The Rest of the Cherries: 140 decades of Climate Models vs. Observations
by Roy W. Spencer, Ph. D.
Since one of the criticisms of our recent Remote Sensing paper was that we cherry-picked the climate models we chose to compare the satellite observations of climate variations to, here are all 140 10-year periods from all 14 climate models’ 20th Century runs we analyzed (click to see the full res. version):
As you can see, the observations of the Earth (in blue, CERES radiative energy budget versus HadCRUT3 surface temperature variations) are outside the range of climate model behavior, at least over the span of time lags we believe are most related to feedbacks, which in turn determine the sensitivity of the climate system to increasing greenhouse gas concentrations. (See Lindzen & Choi, 2011 for more about time lags).
Now, at ZERO time lag, there are a few decades from a few models (less than 10% of them) which exceed the satellite measurements. So, would you then say that the satellite measurements are “not inconsistent” with the models? I wouldn’t.
Especially since the IPCC’s best estimate of future warming (about 3 deg C.) from a doubling of atmospheric CO2 is almost exactly the AVERAGE response of ALL of the climate models. Note that the average of all 140 model decades (dashed black line in the above graph) is pretty darn far from the satellite data.
So, even with all of 140 cherries picked, we still see evidence there is something wrong with the IPCC models in general. And I believe the problem is they are too sensitive, and thus are predicting too much future global warming.
James Sexton says:
September 22, 2011 at 5:20 pm
“Wouldn’t the title be more apt if it read “Spencer b-slaps pedantic critics!”?”
You are projecting, James, which is forgivable in this case. Spencer’s style is to get all the critics in a circle on the floor and use tinker toys, if necessary, to explain the points to them.
Theo Goodwin says:
September 22, 2011 at 8:12 pm
James Sexton says:
September 22, 2011 at 5:20 pm
“Wouldn’t the title be more apt if it read “Spencer b-slaps pedantic critics!”?”>>>
You are projecting, James, which is forgivable in this case. Spencer’s style is to get all the critics in a circle on the floor and use tinker toys, if necessary, to explain the points to them.>>>
I don’t give a tinker’s d***, nor do I mind
If he b-slaps them in circles… or in lines
Watever it takes to make the sound of their noise
About needing yet more high tech toys
For studies and papers and getting editors sacked
For publishing facts! Actual facts!
Its not about science, because science is lax
Unless it produces something to tax!
“DM says:
September 22, 2011 at 6:40 pm
Climate models have to be based on some physical theories and hypotheses, otherwise you couldn’t put together the code for them to run. The fact that the climate models aren’t doing a good job suggests there’s something missing/wrong with the underlying theories and hypotheses”
I agree completely! So exactly what is missing or wrong? How do they come up with 3 C for a doubling of CO2? My understanding is that they assume the CO2 alone causes a 1 C increase. Then they add positive feedbacks and say that warmer air causes more water vapor and since water vapor is the strongest greenhouse gas, the 1 C increases to 3 C. But this violates one of the best known chemical laws, namely Le Chatelier’s Principle. It seems as if the feedbacks are negative due to more clouds, assuming CO2 initially actually does cause warming as the IPCC claims. A valid question naturally would be whether or not Le Chatelier’s Principle applies here. I would suggest that all evidence to date suggests that it does apply and the IPCC models are wrong because they did not incorporate this basic law into their models.
80,000 years from now, as the Next Ice Age is nearing an end, Kruug the Caveman attempts 14 times to reproduce the pattern on an Etch-A-Sketch he unearthed in his cavernous dwelling.
Frustrated and disgusted, he takes his mighty Club to it, muttering to himself to stick to painting murals from now on. His wife agrees, and presents him with a hearty slice of roasted Gore Boar.
Mann Dog wags his strangely bent tail, anxiouly awaiting a scrap or two.
Why bother with computer games? Real temperatures are available from UAH and RSS satellites that measure Oxygen microwave radiation.from lower troposphere.
Theo Goodwin says:
September 22, 2011 at 3:27 pm
“Do you really think that what you use to launch and manage satellites is a model?”
Theo: Your question and subsequent comment display serious philosophical and practical problems. ANY prediction of the future is based on a model, whether explicit or implicit. Furthermore, this model must be a simplification of the real system about which the prediction is made — otherwise we would need a full duplicate universe (but run in faster time…).
Now physical “laws” may be applied to that model, as in Newton’s laws applied to AJ’s orbital systems. But that in no way means that absolutely accurate predictions can be made. (We now know that Newton’s laws are only an approximation, and they failed to predict key features about, for example, Mercury’s orbit.)
Maybe AJ’s model treated the Earth as a perfect sphere of uniform density. For all of my engineering purposes, that model is sufficient for my needs. It certainly is not good enough for the purposes of the GRACE satellite, which depends on gravitational variations. It’s not good enough to predict the precession of the Earth’s axis. Whether AJ’s models assume this or not, their model of the earth and its gravitational attraction is certainly an imperfect simplification.
There are other factors that need to be modeled as well. Satellites in low-earth orbit experience slight atmospheric drag. Do the models assume this drag is constant? (It’s not.) If not, how well can they predict how it changes? Will the drag have a tendency to twist the satellite? How will this affect the future orbit?
AJ says his models are insufficient to predict the trajectory to sufficient accuracy more than about a week out. Hmm — sort of like weather models.
Our weather/climate models are good enough to predict reliably that summers will be warmer than winters, that days will be warmer than nights. I don’t believe they have much long term predictive ability as to weather a given summer in a given region will be warmer or cooler than average, or whether a decade will be warmer or cooler than average globally. It’s all about how good the models’ predictive abilities are compared to the quality of information we need out of them.
davidmhoffer says:
September 22, 2011 at 6:04 pm
Excellent point. Cuts right to the basics in science of philosophy famously stated by Karl Popper. In particular the concept of falsifiability. Briefly, for those unfamiliar with it, a scientific hypothesis need not be provable but it must be able to be, at least in principle, disproven . Proof is often impossible but disproof is often accomplished by a single observation. Popper constructed a famous example “The Black Swan”. He hypothesized that all swans are white. This can never be proven because no matter how hard you unsuccessfully search for a non-white swan you can’t possibly guarantee that didn’t miss one somewhere. On the other hand a single observation of a black swan will falsify the hypothesis. Hypotheses eventually become theory as observations in support of it accumulate without any observations that falsify it.
So called climate change is a hypothesis. The notorious global circulation models are nothing more than the hypotheses stated in the language of computers and mathematics. They are not tests of the hypotheses but rather tools to extract predictions from the hypotheses. The predictions are then tested by comparison actual observation. The hypotheses will never become facts but they may someday become theory. More likely IMO is they’ll become hypotheses that were tested and falsified. Many would argue they have already been falsified.
err… that should have been “philosophy of science” in my previous not “science of philosophy”
Dave Springer check your email
Athony-
Can this be put in the “Climate FAIL Files?”
Curt says:
September 22, 2011 at 8:56 pm
Theo Goodwin says:
September 22, 2011 at 3:27 pm
“Do you really think that what you use to launch and manage satellites is a model?”
“Theo: Your question and subsequent comment display serious philosophical and practical problems.”
Nope, sorry, but I am right on the money.
“ANY prediction of the future is based on a model, whether explicit or implicit. Furthermore, this model must be a simplification of the real system about which the prediction is made — otherwise we would need a full duplicate universe (but run in faster time…).”
You said absolutely nothing there except prediction is based on a model. Explain the concept of prediction you are using. You seem to think that a model is a substitute for the system studied or a simplification of that. Computer models are not toy airplanes.
How do you apply physical laws to a model? Do you even know the difference between physical hypotheses and a model? Can you state those differences?
“Maybe AJ’s model treated the Earth as a perfect sphere of uniform density. For all of my engineering purposes, that model is sufficient for my needs. It certainly is not good enough for the purposes of the GRACE satellite, which depends on gravitational variations.”
I understand that you have a sophisticated computer program that you use and I am sure you put it to good use. However, when you attempt to use the terminology of ‘model’, ‘theory’, and ‘physical hypothesis’, you are not making sense.
“Our weather/climate models are good enough to predict reliably that summers will be warmer than winters, that days will be warmer than nights.”
You have not yet shown that you know what a scientific prediction is. Give it a try. Look up Kepler’s Three Laws on Wikipedia and read about them. You can pick up the terminology of ‘physical hypothesis’, ‘scientific prediction’, and the related terms there.
Theo Goodwin says;
I hear that Trenberth has bought a yellow submarine and that all his friends live there too.>>>
Well only his good friends. Word has it he sent poor Wolfgang Wagner out for a…uhm…. swim.
Theo Goodwin says (at 6:18 above)
“the fact that some [models]…agree well with ENSO…should be of no interest to scientists. Unlike sets of physical hypotheses, aka physical theories, the physical components of a computer model have no cognitive component of their own,,,Each model has to stand or fall on it’s own…”
Is that true? I thought that the current models purported to emulate physical theory of heat transfer, which are reliably established independently. Such models may be inadequate reflections of otherwise well-established theory. They may depend upon inadequate databases of heat content or flow. They may be parameterized, and dependent on insufficient datasets from which to infer proper parameters, But are they so holistic that the greater agreement of data with some models than with others provides us with no useful information?
My experience is with models of molecular dynamics in photophysical systems. Climate is no doubt much more complicated, although more susceptible to fudging. But don’t intermediate results allow for discrimination between better and worse models? If not, why not? I am not familiar with the details of the current 13 models included by the IPCC , but I have been lead to believe that they attempt to model energy flow according to conventional thermodynamics. If so, why can’t the physics incorporated in a particular model be validated when it better agrees with the data? I routinely compared experimental data on molecular photophysics with the predictions of various models of molecular rotation in order to determine which model better explained the data. Why can’t climate data be used to isolate the components of climate models which more closely approximate observation?
I may very well agree with Theo’s position on climate models. I do not intend my observations as opposition, But I also do not understand the argument that the relative superiority of some models over others does not allow us to conclude that the better models move incrementally in a more defensible direction.
davidmhoffer says:
September 22, 2011 at 8:37 pm
Very good!
@Theo Goodwin says:
September 22, 2011 at 7:51 pm
Maybe I’m being too generous here, but persumably the underlying computer code for climate models isn’t just spewing out random numbers that someone then sit’s down and tries to make sense of. Presumably, the code is based on the idea the if things are currently in state X, then event Y occurs then you end up in state Z. In order to get from state X to Z, you’ve got to have some sort of theory or hypothesis about how that change occurs. I’m sure that some of the theories are based on descriptive extrapolations from observed data as you say, but hopefully (he says naively) there are at least some components of those models that have a fundamental physical theory/hypothesis behind them, even if they’re not good ones. What those theories might be, I have no idea. I’m a statistician, not a climate ‘expert’ (as was once pointed out to me on a statistics listserv when I once queried whether anyone else was nervous about the fact that the surface temps from climate models presented in IPCC 4 appeared to have latitudinal biases).
Any model that doesn’t have good predictive ability, is still a model in my books, just not a very useful one. Maybe we’re arguing over semantics, but I’d contend that any computer code that is used to make a prediction or extrapolation, has got to be based on something, where that something would be regarded as a ‘model’ (good, bad or otherwise) in my part of the scientific world.
Forgive me, I am not a climate scientist and so don’t know all the intracacies of models, but why are we just eyeballing the graph and saying the models look off? Is there not some mathematical standard we can use to say how close a model is to the observations? For example, I imagine that the integral of the absolute value of their differences would give a pretty good estimate (that is if f(t) is the graph of the observations and g(t) is the graph of some model, then integrate over |f(x)-g(x)|. It would probably be best to normalize the horizontal axis so that you are integrating over a distance of length 1; this way it can be easily compared to other models over different time lines). This idea would probably have to tweaked at best before working, but it seems to me that there should be some scientific way to measure the accuracy of a model that’s better than just looking.
AJStrata wrote;
“I find this inability to determine models bogus when they fail to model reality accurately really sad”
EXACTLY !!!!!!!!!!!!
These folks in the climate science world have convinced themselves that the models are TRUTH and the observations which don’t match the models must be WRONG. Truly this is very sad.
Here is just a short list of the things that engineers in the aerospace field do not resolve with models;
1) The performance of real hardware when exposed to the vibration experienced during a launch
2) The susceptibility to and emission of EMI (Electromagnetic Interference) from electrical circuitry
3) The actual forces that a structure will sustain before it breaks
4) Etc. Etc.
In all of these cases a Qualification Model (“QM”) of the design is fabricated and actually subjected to the expected environment it will experience when on orbit. Typically the QM is exposed to forces that exceed those it will actually see when on orbit. If it passes, an exact copy (possible because of extreme measures taken while building this flight model (i.e. “FM”) including tracing the lot numbers of materials used to fabricate the FM) is built using extremely well documented procedures.
Even after all of this attention to detail the success rate is much less than we would like it to be.
For example after following all of these procedures ~ 2 out of ~150 Space Shuttle were not completed because of a “mission failure”..
In the engineering field we have a saying; “If your hardware does not perform as predicted by your model you need to improve your model”.
Nothing I see in the literature of climate science seems to acknowledge that this saying might have some merit.
Cheers, Kevin.
Curt says:
September 22, 2011 at 8:56 pm
Theo Goodwin says:
September 22, 2011 at 3:27 pm
“Do you really think that what you use to launch and manage satellites is a model?”
Theo: Your question and subsequent comment display serious philosophical and practical problems. ANY prediction of the future is based on a model, whether explicit or implicit. >>>
Well Theo is more than capable of responding on his own (and I see he did) but Curt, you’re missing the point.
I can write computer code and create a beautiful representation of the earth from space, complete with moving cloud cover, day/night, orbiting moon, even tides that you can zoom in on and watch rise and fall on the coastlines. That’s a model.
But it isn’t a SCIENTIFIC model. It is a representation that I dreamed up out of my head. It might look realistic, it might even fool most of the people most of the time into thinking it is a “real” model of the earth. But it isn’t. It is just, at best, an “artists represenation”.
It has no basis in science. It has not predictive value. None. I cold create a script for ,my “artists representation” that shows the ice caps melting and the oceans rising. That would indeed be a model. It would also be a cartoon.
A scientific model would represent all those factors visually, and might even look just like the cartoon for all I know. But for any scientist to claim that his/her computer simulation of anything is based on science, it must adhere to the laws of physics and portray to the extent known the interation of forces and energy between the elements of the system as a whole. AJ’s “model” is no “artists representation”, it is simply a computer program that calculate the position of things like satellites based on the known and MEASURED factors that can be used to CALCULATE with some precision, the position of the satellite at some given point in time.
That is NOT what Trenberth and company mean when they say a “model”. As per Wolfgang
Wagner’s resignation, which Trenberth forced for the sin of allowing Spencer and Braswell to publish ACTUAL DATA that conflicts with the models, they believe that actual results should no be published with the input and permission of “modelers” who create representations of the earth’s climate based on….
what appears to be little more than an artists representation. Trenberth and NCAR have published a paper “showing” they say, that the “missing heat” is being sequestered in the depths of the ocean…with no corroborating data, no physical process to explain how it got there, no theory as to why it stays there, no explanation of why the thermal expansion they’ve warned about based on their previous “models”…. er, uhm cartoons…. hasn’t happened as a result, why there’s been no massive out gassing of CO2 as a result…. shall I go on? There’s no science in these models, they’re just… an artists representation.
In the meantime, at the top of this post, is a graph showing how much energy is being lost to space, and when, by actual satellite measurements.
Put that science into the models and program them to produce prediction based on the physical facts, and you’ll have something more akin to what AJ is talking about. Call it a scientific model, an accurate model, a model with predictive capabilities…cal it what you will, but it differs completely from a “model” like that of NCAR which is no more based on science than my cartoon earth that LOOKS like it does.
Whoops again;
~ 2 out of 150 Space Shuttle missions were completed without a “mission failure”.
should be;
2 out of ~150 Space Shuttle were not completed because of a “mission failure”.
davidmhoffer says:
September 22, 2011 at 8:37 pm
Pedant alert ahead!
> I don’t give a tinker’s d***, nor do I mind
The phrase is “worth a tinker’s dam.” When a tinker mended a pot or pan he’d make a clay dam to keep the molten tin in place. After the patch was applied the now worthless dam was knocked off and discarded. Being worthless, it became available as a comparison to similarly worthless items.
I’ve said this for a while. Why would you even expect the climate sensitivity number to remain fixed? What rationale is there for the Earth responding the same to a change in forcing no matter the state of currents, plate positioning, circulations, et cetera? Where has the idea of a single constant value for sensitivity ever been supported by a scientific study? What if the equation ΔT=ΔF*S is not actually right? This could easily be the case if ΔT affects S.
Theo Goodwin:
At September 22, 2011 at 3:27 pm you assert:
“The proof is in the pudding. If you doubt what I have said then simply ask a Warmista modeler for the physical theory which is embodied in his model. He won’t be able to show it to you. All he can offer you is his model. Warmista have no physical theory of Earth’s climate.”
I am not a Warmista but I point out that you are mistaken.
The underlying ‘theory’ of all the climate models is that global climate is governed by radiative forcing. More properly,it is an assumption (i.e. not a theory) that radiative forcing governs global climate.
Indeed, Roy Spencer implicitly accepts this assumption in his above article when he says of the climate models;
““And I believe the problem is they are too sensitive, and thus are predicting too much future global warming.”
However, there is no empirical support for the assumption and much empirical evidence refutes it; e.g. ~20% increase in radiative forcing from the Sun over the last 2.5 billion years has had no discernible effect on global climate.
So, there is evidence that global climate is not solely a function of radiative forcing, but the models assume radiative forcing governs global climate. The recent paper by Meehle et al. is relevant to this. That paper assumes Trenberth’s “missing heat” is hiding in the deep oceans. The “missing heat” only exists because of increased radiative forcing. But if increased radiative forcing leads to heat ‘hiding’ in the oceans then that ‘hiding’ prevents radiative forcing having a direct effect on global climate at less than millennial time scales,.
So, the paper by Meehle et al. contradicts the basic assumption underlying all the climate models.
The real problem is that the climate is a complex system and – in common with other complex systems – may adapt in response to a change in input and/or output. For example, the human body is a complex system that loses weight if subjected to increased work (such as military or sports training) without increased food input, but there are people who assume body weight is solely a function of food intake. Similarly, there are people who assume global climate is solely a function of radiative forcing and all the climate models are representations of that untrue assumption.
Richard
I have no idea what Roy Spencer is feeding that black swan, but it sure looks to be a very healthy bird!
Theo Goodwin says: September 22, 2011 at 7:51 pm
“I began managing models for large corporations about forty years ago. I can assure you that no computer model, however complex, amounts to a theory. After all, there are beautiful computer models of the stock market. And they are incredibly useful analytic tools. But surely you don’t think for one minute that someone actually has a theory of the stock market that consists of reasonably well-confirmed hypotheses that enable accurate prediction of stock market behavior.”
Hi Theo and all. Good computer models often embody a theory and it is important to understand what a model attempts to reflect before declaring it void where it does not match a particular observation. I design stock market models for a living. They embody a theory of markets. They are profitable. They do NOT predict general market direction and therefore do NOT predict market crashes.
Questions: Is my model “wrong” because it does not predict crashes?
Is my model “wrong” because it loses money some years?
Considering this context, declaring a GCM wrong because it does not reproduce ENSO when it was never intended to seems misplaced. Declaring a GCM wrong over short time frames is perhaps too hasty as well. The right/wrong fixation may make sense if an appropriate question and an appropriate timeframe is being considered. Possibly better is to ask “is a given model useful with respect to what it was designed for?”. (Please note that I am not declaring models infallible nor commenting on Dr Spencer’s work nor attacking Theo Goodwin.)
AJStrata says:
September 22, 2011 at 2:43 pm
Get back on the soapbox. You are 100% right. I cannot believe the stupidity of people like ‘sharperoo’. This inane believe that model, no matter how complex, can accurately predict outcomes of a genuinely chaotic system is beyond me. Mathematicians have stated, without exception, that natural systems are chaotic and as such CANNOT BE PREDICTED.
First law of climate science, if the value of any given product differs between reality and the model , it is reality which is in error .
Ammonite small problem its these very models upon which dramatic changes to people life styles are called for with the spending of vast sums of money. Its not some minor academic argument over a piece of research in some journal. These are what the IPCC and friends are using to drive froward their political goals . The irony its the engagement with the public and the body politic that some climate scientists desperately wanted that has lead to the investigation of their ‘research’ which has seen it fall to bits .