Guest Post by Willis Eschenbach
David Rose has posted this , from the unreleased IPCC Fifth Assessment Report (AR5):
‘ECS is likely in the range 1.5C to 4.5C… The lower limit of the assessed likely range is thus less than the 2C in the [2007 report], reflecting the evidence from new studies.’ SOURCE
I cracked up when I read that … despite the IPCC’s claim of even greater certainty, it’s a step backwards.
You see, back around 1980, about 33 years ago, we got the first estimate from the computer models of the “equilibrium climate sensitivity” (ECS). This is the estimate of how much the world will warm if CO2 doubles. At that time, the range was said to be from 1.5° to 4.5°.
However, that was reduced in the Fourth Assessment Report, to a narrower, presumably more accurate range of from 2°C to 4.5°C. Now, however, they’ve backed away from that, and retreated to their previous estimate.
Now consider: the first estimate was done in 1980, using a simple computer and a simple model. Since then, there has been a huge, almost unimaginable increase in computer power. There has been a correspondingly huge increase in computer speed. The number of gridcells in the models has gone up by a couple orders of magnitude. Separate ocean and atmosphere models have been combined into one to reduce errors. And the size of the models has gone from a few thousand lines of code to millions of lines of code.
And the estimates of climate sensitivity have not gotten even the slightest bit more accurate.
Can anyone name any other scientific field that has made so little progress in the last third of a century? Anyone? Because I can’t.
So … what is the most plausible explanation for this ludicrous, abysmal failure to improve a simple estimate in a third of a century?
I can give you my answer. The models are on the wrong path. And when you’re on the wrong path, it doesn’t matter how big you are or how complex you are or how fast you are—you won’t get the right answer.
And what is the wrong path?
The wrong path is the ludicrous idea that the change in global temperature is a simple function of the change in the “forcings”, which is climatespeak for the amount of downward radiation at the top of the atmosphere. The canonical (incorrect) equation is:
∆T = lambda ∆F
where T is temperature, F is forcing, lambda is the climate sensitivity, and ∆ means “the change in”.
I have shown, in a variety of posts, that the temperature of the earth is not a function of the change in forcings. Instead, the climate is a governed system. As an example of another governed system, consider a car. In general, other things being equal, we can say that the change in speed of a car is a linear function of the change in the amount of gas. Mathematically, this would be:
∆S = lambda ∆G
where S is speed, G is gas, and lambda is the coefficient relating the two.
But suppose we turn on the governor, which in a car is called the cruise control. At that point, the relationship between speed and gas consumption disappears entirely—gas consumption goes up and down, but the speed basically doesn’t change.
Note that this is NOT a feedback, which would just change the coefficient “lambda” giving the linear relationship between the change in speed ∆S and the change in gas ∆G. The addition of a governor completely wipes out that linear relationship, de-coupling the changes in gas consumption from the speed changes entirely.
The exact same thing is going on with the climate. It is governed by a variety of emergent climate phenomena such as thunderstorms, the El Nino/La Nina warm water pump, and the PDO. And as a result, the change in global temperature is totally decoupled from the changes in forcings. This is why it is so hard to find traces of e.g. solar and volcano forcings in the temperature record. We know that both of those change the forcings … but the temperatures do not change correspondingly.
To me, that’s the Occam’s Razor explanation of why, after thirty years, millions of dollars, millions of man-hours, and millions of lines of code, the computer models have not improved the estimation of “climate sensitivity” in the slightest. They do not contain or model any of the emergent phenomena that govern the climate, the phenomena that decouple the temperature from the forcing and render the entire idea of “climate sensitivity” meaningless.
w.
PS—I have also shown that despite their huge complexity, the global temperature output of the models can be emulated to a 98% accuracy by a simple one-line equation. This means that their estimate of the “climate sensitivity” is entirely a function of their choice of forcings … meaning, of course, that even on a good day with a following wind they can tell us nothing about the climate sensitivity.
RC Saumarez says:
September 15, 2013 at 1:31 pm
I keep waiting for you to do that, RC … but to date it’s been nothing but bluster on your part. You haven’t falsified a single one of my statements or findings. You haven’t shown that anything I’ve said was wrong.
Instead, you seem obsessed with what to call what I’ve done … and you seem quite impressed with that.
w.
PS—Thanks for the link to your post, which you claimed was written “to correct the complete rubbish [I] wrote about signal processing”. It was an interesting post, and I thanked you for it in the comments.
However, since your post doesn’t mention me anywhere, it’s not clear where you got the idea that it “corrects the complete rubbish I wrote”. Perhaps you could re-read your post and point out where you mentioned me and my rubbish, so we can all see what you’re referring to.
RC Saumarez
I wouldn’t bother with him, he is either a troll or is just too pompous to converse; I guess monologue is how usually communicates.
I agree with many of the points you make. I think you should, if you’re interested, look at the video I linked to in one of the comments to Richard. It is an excellent presentation by Dr C. Essex and highlights probably the most fundamental issues with the current batch of GCMs – and that cannot be overcome by ever more efficient code or programming paradigms. As a far as I know GCM modellers are at the forefront of parallel programming for sequential algorithms but when the “quality bottleneck” is hardware you’re fighting the wrong war. The main issues is resolution, and up-scaling manifolds in order to solve the Navier-Stokes equations. But i was shocked to see how they deal with numerical instabilities. I can’t believe all these leaps in the design and implementation of code were made without looking at these fundamental issues – without addressing those outlined by Essex they’re always going to be crap!
RC Saumarez:
This thread is not about me and I see no reason to answer any personal questions, especially when you could have had an answer by reading references mentioned in the thread.
And that is my last post addressed to you. I despise trolls.
Richard
@RichardSCourtney.
Are you the Richard S Courtney whom the Desmogblog is rather rude about?,
DPhil Cambridge? Does Cambridge grant a DPhil?
If I have mis-identified you, please make this clear, because having scanned RichardsCoirtney’s publications on the web, I cannot understand how you can claim to be an authority on mathematical modellin.
Ig you are not this RichardsCourtney, I offer my humble apologies.
cd:
I see you have resorted to the usual anonymous troll tactic of poisoning the well.
And tag team trolling, too.
Richard
To Anthony Watts,
cc: Willis Eschenbach.
I am very happy to write a post to discuss model identification. You will remember that after I thought that W’s post on signal processing lacked a certain understanding of the subject, je told me to put up or shut up.
You invited me to write a post, which I did. I went out of my way to be inoffensive in the post, but a number of readers with a signal processing background got the point.
If you would like me to do so again, I would of course be deligheted to do so.
RC Saumarez
Desmogblog
My opinion of you has just crashed. Anyone who gets their opinions, or seeks information from such a poisonous website about anybody, deserves disdain for they seek it for others.
cd says:
September 15, 2013 at 1:44 pm
True ‘dat … but if I say nothing, there are people who might believe him.
Thanks, cd. I very, very rarely look at videos for a couple of reasons. First, I read very fast, and so the videos all move at what to me is an intolerably slow pace.
Second, I can’t skim through the video to find the important parts, as I can with text.
Third, I’m far too involved when it’s a video, and I find myself nodding and agreeing without the kind of critical thought that I apply to text.
Fourth, I don’t have 52 minutes to blow on anything unknown.
So while I’ll pass on the invitation, don’t take it personally. It’s just my idiosyncrasy, nothing to do with you or with the subject matter.
Best regards,
w.
Richard, you said, “You need to look up finite difference analysis (FDA) and finite element analysis (FEA).” These types of mathematical processes do form part of the code strings in climate models. Is that bad or good in your estimation?
Latitude says:
September 15, 2013 at 11:41 am
Latitude, I suspect you’re conflating the half-life of a pulse of CO2 emitted into the atmosphere, with the residence time of the average CO2 molecule. The latter is on the order of five years or so, while the former is much longer.
How much longer? There’s not a lot of agreement on that.
w.
Here are some of Eschenbach’s responses during this thread
Actually, the problem with fusion is not the science—it’s the engineering. The clim*ate question is purely theoretical. But the fusion challenge is entirely physical. We’re trying to cage the sun in a bottle … and as a result, the lack of progress there is quite different from the lack of progress in narrowing the bounds on climate sensitivity.
w.
For heavens sake, Pamela … are you drunk-blogging or something? Your fantasies about what I want are a joke—they have nothing at all to do with me. You’ve totally misunderstood my post, you still don’t seem to have grasped what my model does, and now you are simply wallowing in your sick ideas about what I want and who I am. I’m not who you think I am, not by a thousand miles.
I won’t hold it against you, but for goodness sakes, next time leave all of that kind of pathetic, obsessive personal stuff out entirely and confine yourself to the science. You’re just embarrassing yourself with that puerile nonsense.
w.
Pamela, when you respond to me and say “you must be” this and “you want” that, that’s not a “vignette”.
That’s an accusation about me, and in this case a very ugly and unpleasant accusation that had nothing to do with me.
So I’m sorry, but your “explanation” doesn’t hold water. An apology is in order, not a justification of your unwarranted attack.
w.
PS—Citing an entire Chapter of an IPCC report? Is that your idea of a proper citation? My high-school chemistry teacher would have thumped me with her red pencil if I tried that nonsense. If you have a point you wish to back up, you need to cite chapter and verse.
As it stands, you’re no better than the Bible-thumpers of my childhood, who would stand up in the tent and when someone asked a question would hold up the Bible and shout “The answer’s right here” … perhaps the answer is somewhere in the entire chapter you just cited, but I’m not going to try to guess just which paragraph you’re talking about.
RC Saumarez says:
September 15, 2013 at 10:24 am
Having revisited the “cold equations” , I have to agree with Pamela Gray.
Here we go again with vague claims … you have to agree with Pamela Gray saying WHAT?
All your equations are is a linearisation of the SB effect, which results in a simple ARMA difference equation, This can rougly emulate teperature, as is well known and has been pointed out by many workers, including Mann, McIntyre.
The point of a GCM is that it encapsulates mechanisms. If A happens the B will follow that will precipitate C and so on. This means that in such a model parameters that relate to details of climate can be examined for their effects in the future (Note: I am not saying that they do this with any particular accuracy).
You (and Pamela) seem to have totally misunderstood the idea of a “black box analysis”. Here are three posts that will give you a better idea what I’m talking about.
Zero Point Three Times the Forcing
Life is Like a Black Box of Chocolates
Climate Model Sensitivity Calculated Directly
Therefore I think you are wrong in your assertion that your “model” performs as well as GCM. Your model isn’t a model – it is a curve fitting exercise, while GCMs attempt to capture mechanisms.
Ummm … come back after you’ve read the three posts, and you’ve understood the idea of a “black box analysis”. At present, we’re talking on entirely different levels, and that doesn’t work.
As an aside, if you are dealing with distributions of variables, as you discuss in you “cold Equations”, it seems to me that your mathematics does not capture this.
I’m sorry, but could you be a bit more vague? That’s almost meaningful …
w.
Your specious claim is that in the first paragraph the “you” clearly refers to me, but in the second paragraph it doesn’t?
Hogwash.
Clearly, you are referring to me all the way through, and now, rather than apologize as a decent human would, you’re trying to weasel out of it with that bogus excuse?
You don’t seem to get it, and yes, I do mean “you”. Here’s a protip—that’s why people use the word “you”, because they mean the person they are addressing. Otherwise they say “him”, or “someone”, or “her”, because “you” means … well … you.
You have insulted me, whether deliberately or not, and now you want to justify it on some bogus excuse, that it would have hard to phrase your “vignette” so that it wouldn’t be insulting … really?
PS—if you think that a “black box model” is just a set of transfer functions, you’ve missed the purpose of the black box analysis entirely. The purpose is to understand what the system in question is doing functionally. In this case, my analysis showed that the models are simply lagging and resizing the inputs. This is an important finding, as it builds on and completely explains Kiehl’s observation that models with high forcings have low sensitivities.
And all of your handwaving about transfer functions and differential equations and how wrong I am doesn’t change those findings at all. RC, you haven’t shown that one single thing that I found in my black box analysis of the models is wrong … and that seems to be bothering you a lot, if your agitation is any gauge.
I realise that you have free reign on this blog to post anything you like and you have special license to replay to anybody who displeases you.
Nevertheless, there are people who comment on this blog who have considerably more scientfic education and experience than your good self.
I am unable to undersatnd what makes you think that you habe any authority to make the pronouncements that you do.
If you have to go out of your way to be inoffensive, then you have a problem.
@CD,
I am not saying that Desmogblog’s opinion is worth having.
I am trying to identify Richard S Courtney. I asked him if he had any degrees or any publications that would back up his assertions. If he is the one that is pin-pointed there, one can find his papers. Those that I can find do not point to any particular expertise in the solution of partial differential; equations.
As, I said, if he is the wrong Richard S Courtney, I apologise profusely. He is at perfect liberty to refute this
rogerknights says:
September 15, 2013 at 1:24 pm
Pamela Gray says:
September 15, 2013 at 12:09 pm
Hey Joe! The same thing [a spoken ad starting automatically] has been happening to me! Twice this morning! It just started happening today on this thread. Weird.
Me too. I don’t think it’s a browser problem. I think WordPress has fiddled with something.
__________________
Yes, something.
RC Saumarez says:
September 15, 2013 at 2:08 pm
Where are you going with all this. You seem a little disgruntled to say the least; and seem more interested at times in discrediting the man rather than arguments.
1) Is it because you feel entitled to peoples’ agreement?
2) That the given review is wrong? Is the premise wrong or are the arguments wrong?
3) You’re here to sabotage the thread?
RC Saumarez
You have no right to demand thta anyone disclose who they are. If you have done so then more fool you.
BTW, qualifications have nothing to with it. I have ample amount of qualifications to argue on most scientific issues but that doesn’t make right – it’s my arguments that do. In short, what do his qualifications, or who he is, have to do with anything. He’s made his case that is all you need here!
@ur momisugly RC Saumarez,
I’m really unsure of your point. If WIllis demonstrated quite some time ago, that the models’ outputs reduce to a simple ‘black box’ equation, either you see that his analysis is valid, or disagree and can prove it.
@Jeff Alberts.
I suggest you read what W has posted in response to people.
The background to this dispute is that I have a PhD from a signal processing lab and can distinguish something that is sensible from something that is not.
W produced a rather strange post on filtering and correlation and I commented on it. I received a diatribe of the nature shown in my post above in which I was told to put up or shut up. I was invited to write a post, which I did. In this I was careful not to throw stones at Mr Eschenbach but to be objective an informative.
My point is this.
WUWT that is a premier science blog and some very good people write posts on it. Mr Eschenbach writes extensively on mathematical modelling, signal processing and many other topics. Those of us who have expertise in these fields make some critical comments that would be regarded as polite, but effictive, in academia and we are met with a volley of abuse.
My personal opinion as a PhD with some experience is that Eschenbachs’s posts on modelling and signal processing are naive. This stems from a lack of training and experience in the subject that he posts on.
Pamela Gray:
I said you had crossed the line, and I had had enough of answering you. But in the light of the RC Saumarez troll’s attacks I suppose I must make some response to your post at September 15, 2013 at 2:05 pm which asks me
Oh dear, that is not even wrong
http://en.wikipedia.org/wiki/Not_even_wrong
The model is constructed as a framework and iterates to stability at its nodes.
As I said, you need to look up FDA to understand what the models are doing.
The code determines what occurs in the model.
FDA is neither good or bad. There was a time when I used FEA (a similar model method to FDA) for stress analysis. It is a method. Formulate a good model and it enables analyses of surprising accuracy, precision and reliability. Do it wrong and the result is either nonsense or fails to achieve stability.
So, your question as to whether the models using FDA being good or bad does not have an answer (it is like asking the name of the Pope’s wife) because it depends on the model, how it was formulated, and the processes coded into it. If a process is incorrectly formulated or is not included then it cannot be known if the effect of that is significant or not because nobody can know what the model actually does in achieving stability through its iterations (just as the name cannot be known of a person who is not known to exist).
Richard
@ur momisugly Luther Wu.
I would be delighted to do so if I am invited.
cd:
It seems I owe you an apology.
You made a comment which I understood to be your supporting the behaviour of RC Saumarez.
Your subsequent posts demonstrate that I was mistaken in that understansding.
I completely apologise for my erroneous assumption, resulting accusation, and any offence thus caused.
Richard
RC Saumarez says:
September 15, 2013 at 2:08 pm
Yes, those are indeed my responses. I haven’t any idea why you felt the need to repost them, it doesn’t change them at all. That’s in fact what I wrote, although poorly re-formatted due to your crude copying and pasting skills .
But reposting my responses doesn’t falsify them, which is what you’d need to do to get any traction. Simply reposting them and then following that with your considered opinion about their value in toto doesn’t advance the discussion at all.
To advance the discussion, you’d actually have to start either answering the questions that I’ve asked you, or showing that particular parts of what I’ve said are wrong.
But you’ve done neither one of those. Instead, you’ve just rolled up what I said into a ball, and told us that you disapprove of it and of me, and that you think I’m wrong because … well, because you think I’m wrong.
… and this is supposed to convince us that you have the inside track on the truth?
Look, RC, I know you’re a smart guy. But if you want to get any traction around here, you would be well advised to begin by picking something that I said and showing exactly how it is wrong.
But instead of doing that, you’ve said things like:
How am I supposed to respond to that, RC? You haven’t said what my error is or where it occurred, but by God you’re going to wave your hands and proudly tell everyone that I’m wrong … do you see how ineffective that is? You haven’t falsified a single thing I said in the Cold Equations, all you’ve done is made a vague, unsubstantiated, unspecified accusation … and that’s it.
I’m never happy to be proven wrong. However, that’s part of science, and whether I like it or not, it happens.
But it doesn’t happen by you declaiming that I made some unspecified mistake somewhere in a post I wrote two years ago.
And it doesn’t happen when you claim you wrote a whole post specifically to ““to correct the complete rubbish [I] wrote”, and when we look at the post we find nothing about any rubbish I wrote, either in the post itself or in the comments.
All that kind of thing does is makes you look foolish.
Finally, you say that if I put my head up above what you all the “parapit”, I should expect to be shot at … my friend, I’ve been shot constantly since I started writing for this blog. And fortunately, like your shots, most of them have missed completely, and for the same reason—you’ve been so busy telling folks that you are right, right, right, that you’ve neglected to specify exactly why and where I’m wrong.
As in this case, where you wrap up everything I’ve said, and then claim it’s all worthless, root and branch … indeed it might all be worthless, RC, although I doubt it, but that’s not how to go about showing anything.
Because all you’ve shown us by that action is that you really, really don’t like me … but I think everyone already knew that.
w.
@richard S Courtney
It’s been a long time since I’ve used finite element analysis, but I seem to remember that this is based on variational principles. Finite difference methood is based on a Maclaurin’s/Taylor’s series.
Please correct me if I’ve misunderstood this.
Your humble troll
Steven Mosher says:
September 15, 2013 at 12:33 pm
Thanks for that, Mosh. It’s a fascinating reference. As far as I know, the first estimates from computer models were from 1979, but there may have been earlier ones.
w.
@ur momisugly Willis Eschenbach,
If Anthony Watts invites me to write a post, i will do so,