The one I gave was circulating a year or so ago, and nobody could supply the provenance for that one.
BTW, your formula disagrees with Trenberth's calculations at:
http://www.aps.org/units/fps/newsletters/200904/trenberth.cfm
where he gives a figure of 0.9 w/m2 for radiative forcing, in the most incredible exposition of physics that I have ever seen!
Is the Global temperature record correct? Looking at the CET you would expect to see the large increase in global temperatures showing to some degree but it hasn't.
http://wattsupwiththat.com/2013/10/30/how-long-before-we-reach-the-catastrophic-2c-warming/#more-96543
Check this graph of temperature vs monitoring stations.
http://www.uoguelph.ca/~rmckitri/research/nvst.html
Perhaps this can explain the standstill. The cull takes out stations in the colder regions so the global temperature record shows a rise. After the cull the situation is relatively stable and the stations are reflecting the true global temperature which is not rising. This would explain why we had a massive step in the global temperature record and the current pause.
It would also explain why the increase in storms etc. that might be expected from a large rise in global temperatures never happened. The Earth just did not warm as per the records, the warming is artificial, a product of poor science.
Many thanks, and congratulations for rendering statistical theory comprehensible to the layman. In a sane world there would be Horizon programmes about this paper, and science correspondents would be fighting to interview Doug.
In the real world, he was invited to the 2010 Guardian debate (which he mentions towards the end of the paper) as the token sceptic. (Steve McIntyre was added at the last moment thanks to the generosity of some BH readers, who put up his fare). Then the Guardian cut his contributions out of the audio recording of the debate. Doug provided the missing extracts; and a full transcript is available at
https://sites.google.com/site/mytranscriptbox/home/20100714_gn
deltaF = 5.35ln(400/270)
Blimey, I haven't seen that one before. EM please could you give the provenance for that formula? The one I gave was circulating a year or so ago, and nobody could supply the provenance for that one.(...) | Roger Longstaff
That formula is a favourite of EM's. He quotes it as if it has the same sort of validity as formulas like
s = u t + ½ f t².
Some time back I asked where the "CO2 forcing is logarithmic" story came from, as simple formulas usually come from analysing a simple model of some sort and it seemed surprising that its origin did not seem to be universally known.
In the end I did track it down*. The logarithm comes merely as an approximation to a set of numerical values produced from a computer model (unverified, as normal in climate science) for computing 'radiative forcing' for different CO2 concentrations.
Quoting that formula as if it were an accurate representation of reality illustrates EM's difficulties in distinguishing between simple formulas and barely understood reality (or not at all understood reality in some cases - possibly this one).
* Myhre et al., New estimates of radiative forcing due to well mixed greenhouse gases
Re the log relationship of absorption of radiation to concentration - Look up Beer's Law.
It works I promise you (I used to make and design gas analysers using IR absorption as the measuring technique). Just how that applies to CO2 and global warming may be a different matter of course.
Mike A - er, yes. But here they are talking about "radiative forcing" not simply absorption. As someone on BH said a while back