Willis Eschenbach sent in the following information about the Coupled Model Intercomparison Project (CMIP) project. I’ve not checked the analysis myself, but it is an interesting topic and well worth a separate thread. There are some other pretty good posts like this. If people want to suggest some back posts for individual threads, it just takes a couple of minutes for me to transpose them and I’m happy to do so.
Willis writes:
Dana provided a very relevant link to the Coupled Model Intercomparison Project (CMIP) project. Here’s Figure 1 from the project, showing the “control run” for each computer. (Before they try to hindcast the past or forecast the future, they do a “control run”. This is described by the CMIP as a run with constant forcing, rather than a run where the CO2 is changing over time.) Here are the results from that control run:
SOURCE http://www-pcmdi.llnl.gov/projects/cmip/overview_ms/control_tseries.pdfNow, before we start talking about subtle changes in temperatures from subtle changes in atmospheric gases, you’d imagine that all of the GCMs would at least be able to give an accurate figure for the current global temperature … but noooooo …
Instead, we find that when presented with identical forcings, the various GCMs give results for the current global temperature that vary from 11.5° to 16.5°C.
Why does this matter? Well, we’ve been discussing the fact that the change in forcing is expected (according to DanàÆàs reference, which agrees with my own forecast) to be on the order of 1 watt per square meter over the next fifty years. So we’re asking the computers to be accurate to this size of forcing, accurate to within 1 w/m2.
The difference in the radiation temperatures from the highest to the lowest GCMs global control runs (from 16.5°C to 11.5°C), on the other hand, is 27 watts per square meter … clearly, some of them are very, very wrong. Heck, look at the top red line. It changes by 1°C, that’s 5.5 w/m2, when the forcings haven’t changed at all. Can this GCM tell us anything about a 1 watt/m2 change over 50 years? I don’t think so …
What is the IPCC response to this? You’ld think that with the Fourth Annual Report coming up, that by now they would have created some standards and would only consider models that met those standards … but nooooo … that might offend somebody, I guess. So they just average them all together, the good, the bad, and the ugly, and call it an “ensemble” to make it sound scientific. Their only standard seems to be that the computer program be a) really complex, and b) have lots of people working on it. Other than that, anything goes.
To me, that’s not science, that’s intellectual dishonesty.
Perhaps the biggest scam in all of this is that they tune their models to reproduce the past temperature trends, and then claim that they are ready for prime time because they can reproduce the past … folks, the fact that they can reproduce the past means absolutely nothing. They have tuned their models to reproduce the past, it would be embarrassing if they couldn’t, but it doesn’t prove a thing.
The real way to test GCMs against the past is to compare other metrics than temperature trends, to compare such things as the standard deviation, interquartile range, derivatives, skewness, kurtosis, outliers, and other statistical measures, for their hindcasts with the actual reality. When we do this, we find that although they may be able to roughly reproduce the temperature trends, they do it by predicting patterns of temperature that are widely different from reality “¢’¬? temperatures that wander all over the place, or temperatures that hardly change at all month to month, or temperatures that change monthly more than has ever been seen in the historical record. Yes, they reproduce the trends … but that is by no means enough.
In fact, even if they could reproduce all of the above, that may not be enough. As the performance of computer models in related fields such as the stock market has proved over and over again, enough so that it is a required disclaimer in US brokers advertisements, “Past performance is no guarantee of future results, and there can be no assurance that any investment vehicle will perform as well as the prior performance results illustrated herein.”
… and the same disclaimer should be required on all GCMs … but I digress …
I am always amazed by the unbelievable credulity of people who, knowing full well that computers cannot forecast next week’s weather, nonetheless believe that they can forecast next century’s climate. Folks, they can’t even get today’s global temperature right, they’re all over the map, temperatures from 11.5°C to 16.5°C … do you really think they can forecast next century’s climate?
Climate Audit 
72 Comments
This quote sums it up for me – ‘scam’, ‘dishonesty’. I refer people to Lee’s post #61 here . You can’t be taken as credible if you fill posts with such taunts. You can think modellers wrong, but ‘scam’ and ‘dishonest’? No, the gulf will just widen :(John A: Go away and play somewhere else. You’re out of your intellectual depth.
Transfer:
Great post Willis,
But there are a few things you left out, like flux adjustments for example. What do you do when you try to link (couple) two models together i.e. a model of the atmosphere and a model of the ocean and after only a short period of simulation they diverge at their interface and the output becomes unstable? Answer you ignore the laws of physics (those Inconvenient Truths like conservation of energy and momentum) and you fiddle the energy/momentum balance at the interface through “flux adjustments’. Now that’s a ephemism if every I heard one. Now before anyone jumps in and says the latest GCMs don’t do this anymore then read these here and here and Google for “GCM flux adjustment’ for others.
KevinUK
Comment by KevinUK “¢’¬? 26 August 2006 @ 8:56 am | Edit This
PH,
He posted
‘intellectual dishonesty’ and
‘Perhaps the biggest scam in all of this is that they tune their models to reproduce the past temperature trends, and then claim that they are ready for prime time because they can reproduce the past’
and althougth as someone who believes in the predictions of GCMs you might consider these taunts, to me (someone who has worked in the area of complex mathematical modelling in the past) these are perfectly reasonable statements. The bit in brackets by the way is not an attempt at an appeal to authority but rather to let you know that I am qualified to be able to make some of the statements I’ve made on GCMs in the past.
Just as my ‘friends’ in Exeter (close to where your farm is?) do, I also used to use supercomputers (Cray 2) to model complex 3D transient heat transfer and fluid flow processes in nuclear reactors. Now sadly I only have my ‘fag packet supercomputer’ with which I have recently estimated the future skill of GCMs based (to a level of 90% confidence) on the two data points provided to me by Steve Bloom concerning the increased skill of hurricane track prediction models. Steve is obviously struggling to repeat my calculations and so verify them, but to be fair to Steve doesn’t have access to my very sophisticated and complex ‘fag packet supercomputer’.
KevinUK
Steve
It appears that when you are transfering posts, you are loseing the links that where in the original post. I suspect this is because you are just cutting and pasting the text. If you want to maintain the anchor tags/URL in th eoriginal post then just view the underlying HTML using ‘View’ then ‘Source’ and then cut and paste from from that instead.
KevinUK
Interesting post Willis, although I am not quite so pessimistic about the General Circulation Models are you are. Back in the late 60s and early 70s, economists were building a macro (economic) model of the US: the MIT-FRB (Federal Reserve Board) model. This was quite involved although nothing like the GCMs, and to quote myself (when applying for a FRB job tweaking the model), “They don’t work.” (I didn’t get the job.) So with each year and each unsuccessful forecast, the models added and few new variables and a few new equations (and simulation routines, and lagging structures, and adjustment procedures, and etc.). This reaction was likened, by a cynical observer, to the “Vietnam Doctrine”: if it doesn’t work, do more of the same.
Now it appears to this very unsophisticated observer that the GCMs are “doing more of the same.” Maybe not adding new markets to the model, but the functional equivalent. Now for most of the 70s the simpler macro models out-forecasted the MIT-FRB model [analogous to the issue of fitting and forecasting global temperature with a simple time-series model], which had changed its name to the MIT-FRB-Wharton (or Penn) model and, I believe, goes under something completely different these days. But over time the big model began to work, at least well enough for the Fed and others to pay attention to its forecasts. So maybe the 5 C model difference is just a necessary step in building a more reliable model. Heck, we should be encouraged by the differences: it implies that not all the models think the same (or if you are more cynical, not all modelers make the same mistakes).
Like many I am not ready to rely on the GCM forecasts for policy decisions. And where I am maybe more cynical than you is regarding backcasting or hindcast (ex post forecasting). The GCMs have been tested too many times against the known data to trust any ex post forecast to be honestly blind to values of the ex post forecast period. Think of the models and the models as some grand neural net which has been trained on the recorded history of the climate. We can’t, as practical matter, go back and ask this neural net to remove all the effective learning form data from, say 1995-2005 — well, unless someone forecasts a string, say, six or seven 0.5 C all positive or negative changes. Neural nets are just too nonlinear for that. So the test must be with real forecasts, and unless they models want to wait 20-30 years, the test can’t be average, annual, global temperatures, but the regional temperatures and seasonal (or even finer) patterns.
It ain’t happened yet, but be careful about betting against in the next 5-10 years.
Steve and Willis,
A thought has justed occurred to me based on something which Willis has posted above.
“What is the IPCC response to this? You’ld think that with the Fourth Annual Report coming up, that by now they would have created some standards and would only consider models that met those standards … but nooooo … that might offend somebody, I guess. So they just average them all together, the good, the bad, and the ugly, and call it an “ensemble” to make it sound scientific.”
If you did reject the predictions of some of these models and didn’t go for an ‘ensemble’ then you wouldn’t be able to claim that you had a ‘consensus’ would you? And we all know that’s a very important word to the media and the politicos.
KevinUK
Re #1, Peter, thanks for your comment. You say that for me to:
* describe the actions of the IPCC as “intellectual dishonesty”, or to
* describe it as a “scam” for modelers to tune their models to the past, then claim that the fact that their models represent the past means that they can be relied on to forecast the future
is inappropriate.
Perhaps you are correct, Peter. However, I’m getting very tired of the circumlocution practiced by the climate science fraternity. I received comments on a paper sent to a reviewer the other day. In that paper I said of Mann and Jones,
The reviewer said:
Say what? It’s derogatory and accusatory to accuse Mann and Jones of actually knowing that temperature series are usually neither normal nor stationary? I can see that it might be wishful thinking to accuse them of knowing that, but “derogatory and accusatory”?
This kind of coddling of other scientists in the field is one of the reasons that Michael Mann still has a job “¢’¬? a lot of other climate scientists seem to want to make nice, rather than to come out and say that the man with the “CENSORED” directory is being dishonest.
I don’t believe in that kind of circumlocution. Me, I’m a kid from a cattle ranch, where folks used to say “Son, you can piss on my boots … but you can’t convince me it’s raining.”
At this point in history, for the IPCC not to test the models, to simply take all of the models and call them an “ensemble”, is intellectually dishonest. I could understand them doing it at the time of the First Assessment Report. But we’re 15 years beyond that. Many people have called for such testing. For the IPCC to pretend that using an “ensemble” of models is somehow better than TESTING THE DAMN THINGS is intellectual dishonesty.
For the modelers to tune their models to the past, and then try to sell their model’s value on that basis, is a scam. Perhaps you don’t understand the gravity of what the modelers are doing. If a stock broker or a mine promoter did the same thing, they could be put in jail for fraud “¢’¬? you can’t make those kind of unsubstantiated claims, it’s taking advantage of people’s credulity. That’s why it’s a scam, and that’s why brokers have the disclaimer in their advertisements.
Here’s the deal. The modelers set up a model with what they know about natural forcings. It doesn’t resemble reality.
So they add in human-generated CO2. But it still doesn’t resemble reality.
So they add in “positive feedback”, and tune that feedback to make the CO2 have five times the effect it would otherwise have. Oops, still doesn’t resemble reality.
So they add in aerosols, and tune their effect to make the model output finally have some vague semblance of reality … despite the fact that we don’t even know what the effect of the aerosols might be.
(As an aside, the IPCC says “While the radiative forcing due to greenhouse gases may be determined to a reasonably high degree of accuracy (Section 6.3), the uncertainties relating to aerosol radiative forcings remain large, and rely to a large extent on the estimates from global modelling studies that are difficult to verify at the present time. The range of estimates presented in this section represents mainly the structural uncertainties (i.e., differences in the model structures and assumptions) rather than the parametric uncertainties (i.e., the uncertainties in the key parameters)”. In other words, the models are tuned by changing the aerosol’s effects … and then these tuned effects are used to estimate the aerosol effects. This, of course, is a separate scam, that of assuming what you are trying to prove … but I digress …)
Then they say “Our model proves the AGW theory is correct, because we have to add in anthropogenic factors to make it resemble reality” …
Peter, I don’t know any nicer way to put it.
That’s a scam.
Now I could pretend that the modelers are simply making an honest error. But they’re not. This is 2006, they’ve seen all of these issues discussed. I’ve asked Gavin Schmidt a number of times how his team assesses the skill of their forecasts. He refuses to answer. You might come up with all kinds of innocent explanations for that. Gavin says he won’t answer because I’ve been mean to him … nope, it’s the other way around. I’ve been mean to him because he refuses to answer. And he refuses to answer because his model, like all the rest, does very poorly on the tests.
Here’s an analysis of the models used in the paper “Amplification of Surface Temperature Trends and Variability in the Tropical Atmosphere”, B. D. Santer et al. (including Gavin Schmidt), Science, 2005 This paper was notable for its claim that since three sets of different observations differed from the models … that the most likely explanation was that observations were wrong …
But when we look at the models involved, here’s the temperatures they forecast:
LEGEND: The colored boxes show the range from the first (lower) quartile to the third (upper) quartile. NOAA and HadCRUT (red and orange) are observational data, the rest are model hindcasts. Notches show 95% confidence interval for the median. “Whiskers” (dotted lines going up and down from colored boxes) show the range of data out to the size of the Inter Quartile Range (IQR, shown by box height). Circles show “outliers”, points which are further from the quartile than the size of the IQR (length of the whiskers). Gray rectangles at top and bottom of colored boxes show 95% confidence intervals for quartiles. Hatched horizontal strips show 95% confidence intervals for quartiles and median of HadCRUT observational data.
This is called a “notched” boxplot. The heavy dark horizontal line shows the median of each dataset. The notches on each side of each median show a 95% confidence interval for the median. If the notches of two datasets do not overlap vertically, we can say with 95% confidence that the two medians are significantly different. The same is true of the gray rectangles at the top and bottom of each colored box. These are 95% confidence intervals on the quartiles. If these do not overlap, once again we have 95% confidence that the quartile is significantly different. The three confidence ranges of the HadCRUT data are shown as hatched bands behind the boxplots, so we can compare models to the 95% confidence level of the data.
Now, without even considering the numbers and confidence levels, which of these model hindcasts look “lifelike” and which don’t? It’s like one of those tests we used to hate to take in high school, “which of the boxplots on the right belong to the group on the left?”
I’d say the UKMO model is really the only “lifelike” one. The real world data (NOAA and HadCRUT) has a peculiar and distinctive shape. The colored boxes of the data are short, with numerous widely spread outliers at the top, and a few outliers bunched up close to the bottom. UKMO reproduces this well. M_medres is a distant second, and none of the others are even close.
Let me digress for a moment here and remind you of the underlying assumptions of the climate modellers. In a widely-quoted paper explaining why climate models work (Thorpe, Alan J. “Climate Change Prediction — A challenging scientific problem”, Institute for Physics, 76 Portland Place London W1B 1NT), the author states (emphasis mine):
Well, that all sounds good, and if it worked, it would be good. But the huge differences in the model hindcasts clearly demonstrate that in all except perhaps one of these models:
o The average and statistics are not described correctly
o The results are not constrained by anything, including constrained by the overall radiative budget.
o The various climate forcings and factors are not accurately represented in the models
The modelers know this full well, and yet they still claim that their models have value.
That’s a scam.
w.
It is my impression that use of ensemble averages of several computer calculations that are based on deterministic models and equations is unique to the weather (NWSs) and climate-change (AOLGCMs) communities in all of science and engineering. I can be easily corrected on this point if anyone can provide a reference that shows that the procedure is used in any other applications. (The use of monte carlo methods to solve the model equations is not the same thing). The use of ensemble averaging and the resulting graphs of the results makes it very difficult to gain an understanding of the calculated results; rough long-term trends are about all that can be discerned from the plots.
Consider the following as an example. Let’s say the results presented above represented calculations of the temperature in an engineering system and the temperature is used as a signal for a safety device; the release from or addition to the system of a gas, for example. Let’s further say the incorrect response of the system had potentially hazardous effects on the health and safety of the public. This is not a completely hypothetical example, by the way.
Under these conditions the above results are clearly unacceptable. For this kind of application the differences between the various calculations would be resolved, exactly and unambiguously, down to several places of decimals. If refinements of the calculations should not result in this degree of agreement, an experiment would be conducted in order to determine which, if any, of them was correct.
It is my understanding that the response shown in the figure is assigned to ‘climate is chaotic’. However, I have not been successful in finding documentation in which it has been shown analytically that the system of continuous equations used in any AOLGCM model has the chaotic properties that seem to be invoked by association and not by complete analysis. Strange-looking output from computer codes does not prove that the system of continuous equations possess chaotic characteristics. Output from computer codes might very likely be results of modeling problems, mistakes, solution errors, and/or numerical instabilities.
Invoking/appealing-to an analogy to the Lorenz continuous equations is not appropriate for any other model systems. The Lorenz model equations are a severely truncated approximation of an already overly simplified model. The wide range of physical time constants and potential phase errors in the numerical solutions almost guarantees that aperiodic behavior will be calculated.
Corrrections to the technical issues mentioned here will be appreciated.
The theory behind the ensembles is that the parameterizations are likely to be randomly wrong and that there are no systemic mis-paramterizations. One of the really interesting aspects of a recent realclimate thread on Soden and Held was the seeming astonishment of the GCM modelers when one of the amateur blog posters pointed out that the cloud feedback in ALL of the GCMs was strongly positive, whereas the proponents seemed to expect that some would be positive and some would be negative.
If I were parsing the GCMs for a systemic problem, the first place that I would look is in the atmospheric absorption of solar near infra red by water vapor and clouds – with some problems in the liquid phase of clouds, where there seem to difficulties in the droplet size distributions. It looks fraught with problems on the specialist side. I would like to spend some time seeing exactly how they determine how much NIR is absorbed by aerosols and how much by clouds and how sold these allocations are. IF some portion of the NIR absorption is a negative feedback (due to clouds) as opposed aerosol forcing which is specified, one could picture how the degree of feedback could end up being quite different with relative modest changes in weakly known parameterizations.
The surprise of the modelers at the uniformity of the postive cloud feedback in the GCMs was startling anyway. You’d think that they’d have already known this.
Additional discussions of the models are available in the threads here, here, and here.
The Global Average Temperature is a solution meta-functional; it maps every calculation of every model in a given code to a single number. Solution meta-functionals have proven potential and capability to hide an enormous number of actual problems in the calculation. It is an extremely weak metric for objective evaluation of the correctness of a calcuation. It’s interesting that randomly wrong mis-paramterizations do not get washed out even for such a weak solution metric.
“The Seven Deadly Sins of Verification:
(1) Assume the code is correct.
(2) Qualitative comparison.
(3) Use of problem-specific settings.
(4) Code-to-code comparisons only.
(5) Computing on one mesh only.
(6) Show only results that make the code “look good.”
(7) Do not differentiate between accuracy and robustness.”
I’m with Dan Hughes, getting the global average right is not very meaningful. You could have alligators at the poles and polar bears at the equator but get the right average. A season by season comparison between a single model and the measured temperature distribution over the earth’s surface would be a more interesting and revealing plot.
Re 13, even the modelers agree that the models do very poorly at predicting things on a regional basis …
w.
Re: 6
KevinUk,
I am still chuckling thinking about a consensus of computer models. I am certain that there are people who find no humor in this, but I certainly do.
So they build a model, the model runs and it is out by +/- 3.5 degrees. Then they plug a bunch of assumptions into it and add a bunch of fudge factors to get the average up to earth’s average 15 degree temperature. Then they tweak the model so that it shows an increase of 0.6 degrees over the past century and, viola, we have a fully functioning climate model.
Then the new fudginator models are run again and they show an increase of 1.5 degrees to 9.1 degrees over the next century and we are supposed to shut down the coal industry and get rid of our cars, trucks and trains. Unbelievable.
here is another report that compares climate models (thanks to dano):
An Appraisal of Coupled Climate Model Simulations
K. AchutaRao et al.
Laurence livermore Laboratories
August 16, 2004
Click to access model_appraisal.pdf
see also:
http://home.casema.nl/errenwijlens/co2/tcscrichton.htm
HANSEN AND MODEL RELIABILITY
James Hansen of NASA has a strong defense of model reliability at http://www.giss.nasa.gov/edu/gwdebate/
In this paper, he argues that the model predictions which have been made were in fact skillful (although he doesn’t use that word.) In support of this, he shows the following figure:
(Original caption)Fig. 1: Climate model calculations reported in Hansen et al. (1988).
The three scenarios, A, B, and C, are described in the text as follows:
I became curious about how that prediction had held up in the years since his defense of modeling was written (January 1999). So I started looking more closely at the figure.
The first thing that I noted is that the four curves (Scenarios A, B, C, and Observations) don’t start from the same point. All three scenarios start from the same point, but the observations start well above that point … hmmm.
In any case, I overlaid his figure with the very latest, hot off the presses, HadCRUT3 data from Phil Jones at the CRU … and in this case, I started the HadCRUT3 curve at the same point where the scenarios started. Here’s the result:
Fig. 2: Climate model calculations reported in Hansen et al. (1988), along with HadCRUT3 data.
A few things are worthy of note here. One is that starting the scenarios off at the same point gives a very different result from Hansen’s.
The second is the size of the divergence. Scenario C, where greenhouse gases stop increasing in 2000, can be ignored “¢’¬? obviously, that didn’t happen. Looking at the other scenarios, the observed temperature in 2005 is a quarter of a degree C below Scenario B, and 0.6°C below Scenario A.
Finally, the observations have mostly been below both all of the scenarios since the start of the record in 1958. Since according to Hansen Scenarios A and C were “meant to bracket plausible rates of change”, I would say that they have not done so.
A final note: I am absolutely not accusing James Hansen of either a scam or intellectual dishonesty, he clearly believes in what he is saying. However, he has shaded his original conclusions by starting the observational record well above where the three scenarios all start.
Mainly, the problem is that the world has not continued to heat up as was expected post 1998, while his Scenarios A and B did continue to warm. The post-1998 climate actually is acting very much like his Scenario C … except, of course, that the CO2 emissions didn’t level off in 2000 as in Scenario C.
w.
One of the really interesting aspects of a recent realclimate thread on Soden and Held was the seeming astonishment of the GCM modelers when one of the amateur blog posters pointed out that the cloud feedback in ALL of the GCMs was strongly positive, whereas the proponents seemed to expect that some would be positive and some would be negative.
for referance:
http://www.realclimate.org/index.php/archives/2006/08/climate-feedbacks/
It’s worth noting on this thread that the requirements of the CMIP were a major factor in the IPCC’s decision to persist with its flawed emissions scenarios for the forthcoming Assessment Report. Soon after David Henderson and I drew attention to serious deficiencies in the scenarios, Dr. John Mitchell of the UK’s Hadley Centre wrote to Dr. Pachauri, Chair of the IPCC, on behalf of the Working Group on Climate Models, as follows:
“The World Meteorological Organisation JSC/CLIVAR Working Group on Coupled Models (WGCM), which includes representatives from almost all the major climate modelling centres contributing to the Third Assessment Report, recently considered what work needed to be done to ensure that the best possible modelling advice will be available for the next IPCC Assessment, due to report in 2007. For each emission scenario, it is necessary to run an ensemble of simulations to define the uncertainty due to natural variability, and to do this with as many different models as possible to define the range of uncertainty in modelling the earth system. These uncertainties mean that there is little scientific justification in running new scenarios since the resulting climate change outcome is unlikely to be indistinguishable (sic) from existing scenarios with similar radiative forcing. Hence the WGCM unanimously urge IPCC to retain the current SRES scenarios without change …
“We appreciate that small changes in the emission scenarios may require large economic and social changes, and that the effect of the social and economic changes could be assessed in time for the next report. However, unless the accompanying changes in radiative forcing are likely to produce detectable changes in climate, we believe that it is better not to try and run new model experiments, but to stick to the scenarios used in the TAR. This will allow a better definition of the range of uncertainty in projected changes due to model uncertainty and natural variability, which are likely to dwarf any difference due to tweaking the existing emissions scenarios. This we believe will provide the best scientific basis for the next IPCC assessment” (J F B Mitchell to Rajendra Pachauri, 30 October 2002: the letter was tabled at the IPCC Bureau Session in Geneva on 10-11 December 2002).
In my presentation to the IPCC Expert Meeting on Emissions Scenarios in Amsterdam on 10 January 2003, I contested the claim made by Dr. Tom Wigley of NCAR earlier in the meeting that the IPCC’s scenarios captured the total range of possible variation “in terms of the 2100 forcing pattern”. Wigley had argued strongly against “tweaking” the SRES scenarios.
On 28 June 2003, without reference to me and without responding to my arguments, Wigley wrote to Senator John McCain, then Chairman of the Senate Committee on Commerce, Science and Transportation as follows (these are the first and last sentences of Dr. Wigley’s letter): “I am writing to you as one of the world’s leading climate scientists to alert you to misrepresentations of my work … I note also that work I have been involved in has shown that the use of Purchasing Power Parity rather than Market Exchange Rates does not affect the future emissions in any important ways, contrary to the claims of Castles.”
Dr. Wigley’s letter was included in material that Senator McCain circulated to his colleagues on the Committee on 28 July 2003, under cover of a letter asking them to “consider this information as the Senate works to implement meaningful climate change legislation.” I didn’t know of Wigley’s letter or Senator McCain’s use of it until after the IPCC’s formal press release of 8 December 2003, now posted on the Panel’s website, which says that “In recent months some disinformation has been spread questioning the scenarios used by the IPCC”, and refers to David Hederson and me as “so called “two independent commentators”‘.
In a submission of 8 April 2006 to the UK’s Stern Review on The Economics of Climate Change, Erwin Diewert, Professor of Economics at the University of British Columbia, said that “Castles and Henderson are right to criticize the first part of the SRES modeling strategy, which relies on market exchange rates to calculate per capita real income differences between countries… The difference between PPPs and market exchange rates can be very large so their criticism is not a negligible one.”
Professor Diewert’s expertise in the conceptual and practical issues involved in economic measurement is legendary. Last year he received the Julius Shiskin Award for innovation in economic statistics research and application, the citation stating that he had “developed, and adapted for implementation, … theoretically-improved measurement methods in direct support of statistical agencies in the United States, Canada, Australia, the United Kingdom, France, Germany, Sweden and New Zealand, and of the United Nations, the World Bank, and the European Central Bank, among others.”
In general on this subject, 1) *everyone* know the models need a lot more work, and 2) the models are hardly the sole basis for the conclusion that we have a problem with anthropogenic forcings, in particular GHGs.
Re #8: I think you’re reading way too much into the off-hand comment of one scientist.
Re #10: Willis, I started in on trying to understand what you were talking about here, and then it occurred to me that I’ve caught you twice engaging in elaborate analyses (Arctic sea ice in Warwick Hughes’ Coolwire 13 and just now hurricanes on this blog) that turned out to be entirely baseless for reasons that should have been (and I think were) obvious to you at the start. So no, I’m not going to spend any effort on this sort of thing from you. Now go ahead and accuse me of more "foul nastiness" if you want.
(P.S. — Nice graphing software! What is it?)
Re #16: This makes me wonder how much insurance would be sold if much of it wasn’t mandatory.
#14
Then what’s the value of the models? To paraphrase the ecology movement, we live locally not globally. Unlike an orange where “global” temperature means something, it’s just a mathematical construct in the case of the earth.
#21
If not the models, then just what is it that tells us we have a problem with GHGs? In normal science you need a theory to connect A to B, i.e., the models. Without a physical model all you have is an observation that A and B are correlated. That can mean: A causes B; B causes A; an unknown C causes A and B. Without a valid model tested against data you can’t decide which is true.
Re (my former) #21: Censored again, it would seem. I suppose it’s a practical necessity for Steve M. to allow John A. to have a certain amount of fun in exchange for his labor on the site, but if so he should at least admit that’s what’s going on.
For more info on the cloud feedback issue, see here.
KevinUK, thanks for the excellent post. I’ve transferred it over here from “National Post Today.” You said:
Flux adjustment is indeed still more common than not. “An Appraisal of Coupled Climate Model Simulations” looks at fourteen models that will be used in the FAR. Of these, 7 of them use “flux adjustment”, three don’t, and four are atmosphere only models that are combined with sea temperature observational data.
The ones that did not use flux adjustment did much worse at reproducing sea surface temperatures (r^2 = 0.82) than those that used flux adjustments (r^2 = 0.96) … amazing how accurate you can be when you can just tweak the adjustments ’til they fit …
w.
Re #20, Dr. Castles, thank you for your contribution.
I must admit, the irony using the CMIP (which will not disqualify even the worst model from being used in the Fourth Assessment Report) to justify the continued use of their flawed MER metric instead of the proper PPP metric could serve as a poster child for the entire IPCC process …
w.
Willis what is the climate sensitivity for CO2 doubling of the UKMO model?
Re #20: Not all economists agree as to the significance, apparently.
Model Information of Potential Use to the IPCC Lead Authors and the AR4.
UKMO-HadCM3
28 July 2006
0.83 K/Wm-2
That’s 3 K/2xCO2
There’s that name Wigley again.
RE # 27 by Steve Bloom “Not all economists agree as to the significance.” I think this is a typical case of cherry picking.
Re #21 and #27, Steve Bloom, you don’t seem to get it. Unsubstantiated allegations mean nothing on this site. You say:
You have already been asked once by someone else to back up your claim that models are not the sole basis to support the conclusion that we have a problem with GHGs, and you have provided nothing … I repeat that request.
Steve, you’ve never “caught” me in your life. While I have certainly made errors in my life and my work, I’ve done nothing to “catch”. I have asked you before what the problem in Coolwire 13 was, and received no reply (citation available upon request). Nor have you pointed out anything that makes my hurricane analysis “entirely baseless”, anyone can re-read the relevant pages and see that.
However, I must confess that I am overjoyed that you are “not going to spend any further effort” discussing my contributions, best news I’ve had in a while … but we’ll see if it’s true, or just another of your claims for which you provide no evidence …
There are three graphs on this page, not sure which one you’re referring to. The one at the top of the page is from the original document. The boxplot of temperatures is done in “R”. The overlay of the Hansen graph was done using “Vectorworks”, which is the Macintosh equivalent of Autocad. Vectorworks has its own internal programming language (a subset of Pascal with graphics primitives). I wrote a special program for the purpose that reads a text file of numbers (temperatures in this case), puts a dot at each data point, and connects them with a polygon … I have the program do this on top of the imported bitmap. Then I added the “HadCRUT3” legend manually.
Steve, truly, I do wish you’d put up or shut up regarding evidence for your claims. My sense is that you’re an intelligent person. But from what I read here and on other blogs, I’m not the only one who has noticed that while you are very quick to make unpleasant accusations, you are extremely slow to back them up. And again, from what I read here and on other blogs, this has resulted in your contributions being largely ignored.
Unless having your vote cancelled in that fashion is what you intend, you should seriously consider finding some facts to support your claims.
Finally, regarding your claim in #27 that not all economists agree, you forgot to mention what it was they don’t agree on. It can’t be that PPP is preferable, since John Quiggin says on the page that you cited:
(It’s interesting that you should cite that page, Steve, since I was a participant in the discussion and recall it well. Here’s how it went from there …)
OK, John Quiggin disagreed about how much difference it will make, but even he says PPP is preferable. I responded to John’s posting by saying:
Prof. Quiggin stepped right up to the plate and replied that … well, no, that’s just in my dreams. Actually, Prof. Quiggin took a page from Steve Bloom’s book and didn’t reply to a single one of my questions … bad professor … no cookies …
So I’m afraid that as usual, Steve, on further investigation your citation means nothing. Ian Castles has quoted an extremely reputable economist, winner of several awards, to back up his position. You have quoted an economics professor who, like you, doesn’t even have the … … well, since this is a family blog, let me call them “principles” … to back up his own position.
The questions I posed in my post to Prof. Quiggins are as germane today as when I wrote them (January ’06). Professor Quiggin, to the best of my knowledge, has still not answered them. You have provided nothing to disagree with either Ian Castles or myself.
It sounds to me, Steve, like either (a) you are opposed to using PPP, or (b) you just want to tear down Ian Castles. Which one is it, or is there a choice (c) that I’m missing here?
w.
Hans, thanks for the info in the link you give in #28. I loved the part where the form asks about the:
Learn something every day … to make their model work, they have to create an artificial island at the North Pole … really, you couldn’t make this stuff up.
It seems to me that if your model leads to a “singularity”, this means there must be something fundamentally wrong with the model. After all, these models are supposed to be based on physical principles, and if the reality doesn’t require an island to prevent a “singularity”, then why should the model? The conclusion for me would be to fix the model, not to create an “artificial island” …
But what do I know? …
w.
I am sure that Ian Castles will reply for himself, but if Steve bloom reads the piece by Diewert he will see that it is precisely a discussion of Quiggin’s views on the SRES scenarios. Moreover it gives Quiggin quite a lot of credit for suggesting a better way of modelling convergence, which is to model energy convergence directly instead of via GDP estimates. Diewert is undoubetdly one of the world’s foremost experts on index number issues and he shows quite clearly how using MERs can lead to very substantail divergences from using the undoubtedly preferable PPPs. Nordhaus has done the same. I don’t understand why the IPPC just don’t get on and do it properly and then no one can make a fuss.
but willis,
is this the same UKMO model that has the best statistical behaviour. If so what is the cause of the extreme climate sensitivity. How can the strong positive feedbacks of UKMO agree with the strong negative feedbacks you and doug find in the emperical cases?
Could we not call the north pole an island of ice?
About cloud feedback:
There are a few recent studies about cloud behaviour which challenge the positive cloud feedback included in (near all) current climate models.
Chen and Wielicki (2002) observed satellite based cloud changes in the period 1985-2000, where an increasing SST (+0.085 C/decade) was accompanied with higher insolation (2-3 W/m2), but also higher escape of heat to space (~5 W/m2), with as net result 2-3 W/m2 TOA loss to space for the 30N-30S band. This was caused by faster Walker/Hadley cell circulation, drying up of the upper troposphere and less cirrus clouds.
In 2005, these findings were expanded by J. Norris with surface based cloud observations in time (from 1952 on for clouds over the oceans, from 1971 on over land) and latitudes. There is a negative trend for upper-level clouds over these periods of 1.3-1.5%. As upper-level clouds have a warming effect, this seems to be an important negative feedback.
J. Norris has a paper in preparation about cloud cover trends and global climate change.
On page 58, there is a calculation of cloud feedback, assuming that the observed change in cloud cover is solely a response to increased forcing. The net response is -0.8, which is a very strong negative feedback… Of course this is the response, if nothing else is influencing cloud properties/cover, but important enough for further investigation.
Even internal oscillations, like an El Nino (1998) leads to several extra W/m2 more net loss of energy to space, due to higher sea surface temperatures. Thus IMHO, if models include a (zero, small or large) positive feedback by clouds, they are not reflecting reality.
#31 — “or is there a choice (c) that I’m missing here?”
c. both
Steve Bloom, thank you for citing John Quiggin’s piece on “Castles and Henderson, again”. Willis, thank you for your pertinent comments in #31, and mikep, thanks for your’s in #33. The failure of the IPCC to adopt PPP is indeed doing considerable collateral damage to the work of the World Bank, the UNEP and the Millennium Ecosystem Assessment, all of which are under the influence of the IPCC milieu. It is also highly damaging to the UNFCCC reporting system (e.g. Germany & France report their emissiions intensities in MERs, and the European Commission report in PPPs. Developing countries contribute to the expert reviews, and thus get lessons on how to cheat with statistics).
Now for some background. Before he’d made the “Castles and Henderson, again” posting, Professor Quiggin had (unbeknown to me) made a submission to the UK Stern Review on “the Castles-Henderson critique of the IPCC.” When I read this submission and noted obvious errors, I wrote to Erwin Diewert seeking his comments. Following discussion with both of us, Diewert sent Quiggin a comment on his paper, which JQ posted on his website on 31 March.
I only quoted the first of Professor Diewert’s conclusions before, so let me now quote all three conclusions:
“Castles and Henderson and right to criticize the first part of the SRES modelling strategy, which relies on market exchange rates to calculate per capita real income between countries. It would be much better to use ICP PPPs for this first part of the SRES modelling strategy. The difference between PPP’s and market exchange rates can be very large so that their criticism is not a negligle one.
“Quiggin is right to implicitly criticise the entire SRES modelling strategy. It would be simpler to abandon the two stage modeling strategy and make direct comparisons of energy intensities across countries and assume energy eonvergence rather than real income convergence.
“Either way, the SRES model should be reestimated.”
So all of us — Diewert, Quiggin, Castles & Henderson — agree that the IPCC got it wrong, and all of us agree that the correct procedure is to use PPPs.
As a result of comments made by John Quiggin when posting Erwin Diewert’s paper, I wrote again to Erwin to seek clarification. On 1 April he wrote confirming what I thought was his view – “whether the assumption of gradual convergence in per capita energy use is any more sensible than the assumption of gradual per capita real income convergence is a different matter. I (Diewert) am extremely sceptical on the prospects of either type of convergence occurring.”
At my suggestion, Erwin made several revisions to his paper. He added references to two more Castles & Henderson papers, and also cited a paper by Stegman and McKinnon, 2005 (“Convergence and Per Capita Emissions”, Brookings Discussion Papers in International Economics No. 167) in a footnote which said “The convergence hypothesis [which underlies most of the IPCC scenarios] is somewhat questionable to say the least … However, discussing the merits of various convergence hypotheses is not the subject of this note.” The abstract of S&M (2005) says that:
“The focus of the paper is on per capita carbon emissions from fossil fuel use because this is the basis of many projections as well as a variety of policy proposals. We also present evidence on GDP per capita, energy intensity of output and the emissions intensity of energy supply. We find strong evidence that the wide variety of assumptions about “convergence’ commonly used in emissions projections are not based on empirically observed phenomena.”
So apart from using the wrong conversion method, the IPCC modellers adopted assumptions that were “questionable to say the least.”
On 8 April Erwin Diewert wrote to John Quiggin, “cc” to me, saying “Ian Castles sent me several very useful comments on my note and so I have revised it a bit in the light of his comments. I would appreciate it if you could replace the previous version of my comment on your website with the attached version.” Quiggin has not replaced the earlier (27 March) version of his paper on his website, nor has he added the reference to the C&H papers to his own comment as he promised to do, nor has he advised the Stern review that his comment on C&H was only in draft form (as he stated on his blog).
The revised (8 April) version of Erwin Diewert’s comment is posted on Professor Diewert’s own website and also on the Stern Review website.
On the more general claim by Steve Bloom that “not all economists agree, apparently”, it should be noted that the proportion of PPP sceptics among economists is minuscule compared with the proportion of AGW sceptics among scientists. In 1999 the World Bank reported to the UN Statistical Commission that “there is unanimous agreement among researchers and theoreticians [that] proper cross-country comparisons can only be made once values have been adjusted to eliminate differences in price levels using purchasing power parities” (Document E/CN.3/1999/8/Add.1).
In a paper recently published in Climatic Change (“How Should Emissions Projections be Evaluated?”, March 2006, vol. 75: 1-7), economist Alison Stegman of Australia’s Macquarie University, a senior research analyst with the Brookings Institution, says that “The Castles and Henderson critique of the use of exchange rates to convert international output is a relatively straightforward statistical issue which has widespread acceptance in statistical agencies” and quotes the explicit provision “in the United Nations System of National Accounts.”
In fact, it is not just the United Nations System: the SNA is published jointly by the UN, the World Bank, the IMF, the OECD and the Commission of the European Communities, and the heads of all of these organisations jointly signed the Foreword recommending it for use by national and international statistical agencies. I was the Australian member of the Statistical Commission at the meeting at which the SNA was welcomed and unanimously approved.
In announcing last year that it was adopting the practice of using PPP rates for converting national GDPs into a common currency, the US Energy Information Administration cited “the internationally agreed System of National Accounts 1993, to which the United States is a signatory” (International Energy Outlook 2005, p. 13). This raises the question “Why does the UN’s IPCC, not follow the UN’s (and others’) internationally-agreed System of National Accounts. Professor John Reilly of the MIT Joint Program on the Science and Policy of Global Change gave the probable answer to Terence Corcoran of the Toronto “National Post” in 2002:
“The SRES group are not in general trained economists but rather energy systems modelers, quasi-engineers cum economists, operations research types, etc. And in my experience they have a hazy notion of the PPP but not a complete understanding. The IPCC has used PPP when it suited their purpose: they wanted damages to be larger, so they converted dollar based damages in developing countries using PPP so they were comparable to damages in developed countries. But then they have made cross country income comparisons based on exchange rate conversion because it made the income differences appear larger. So, they know enough about the conversion to use it to their advantage, but not enough to be consistent in how they use it if it gets in the way of making a point.”
Steve B, does the Sierra Club support the IPCC’s failure to follow the requirements of the internationally-agreed System of National Accounts?
Further to #31 – **2) the models are hardly the sole basis for the conclusion that we have a problem with anthropogenic forcings, in particular GHGs.**
I asked Judith Curry the following –
*In your presentation, you indicate that Knutson et al (2006) specifically attributed the increase in global tropical sea surface temperatures to greenhouse warming(and therefore the increase in hurricanes)
Can you clarify if this was scientifically measured or was this statement a result of a computer model or simulation?*
Her reply: *Knutson paper was computer simulation*
In other words – more modelling.
This was used in her Congressional Testimony.
re: #5
To my mind the Federal Reserve has an impossible task of making monetary decisions based on projections of the economy 6 to 18 months out. That they are doing a better job of this is not obviously evident from the reoccurring business cycle episodes we continue to experience. Perhaps they are motivated less of late by political concerns, but some would say the Fed continues only to exaggerate the natural swings. (That is also why even if we understood the climate well enough to affect it, I would be greatly concerned that our dilemma would be the same as the Feds, including politics and overreaction).
I would think that modeling the climate (or at least determining whether one could model it reasonably well) would be more straight-forward than modeling the economy, but that would be the case only if that modeling relied on the applicable physics and that it included most of the applicable physics. My question to those more familiar with the inputs into the climate models and the adjustments to “tune” them would be what precautions if any are taken to avoid data mining (without some kind of Bonferroni corrections) and over fitting.
In comment #18 of this thread Willis Eischenbach in Figure 2 shows the out-of-sample performance of actual temperatures (HADCRUT3) versus various computer model scenarios for recent times with results that could be readily indicative of an over fitted model. Now when financial models show over fitting out-of-sample the model owner will sometimes turn attention to a better model (based on the initial model) that will work better because of some newly recognized investing principle. They show how it would have fit better with older in-sample and newer out-of-sample results. This model than is used until sufficient out-of-sample results show it to fail. For the naàÆà⮶e investor this process could continue ad infinitum. Is this the case with models of the climate? Now I can see legitimately using more well-proven physics or newly determined and well-proven physics relationships in the models as a matter of fine tuning as long as they could not have made alternative selections and thus allowing some arbitrary fitting. How much of the fine tuning of these climate models is done using well-proven physics, how much is done by selecting certain physical relationships from an array of possible ones and how much is done simply with past data adjustments?
In comment #17 by Hans Erren, he lists a link (with thanks to dano) that shows what I believe are unforced climate models calculations of mean temperatures over periods of 100 to 200 years. While the graphs are meant to show how little the models drift under these conditions as I assume is the same as those that Willis Eschenbach presented above, these graphs seem to show greater oscillations over 20 to 50 years (than those in Eschenbach’s graph) where one can see temperature changes of up to 0.6 degree Kelvin. Has there been any discussion of the shorter term model excursions as opposed to longer time period trends or am I not understanding what is being presented?
As an aside I read two of Steve B’s links, one in #23 on cloud feedback and one in #27 on the IPCC energy/fossil fuel consumption model, and walked away with much uncertainty about both issues. This seems quite typical of links that I read in supposed support of the whole spectrum of views presented at this blog and related to climate”¢’¬?uncertainty.
Ferdinand, thank you for the references in #36. You say:
I have always been astounded that people actually think that more clouds would make the world warmer. Practical experience is my key here.
My experience with clouds is that they have opposite effects day and night.
During the day, when a cloud comes over, I feel cooler. Much cooler.
During the night, when a cloud comes over, I feel warmer. A little warmer.
This practical experience is supported by the physics. During the day, the cloud is reflecting away some large portion of the solar irradiation. This irradiation averages about 650 w/m2, and is more in the tropics. So a cloud blocks out hundreds of w/m2 when it comes over during the day.
During both the day and the night, on the other hand, the cloud increases the greenhouse effect. How much? Depends on the clouds … but not hundreds of watts per m2.
Finally, along with clouds we get parasitic losses to the greenhouse (conduction, convection, evaporation, condensation, hydrometeors), all of which tend to cool the earth.
My conclusion, therefore, based on just looking at the world, is that clouds cool the planet.
NASA agrees, saying:
SOURCE
So my question is, if we found out in the 1980s that clouds cool the world, that is to say they are a negative feedback … how come the models have positive … oh, never mind.
w.
Re # 32.
‘It seems to me that if your model leads to a “singularity”, this means there must be something fundamentally wrong with the model. After all, these models are supposed to be based on physical principles, and if the reality doesn’t require an island to prevent a “singularity”, then why should the model? The conclusion for me would be to fix the model, not to create an “artificial island” …’
The singularity in the model is probably due to the use of a spherical coordinate system. There is nothing fundamentally wrong then and the use of an artificail island might be a quite good method. Try to get some background in numerical modelling before you make such misleading statements.
Re 42, gb, you say the singularity is “probably from a spherical coordinate system”?
Probably?
Of course it is from a spherical coordinate system, no probably about it. However, there are much better ways to deal with such a singularity than an artificial island.
For example, a local orthogonal grid over the Arctic Ocean can be used to avoid the singularity. Alternatively, the singularity can be displaced onto a local landmass such as Russia or Canada. This is the solution used in the CCSM3.0 GCM, for example.
The GDFL model, to take another example, has two “North” Poles, one in Canada and one in Russia. Another option is to use a rotated lat/long grid for the Arctic and North Atlantic Oceans, and a regular grid for the other oceans. Or, some models such as the CCSR/NIES use antipodal poles in Greenland and Antarctica.
Finally, some GCMs use a geodesic grid to avoid the singularities. This has some computational advantages, and avoids the problem of having gridcells of greatly differing sizes that shrink somewhere to a singularity “¢’¬? a geodesic grid has no singularity at all.
What surprised me was the introduction of an artificial island in the Arctic Ocean to solve the singularity problem. Why? Well, to quote from one of the developers of the OCCAM GCM model:
Perhaps the most interesting work being done in the field involves unstructured grids, some workers in the field feel that structured grids are out of date. Unstructured grids do not have regular mesh sizes. These have the advantage of not having singularities anywhere, plus they can have smaller grid sizes in areas of special interest. Finally, they can be dynamically configured and used in FEA, finite element analysis.
Clearly, the direction of the research being done is exactly what I said, which is to fix what is wrong with the model by removing the singularity, rather than to “solve” the singularity with an artificial island.
Anyhow, thanks your posting. As for your condescension, no thanks, you should probably save it for someone who needs it …
w.
Re #41: Of course they’re a negative feedback in an absolute sense. As the quoted text points out, life would be rather toastier if that wasn’t true.
But anyway, the issue is whether they will be a positive or negative feedback in the context of further warming. In that regard, the same NASA article you linked to goes on to say:
“Given the large impact of clouds on the radiative energy balance, the critical question now becomes: What effect will clouds have on surface temperatures if global climate changes in the next century? No one knows. Clouds could act to dampen any greenhouse gas warming by increasing cloud cover, increasing thickness, or by decreasing in altitude. Conversely, clouds could act to increase warming of the planet if the opposite trends occur. In fact, the climate is so sensitive to how clouds might change, that our current best models of global climate can vary in their global warming predictions by more than a factor of three depending on how we try to model the clouds.”
Re #40: Ken, if it’ll make you happier I’ll try to stick to linking stuff that expresses more certainty.
Re #22 and (part of) #31: It’s not just the models and the hockey stick. The relevant portions of the TAR cover the evidence. Willis has certainly read this before, but for anyone who hasn’t be sure to read all of B using the page turning button at the bottom. Of course this information is now seven years old and the AR4 will add to it considerably.
(I’ll post the Arctic stuff tomorrow, Willis.)
Steve Bloom,
Your link to Section B of the TAR establishes that the earth has warmed over the last century. I don’t think that anyone on this site disputes that. However it does not include anything specifically about attribution.
It does, however, include the hockeystick graph, and the following statements:
It is likely that the rate and duration of the warming of the 20th century is larger than any other time during the last 1,000 years. The 1990s are likely to have been the warmest decade of the millennium in the Northern Hemisphere, and 1998 is likely to have been the warmest year.
… statements which we know, thanks to Steve M, the NAS Panel and Wegman, to be unsupportable.
So if it isn’t the models, and it isn’t the hockeystick what is the support for AGW?
On cloud feedback, it is worth re-reading Ou’s consideration of this topic where he concluded that clouds had a negative feedback from a completely different approach than the GCMs discussed here . Holloway’s view on oceans as modeled in GCMs discussed here is also relevant.
#15, brooks
“I am still chuckling thinking about a consensus of computer models. I am certain that there are people who find no humor in this, but I certainly do.”
I’m glad you found it amusing as I meant it to be.
Now what I meant by this post was that if the IPCC decided to remove certain models from the ensemble based on their poor performance then I think that most likely those models which were rejected (and the governments from those countries that fund them) would dispute the predictions of those models which were not rejected. It is in the interests of the IPCC to maintain harmony within the climate modelling fraternity so that it can claim that there is a consensus. If some models were not included then I could easily see a situation developing in which some (at least) of the rejected models would start to show predictions that contradicted the so called ‘consensus’ from the non-rejected models. That would be bad news for the IPCC so not surprisingly it has decided to continue to include them all.
KevinUK
The following is from the August 22nd issue of Eos.
Re #46: OK, for attribution you’ll want to read C, D and E beginning here.
Re #50:
Steve B., the IPCC is applying “best guesses”, more than attribution…
As already mentioned in a lot of other occasions, the influence of anthropogenic aerosols is far from sure (even the sign may be wrong…) and anyway overestimated. And cloud changes give a negative feedback for warming oceans (with an order of magnitude larger change in radiation than from GHGs in the same time frame), while they give a positive feedback for solar changes… Thus the real influence (“attribution”) of the different actors (solar, volcanic, GHGs and aerosols, with their individual feedbacks) is far from exactly known.
Re #50: Chen and Wielicki from 2002 yet on the cloud stuff, eh? I thought we had established elsewhere that even they don’t think those results are valid any longer. You have read their more recent papers, right? But anyway, if your approach is to a priori leave enough room for a large insolation forcing, then I suppose you have to reject the IPCC.
Re 50, Steve, the B and C sections you referred to have nothing to do with attribution. The section that deals with attribution, Section E, bases all of its claims on climate models … and if you believe those models, I have a bridge in New York to sell to you, great deal, cheap …
The overriding truth about the climate is “we don’t know”. Everyone involved in the field should practice saying that, it’s the most important phrase in science. Only when we admit that we don’t know do we have a chance of learning.
The IPCC takes a shot at this at the end of Section E, where they say:
Having actually made some true statements, including the fact that the differences between the models are huge, and bearing in mind that all of their attribution statements are model-based, you’d think they’d close by saying “we don’t know” … but nooooo, this is the IPCC, so they say:
Ho, ho, ho … that’s what’cha call a “non sequitur”, a conclusion that doesn’t follow from the underlying facts.
w.
Re #38: Ian, the issue isn’t whether there’s a difference but whether it’s important enough to make a big stink about PPP bit being used in the AR4. Obviously there is some very substantial disagreement among economists on that point, which is what I was trying to say in the comment to which you responded.
I did a quick google of PPP MER IPCC and got some interesting hits. One of them, a 2003 AEI-Brookings Joint Center study, was especially interesting since it probably explains why the US government decided not to care about this issue. The executive summary reads:
“Critics of the Intergovernmental Panel on Climate Change’s Special Report on Emission Scenarios claim that the use of market exchange rates rather than purchasing power parity has led to a significant upward bias in projections of greenhouse gas missions, and hence unrealistically high future temperature. Rather than revisit the debate on the choice of exchange rates, we address a much simpler question: does the choice make a difference when it comes to projecting future temperature change? Employing a computable general equilibrium model designed to examine a variety of issues in the climate debate, we find that the answer is yes, but the difference is only minor.”
I would also note that you cited the US Energy Information Administration as having adopted PPP just *last year*. I think that says a lot about the urgency with which this issue is viewed.
In any case, it sounds as if there won’t be any question about the change being made for the AR5. If it makes you feel any better, I’m not very happy about the AR4 either.
Re 54, Steve Bloom, thank you for your interesting comment. You will no doubt be surprised to hear that I agree with you “¢’¬? I don’t think too many people believe that PPP will make a huge difference overall.
However, my point in my answer to John Quiggins (above) still stands, which I will repeat here:
Thus, while the errors will average out overall, the models are not using averages “¢’¬? they are sensitive to not only how much GHGs are being emitted, but where they are being emitted. That’s why PPPs are far superior to MERs.
w.
Steve, The AEI Brookings Joint Center “study” was the WORKING PAPER which the IPCC cited in its press release of 8 December 2003 dismissing Castles & Henderson. In other contexts, the IPCC claims to place great weight on peer review. The Lords Committee said that C&H “have helped to generate a valuable literature that calls into question a whole series of issues relating to the SRES, not just the issue of MER versus PPP” and called for “a wholesale reasppraisal of the emissions scenario exercise.” They emphasised that this was not just a matter of making ‘adjustments for improved data’, but that “There is a need to reconsider the economic basis on which the scenarios are constructed.”
The Stegman paper in “Climatic Change” which I cited has appeared since the Lords Report. She claims that “the SRES authors themselves … provide evidence that if PPP converted data were used to generate economic growth projections the resulting emissions projections in the SRES would be affected”; that in the McKibbin et al analysis “the mistaken use of exchange rate converted data [in the SRES] results in total emissions that are … 40% higher by 2100, than when PPP converted data is used”; and that “it is not clear what the impact on emissions of using exchange rate converted data is in the SRES because the report is not transparent in its assumptions and methodology.” She concludes:
“… The results do not negate the need for a large scale review on the grounds of statistical inaccuracies, questionable methodological assumptions and empirical inconsistencies. The IPCC has not demonstrated that the SRES emission projections have a sound economic foundation. Because these emissions projections are used as inputs in models of temperature and climate impacts, these in turn do not have a sound economic basis.”
I didn’t know that the US Government “decided not to care about the issue.” If that decision WAS made, did anyone consider consulting an expert in national accounts or productivity measurement?
Re #52,
Steve B., as far as I know, Chen and Wielicki never retracted their findings. Their comment on the findings still can be read at the NASA pages. And have a look at one of the following pages, “The Trouble with Models”: The difference between models and observations is at times of the same order as what is expected from 2xCO2 (4 W/m2).
Further, as already mentioned, the Wielicki and Chen findings were confirmed (2005) by ground based cloud observations for a longer time frame and higher latitudes by J. Norris.
That climate models don’t capture the observed changes in cloud cover is also confirmed by Allan and Slingo for the HadCM3 model.
And the same model underestimates solar influences with at least a factor 2, see the attribution work of Stott ea.. This may be even worse, as they used a fixed (high) contribution of human made aerosols.
This all is about observed cloud/radiation changes. While the IPCC forcings may be right (except for human made aerosols), the feedbacks implied in current models underestimate solar influences and overestimate GHG and aerosols influences, both due to cloud feedbacks, obviously implemented with the wrong sign. Thus the IPCC range for 2xCO2 (and consequently for all scenario’s) is hugely overestimated…
Re #31: OK, Willis, the Arctic stuff as promised (this time with excerpts):
Willis concluded in Warwick Hughes’ Coolwire 13 that 1) media claims that 2005 was in fact a record Arctic sea ice low were false and 2) there is nothing unusual abut present Arctic temperature trends.
I refuted the first point on Warwick’s blog:
"To expand briefly about my point on the sea ice extent, there are two data sets maintained, one by NSIDC and the other by UIUC. While based on the same raw satellite data, they use very different metrics for deciding whether a given satellite pixel has sea ice or not. NSIDC states their metric on their site and UIUC does not, but a quick comparison of contemporaneous sea ice extent graphics on the two sites makes it apparent that the metrics are quite different. This is fair enough as far as it goes, since there’s nothing intrinsically more valid about using, e.g., 10% versus 15% coverage to determine whether a given pixel has sea ice or not. Of course these different metrics result in similar but not identical anomalies, the graphs of which are also available on each site.
"Another consequence is that the two methodologies will sometimes show different record years, which is what happened this year when NSIDC showed a record and UIUC did not. In any event, the media coverage in September was about the new record set by the NSIDC data, but Willis used the UIUC data to refute it. It was only by a very unlikely coincidence that I happened to know about the difference in metrics. There was obviously some sort of discussion to be had contrasting and comparing the two data sets and talking about the overall trend, [snip]"
I think I wrote something with a bit more detail elsewhere, but the foregoing will do. (Speaking of Arctic sea ice, the NSIDC has established a special page so folks can monitor this year’s exciting finish to the melt season.)
On the Arctic temperature issue, it turns out Willis cherry-picked his data by using 70 degrees latitude to define the Arctic rather than the standard 60 degrees. William Connelley nailed him on it, but for the details we’ll need to wait a few days since it was on William’s old blog and I can’t access those archives. But OTOH it’s something for Willis to look forward to.
I dunno Bloom. I’m sure Willis will respond, but I read your post, I looked at Willis’ Coolwire post, then I re-read your post, and I re-read Willis’ Coolwire post, and I can’t see that you’ve refuted anything.
Are you saying that one dataset says the 2005 sea-ice was a record low, and the other doesn’t? And that one data-set says current arctic temperatures are unusual, and the other doesn’t?
Of course, my #56 is addressed to Steve Bloom, not Steve McIntyre. And I should have mentioned that Alison Stegman’s research has been funded by the Australian Greenhouse Office. You’d think that the IPCC might feel some twinge of embarrassment that its use of MERs is condemned by economists of the calibre of Professors Sir Partha Dasgupta (Cambridge); Alan Heston (University of Pennsylvnia, Chair of the Technical Advisory Group to the International Comparison Programme, in which he has been involved for forty years); Warwick McKibbin (Australian National University and Brookings Institution); Richard Tol (Universities of Hamburg, Princeton and Vrije); Colin Robinson (University of Surrey); Erwin Diewert (University of British Columbia); and Angus Maddison (world’s leading expert in the measurement of economic output between countries and across the ages). But apparently not.
Re #58:
Steve B., why is the “standard” 60 degrees? As far as I know, 66.66 degrees is where the polar circle is, thus Willis is closer to the definition than the “standard”?
Btw, I made some oversight of all circumpolar (over 67 degree) station data some years ago. It revealed that most (70%) of the stations had their highest temperatures in the 1930-1940 period, with a decrease thereafter and an increase again, not/just reaching the 1930-1940 period. Only 30% of the stations (mainly Eastern Siberia, Alaska and Western Canada) had recently higher temperatures. This may have changed in recent years…
Another item: cloud changes in the Arctic (again) act as negative feedback for increasing temperatures. While in spring/summer more clouds reflect more sunlight, thus decreasing insolation/warming, the trend is reversed in winter: less clouds allow for more heat escape to space, thus giving more cooling. The net result is that near as much water is refrozen in winter than is melted in summer, causing the winter ice trend to decline less than the summer trend. Again, that is not predicted by climate models… See the comment in Cicero:
I found the graph at the top along with other data from the comparison in a paper from Curt Covey et.al (2003) “An overview of results from the Coupled Model Intercomparison Project” here.
~ghw
While I know almost nothing about GCMs, the discussion here, and my experience with both physical and statistical modeling in general, make me wonder about the, well, “standards” associated with climate modeling. I used to imagine that in such a luxuriously funded area as climate research it would be possible to maintain high standards, to do modeling “right” (i.e. identify and characterize all sources of variability/uncertainty: Measurement error; Deterministic structure; Model error; Natural variability; etc.). Coming from an area with far less resources, I always somewhat envied the GCM folks.
However, now I am wondering whether the climate modelers squandered their good fortune. It seems that, perhaps, instead of trying to do a superb job with ingenius, simple, tractable, models (i.e. ones reflecting only the most important features of the climate system; these are admittedly hard to construct — you have to know a lot about the natural system — and I have no idea if such a model even exists), the modelers instead created GCMs that are incredibly complicated in every way, involve huge amounts of brute-force computation, and, if I understand correctly, are consequently intractable. Of course the modelers love them despite their opacity (modelers always do); the rest of us have to accept (or reject) the results as a matter of faith (over time, of course, the truth becomes apparent).
It makes me realize that there are advantages to a tight budget: It forces you to be careful (and explicit), as well as clever, about precisely which observables the model will reproduce; It drives you to simple, tractable, models; and it requires you to characterize quantitatively how the “critical three variables” (model results; measured values; and the true value of the variable of interest), correspond to one another, because that’s how you optimize the data collection to get the best precision given the budget. The result is transparency: You — and everyone else — know exactly how bad your model is, and that is exactly the point. I’ve heard it said that the role of the modeler is to quantify ignorance, not create knowledge.
Finally, while it is silly to criticize a model for not precisely reproducing every conceivable statistic, it is still reasonable to demand that a model provide criteria for self-assessment of each statistic it produces. As with any product, you need a quantitative basis for determining whether the model meets the manufacturer’s specifications (and these need to be written down a priori). Without that, all one has to rely on is a sense of “smell” — and by the time things really stink, it may be too late.
Further to my #60, I was remiss in not including in the list my distinguished co-author David Henderson, sometime Head of the Economics and Statistics Department at the OECD; Professor of Economics at University College London; Fellow of Lincoln College, Oxford; Director of the Economics Department at the World Bank; national civil servant in HM Treasury and the UK Department of Aviation; and Visiting Fellow or Professor at institutions in Britain, France, Belgium, Australia and New Zealand.
I should also have mentioned the other co-authors of our submission to the UK Stern Review: Sir Ian Byatt, Chairman of the Water Industry Commission for Scotland, Senior Associate with Frontier Economics, Honorary Professor at Buckingham University and former Deputy Economic Adviser to HM Treasury; Lord Nigel Lawson, former British Chancellor of the Exchequer; Ross McKitrick, Associate Professor of Economics at Guelph University, Ontario; Julian Morris, Executive Director of the International Policy Network and Visiting Professor at the University of Buckingham; Sir Alan Peacock, Honorary Professor of Public Finance at Heriot-Watt University and a former Chief Economic Adviser to the UK Department of Trade and Industry; and Lord Skidelsky, Professor of Political Economy at the University of Warwick and author of the award-winning biography of John Maynard Keynes.
I did mention Colin Robinson, Emeritus Professor of Economics at the University of Surrey, but should also have noted that he is a recipient of the International Association for Energy Economics award for ‘Outstanding Contributions to the Profession of Energy Economics and its Literature.’ And I should also have included Professor William Nordhaus of Yale, who was the keynote speaker at the IPCC Expert Meeting on Emissions Scenarios in Washington, D.C. in January 2005, and Professor Peter Dixon of Monash University, Melbourne, whose research with Maureen Rimmer of that University was presented at an economic modellers’ conference at Lubeck, Germany in June 2005. The work of Dixon & Rimmer, as of McKibbin, Pearce and Stegman, was supported by the Australian Greenhouse Office.
For the sake of completeness, and with apologies to those who don’t care, I also record that I am a former Head of the Australian Bureau of Statistics and of the Australian Department of Finance, a former President of the International Association of Official Statistics, and a former Executive Director and Vice-President of the Academy of the Social Sciences in Australia. My review of the Eurostat OECD-PPP Programme was described in the OECD’s PPP methodological manual (June 2005) as “an important milestone in its history. Most significantly, it confirmed the usefulness of PPPs and fostered a better understanding of their respective responsibilities and roles between Eurostat, the OECD and participating countries.”
Finally, let me quote from a message that was widely circulated in 2001 by Dr. Jacob Ryten, former head of economic statistics at Statistics Canada and consultant to the United Nations Statistical Commission on the International Comparison Programme:
“Both Ian Castles and I [Jacob Ryten] were commissioned to look into the worldwide conditions of PPPs and to recommend what should be done about them if the estimation thereof had fallen into disrepair. In our two reports, which were written independently, we discussed the many improvements required to bolster the use and credibility of what we both rated an indispensable tool for inter-country economic comparisons. We also argued – persuasively, I thought at the time – that whatever blemishes could be attached to the latest PPP rates or to the time series of PPPs, they were of little consequence when compared to the distortions entailed by comparisons based on straight exchange rates…. If indeed there are very serious doubts about PPPs, the only course of action open is to abstain entirely from making comparisons involving GDPs per capita.”
It’s hard to bear that two US economists found that projections of GHG emissions weren’t much affected by whether PPP and MER measures don’t much affect projections of emissions, but I think I can live with it.
Re #56, I shouldn’t have added the square-bracketed [in the SRES]. McKibbin, Pearce and Stegman were comparing the hypothetical (and incorrect) use of MERs in the McKibbin & Wilcoxen G-Cubed model with the output of the same model when the parameters were correctly specified. They found that the effect on emissions was three times as great as in the Manne/Richels exercise cited by the IPCC.
Re 58, Steve Bloom, [snip]
You have finally said what it was that you were on about re Coolwire 13. You show that there are two polar datasets, the UIUC dataset and the NSIDC dataset. One showed a record, and one didn’t. Fair enough, and I’m glad to hear of it, tho’ it doesn’t make much difference to my analysis. [snip] I had no idea that there were two datasets,[snip] . You yourself say you only discovered that there were two datasets through an "unlikely coincidence", [snip]
[snip]
w
I am bumping number 34 because I think it’s important:
Where do I find the details of cloud behaviour, aerosol sensitivity, solar sensitivity and greenhouse sensitivity for the UKMO “model”?
Is the UKMO model a model or a gridded set of observations?
http://www.grida.no/climate/ipcc_tar/wg1/056.htm
Re #66, so you’ve not, until now, heard of the NSIDC?
Re #66: Willis, [snip] BTW, I said that it was an unlikely coincidence that I knew about it simply because it’s not a point that I ever researched.
A current google of "Arctic ‘sea ice’ data" finds two pages filled mostly with references to the NSIDC data, with UIUC appearing only near the bottom of the second page. [snip]
Steve B and Willis – please don’t bring the Coolwire brawl over here. I’ve tried to snip even-handedly.
Re #59: “Are you saying that one dataset says the 2005 sea-ice was a record low, and the other doesn’t?” Yes, and as I noted there’s even a fair discussion to be had about that.
Willis,
Excellent conclusion – and little wonder – the GCM’s suffer from the same problems economic modelling has – attempting to model phenomena that cannot be intrinsically modelled. Economic modelling attempts to predict what individual humans will do in an aggregate sense. Impossible.
Yet economic modelling continues to be used by the Keynesians to guide policy despite the facts that it is gobbledygook. So I suspect the GCM’s will also be used to guide policy despite the facts you summarise above.
As Captain Binghampton of McHale’s Navy would have said “I could just scream!”
One Trackback
[…] CMIP Control Runs :: Issues with the Computer Model Intercomparison Project model runs. Testing … testing … is this model powered up? :: Are the models “lifelike”? Most of them are not. Knobs :: Problems with climate model representations of net cloud forcing. Which way to the feedback? :: Feedback processes in the Eastern Tropical Pacific. Top Secret NOFORN Restricted Access Climate Model Results :: Problems with getting the CCSM3 model data. First Light on the Ozone Hockeystick :: Ozone forcings used by the CCSM3 to hindcast 20th century temperatures, and their problems. […]