Guest post by Pat Frank
The summary of results from version 5 of the Coupled Model Intercomparison Project (CMIP5) has come out. [1] CMIP5 is evaluating the state-of-the-art general circulation climate models (GCMs) that will be used in the IPCC’s forthcoming 5th Assessment Report (AR5) to project climate futures. The fidelity of these models will tell us what we may believe when climate futures are projected. This essay presents a small step in evaluating the CMIP5 GCMs, and the reliability of the IPCC conclusions based on them.
I also wanted to see how the new GCMs did compared to the CMIP3 models described in 2003. [2] The CMIP3 were used in IPCC AR4. Those models produced an estimated 10.1% error in cloudiness, which translated into an uncertainty in cloud feedback of (+/-)2.8 Watts/m^2. [3] That uncertainty is equivalent to (+/-)100% of the excess forcing due to all the GHGs humans have released into the atmosphere since 1900. Cloudiness uncertainty, all by itself, is as large as the entire effect the IPCC is trying to resolve.
If we’re to know how reliable model projections are, the energetic uncertainty, e.g., due to cloudiness error, must be propagated into GCM calculations of future climate. From there the uncertainty should appear as error bars that condition the projected future air temperature. However inclusion of true physical error bars never seems to happen.
Climate modelers invariably publish projections of temperature futures without any sort of physical uncertainty bars at all. Here is a very standard example, showing several GCM projections of future arctic surface air temperatures. There’s not an uncertainty bar among them.
The IPCC does the same thing, brought to you here courtesy of an uncritical US EPA. The shaded regions in the IPCC SRES graphic refer to the numerical variability of individual GCM model runs. They have nothing to do with physical uncertainty or with the reliability of the projections.
Figure S1 in Jiang, et al.’s Auxiliary Material [1] summarized how accurately the CMIP5 GCMs hindcasted global cloudiness. Here it is:
Here’s what Jiang, et al., say about their results, “Figure S1 shows the multi-year global, tropical, mid-latitude and high-latitude mean TCFs from CMIP3 and CMIP5 models, and from the MODIS and ISCCP observations. The differences between the MODIS and ISCCP are within 3%, while the spread among the models is as large as ~15%.” “TCF” is Total Cloud Fraction.
Including the CMIP3 model results with the new CMIP5 outputs conveniently allows a check for improvement over the last 9 years. It also allows a comparison of the official CMIP3 cloudiness error with the 10.1% average cloudiness error I estimated earlier from the outputs of equivalent GCMs.
First, here are Tables of the CMIP3 and CMIP5 cloud projections and their associated errors. Observed cloudiness was taken as the average of the ISCCP and Modis Aqua satellite measurements (the last two bar groupings in Jiang Figure S1), which was 67.7% cloud cover. GCM abbreviations follow the references below.
Table 1: CMIP3 GCM Global Fractional Cloudiness Predictions and Error
| Model Source | CMIP3 GCM | Global Average Cloudiness Fraction | Fractional Global Cloudiness Error |
| NCC | bcm2 | 67.7 | 0.00 |
| CCCMA | cgcm3.1 | 60.7 | -0.10 |
| CNRM | cm3 | 73.8 | 0.09 |
| CSIRO | mk3 | 65.8 | -0.03 |
| GFDL | cm2 | 66.3 | -0.02 |
| GISS | e-h | 57.9 | -0.14 |
| GISS | e-r | 59.8 | -0.12 |
| INM | cm3 | 67.3 | -0.01 |
| IPSL | cm4 | 62.6 | -0.08 |
| MIROC | miroc3.2 | 54.2 | -0.20 |
| NCAR | ccsm3 | 55.6 | -0.18 |
| UKMO | hadgem1 | 54.2 | -0.20 |
| Avg. 62.1 | R.M.S. Avg. ±12.1% |
Table 2: CMIP5 GCM Global Fractional Cloudiness Predictions and Error
| Model Source | CMIP5 GCM | Global Average Cloudiness Fraction | Fractional Global Cloudiness Error |
| NCC | noresm | 54.2 | -0.20 |
| CCCMA | canesm2 | 61.6 | -0.09 |
| CNRM | cm5 | 57.9 | -0.14 |
| CSIRO | mk3.6 | 69.1 | 0.02 |
| GFDL | cm3 | 71.9 | 0.06 |
| GISS | e2-h | 61.2 | -0.10 |
| GISS | e2-r | 61.6 | -0.09 |
| INM | cm4 | 64.0 | -0.06 |
| IPSL | cm5a | 57.9 | -0.14 |
| MIROC | miroc5 | 57.0 | -0.16 |
| NCAR | cam5 | 63.5 | -0.06 |
| UKMO | hadgem2-a | 54.2 | -0.20 |
| Avg. 61.2 | R.M.S. Avg. ±12.4% |
Looking at these results, some models improved between 2003 and 2012, some did not, and some apparently became less accurate. The error fractions are one-dimensional numbers that are condensed from errors in three-dimensions. The error fractions do not mean that some given GCM predicts a constant fraction of cloudiness above or below the observed cloudiness. GCM cloud predictions weave about the observed cloudiness in all three dimensions; here predicting more cloudiness, there less. [4, 5] The GCM fractional errors are the positive and negative cloudiness errors integrated over the entire globe, suppressed into single numbers. The total average error of all the GCMs is represented as the root-mean-square average of the individual GCM fractional errors.
CMIP5 cloud cover was averaged over 25 model years (1980-2004) while CMIP3 averages represent 20 model years (1980-1999). GCM average cloudiness was calculated from “monthly mean grid-box averages.” So a 20-year global average included 12×20 monthly global cloudiness realizations, reducing GCM random calculational error by at least 15x.
It’s safe to surmise, therefore, that the residual errors recorded in Tables 1 and 2 represent systematic GCM cloudiness error, similar to the cloudiness error found previously. The average GCM cloudiness systematic error has not diminished between 2003 and 2012. The CMIP3 models averaged 12.1% error, which is not significantly different from my estimate of 10.1% cloudiness error for GCMs of similar sophistication.
We can now proceed to propagate GCM cloudiness error into a global air temperature projection, to see how uncertainty grows over GCM projection time. GCMs typically calculate future climate in a step-wise fashion, month-by-month to year-by-year. In each step, the climate variables calculated in the previous time period are extrapolated into the next time period. Any errors residing in those calculated variables must propagate forward with them. When cloudiness is calculationally extrapolated from year-to-year in a climate projection, the 12.4 % average cloudiness error in CMIP5 GCMs represents an uncertainty of (+/-)3.4 Watts/m^2 in climate energy. [3] That (+/-)3.4 Watts/m^2 of energetic uncertainty must propagate forward in any calculation of future climate.
Climate models represent bounded systems. Bounded variables are constrained to remain within certain limits set by the physics of the system. However, systematic uncertainty is not bounded. In a step-wise calculation, any systematic error or uncertainty in input variables must propagate as (+/-)sqrt(sum(per-step error)^2) into output variables. Per-step uncertainty increments with each new step. This condition is summarized in the next Figure.
The left picture illustrates the way a GCM projects future climate in a step-wise fashion. [6] The white rectangles represent a climate propagated forward through a series of time steps. Each prior climate includes the variables that calculationally evolve into the climate of the next step. Errors in the prior conditions, uncertainties in parameterized quantities, and limitations of theory all produce errors in projected climates. The errors in each preceding step of the evolving climate calculation propagate into the following step.
This propagation produces the growth of error shown in the right side of the figure, illustrated with temperature. The initial conditions produce the first temperature. But errors in the calculation produce high- and low uncertainty bounds (e_T1, etc.) on the first projected temperature. The next calculational step produces its own errors, which add in and expand the high and low uncertainty bounds around the next temperature.
When the error is systematic, prior uncertainties do not diminish like random error. Instead, uncertainty propagates forward, producing a widening spread of uncertainty around time-stepped climate calculations. The wedge on the far right summarizes the way propagated systematic error increases as a step-wise calculation is projected across time.
When the uncertainty due to the growth of systematic error becomes larger than the physical bound, the calculated variable no longer has any physical meaning. In the Figure above, the projected temperature would become no more meaningful than a random guess.
With that in mind, the CMIP5 average error in global cloudiness feedback can be propagated into a time-wise temperature projection. This will show the average uncertainty in projected future air temperature due to the errors GCMs make in cloud feedback.
Here, then, is the CMIP5 12.4% average systematic cloudiness error propagated into everyone’s favorite global temperature projection: Jim Hansen’s famous doomsday Figure; the one he presented in his 1988 Congressional testimony. The error bars were calculated using the insight provided by my high-fidelity GCM temperature projection emulator, as previously described in detail (892 kB pdf download).
The GCM emulator showed that, within GCMs, global surface air temperature is merely a linear function of net GHG W/m^2. The uncertainty in calculated air temperature is then just a linear function of the uncertainty in W/m^2. The average GCM cloudiness error was propagated as:
“Air temperature” is the surface air temperature in Celsius. Air temperature increments annually with increasing GHG forcing. “Total forcing” is the forcing in W/m^2 produced by CO2, nitrous oxide, and methane in each of the “i” projection years. Forcing increments annually with the levels of GHGs. The equation is the standard propagation of systematic errors (very nice lecture notes here (197 kb pdf)). And now, the doomsday prediction:
The Figure is self-explanatory. Lines A, B, and C in part “a” are the projections of future temperature as Jim Hansen presented them in 1988. Part “b” shows the identical lines, but now with uncertainty bars propagated from the CMIP5 12.4 % average systematic global cloudiness error. The uncertainty increases with each annual step. It is here safely assumed that Jim Hansen’s 1988 GISS Model E was not up to CMIP5 standards. So, the uncertainties produced by CMIP5 model inaccuracies are a true minimum estimate of the uncertainty bars around Jim Hansen’s 1988 temperature projections.
The large uncertainty by year 2020, about (+/-)25 C, has compressed the three scenarios so that they all appear to nearly run along the baseline. The uncertainty of (+/-)25 C does not mean I’m suggesting that the model predicts air temperatures could warm or cool by 25 C between 1958 and 2020. Instead, the uncertainty represents the pixel size resolvable by a CMIP5 GCM.
Each year, the pixel gets larger, meaning the resolution of the GCM gets worse. After only a few years, the view of a GCM is so pixelated that nothing important can be resolved. That’s the meaning of the final (+/-25) C uncertainty bars. They mean that, at a distance of 62 years, the CMIP5 GCMs cannot resolve any air temperature change smaller than about 25 C (on average).
Under Jim Hansen’s forcing scenario, after one year the (+/-)3.4 Watts/m^2 cloudiness error already produces a single-step pixel size of (+/-)3.4 C. In short, CMIP5-level GCMs are completely unable to resolve any change in global surface air temperature that might occur at any time, due to the presence of human-produced carbon dioxide.
Likewise, Jim Hansen’s 1988 version of GISS Model E was unable predict anything about future climate. Neither can his 2012 version. The huge uncertainty bars makes his 1988 projections physically meaningless. Arguing about which of them tracks the global anomaly trend is like arguing about angels and heads of pins. Neither argument has any factual resolution.
Similar physical uncertainty bars should surround any GCM ensemble projection of global temperature. However, the likelihood of seeing them in the AR5 (or in any peer-reviewed climate journal) is vanishingly small.
Nevertheless, CMIP5 climate models are unable to predict global surface air temperature even one year in advance. Their projections have no particular physical meaning and are entirely unreliable.
Following on from the title of this piece: there will be no reason to believe any – repeat, any — CMIP5-level future climate projection in the AR5.
Climate projections that may appear in the IPCC AR5 will be completely vacant of physical credibility. Attribution studies focusing on CO2 will necessarily be meaningless. This failure is at the center of AGW Climate Promotions, Inc.
Anyone, in short, who claims that human-produced CO2 has caused the climate to warm since 1900 (or 1950, or 1976) is implicitly asserting the physical reliability of climate models. Anyone seriously asserting the physical reliability of climate models literally does not know what they are talking about. Recent pronouncements about human culpability should be seen in that light.
The verdict against the IPCC is identical: they do not know what they’re talking about.
Does anyone think that will stop them talking?
References:
1. Jiang, J.H., et al., Evaluation of cloud and water vapor simulations in CMIP5 climate models using NASA “A-Train” satellite observations. J. Geophys. Res., 2012. 117, D14105.
2. Covey, C., et al., An overview of results from the Coupled Model Intercomparison Project. Global Planet. Change, 2003. 37, 103-133.
3. Stephens, G.L., Cloud Feedbacks in the Climate System: A Critical Review. J. Climate, 2005. 18, 237-273.
4. AchutaRao, K., et al., Report UCRL-TR-202550. An Appraisal of Coupled Climate Model Simulations, D. Bader, Editor. 2004, Lawrence Livermore National Laboratory: Livermore.
5. Gates, W.L., et al., An Overview of the Results of the Atmospheric Model Intercomparison Project (AMIP I). Bull. Amer. Met. Soc., 1999. 80(1), 29-55.
6. Saitoh, T.S. and S. Wakashima, An efficient time-space numerical solver for global warming, in Energy Conversion Engineering Conference and Exhibit (IECEC) 35th Intersociety. 2000, IECEC: Las Vegas. p. 1026-1031.
Climate Model Abbreviations
BCC: Beijing Climate Center, China.
CCCMA: Canadian Centre for Climate Modeling and Analysis, Canada.
CNRM: Centre National de Recherches Météorologiques, France.
CSIRO: Commonwealth Scientific and Industrial Research Organization, Queensland Australia.
GFDL: Geophysical Fluid Dynamics Laboratory, USA.
GISS: Goddard Institute for Space Studies, USA.
INM: Institute for Numerical Mathematics, Russia.
IPSL: Institut Pierre Simon Laplace, France.
MIROC: U. Tokyo/Nat. Ins. Env. Std./Japan Agency for Marine-Earth Sci.&Tech., Japan.
NCAR: National Center for Atmospheric Research, USA.
NCC: Norwegian Climate Centre, Norway.
UKMO: Met Office Hadley Centre, UK.
Thanks everyone for your interest and comments. Regrets for the delay in replying . . . it was a long day at work.
But forthwith:
Maus, GCM cloudiness error is an accuracy error, not a rounding error. By “viewed appropriately” I suppose you mean fancy can take flight when there are no measurement constraints. In that arena we can find philosophy, religion, and, all too often, politics.
Stephen maybe a letter to GISS or NCAR?
Mike B the K, the GCM temperature projection emulator noted above will produce a GCM-equivalent GHG temperature projection for approximately zero cost.
Terry, the CMIP5 GCMs were tested against physical measurements. The modelers have made a predictive representation, validating an error analysis. But I agree with you that GCMs are unfalsifiable and do not make predictions in the scientific sense.
Mr. Lynn, thanks for giving me the opportunity to make an important point. Nothing I’ve posted says that human-caused climate warming is not happening. The point of the essay is that the analysis shows climate models are unable to resolve anything at the few W/m^2 level of GHG forcing.
Science is not about what’s happening; nor is it about “the universe.” It’s about what we know about what we’ve observed. We’ve observed the climate has warmed. But climate models are completely unable to tell us why that’s happened.
We’ve also observed that there’s nothing whatever unusual about any weather extremes, about the rate or the duration of the recent warming, or about any weather trends. In other words, there is no scientific or empirical support for any discrete alarm. But, again, that doesn’t mean the recent warming isn’t due to the extra GHGs. We plain don’t know.
Mark given serious advances in physics, climate models may eventually be able to tell us how climates behave, even if they can’t predict exactly how our particular climate will behave in detail. It’s worth pursuing climate science. It’s just that so many climate scientists have abandoned integrity for a righteous and sentimentalized advocacy. Think baby, bathwater.
Thanks, Theo. 🙂
Terry Oldberg says:
August 24, 2012 at 9:15 pm
davidmhoffer:
If you were to more fully explain what the point of your comment was, perhaps I could address it.
>>>>>>>>>>>>>>>>>>>>>
OK, I’ll take one more crack at it.
You’re position is completely correct. Now ask yourself, of what value is it to be correct if nobody understands you? How many people can you persuade that the IPCC reports are garbage if they cannot understand your explanation?
I’m in the persuasion business. Someone asks me what I do for a living and I say that I architect and implement large scale, highly resilient, data management infrastructures. The vast majority of people look at me like I just grew an extra head. So I tell them I sell really big computers like the kind governments and really big companies use. Both descriptions are accurate, but one almost everyone thinks is gibberish and the other almost everyone understands. Which answer I give depends on the technical background of the person I am talking to.
In the climate debate, trying to accurately differentiate between predictions and projections for most people, is gibberish. Pointing out that the IPCC reports said certain things were going to happen by now, and they haven’t, is accurate, and people understand it. Let the IPCC’s advocates respond to explain that they used projections not predictions, and most people will just tune out.
So, you want to be technically accurate but tuned out by almost everyone? Or do you want to persuade the majority of people that the IPCC science is bunk? If the latter, then choose your arguments wisely and couch them in terms that the average reader can understand.
davidmhoffer:
Thanks for the clarification. Your question of how to explain this to the masses so they understand it and take action to throw out the corrupt or incompetent “scientists” is a good one. I don’t know the answer. If you have any ideas, I’d like to hear them.
I have a background in the planning and management of scientific studies. With this background, I know that if a purportedly scientific study lacks a statistical population then this study is not scientific, for claims made by the models that are a product of this study are insusceptible to being falsified by the evidence. Rather than being scientific, the methodology is dogmatic. Adoption of a dogmatic methodology ensures a failure to learn and consequential waste of the money that is invested in the research. However, a dogmatic methdology disguised as a scientific methodology is ideal for the purpose of duping people into false beliefs.
Four years ago, out of idle curiosity I began to poke around in climatology. Within half an hour, Web searches had produced enough evidence for me to be pretty sure that the methodology of the research was not scientific; I also found that the IPCC was representing the research to be scientific. In the report of AR4, I was unable to find reference to the statistical population. Also, an IPCC reviewer named Vincent Gray reported that he had informed the IPCC that its models were insusceptible to statistical validation; the IPCC had responded by obfuscating the issue using the strategy of ambiguous reference by terms in the language of its assessment reports. Elements of this strategy included: a) using “prediction” and “projection” as synonyms b) using “evaluation” and “validation” as synonyms and c) exploiting ambiguity of reference by the word “science” in which the word references the idea of “demonstrable knowledge” and “the process that is operated by people calling themselves ‘scientists’.” Subsequently, I did a more thorough investigation and published my findings under the peer review of Judith Curry ( http://judithcurry.com/2011/02/15/the-principles-of-reasoning-part-iii-logic-and-climatology/ ).
I’ve seen this situation before. Twenty-seven years ago, after accepting the lead role in an area of engineering research, I discovered that past research in this area defined no statistical population. To remedy the situation would require modification of an industrial process but to do so would require admission of past error on the part of high level mucky mucks. When I attempted to remedy the error, I experienced unanimous opposition from the mucky mucks. Today, twenty-seven years later, this error remains in place where it threatens the lives of people who travel on aircraft or work in nuclear power plants. Many of the mucky mucks went on to even higher levels of glory. My family and I suffered years of hardship from my temerity in taking on the mucky mucks.
Hans Christian Anderson’s story “The Emperor’s New Clothes” provides a metaphor for situations such as this. As the emperor marches naked through the village, his ministers profess to see him clothed in the nonexistent golden cloth for which he has paid swindlers but a child, being free from the pretensions of the adults, crys out “the emperor has no clothes!” Apparently unaffected by this disclosure, the king continues to march gloriously along in his nonexistent golden clothes.
gregole, you’re right about clouds. They are a huge unknown. GHGs increase the energy content in the atmosphere, which may well influence density and types of clouds. But climate science isn’t advanced enough to know how.
As it’s turned out, the oceans are another huge unknown. Carl Wunsch has noted that ocean models don’t converge, meaning they give no unique solution. But he’s said modelers ignore that because the results “look reasonable.” Those non-converged ocean models enter into GCM climate predictions and are part of the reason they’re so inaccurate. CW has also noted that the thermohaline model is fine for cocktail party conversation, but doesn’t describe ocean dynamics.
So, there’s a lot to do, but all that’s really hard work. Instead of hard work, many climate scientists have gotten lazy, preferring to play video games and issue portentous statements.
Jim D, the GCM model outputs were annual average cloudiness. The errors are error in average annual cloudiness. Why isn’t it correct, therefore, to calculate the uncertainty over annual time-steps?
AlaskaHound, everything is in terms of annual averages: averaged over 365 days, 365 nights. Average excess energy, average feedback error.
Philip Bradley, as long as they’re comparing their model outputs to physical measurements, it’s valid to calculate and propagate an error.
george e smith, if you look at the SI to my Skeptic article, here (892 kb pdf), you’ll find that the 10.1% error was derived by integrating GCM hindcasted vs observed global average cloudiness. I.e., it’s an empirical error estimate.
Global average cloudiness turns out to be fairly constant from year-to-year. Total cloud cover has been available since about 1983 with the GMS series satellite. Here’s an overview of the US part of the International Satellite Cloud Climatology Project (ISCCP).
Allan MacRae, right on! And as long as they’re publishing comparisons between GCM hindcasts and target observations, it’s worth calculating the errors and propagating the uncertainties to drive home the point that GCMs are unreliable and their predictive robusticity is mere politics.
Ian W, “Personally, I would like to see strict Validation of these models …” Good luck getting that program started. 🙂
Robert Brown, please recall that the analysis is concerned with GCMs rather than with climate. Clearly GCMs are complex beasts, as you say, “systems of parametric coupled nonlinear differential equations”
However, with respect to GHG-induced air temperature, GCM outputs are no more than linear extrapolations of GHG forcing. That is made clear by the very good emulation of GCM air temperature projections using an equation with only one degree of freedom — increasing GHG forcing. All the internal complexity of a GCM is somehow lost in air temperature projections. Their output is linear. That being true, one has an empirical justification to forward propagate GCM temperature error using the same linear function that successfully emulated the GCM forward projection of air temperature itself.
Wonderful comments Pat- very informative.
Regarding hindcasting of models, you may have seen this exchange with Douglas Hoyt on aerosols. I would be interested in your opinion. I continue to correspond from time to time with Douglas, and would be pleased to invite him onto this thread if you so requested.
http://wattsupwiththat.com/2012/04/28/tisdale-a-closer-look-at-crutem4-since-1975/#comment-970931
markx says:April 29, 2012 at 9:56 am
ferd berple says: April 28, 2012 at 12:09 pm
Climate scientists complain when someone outside of climate science talks about climate science, but ignore the fact that climate science is no qualification to build reliable computer models.
Markx: IMHO, this one of the most important observations made within these pages.
___________
Allan:
Agree – when you build a mathematical model, you first try to verify it. One method is to determine how well it models the past (“hindcasting”).
The history of climate model hindcasting has been one of blatant fraud. Fabricated aerosol data has been the key “fudge factor”.
Here is another earlier post on this subject, dating from mid-2009.
It is remarkable that this obvious global warming fraud has lasted this long, with supporting aerosol data literally “made up from thin air”.
Using real measured aerosol data that dates to the 1880’s, the phony global warming crisis “disappears in a puff of smoke”.
Regards, Allan
http://wattsupwiththat.com/2009/06/27/new-paper-global-dimming-and-brightening-a-review/#comment-151040
Allan MacRae (03:23:07) 28/06/2009 [excerpt]
FABRICATION OF AEROSOL DATA USED FOR CLIMATE MODELS:
Douglas Hoyt:
The pyrheliometric ratioing technique is very insensitive to any changes in calibration of the instruments and very sensitive to aerosol changes.
Here are three papers using the technique:
Hoyt, D. V. and C. Frohlich, 1983. Atmospheric transmission at Davos, Switzerland, 1909-1979. Climatic Change, 5, 61-72.
Hoyt, D. V., C. P. Turner, and R. D. Evans, 1980. Trends in atmospheric transmission at three locations in the United States from 1940 to 1977. Mon. Wea. Rev., 108, 1430-1439.
Hoyt, D. V., 1979. Pyrheliometric and circumsolar sky radiation measurements by the Smithsonian Astrophysical Observatory from 1923 to 1954. Tellus, 31, 217-229.
In none of these studies were any long-term trends found in aerosols, although volcanic events show up quite clearly. There are other studies from Belgium, Ireland, and Hawaii that reach the same conclusions. It is significant that Davos shows no trend whereas the IPCC models show it in the area where the greatest changes in aerosols were occurring.
There are earlier aerosol studies by Hand and Marvin in Monthly Weather Review going back to the 1880s and these studies also show no trends.
___________________________
Allan:
Repeating: “In none of these studies were any long-term trends found in aerosols, although volcanic events show up quite clearly.”
___________________________
Here is an email just received from Douglas Hoyt [my comments in square brackets]:
It [aerosol numbers used in climate models] comes from the modelling work of Charlson where total aerosol optical depth is modeled as being proportional to industrial activity.
[For example, the 1992 paper in Science by Charlson, Hansen et al]
http://www.sciencemag.org/cgi/content/abstract/255/5043/423
or [the 2000 letter report to James Baker from Hansen and Ramaswamy]
http://74.125.95.132/search?q=cache:DjVCJ3s0PeYJ:www-nacip.ucsd.edu/Ltr-Baker.pdf+%22aerosol+optical+depth%22+time+dependence&cd=4&hl=en&ct=clnk&gl=us
where it says [para 2 of covering letter] “aerosols are not measured with an accuracy that allows determination of even the sign of annual or decadal trends of aerosol climate forcing.”
Let’s turn the question on its head and ask to see the raw measurements of atmospheric transmission that support Charlson.
Hint: There aren’t any, as the statement from the workshop above confirms.
__________________________
IN SUMMARY
There are actual measurements by Hoyt and others that show NO trends in atmospheric aerosols, but volcanic events are clearly evident.
So Charlson, Hansen et al ignored these inconvenient aerosol measurements and “cooked up” (fabricated) aerosol data that forced their climate models to better conform to the global cooling that was observed pre~1975.
Voila! Their models could hindcast (model the past) better using this fabricated aerosol data, and therefore must predict the future with accuracy. (NOT)
That is the evidence of fabrication of the aerosol data used in climate models that (falsely) predict catastrophic humanmade global warming.
And we are going to spend trillions and cripple our Western economies based on this fabrication of false data, this model cooking, this nonsense?
*************************************************
LazyTeenager says:
That is a completely clueless thing to claim.
No, it isn’t. Pat is correct.
The models are calculations not physical measurements. You are supposed to infer their uncertainties differently, by:
1. studying the spread of the different simulations
2. comparing the simulations with actual measurements
You left out:
3. By running sensitivity analyses, in which the uncertainty of the input data and the assumptions of the model are propagated thru the model in a way that determines the range and probability distribution of the model results that may arise from those factors. The spread of the results of such sensitivity analyses is one component of the error bars that should be presented on model prodictions. You would need to do this for each of the scenario runs in 1)
Such analyses are almost never performed, when performed are almost never published, and when published are almost never publicized. The cumulitive effect is indistinguishable from never.
http://www.scribd.com/doc/14434852/Terrestrial-Sources-of-Carbon-and-Earthquake-OutGassing-Thomas-Gold
They also noted that at the epicenter of the earthquake of May 14, 1970, an intensive release into the atmosphere of H2,CO2 and He was observed for a period of 40 days.Thus the observations in southern Dagestan give strong support to the view that there are deep and presumably abiogenic sources of hydrocarbons in the crust; that faults and earthquakes play a role in the escape of gases from great depths; and that gas emissions over regions of hundreds of km can all be affected by a single event. The authors suggest indeed that lithospheric outgassing of carbon through faults may be a major factor in the supply of surface carbon .
Anthony, in your reply you rightly point out that weather forecast errors grow with time, but this is completely different. I’m in no way saying the uncertainties shouldn’t grow with time, but the critical issue is *how* they grow. In any case my main question was about the collapse of the A,B, C scenarios onto one another, which makes no sense I can figure out. Frank doesn’t seem to have answered either question.
As for Antarctica — don’t worry, I’m not likely to buy anything you are selling!
Eric Steig;
In any case my main question was about the collapse of the A,B, C scenarios onto one another, which makes no sense I can figure out.
>>>>>>>>>>>>>>>>
What collapse? figure b is scaled so that the error range can be depicted in comparison to the scenarios themselves.
Terry Oldberg;
I sympathize with your predicament because I’ve been in the same situation more than once. Speaking Truth to Power is a risky business, and as you learned yourself, doing so is what gave rise to that oft repeated phrase “don’t shoot the messenger”. My advice to you is to craft the message to the audience. If you are one on one with someone who has the background in stats to understand the issues, by all means. If you are talking to a wider audience, “they said X would happen and it didn’t” is more effective and not innacurate.
P.S. to Pat:
I am not convinced that CO2 has ANY significant impact on gloabl temperature.
In 2008, I published that dCO2/dt varied ~contemporaneously with average global temperature, and atmospheric CO2 concentration lagged temperature by ~9 months. This CO2-lags-temperature observation is consistent with longer lag times observed (on longer cycles) in the ice core data.
I concluded in 2008 that temperature drives CO2, not the reverse. This conclusion is opposite to the conventional “wisdom” of BOTH sides of the rancorous global warming (CAGW) debate. I now predict that within ten years, temperature-drives-CO2 will be the newly accepted scientific premise of the climate science community.
Full article at :
http://icecap.us/index.php/go/joes-blog/carbon_dioxide_in_not_the_primary_cause_of_global_warming_the_future_can_no/
My paper was published two years before Murry Salby presented the same conclusion, with more supporting evidence, at:
______
Finally, I conclude that the entire global warming crisis has been a huge waste of scarce global resources (see prediction #3 above) that could have been used to solve real environmental and social problems that exist today, such as providing clean drinking water and sanitation systems in the third world.
In the ~25 years that the world has obsessed with alleged dangerous humanmade global warming, over 50 million children below the age of 5 have died from drinking contaminated water. That is more people (of all ages and from all sides) than were killed in World War 2. That is the “nominal” cost of the misguided obsession with global warming mania. And that reality also concerns me.
Regarding error bars, predictions need error bars. However, projections don’t have them.
By the way, in my initial response I neglected to thank Anthony for posting my essay. Forthwith the repair: thank-you Anthony! You’re a real sanity-saver!
Think of it – how would we all air our dismay, quantitative or otherwise, about what’s going on without Anthony and his like giving us public digital space? Without them, one is faced with either frustration-induced insanity or, in despair, abandoning the contest entirely — focusing down to protect one’s mental health.
So, thank-you Anthony. You’re doing more for personal sanity than any group of psychologists, ever.
And now . . . 🙂
Eric Steig, error bars grow when uncertainties are propagated into subsequently calculated properties.
When those calculated properties, e.g., climate variables, are projected forward by further calculations, the prior uncertainties are also propagated forward.
The standard method for propagated uncertainties in any serial calculation is sqrt{sum-over_i [(uncertainty)^2]_i]}, where “i” is each calculational step. See Bevington and Robinson “Data Reduction and Error Analysis for the Physical Sciences,” Chapter 3, Error Analysis, and especially pp. 39ff.
Hanson’s three lines don’t really collapse into one another. They’re as far apart in the “b” side as they are in the “a” side. The reason they look so close is that the ordinate of the plot in “b” has become ~40x wider than in “a” to accommodate the uncertainty bars.
Anthony is correct. Forecasts get more uncertain into the future, just as hindcasts get more uncertain into the past. Tree-ring paleotemperatures have no uncertainty at all, of course, because they’re not temperatures.
But now that you’re here, I’d like to ask you a temperate question: are you aware of the temperature-sensor measurement error analysis of H. Huwald, et al. (2009) “Albedo effect on radiative errors in air temperature measurements Water Resources Res. 45, W08431? They used sonic anemometers to measure the albedo-induced systematic error in an unaspirated R.M. Young probe mounted in a glacial snow field in the Swiss Alps, across two winters.
The measurement errors they found are non-normal and long-tailed to the high T side. The errors lower the resolution of any measurement data set to about (+/-)0.5 C. How did you discount the analogous systematic error that should have been in the measurement data sets you used for your Antarctic temperature reconstruction?
Their photographs, by the way, also showed wind-encrusted snow blocking the louvers of the RM Young shield. The consequent blockade of air-flow is another source of systematic error; one also relevant to Antarctica.
Bill Illis, thanks for the link to Troy Master’s paper on cloud feedback. Once again, it seems, the science proves to be unsettled. Sorry to say, it’s no surprise that Andrew Dessler “went to town” on it (in the Phil Jones sense).
Here is the zip-file supplement. FWIW, I downloaded the zip and scanned that copy with up-to-date Norton antivirus. It verified as virus-free.
Arno Arrak thanks for the description of Ferenc Miskolczi’s results. He’s come up with an interesting theory and I wish him a lot of luck. On the other hand, my essay wasn’t “boosting” the CMIP5 GCMs. It offers an error analysis.
Rud Istvan, as you know my analysis is very basic and incomplete. It looks only for an average lower limit of uncertainty.
A more thorough-going uncertainty analysis pertaining to clouds would compare GCM-projected and observed clouds at altitude and latitude globally, use data and theory to translate the differences into cloud feedback energy error, and use that in turn to construct an uncertainty estimate that reflected cloud error more specifically.
Maybe you’ll be in a position to give that a try, assuming all that data is actually in hand.
davidmhoffer, the IPCC also call their climate projections, “story lines,” which is an especially apt description. But you’re right, whatever they call them, the IPCC presents them as predictions. Predictions that they term “highly likely.” They are disingenuous to the max, and you’re also right that the intent is entirely polemical; meant to influence legislators and the public who are generally untrained in science.
george e smith, the if you can’t build it you don’t know it engineering approach to climate science! 🙂 Your point is well-taken, but after all no one can build a physical model of Earth. So, computer models are a necessity (as they are in all of science and engineering these days). Your implicit point is completely right that, given the large unknowns, much more modesty is required than is shown by climate modelers of the IPCC-stripe.
Philip Bradley, you put your finger right on a nub problem of the field, IMO.
davidmhoffer, you’re exactly right. GCM projections are offered as predictions, and they have turned out to not correspond to observations. That is a prima facie case of falsification.
It’s true, too, that the projections are not true predictions. They’re more like naive existential statements — the moon is made of green cheese. Such statements can be falsified by examination, but neither the statement nor the falsification constitute doing science.
As you know, all the supposed uncertainty limits on the AR4 WG1 SRES picture you linked merely show statistical model variability, given various inputs. They’re derived from numerical variances, not from physical errors or uncertainties. The lack of physically real uncertainties makes those projections physically meaningless. It can’t be that no one among the modelers or in the IPCC is unaware of that.
Terry Oldberg, in science a prediction is not, “an extrapolation from an observed condition to the unobserved but observable outcome of a statistical event.” Rather, it’s a deduction from falsifiable theory of an experimental or observational outcome. In terms of prediction, statistics has some scientific validity if it is applied in the necessitating context of a theory, such as in the distribution of photons in a double-slit experiment of QM. But direct statistical extrapolation of an observation is induction, which is a-scientific.
On the other hand, as you pointed out, projections are not predictions unless — following Steve McIntyre’s observation about climate and weather — they are. 🙂
LazyTeenager, the head post is about physically real uncertainties, not about numerical variances which are obtained by “studying the spread of the different simulations.” Strictly numerical variances have zip to do with physical uncertainty. Your number 1 objection is irrelevant.
Your number 2, “comparing the simulations with actual measurements,” is exactly what I did. You have no complaint here.
And you forgot number 3: Physical uncertainty estimates derived by propagating errors and (parameter) uncertainties through the model. I’ve never seen such a study published.
LazyTeenager next post: not correct. Serial calculations of physical quantities require serially propagated physical errors.
Allan MacRae, thanks. 🙂 I missed the aerosol thread, but seem to recall reading about Doug Hoyt’s aerosol work previously; maybe on the website he used to maintain.
Doug is very welcome to comment here; he’s always been a rational voice. In any case, everyone is welcome to comment here.
JJ, I should have just linked to your comment when replying to LazyTeenager. 🙂
Eric Steig, I’ve now replied to both of your questions. Also note davidmhoffer’s explanation. He’s exactly right.
If a “critical issue” is “*how*” physical uncertainties grow in a climate projection, why has no one published on that, ever, anywhere, any time?
Note my response to Robert Brown, in which it’s pointed out that GCMs merely linearly propagate GHG forcing, demonstrated here. That makes error propagation very straight-forward.
Allan MacRae, regarding CO2, I tend to agree in that there certainly has been no evidence of any influence on temperature.
In May 2010, JeffID was kind enough to post study I did on 20th century climate sensitivity. It had turned out there was a cosine component in global temperatures that had appeared when the land surface air temperatures were subtracted from the global surface air temperatures (Figure 1 here). The periodic temperature dependence represented by the cosine function was present in the SSTs, which imposed the thermal signal globally.
The entire air temperature anomaly trend since 1880 can be fit with that cosine (period 60 years) plus a positive-slope straight line. That line represents the warming trend. Looking in detail at the straight line component, a small increase in slope can be found for 1951-2010 relative to 1880-1950. Assigning the positive difference to GHG produced an estimate of climate sensitivity = 0.34 C/doubling of CO2, or 0.1 C/W-m^-2 of extra forcing. Not much to worry about.
Your comments are exactly on target regarding the huge waste of money and the terrible immorality of global warming environmentalist extremism. The waste of resources and the enriching of PR groups and eNGO lawyerly fat cats at the expense of amelioration of real problems, with the willfull collusion of scientific societies, will be remembered as the most widespread and systematic betrayal of ethics, ever.
Pat Frank:
Thank you for for sharing your views on the important issues that you have raised with respect to my various remarks. My response follows.
It sounds as though you have concluded from my use of the term “statistical event” that my remarks pertain only to statistical models. This conclusion is mistaken. My remarks pertain to models that are purely mechanistic, to models that are purely statistical and to models that mix the statistical and mechanistic approaches to building models. The three types of model have in common that they rely upon the existence of observed statistical events for the statistical testing of them. The statistical and hybrid approaches rely upon the existence of observed events for the additional purpose of building the associated models.
As I define the terms, a “condition” is a condition on the Cartesian product space of a model’s independent variables and is a proper subset of the tuples in this space. A meteorological example of a “condition” is “cloudy.”
An “outcome” is a condition on the Cartesian product space of a model’s dependent variables and is a proper subset of the tuples in this space. A meteorological example of an “outcome” is “rain in the next 24 hours.”
A condition and an outcome describe a statistical event. For example, “cloudy, rain in the next 24 hours” describes an event. A “prediction” is an extrapolation from a condition to an outcome. For example, it is an extrapolation from the condition “cloudy” to the outcome “rain in the next 24 hours.” At the time a prediction is made, the condition has been observed; the outcome has not been observed but is observable.
The complete set of events that are referenced by a study is a kind of statistical population. Each of the possible outcomes of an event has a relative frequency. In testing a model, one compares the predicted to the measured relative frequencies. If there is not a match, the model is falsified.
The links between the deductive logic, the inductive logic and science can be uncovered with the help of information theory. The English word “science” makes ambiguous reference to the associated ideas. One of these ideas is “demonstrable knowledge.” It is convenient to label this idea by “scientia,” the Latin word for “demonstrable knowledge.”
In information theoretic terms the “scientia” is the mutual information between a pair of state-spaces. One of these state-spaces contains the states that I have called “conditions.” The other state-space contains the states that I have called “outcomes.” The “scientia” is the information that one has about the outcome given the condition. Conversely, it is the information that one has about the condition given the outcome.
It can be shown that the inductive logic differs from the deductive logic in the respect information for a deductive conclusion is missing for the former branch of logic but not the latter. Thermodynamics and quantum mechanics are examples of models for which information about the outcome, given the condition, is missing; both are regarded as “scientific.” Thus, it is inaccurate to claim that a model built under the inductive logic is “a-scientific.” A model is properly described as “scientific” if and only if it conveys “scientia” to its users.
Finally, I’d like to address your contention that the climate has a “climate sensitivity ” (CS). This contention references a pair of state-spaces. One contains all of the possibilities for the change in the equilibrium temperature at Earth’s surface. The other contains all of the possibilities for the change in the logarithm of the atmospheric CO2 concentration. The mutual information between the two state-spaces is not defined, for the equilibrium temperature is not an observable. Thus, the change in the logarithm of the CO2 concentration provides no information about the change in the equilibrium temperature. It provides no “scientia.”
Allan, thanks for the link to the video. Murray Salby ‘s work on the source of atmospheric CO2 looks very important. Has he published it anywhere? I didn’t find anything using Google Scholar, and there’s nothing in his on-line publication list.
Thanks for your comments Pat,
Perhaps not coincidentally, prior to my January 2008 icecap.us paper, my best estimate of “climate sensitivity ” to a doubling of atmospheric CO2 was 0.3C, versus your 0.34C.
Close enough, I guess. 🙂
P.S.
My immediate concern, which I apologize for carping about yet again, is the use of 40% of the huge USA corn crop for gasoline additives. Due to the drought this season, corn now costs over US$8 per bushel – and corn is a staple for many poor people in the Americas.
This situation is simply wrong – it is a monstrous ethical and humanitarian failing, and our leaders in the USA and Canada should have the courage and integrity to end the fuel ethanol mandate immediately.
Pat Frank says: August 26, 2012 at 6:48 pm
Re your question Pat:
I read that Murry Salby was writing a book on this subject.
Since 2008, there has been considerable discussion on climateaudit.org , wattsup and elsewhere on this subject, particularly between Richard S Courtney and Ferdinand Engelbeen. Ferdinand holds to the “mass balance argument” regarding the role of humanmade CO2 emissions in the increase in atmospheric CO2, whereas Richard (and I) say this argument is not necessarily correct.
I also recall someone admonishing me that the C13/C12 ratio disproved my hypo – I recall it took me about twenty minutes of serious investigation to dismiss that notion. Then someone said the C14/C12 ratio would disprove it – and frankly I could not be bothered to check, since I was a bit annoyed at the time by the C13/C12 nonsense. Hope Murry Salby is checking C14/C12.
Here is a more recent (2012) presentation to the Sydney Institute by Salby – it takes a while to load:
http://podcast.thesydneyinstitute.com.au/podcasts/2012/THE_SYDNEY_INSTITUTE_24_JUL_MURRY_SALBY.mp3
I think there are a few people who may have a pretty good grasp of this situation – perhaps Jan Veizer and close colleagues.
Best, Allan
@ur momisugly Pat Frank
“””””…..george e smith, the if you can’t build it you don’t know it engineering approach to climate science! 🙂 Your point is well-taken, but after all no one can build a physical model of Earth. So, computer models are a necessity (as they are in all of science and engineering these days). Your implicit point is completely right that, given the large unknowns, much more modesty is required than is shown by climate modelers of the IPCC-stripe. …..”””””
Don’t have a problem per se Pat, with computer models; I use them constantly for 10 or more hours every day. So my clients can go from my modelled projections; excuse me “predictions”, to a manufactured finished product, with hard tooling; and I have never had one not perform as predicted. The difference is the computer “models” I use ARE built around the applicable Physical laws, and actually, in a hierarchy of complexity and rigor. For example, you don’t need Einstein general relativity, or special, to learn how to drive a car in a fuel economical way. Newton’s laws will suffice..
So I don’t have to build physical models, since the available mathematics, describes the observed Physics, to well beyond the precision needs of my clients’ product needs.
But I put in math representations of the physical objects I plan for my customer to use, placed where I want him to place them, and then run it. I don’t make any prior judgement, as to how they will interract; the computer will show me that. If I don’t like the result, then Ichange something.
Of course it helps to be thoroughly grounded in the theory, so that I can make the changes in an intelligent way, and expect the behavior to be perturbed by about the right amount.
If astronomers can model stars from birth to death, I don’t see why earth’s clouds can’t be modelled so they behave as observed, without having to decide first whether they go into the model as a positive feedback, or negative, or ANY feedback at all. Put them in so they obey known physical laws; not homogenizedanomalies.
george e smith says: August 26, 2012 at 11:18 pm
Hello George,
If I may interject: Your models are apparently built with integrity, whereas the climate models are not.
The climate models have been an exercise in pseudo-scientific fraud, in my opinion.
For example, the climate models falsely assume a sensitivity of climate to a doubling of CO2 that is about an order of magnitude too high (3C versus 0.3C). The modelers then justify this false high sensitivity by assuming false manufactured aerosol data that force-fits the hindcasting of the global cooling that occurred from ~1940 to ~1975. This is perhaps better explained in the 2009 post below.
As for the modeling of clouds, I’m not sure we understand the physics of clouds well enough to model them, even at a first-order approximation of their complexity.
Finally, this entire post assumes that atmospheric CO2 drives temperature. I came to the conclusion in 2008 that temperature drives CO2. I was not the first to reach this conclusion. The following quotation is from “Celestial Climate Driver: A Perspective from Four Billion Years of the Carbon Cycle” (Jan Veizer, Geoscience Canada, Volume 32 Number 1 March 2005):
“Again, while CO2 may act as an amplifying greenhouse gas, the actual atmospheric CO2 concentrations are controlled in the first instance by the climate, that is by the sun-driven water cycle, and not the other way around.”
__________________
http://wattsupwiththat.com/2009/06/27/new-paper-global-dimming-and-brightening-a-review/#comments
Leif Svalgaard (13:21:18) :
Allan M R MacRae (12:54:27) :
“There are actual measurements by Hoyt and others that show NO trends in atmospheric aerosols, but volcanic events are clearly evident.”
But increased atmospheric CO2 is NOT a significant driver of global warming – that much is obvious by now.
Leif: But what has that to do with aerosols?
**************************
Leif, I did say:
“The sensitivity of global temperature to increased atmospheric CO2 is so small as to be inconsequential – much less than 1 degree C for a doubling of atmospheric CO2. CO2 feedbacks are negative, not positive. Climate model hindcasting fails unless false aerosol data is used to “cook” the model.”
Connecting the dots, to answer your question:
The fabricated aerosol data allows climate model hindcasting to appear credible while assuming a false high sensitivity of global temperature to atmospheric CO2.
The false high sensitivity is then used to forecast catastrophic humanmade global warming (the results of the “cooked” climate models).
What happens if the fabricated (phony) aerosol data is not used?
No phony aerosol data -> no false-credible model hindcasting -> no artificially high climate sensitivity to CO2 -> no model forecasting of catastrophic humanmade global warming.
Regards, Allan
Allan, thanks for the further information and the updated video. I’d sure like to see him write a paper on the subject. If his thesis proved out, it would completely undercut the entire paradigm of blame.
He’s also the first researcher I’ve seen mentioning the recent IBUKI satellite results.
george e smith, honestly I don’t have any trouble with computer models either, George. From the sound of it, you’re developing and using well-verified engineering models.
Engineers have to pay close attention to detail developing such models, mapping out their response to high accuracy within the required specifications (boundary conditions). They have to be sure that the models give physically accurate predictions, validated by performance. I have nothing but respect for that business.
One of my brothers is a double-E, and uses engineering models to develop signal-processing ICs that go into microwave devices. Large amounts of money, not to say safety, depend on those working as advertised.
I use models in research, too, and am aways sweating uncertainties. Climate modelers seem to have lost perspective, treating their models as physically complete and accurate, so that nothing more is needed than to map out the response phase-space. In that supposition they’re being naive (or incompetent) to an astonishing degree.
Several years ago, Jerry Browning discussed climate models extensively over at Climate Audit. He described how small errors grow exponentially, making predictions impossible. He also discussed the modelers suppressing sub-grid turbulence by imposing a hyperviscous atmosphere, noting that a “molasses atmosphere” necessitated unphysical parameterizations. The whole climate model business is extremely shaky and it’s beyond knowing how that is so thoroughly unremarked in the profession or the journals.
Terry Oldberg, thanks for the interesting conversation. My comment started from your definition of “prediction” as being the extrapolation of an observation, i.e., your ““prediction” defined as an extrapolation from an observed condition to the unobserved…” That sounds just like inductive inference to me, which is non-scientific. If you had meant prediction to be an, ‘extrapolation from an observed condition to the unobserved by means of deduction from (a scientifically valid) theory,’ I’d have no problem.
As it is, I don’t really know what you mean by “model.”Your description seems to mean an ad hoc mathematical system that reproduces known past behavior? Such models can’t make predictions. On the other hand statistical models can derive probabilities of future behavior, but only based upon the assumption that the future is like the past.
Such models are useful, but they don’t really follow the scientific method. Regarding statistical models, you wrote that, “If there is not a match [between observed and predicted outcomes], the model is falsified.” But whats really falsified in that case is the assumption of regularity. The statistical model itself has not been falsified. One merely needs to adjust the model coefficients to the statistics of the new conditions. In science, the model itself would be falsified if observations contradicted the predictions.
You wrote that, “Thermodynamics and quantum mechanics are examples of models for which information about the outcome, given the condition, is missing …” I presume you mean Statistical Thermodynamics, as that completes the analogy with QM.
But in any case, both theories are completely deterministic in that their states evolve in a completely deterministic way, and their predicted outcomes are also completely determined by theory. In QM, the observable of an individual scattering event can not be predicted, but Heisenberg’s uncertainty statement makes that inability a function of theory, not of information. That is, information cannot be said to be missing when the theory itself specifies that information cannot exist. Beyond the quantum horizon, “information” has no meaning and so supposing it is “missing” is to make a category mistake. One cannot say something is missing when it cannot in principle be present.
In chemical physics, the trajectories of individual atoms can be predicted, e.g., in a Penning trap, and also see, but then, of course, Statistical Thermodynamics no longer apply.
You’re right that I was careless in describing a “climate sensitivity.” The term is common parlance in climate science, and I did use it in the conventionally off-hand manner. Many people have noted that the climate is always out of equilibrium. It would better be called a dissipative system in the far-from-equilibrium sense of Prigogine; one that can transition between many quasi-stable states.
I meant the concept only in terms of a perturbation of a magnitude to cause a response in the local state energy content without causing a state transition. You were right to call me on it.
In other news, Eric Steig has not answered my question about how he discounted systematic temperature sensor measurement error in his study of Antarctic temperature trends.
Pat Frank:
Thank you for the stimulating response. In my definition of the term, a “model” is a procedure for making inferences. As logic contains the principles for discrimination of correct from incorrect inferences, by this definition of the term one ties the idea that is referenced by the term “model” to logic. To do so is to set ones self up for discovery of logical errors in the methodology of an inquiry. To do so is also to set one’s self up for specification of a methodology for a scientific inquiry that is logically flawless. I found the latter feature quite useful in a period in which I earned my living through the design and management of scientific inquiries.
The scope of the deductive logic is generally inadequate for the task, but a generalization of it proves adequate. The generalization replaces the rule that every proposition has a truth value by the rule that every proposition has a probability. This generalization produces the “probabilistic logic.” In this logic, it is easy to show that every inference has a unique measure. The measure of an inference is the missing information in it for a deductive conclusion, the so called “entropy.” This finding arms the searcher for methodological error with the machinery of information theory.
In the philosophical literature, the realm lying outside the deductive logic is called the “inductive logic.” It follows from the existence and uniqueness of the entropy as the measure of an inference that the inductive logic differs from the deductive logic in the respect that the entropy of an inference is non-nil. In thermodynamics, the entropies of the inferences are generally non-nil and thus the logic of thermodynamics is generally inductive.
A caveat applies to what I have just said. The idea of a “probability” implies that every set bearing a probability is crisply defined but as it turns out the crispness of these sets is antithetical to the existence of a quantitative independent variable such as the global average temperature at Earth’s surface. This shortcoming may be overcome through a generalization that incorporates Zadeh’s fuzzy set theory. Zadeh’s theory preserves the idea of entropy while generalizing the idea of probability such that the generalized probability is defined on fuzzy sets.
Using these ideas, I’ve initiated a search for errors in the methodology of the IPCC’s inquiry into global warming. This search has uncovered errors that are fatal to the IPCC’s pretensions to having conducted a scientific investigation. The presence of these errors is, however, obscured by the ambiguity of reference to the associated ideas by terms in the language of climatology. The ambiguity of reference yields negation of Aristotle’s law of non-contradiction. Using the negated law as a false but apparently true premise to a variety of arguments, authors of IPCC assessment reports make arguments whose conclusions seem to the general public to be true but whose conclusions are actually false or unproved. Among the stratagems used by IPCC authors in achieving this end is to conflate the idea that is referenced by the term “projection” with the idea that is referenced by the term “prediction” by using the two terms as synonyms.
I’ve just discovered that four days ago, Michael Tobis called me “tedious” and a purveyor of “falsehoods” in a post that also linked me with pseudoscience, confusionism, and garbage. Apart from that, he offered no actual critique of my cloudiness error analysis.
I’ve invited MT to dispute me here (we’ll see whether the invitation survives moderation), where neither of us will be censored. So far, I’ve had two experiences debating Micheal on blogs he controlled. Neither experience encourages confidence that freedom will prevail on his Planet 3.0
Terry Oldberg, logic has an invaluable part to play in the validity of scientific inferences, it’s true.
Every theory in science must be internally self-consistent in the logical sense as you mean it. Typically, logical self-consistency is achieved by coding the theory in mathematics. A basic conflict is revealed by this within Physics, which is that QM is expressed in discrete mathematics whereas Relativistic Mechanics requires continuum mathematics. The choice of mathematical expression is not arbitrary, of course, nor axiomatic, but rather imposed by the results of experiment. That observational foundation frees the conflict from Philosophy.
The freedom of Science from Philosophy leads me to the source of a need to part ways a bit from your analysis. The various logics you describe are all useful; most especially when assessing polemics. Polemical views are always based in some assumed view, and philosophical analysis is probably the best way to reveal that. Hence your success examining the rhetorical flourishes of the IPCC.
I completely agree with your point that the IPCC tendentiously conflates/distinguishes prediction and/from projection, whenever it suits their fancy. This, their tendency, is mentioned in passing in my Skeptic article.
I also agree that deployment of logic can be very useful in determining the validity and coherence of a scientific project.
With regard to the inadequacy of deductive logic, I believe that most scientists would agree with you. The fact that any theory is incomplete and therefore unable to yield completely specified predictions is well-known. This lack always leaves open the possibility of ad hoc adjustments to avoid falsification.
The persistent availability of ad hoc adjustments strictly removes what you called the “unique measure” of the probability of the expectation value of a theory (the calculated outcome or prediction). Employment of Ockham’s Razor gains its utility here.
When discussing science, it’s necessary to be very careful with the use of terms. Physical “entropy” is the the number of states among which a system may distribute itself. Typically the greatest entropy includes large numbers of degenerate (equipotential) states. But you have used “entropy” in the Shannon sense. I consider that use an illicit absconding of the term, based merely on the equivalent mathematical structure of physical entropy and Shannon’s information loss.
Consider this pair of statements (H/T Bill Wattenberg): “Time flies like an arrow. Fruit flies like a banana.” Equivalent syntax, orthogonal meanings. Such are physical and Shannon entropies.
If I understand your distinction of inductive and deductive logic, your statement is that an inductive inference is “non-nil,” meaning that every inductive inference has a unique measure, which in your terminology seems to also be a unique probability measure.
As I see it, the uniqueness of your probability measure derives entirely from the axiomatic basis from which your inductive inference proceeds. If that’s true, then the non-missing information in your inference stems from the fact that the necessary quantity of information has been assumed. That is, like mathematics, your certainty of consequence — your measure of probabilistic uniqueness — comes from the fact that your conclusions are contained in your assumptions. The logic is thoroughly and completely rigorous, but ultimately nothing is new even if it may be surprising.
Science, however, is anaxiomatic in the sense that no assumptions are invariantly held as true. Science derives its theories irrationally, i.e., despite insufficient information, even though they are required to be expressed logically and coherently, and then tests them using observations. Discoveries of science can therefore, and uniquely, be both new and surprising.
The “missing information,” due to the lack of grounding axioms and leading to theory invention and to deductive logic, is a strength of science. It allows knowledge to progress in the face of ignorance.
Thermodynamics, prior to the Statistical Thermodynamics that grew out of Atomic Theory, was almost strictly an observational science. The Carnot Cycle was derived inductively, for example, and formed the basis for 19th century thermodynamics. That makes its inferences probabilistic in the sense you perhaps mean.
But virtually all of classical thermodynamics can now be derived from Newtonian Physics, Statistical Thermodynamics, and Maxwell’s Theory.
That is, classical thermodynamics is now a deductive inference, even though Statistical Thermodynamics can make what you call non-nil inferences, i.e., unique probabilistic statements.
At the end, I’m not saying your analysis is wrong; only that it is in some cases incomplete with respect to the methodological structure of science.
As an interesting convergence, one might see your “ambiguity of reference yield[ing] negation of Aristotle’s law of non-contradiction” with reference to the IPCC, also in the consistent lack of physical error analysis in the IPCC’s representations of future climate. The representations are presented as physics, but the consistent lack of physical uncertainty bounds makes them physically meaningless. I.e., they appear to be self-contradictory in the Aristotelian sense you describe. My analysis is a first-step attempt to resolve the contradiction.
By the way, Chris Essex & co published a paper showing that global average temperature is not a physical quantity at all, and has no sure physical meaning. You may enjoy it: C. Essex, R. McKitrick, and B. Andresen (2007) Does a Global Temperature Exist?” J. Non-Equilibrium Thermodynamics 32, 1–27 Here’s the abstract page. The usual AGW suspects derided it, but the thermodynamic idea is sound, namely that temperature is a physical observable and not a statistic.
I’d like to offer two points in clarification and three examples. First point of clarification: in logic, there is not a distinction to be made between the physical and Shannon entropies; they are simply different applications of the same concept.
Second point in clarification: the entropy is the unique measure of an inference in the deductive as well as the inductive branches of logic. In the deductive branch, the value of the entropy is nil while in the inductive branch it is not-nil.
First example: Cardano’s theory of fair gambling devices offers a mathematically simple application of the inductive logic. Under the principle of entropy maximization, the entropy of the various ways in which the outcome can occur is maximized, under constraints expressing the available information. In Cardano’s theory there are no constraints. Maximization of the entropy without constraints assigns equal numerical values to the various ways. For example, it assigns 1/6 to each of the 6 ways in which an outcome can occur in the throw of a die.
Second example: Thermodynamics results from an application that is similar but that adds a constraint. The ways in which an outcome can occur are the “microstates” and the constraint is energy conservation. Maximization of the entropy under this constraint (the second law of thermodynamics) assigns equal numerical values to the probabilities of the microstates that satisfy energy conservation, the so-called “accessible microstates.”
Third example: In other applications it is logical to minimize the conditional entropy of one type of inference and to maximize the entropy of a different type of inference. This approach yields, for example, Shannon’s mathematical theory of communication.
Terry, I accept your examples. My only point is that logic is not science, although science uses logic as a critically important tool.
The observational foundations of science means that Shannon entropy is not physical entropy. It’s very central to realize that equivalence in mathematical structures does not equal equivalence of meaning.
In your post of 27 august, you wrote that, “It sounds as though you have concluded from my use of the term “statistical event” that my remarks pertain only to statistical models. This conclusion is mistaken. My remarks pertain to models that are purely mechanistic, to models that are purely statistical and to models that mix the statistical and mechanistic approaches to building models.”
And also that, “A condition and an outcome describe a statistical event. … The complete set of events that are referenced by a study is a kind of statistical population.”
I believe these comments are central to your approach, and would like to express the point of view of a scientist. The point will not contradict your views, but does move science to its proper place outside of them.
That is, the statistical analysis of scientific outcomes is posterior to the construction of the theory and the measurement of the data. The final decision about application of statistics, and the determination of what form the statistics should take, comes after the anticipations of theory and after evaluation of the form taken by the data.
The emergent values of data, in particular, are independent of prior expectations. It is for this reason that science is not, and can not be, axiomatic.
Therefore, whereas “a condition and outcome [can] describe a statistical event,” under the methodology of science, they need not. That is, statistics is not inherently necessitated by science, nor is science a branch of statistics.
The array of outcomes — observables — following an experiment are predicted by a theory that describes the partitioning of energy states, with consequent and dependent material transformations. These, as described by theory, may produce an ensemble of outcomes that is statistically analyzable. However, the condition of able to be analyzed by statistics follows from the content of theory and and the structure of the observation, rather than from any a priori necessity stemming strictly from statistics or logic.
That is, statistics and logic themselves do not determine the structure of science, no matter that they have turned out to be useful in analyzing the structure of science following its emergence.
Therefore, the success of your approach as regards science, starting as it does from logic and a statistical formalism, depends for its power on the pre-existent and independent presence of statistical elements in the theory and results of science. As there are many pre-existent elements in physical theory that have been found to follow a statistical order, from subatomic theory all the way up through Biology, your approach from logic and statistics will have success.
However, it must be kept in mind that the success of your approach to science follows after, and is dependent upon, the pre-presence of statistical elements throughout science. These statistical elements in science arose because they were demanded by the structure of the ensemble of observations.
There is no prior necessity that observations should have taken that structure nor any prior necessity that science should have followed certain logical or statistical rules.Terry, I accept your examples. My only point is that logic is not science, although science uses logic as a critically important tool.
The observational foundations of science means that Shannon entropy is not physical entropy. It’s very central to realize that equivalence in mathematical structures does not equal equivalence of meaning.
Pat Frank:
Thanks for sharing your views.
A point of commonality between the “physical entropy” and the “Shannon entropy” is that both are the unique measure of an inference. Logic contains the rules that discriminate correct from incorrect inferences. Thus, from a logical perspective, there is not a difference between the physical and the Shannon entropy.
Pat Frank:
While one can draw a distinction between the “physical entropy” and the “Shannon entropy,” there is not a difference between the two concepts. In both cases, the concept that is referenced by the term is the missing information about the unobserved state per event.
You’ve raised the issue of the relationship between logic and science. Ambiguity of the English word “science” muddies the waters in attempts at addressing this issue. To clarify these waters, I’ll substitute the Latin word “scientia” for “science.” “Scientia” means “demonstrable knowledge.”
In information theoretic terms, the “scientia” is the information per event that is provided by the model about the unobserved but observable state, given the observed state. The entropy is the information per event that is NOT provided by the model about the unobserved but observable state of an event given the observed state. The entropy and the scientia are related in this intimate way.
Logic is related to entropy and thus to scientia by the facts that: a) its entropy is the unique measure of an inference and b) logic contains the principles that discriminate correct from incorrect inferences. In a 49 year old advance, these principles were discovered: In a set of inferences that are candidates for being made by a model, the correct inference is the one that minimizes or (dependent upon the type of inference) maximizes the entropy.
Academia has not yet caught up with this advance. A consequence is for most model builders to discriminate correct from incorrect inferences in the old fashioned way through the use of heuristics. A result from widespread use of heuristics is for most of the models that we use today to be riddled with logical errors. One has only to poke around in these models a bit to find the errors.
Moderator, the last sentence of my prior post was somehow followed again by the first four sentences (two small paragraphs). If you could please remove them, and this post too, thanks very much.
As of this writing, Michael Tobis has not responded to my invitation to refute the head-post analysis, given that he had called the analysis misleading and me tedious.
Nor has Eric Steig responded to my request that he illuminate how he had apparently discounted 100% of the systematic temperature sensor error in his Antarctic work.
Terry Oldberg
You wrote, “In information theoretic terms, … the entropy is the information per event that is NOT provided by the model about the unobserved but observable state of an event given the observed state.
That’s fine, Terry, but entropy is not “missing information” in physical theoretic terms. It’s about the dispersal of particles, including photons, among available microscopic energy states.
It’s clear that the equivalence of the mathematics makes seductive the redefinition of physical entropy in terms of information. After all, we humans think of knowledge in terms of accurate descriptions of discrete events, and entropy is defined probabilistically.
But the probabilistic description doesn’t mean information is missing. The probability distribution is the information. No other information is possible, therefore no information is missing.
You’re clearly wedded to the information theoretic account of things, as I am to the physical theoretic account of things. Your adherence to information theoretic axiomatics requires that you describe entropy as missing information. But axiomatic conclusions are necessarily self-referential, for reasons noted above.
Physical theory is not axiomatic, in the sense that the reliance on observation means that the system is not closed. Science is not self-referential for that reason.
For all the power of your logic, its application to science is happenstantial. Physical science has developed in a way, such that information theory has an accidental congruence with the structure of science. That accident has conferred an analytical power.
But I believe it’s a mistake to push the analogy too far, to the point of redefining the meaning of physical concepts so that they are extruded into an information theoretic mold. Doing so only sows confusion. Logic is not related to physical entropy. In no physical theory does logic enter as a calculational term.
Pat Frank:
Thank you for taking the time to respond. I have to point out that when you claim that “the probability distribution is the information,” your claim is inconsistent with the information theoretic definition of “information.” When the “information” and the “missing information” are properly defined, a logic emerges from these definitions. One of many applications of this logic is the second law of thermodynamics. As the physicist Edwin Jaynes realized in his 1957 paper “Information theory and statistical mechanics,” the second law can be regarded as an application of the generally applicable principle called “entropy maximization.” Though it was first discovered in empirical data, entropy maximization can be mathematically derived from its roots in measure theory and probability theory