Ducking like a quack

ONCE TWO SCIENTISTS–it hardly matters what sort–were walking before dinner beside a pleasant pond with their friend, a reporter for the Dispatch, when they happened to notice a bird standing beside the water.

“I am a skeptic,” said the first scientist. “I demand convincing evidence before I make an assertion. But I believe I can identify that bird, beyond all reasonable doubt, as a duck.” The journalist nodded silently at this assertion.

“I also am a skeptic,” said the second, “but evidently of a more refined sort, for I demand a much higher standard of evidence than you do. I see no irrefutable evidence to back up your assertion that this object before us is even a bird, let alone positively identifying it as a duck.” The journalist raised his eyebrow sagely.

Read the rest here.


I got that chuckle thanks to an interesting post over at InItForTheGold well worth the read for Part 1, all about styles of climate septics.

The third kind is the “throw spaghetti at the wall” type, the one who will wheel out fifteen half-baked arguments for every one you try to refute. This kind seems the most thoroughly trained in political shenanigans. The approach is as common as it is frustrating. They refuse to play by anything resembling the rules of logic, instead resorting to pure polemics. If you score a point of any sort, they will pretend not to notice. Therefore Very Very not the IPCC.

(Bring any local contributors to mind?)

And well worth the read for Part 2:

Anyway, and here is the point, when you have actual scientific knowledge, information from one source or topic can and should influence your thinking on others. Thus a coherent worldview arises. The naysayers on climate change lack a coherent worldview. Their claim is that climate has some magical properties making it unapproachable by science, and that thus very little can be known. But all they know for a fact is that they themselves know very little. Those of us who know enough are uncomfortable with bits of information that don’t fit in right. Our experience is that if we look closely enough, there is usually a difference in assumption or nomenclature, not inconsistent evidence. We don’t know everything, but what we do know tends to hang together, because we are practicing actual science, not as some philosopher of science would describe it, but as it actually works. Human minds collaborate to produce a robust and coherent view of the world.

Lack of coherence is the hallmark of the run of the mill psuedo skepticism we see so often around here. Anything goes as long as it is Not The IPCC.

122 thoughts on “Ducking like a quack

  1. Coby: 🙂 Yes, I do check the referrals to see if there are interesting links over to my place. But I don’t look at it as much any more, so you’re better off (and always welcome) to drop me an email when you’ve got something interesting going on in the comment section. I read all your posts, but don’t look at all the comments.

    For the simplest climate model:
    Tinkering with the solar constant will give you the right order of change — if you do it correctly. What you have up there isn’t correct, though. The sun’s energy hits a disk, area pi*r^2. The surface of the earth, however, has area 4*pi*r^2. 1 W/m^2 in the solar constant is only 1/4th W/m^2 averaged over the surface of the earth. To represent the 4 Watts per meter^2 of a doubled CO2, which is affecting the full surface of the earth, you need to add 16 W/m^2 to the solar constant. (!)

    Regarding parameters:
    There are several sorts, and it looks like your discussant isn’t distinguishing them. One family are fundamental physical constants of the universe — the Stefan-Boltzmann constant, ideal gas constant, Planck’s constant, and so on. A second family are the elementary parameters describing certain important parts of the earth’s climate system — the earth’s orbital parameters, the solar constant, the thermodynamics of water, the topography of the continents and bathymetry of the ocean, radiative band locations and intensities for CO2, H2O, etc., and so on. The ideal climate model (and ideal weather prediction model) would only have these two sorts of parameters.

    The third family are those generated by the fact that we would need about 10^30 times present most powerful computers in order to run that ideal model. Ideally, in doing dynamics, we would use only molecular viscosity. To do so would require having a global mesh no coarser than 1 mm. That’s impossible (by those 30 orders of magnitude), so we have turbulence parameterizations — something that can be run on a much coarser grid than required for molecular viscosity, but still give pretty good answers for the dynamics. These representations have parameters as well (ideally, very few, and mostly based on the type 1 or type 2 parameters).

    We have several different turbulence parameterizations, from exceedingly simple (1 parameter, constant) to quite involved (several parameters, each of which is a function of the atmospheric/oceanic state). The quality of the parameterization is then checked by running the model in conditions where you know the answer and seeing how closely the results match what is observed, mostly in terms of how well it reproduces the turbulence. The ‘climate sensitivity’ is not one of those observed things, hence in no way can be what is used in selecting parameter values. The parameter values chosen are those that give the best wind speeds, boundary layer thickness, mixing rates, boundary layer wind profile, and so on.

    Like

  2. Just to amplify one part of Robert’s excellent summary. Amongst the second type of parameter Robert mentioned are the absorption/radiative bands of various gasses. Those are known numbers from laboratory experiment. Those bands have been measured very carefully (strength as a function of wavelength). So one has no choice for those parameters. It is not allowed to “tune” them, because they have known values.

    Going to the question of “what is the physical model”, that’s where you really have to read the original paper introducing a model rather than some later paper showing the latest output. From that GISS documentation page I linked to, it’s obvious that GISS E has a rather detailed hydrological model, but you’d have to go back to some early paper to find out exactly how they calibrated any non-obvious parameters in said hydrological model. The calibration there would be analogous to the turbulence model Robert discusses above – that is, it will be calibrated to match relatively short term regional phenomena (annual precipitation patterns and river levels, annual ice melt/regrowth etc.) Long term climate sensitivity is not used to calibrate those parameters.

    Only when a model is complete is it compared against reality. See http://tamino.wordpress.com/2010/01/13/models-2/ Tamino finds that 21 of the 23 models in the IPCC AR4 are pretty good at reproducing reality, while two Canadian models aren’t so good.

    Presumably careful comparison of the differences between the those two physical models and the rest of the batch will illuminate what bit of physics in them is unrealistic. Such a comparison could take a while

    Like

  3. Robert,

    does the third family of parameters include couplings and feedbacks? For your simplest meaningful climate model, which parameters fall into the third family?

    As for the first two families, I thought, perhaps wrongheadedly, that we could assume those parameters were not part of the discussion because, as is pointed out by GFW, one cannot change the vibrational spectrum of water or the speed of light. Maybe we could adjust the distance from the earth to the sun, as per Futurama, but I’m sure that would start another debate altogether.

    If the third family of parameters does include ways in which different processes are coupled and the feedbacks associated with those couplings, then that’s where all the ‘action’ is, right? And since one does not have historical observations to check against a set of parameters like feedbacks, the output of your simulation is going to depend heavily on the physical model one concludes feasible to account for what is happening in the real world, right? This seems to be the major point of contention here.

    GFW,

    ‘Such a comparison could take a while’

    I would assume that this comparison is still in progress since there is disagreement between the major models. I don’t think it has much to do with mac vs. pc.

    Like

  4. Maxwell:
    Glad to see some progress in finding what you mean.

    In the simplest model, there are no parameters from the 3rd family. T = f(S, sigma, r, albedo; 4). 4 is a parameter of geometry (follow the link Coby gave to my summary, my summary includes the link to where Atmoz worked out that the correct number is a little larger than 4). Sigma is the stefan-boltzmann constant, safely in class 1. S is the solar constant, r is the earth sun distance, and albedo is … the albedo. All are in class 2 — things you observe about the earth to characterize it. In getting to this point, we used two laws you don’t have any choice about — the law of conservation of energy, and the law of Stefan-Boltzmann.

    The third family does not include ‘couplings’ and ‘feedbacks’ as you seem to be using the term. Feedbacks are conclusions from analyzing the model results, not terms specified beforehand. To illustrate further, I’ll stay with turbulence, and take the simplest version of it — Bulk Aerodynamics. In this, you say that the stress applied by the atmosphere on the surface (which could be the ocean, or land, or grassy field, …) is proportional to the difference in speed between the air and that other surface. The constant of proportionality is the ‘drag coefficient’. (Other terms show up, density of air, for instance, but they’re demanded physically and you don’t get to choose them.)

    Having started with Bulk Aerodynamics as your parameterization, you have only this one number to find. Granted it might be a different number over different surface types, so you’ll have to examine that possibility as well. You find your drag coefficient by going out to the field and taking observations of wind speeds with height and seeing what value will give you the best accord with the observations. The number is then nailed down. If you’re doing bulk aerodynamics, this is the value you want.

    You’ll notice there is no ‘choosing the feedback’ or ‘choosing the coupling’ present. You don’t have that kind of choice available.

    When people later talk about a ‘feedback’ between boundary layer turbulence and, say, surface temperature or boundary layer thickness, it is either an observation or a paired model run. The observation would be to observe boundary layer turbulence, observe temperature, and start disentangling the causes/effects/and feedbacks. Or to do likewise in a model where you, for instance, nudge the drag coefficient around inside the range that your observations permit. Either way, the feedback or coupling strength you arrive at is a conclusion — not something you specified.

    Emphasize that — you did not, and in fact could not specify a strength for those feedbacks or couplings beforehand. If you could, we would have much better weather prediction models than we do — we know quite well things like the daily range of temperatures, but can’t turn any particular knob to control that. The knobs we can turn are things like drag coefficients, and nobody can say beforehand whether one value that’s permitted by observation will give a better result for diurnal temperature range than some other value also permitted by observation. You can then run further experiments with the drag coefficient to get the best representation of the diurnal cycle, while staying inside the range allowed by your original profile studies. But after this you have exceedingly little freedom to adjust it to give a best fit to the boundary layer thickness, or the other several dozen things people are interested in. The result being, there’s actually very little tuning of parameters for purposes other than their most fundamental use.

    Like

  5. ‘Either way, the feedback or coupling strength you arrive at is a conclusion — not something you specified.’

    But in the physical model for the system, the equations of motion if you will, the information about those feedbacks must exist, right? The simulation is not going to produce a feedback after parallel runs or fudging parameters in the way you mention if the information necessary for that feedback wasn’t present in the physical model. It’s just hard to see it, thus the necessity for the simulation.

    Like

  6. Thanks for the posts Robert, i said earlier i would not know a climate model if it bit me on the arse (as opposed to a sheep) and that statement still stands. However your posts have given me a more greater understanding of the complexities involved so i thank you for that.

    From what you have said is it wrong of me to say that the models are only limited by processing power? Or do you think our knowledge at this stage is also a limiting factor?

    Once again thanks for your contribution.

    Like

  7. Maxwell:
    There’s no ‘fudging’. You look for a simpler way of representing the real processes realistically. It generally does exist. How well you can represent the full reality with a simpler model is a question you do more research on. If you’re interested in this, there is an enormous literature that you can examine.

    The feedbacks are what we (people examining model results, or observations of nature) use to describe nature. Nature doesn’t care about how we choose to describe it, a philosophical point that’s perhaps not well-suited to blog comments. Still, that’s the case. The parameters we’re setting, choosing, or tuning, however, are not and cannot be the feedback magnitudes. I’m not sure why you’re locked on to that concept. I realize that in particle physics, say, the coupling constant for the fundamental forces, or at least for the electromagnetic and weak forces, where a unified theory has been well-tested and demonstrated, is a vital point. But an awful lot of physics is not like elementary particle physics, and climate is one of those many areas.

    crakar:
    Not sure I understand you. There are two things involved. On one hand, I have no doubt that if we had computers that were 10^30 times more powerful than the current ones are, we’d have better climate models. Computational resources are definitely a limit to our ability to model climate. Our knowledge of the fundamental physics greatly outstrips our ability to run the programs.

    On the other hand, I also believe that we don’t need those computers to obtain meaningful results today (or, for that matter, 100 years ago). Much of climate is really rather simple computationally. The simplest climate model, that I wrote up and Coby linked to, is something you can do by hand. (Literally, ignore the calculator or slide rule and you could still do it by hand. Or at least I could. I have my grandfather’s log tables.) Arrhenius arrived at a perfectly good description working by hand in 1896 — including a reasonable sensitivity to CO2 doubling (4 C) and the fact that the pole would show more warming than the equator (already observed).

    So, on one hand, yes — if you gave me a computer 10^30 times more powerful than any that now exist, I could very rapidly get better and more detailed answers to whatever questions we have about climate. Much of our efforts today is devoted to the business I mentioned before — of how to get good answers without having computers powerful enough to let us simply toss in the fundamentals we’re all sure of and let them run.

    On the other hand, I also don’t think we need such computers to get usable answers today (or 20 years ago, or 100 years ago). Our simpler representations, the parameterizations, are generally pretty good, even if we know they are missing some of the details.

    Like

  8. By your answer it would appear that you have understood me Robert, better computer = a better model which = a more accurate result and whilst we do not have a full understanding of climate we know enough to produce an accurate enough result. So a better computer would but merely fine tune the result shall we say.

    Like

  9. In case anyone cares (not that I assume anyone does), I am in fact parceling through this discussion and waiting for a chance to chime in something . . . Ok, “profound” might be a strong term . . . but maybe potentially meaningful–although I advise that no one hold their breath.

    Thanks Rob for your contribution. Thanks Max for your persistence. Thanks Coby, again, for the milieu.

    Like

  10. Robert,

    I think I have a better understanding of what you’re saying. But I think I am not doing a good job of explaining the point I am making in distinguishing between a physical model and the simulation of said model. I’ll think about it a bit more and maybe hit you up on your blog about it.

    I was harping on the point, however, because Coby was trying to convince me that computer simulations can ‘create’ new information about a physical model that wasn’t contained in the model already, implicitly or explicitly. I see there is a subtle distinction between ‘interpretation’ and ‘information’, however. So that the ‘information’ in a climate simulation starts with the physical model, but running a simulation allows one to ‘interpret’ how the parameters interact in a way that simply staring at the equations that make up the physical model would not. Is that more correct?

    I don’t think I quite understand your point concerning Nature not caring which way we describe it. Is that in reference to my language of couplings and feedbacks? Or is it some kind of underlying point you just wanted to make in that context? It was a bit confusing to me and I don’t understand how it is supposed to better inform me in this conversation. It is obviously true and I understand what it means in the context of my own work, but I don’t see the connection here.

    Do you have any good references for how the physical models that climate simulations use are formulated or tested? Thanks for your patience.

    Like

  11. Maxwell,

    So that the ‘information’ in a climate simulation starts with the physical model, but running a simulation allows one to ‘interpret’ how the parameters interact in a way that simply staring at the equations that make up the physical model would not. Is that more correct?

    That’s a pretty good way of saying it. Let me give you an example from engineering simulation that will show how you can get a result from basic physics input that is completely unanticipated. We understand very well how things like steel cables and concrete react to forces on them. We can then build a model of a bridge and predict very well how much it will flex under a certain weight of traffic. We also understand fluid motion pretty well including the turbulence Robert’s been talking about. With a bridge, we can simulate wind and the resultant turbulence in much more detail than we can with the entire planet. With sufficient detail, we learn that the turbulence behind the bridge structures can do something really nifty. When the turbulence is behind a structure with two sides, it can perform “alternating vortex shedding”. And it turns out that each time a vortex is shed, it effectively gives a little nudge of force against the structure. Like I said above, we have a great understanding of how steel and concrete will react to those nudges, and ordinarily they’d be trivially small. But there’s this thing called harmonic motion. It’s what a plucked guitar string (or piano wire, or whathaveyou) does. We can easily predict the harmonic frequencies of a single cable, but it’s harder to predict the harmonics of a complex structure. A simulation will reveal them however. The final point to all the above is that bridge builders are careful to try to minimize vortex shedding, and in particular make sure that the harmonic frequencies of bridges are far from the likely frequency of vortex shedding in any reasonable wind scenario. That the above is not obvious just by looking at the equations for fluid motion and building material elasticity, is demonstrated by this video from before the age of computer simulation. http://www.youtube.com/watch?v=j-zczJXSxnw

    Like

  12. Aside: Some of the comments attached to that youtube show the same “amateur arrogance” (commenter naflodii) or “raving conspiracy lunacy” (commenter JoEiScOoL24) that has been directed at climate science recently.

    Like

  13. Actually, commenter naflodii on that youtube is only partly wrong – aeroelastic flutter is more complex than just vortex shedding, although vortex shedding can definitely be part of aeroelastic flutter. The trick is that aeroelastic flutter brings the forcing into tune with the resonant frequencies of the structure. Resonance is still important, but it’s not a coincidence of external forcing having the same frequency. So my undertanding was incomplete as well. Nonetheless the comment is still arrogant because I’m sure the relevant professors he called “silly” know the above better than he does. Finally, the overall example is still a good one regarding unexpected phenomena emerging from simulation (or from reality, if there wasn’t a good simulation done).

    Like

  14. Maxwell:
    I’m afraid I still don’t see the distinction you’re trying to draw.

    The one I was referring to is one that I’ve borrowed from a philosopher of science friend. You and I talk about feedbacks and couplings. With a bit more work we’ll be talking about the same things, or at least be more aware of what the other means. But nature, reality, doesn’t care about such things at all. You and I, for our convenience in understanding reality, could talk about the ice-albedo feedback, or the roughness-boundary layer thickness feedback, or any of a number of feedbacks that climate modelers refer to. But, push come to shove, reality doesn’t care about this. Reality is just CO2 molecules absorbing and emitting radiation according to quantum mechanics, or passing the energy on by collision with N2 and O2 molecules (again quantum). But the feedbacks that you and I might build in talking about CO2 play no role in what the CO2 molecule does or ‘thinks’. Not a big point in its own right except that it seems to me you have a philosophical distinction you’re trying to draw, and I’m not sure you’re being careful of the difference between how you and I choose to look at things, and how things really behave. Since I’m not sure I know what you mean, though, this isn’t one to worry about.

    For climate modeling, one good text is Introduction to Three Dimensional Climate Modeling, by Parkinson and Washington. There are two reasons I recommend it. First, and most important, I’ve read it and think it’s a good book. Second, I know both authors (so you’re warned of my bias).

    It might be a good idea to buy or borrow a copy of this and take a look. I have a fairly strong feeling that right now you and I are having a language issue. After you read this, you should have a much better idea of how climate modelers speak. You may wind up being more concerned about how we do the work than you are now. If so, fine. Bring up a question or comment in whatever is my most recent ‘question place’ post and I’ll try to respond. I put those up about once a month. (They’re always open.)

    Like

  15. GFW,

    In regards to aeronautical flutter it works the same way, the wings begin to oscillate at their resonate frquency which then feeds back in and makes things worse. This can happen very very quickly. A plane can have its wings flapping like a bird depending on what is bolted on to them and the planes attitude (speed, angle of attack and g forces etc) moments before they break off of course.

    To test this a model just dont cut it, too many variables to be able to simulate the real world. You can get a plane to flutter two ways. First, push it to the limits and hope like hell you survive or two, initiate a flutter when flying straight and level in a safe manner.

    Max are you thinking “if you cant explain it you cant model it” this is one area that troubles me with models.

    Like

  16. To test this a model just dont cut it, too many variables to be able to simulate the real world.

    This impossibility explains the appearance of papers with titles like “Nonlinear mathematical modeling of aircraft wing flutter in transonic range”, of course.

    Like

  17. Thanks Dhog.

    That is of course, what I was trying to get at with my analogy (and as our host has just reminded us, analogies are like models).

    What I was trying to explain is that climate sensitivity, like “amount of flutter a particular structure will experience” is not easily deduced from the plain underlying physics, nor is it an input into a model, but a model running the plain underlying physics will produce it (and yes, there are is active development of models of both).

    Like

  18. Ok two dogs how many test programs for aircraft flutter have you been involved in? None? thought so. So beyond the title i would say that everything about your link is above your comprehension.

    But maybe i am selling you short, just wait a second while i go and ask the scientists that conduct these trials i speak of…………………………………………………………………………………………………………………still there? Good i told him that “i had a friend who claims we can use models to do our testing instead of spending hundreds of thousands of dollars a year”

    His response “What f*^$@&g model” he then went on to say that models have a use but unfortunately they are limited by the virtual bubble world they live in. They are detached and isolated from the real world. A real world where empirical data is king and not some guess work.

    So there you have it two dogs, i will be keeping a closer eye on your turds of disinformation from now on.

    Like

  19. crakar, was this fictional scientist friend of yours related, by any chance, to the fictional mathematician that agreed with Monckton’s fictional calculation of climate sensitivity?

    Sounds like they should know each other, maybe they even appeared in a fictional novel about climate change written by Crichton and accepted as fact by all AGW deniers.

    Like

  20. Well at least you did not call me rotting pond scum, you do however accuse a person of being a fictional character of mine.

    I could give you examples of flutter and of how your beloved models are inaccurate but i feel i would be simply wasting my time as you would deny everything i say. There is some irony in that statement inst there Ian.

    Like

  21. Have no enough cash to buy a building? Worry no more, because this is possible to get the loan to solve such problems. So take a commercial loan to buy all you need.

    [spam neutered but left in so Chris does not appear schizophrenic…at least on this thread! ;-)]

    Like

  22. From the OP “They refuse to play by anything resembling the rules of logic, instead resorting to pure polemics. If you score a point of any sort, they will pretend not to notice.”

    Man, these spambots have such a sense of timing.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s