Data Comparisons

Modeling: Beyond the Spherical Cow.

63 Comments

Judy Curry posted a discussion and viewgraphs from a Leonard Smith talk. I particularly liked this viewgraph and thought some of you might enjoy it as much as I do:
I’ll leave you guessing precisely what made me laugh when I saw this.

63 thoughts on “Modeling: Beyond the Spherical Cow.”

  1. Well… um… the modeling is not very good? I suppose that (rather than the spelling) is the point, from the title of your post, but I’m wondering if I’m missing an inside joke here.

    I did take a brief look at Curry’s post and I found many of her “bullet points” as irritating as bullet points always tend to be, in reports prepared by any kind of consultant, only these seemed to be more poorly worded than average. If communication is the problem, I do not think Leonard Smith is the solution.

    But maybe I’m just feeling cranky today. The general idea of scientists trying not to sound more sure than they really are is a good one, of course.

  2. I was dipping into “Hippocrates’ Shadow” and came across the astonishing fact that medical science does not know why, when you crack your knuckles, it makes a sound. Really. No one knows where that sound comes from, and many people have tried, without success, to figure it out.

    You might think that a science that can’t figure out why cracking your knuckles makes a sound would have nothing useful to contribute to human civilization. But you would be wrong.

    The moral being that while humility and critical reassessment are very good things, there is nevertheless a big difference between limited understanding and total ignorance.

  3. Julio–
    It’s seeing mountains flattened dramatically. Of course we know that moist air traveling over very tall mountains have large effects of transport and local climate. We know the actual height matters. So, assuming the illustration of the model’s estimation of the elevation of Norway is correct, this would seem to fall under one of the cases where, to make the problem tractable, the “cow” was assumed “spherical”.

    I’m guessing Leonard’s underlying point was likely: “Physcics – Schmisics. If the geometry is all wrong, the fact that you conserve mass momentum and energy doesn’t ensure you’ll get anywhere near the right answer. Things like geometry sometimes matter too.”

    As for spherical cow: Sometimes it works; sometimes not. Sometimes “works” means “kinda-sorta”.

    On the more general issue vis. Leonard Smith. I have no idea whether Leonard Smith is “the solution”. I do know that trying to give a summary of a talk based on reading the view graphs only is impossible. So, I wouldn’t ever try that. It appears Judy did.

    Maybe she heard the talk? Anyway, I don’t feel I entirely know what Leonard Smith meant to communicate. I don’t think that’s his fault nor mine. I just don’t expect to extract all that much from viewgraphs.

  4. I was still trying to figure out whether there was supposed to be another mountain in “real height” closer to the (larger) step in the middle.

  5. Robert

    You might think that a science that can’t figure out why cracking your knuckles makes a sound would have nothing useful to contribute to human civilization. But you would be wrong.

    Well… you might.

    The moral being that while humility and critical reassessment are very good things, there is nevertheless a big difference between limited understanding and total ignorance.

    Yes: Humility and critical reassessment are good things. Yes. There is a difference between limited understanding and total ignorance.

    I have to admit I don’t know how the first part of your sentence connects to the second. I also suspect you are trying to make some sort of point, but I have no idea what it might be.

  7. Assumptions can make a difference?!

    Model:
    Hot air rises causing the summer monsoon
    Cold air falls causing the winter monsoon.

    Typical answer: What is the average rainfall in Asia? 379 mm (15″)

    Now Insert the Himalaya mountains

    Now include the Orographic effect

    Impact on annual rainfall:
    Charripunji 11,777 mm (463.7 in)
    Lhasa 426 mm (16.8 in)

    Difference:
    2,765% more rain on wetter/drier sides!

    Consider the consequence of “consensus” on climate!

    PS monsoon means “wind”

  8. Shot? Robert, I have
    a) No idea who might conclude that if we don’t know why knuckle cracking makes sounds, that we don’t know anything about medicine
    b) I have no idea why you think someone, somewhere might conclude that. However, I can only speculate that you think someone might because you have made that sort of mistake in the past.
    c) I have no idea what point you think you are trying to make.
    d) Your apparent inability to make coherent points clearly is not a symptom of my inability to read.
    e) you must have really, really, really bad beside manner.

  10. “e) you must have really, really, really bad beside manner.”

    🙂

    Absolute conviction in the scenario you wish were true, in the absence of any evidence — pretty typical of your reasoning skills, in other words.

  12. Robert (Comment #81835) says

    Really. No one knows where that sound comes from, and many people have tried, without success, to figure it out.

    Apparently Australia’s good ‘ol Dr Karl seems to know:

    Some scientists wanted to learn more about knuckle cracking, so they actually stuck a sensitive microphone onto a finger. They found that there wasn’t just one single sound when you cracked a finger joint – there were actually two separate sounds. The joint space is the space between the bones. There is a liquid in this space, and there are ligaments on each side, holding the bones together. As you pull on the joint, you first drop the pressure in the joint space – and the ligaments get sucked in. Once this pressure gets low enough, a bubble pops into existence – making a popping sound, which is the first of the two sounds.

    If yer intrigued enough you can continue reading at:
    http://www.abc.net.au/science/k2/homework/s95607.htm

  13. “You are wrong on all counts.”

    I don’t find your argument by assertion particularly impressive. Have you already forgotten what you wrote?

    However, I have no opinion about whether or not you are a incompetent engineer. I have no experience of your work habits whatsoever, so I have no facts on which to base such a claim — even if it were in some way relevant to this discussion, which, obviously, it is not.

    I like to have some facts before I reach conclusions. It’s just a little thing I like to call “evidence.” In my reasoning I use it the way you use petty resentment and a vivid imagination.

  14. “Apparently Australia’s good ‘ol Dr Karl seems to know”

    Lots of people — including “good ‘ol Dr Karl” — have opinions. No one really knows. Eventually we’ll figure it out. Nevertheless it is humbling.

  15. Difference:
    2,765% more rain on wetter/drier sides!

    Driving east and west through the Cascades, it’s easy to see the dry side and the wet side. In fact, you can often see the clouds “end” right at the point where the terrain becomes dry. If it is true that models just put in some average elevation in the grid (and I’m not sure it is true– I just laughed at the cartoon), then it would be surprising if they could predict snowfall in mountains, retreats of glaciers, water resource amounts from mountain rain etc.

    Ian

    Once this pressure gets low enough, a bubble pops into existence – making a popping sound, which is the first of the two sounds.

    If correct, that would be cavitation. Wikipedia said it was the leading candidate.

  16. Robert

    “You are wrong on all counts.”

    I don’t find your argument by assertion particularly impressive. Have you already forgotten what you wrote?

    In response to you assertion– which at least appears to be based on no evidence, I observed you were wrong on all counts.

    As you ought to be aware:
    1) I would be the one to know whether my conviction is absolute not you.
    2) I would be the one to know what scenario I “wish were true” and
    3) I would be the one to know what evidence I have.
    And I know that your assertion is wrong on all counts.

  18. Robert,

    The fact that regardless of the immense odds, I actually exist – that’s humbling.
    That I am a white male living relatively comfortably in the ‘west’ – that’s humbling.
    That I have the means and the time to discuss this issue over vast expanses in near real time – that’s humbling.
    Cracking knuckles…interesting, yes; humbling, not so much.

  21. Robert:

    “The moral being that while humility and critical reassessment are very good things, there is nevertheless a big difference between limited understanding and total ignorance.”

    You definitely encompass the later.

    Roy Weiler

  22. Roy Weiler (Comment #81856)-That appears to be the most prominent hypothesis, but not the only one, and it appears to be true that none of them have been definitively shown to be true one way or another.

    Others include the rapid stretching of ligaments, or intra-articular adhesions being broken.

    http://www.jaoa.org/cgi/reprint/102/5/283

    It is worth noting that whatever the cause, there is no evidence whatsoever that cracking your knuckles is connected to arthritis.

    http://www.jabfm.org/cgi/content/abstract/24/2/169

  23. The full text of a very interesting study “‘Cracking joints’. A bioengineering study of cavitation in the metacarpophalangeal joint.” is available for download here: http://osteoarthritis.about.com/gi/o.htm?zi=1/XJ&zTi=1&sdn=osteoarthritis&cdn=health&tm=9&f=10&tt=12&bt=1&bts=0&zu=http%3A//www.pubmedcentral.nih.gov/articlerender.fcgi%3Fartid%3D1004074

    It’s pretty cool. They have a mock-up of a joint, equation for pressure in the joint, discussions of what happens to the gas after cracking, motion pictures of cavitation in the mock-up joint and all sorts of stuff.

    “It is therefore concluded that it is the
    collapse of the vapour bubble which causes the
    crack not its formation. This is borne out by mathematical
    considerations when the physical significance
    of applying tension to the joint is considered. As the
    joint is pulled apart, if the fluid film is very thin
    (i.e. –*I in the equation), a very low pressure is”

    The gas involved is…. CO2! 🙂

  25. Andrew:
    Given that Robert assumed no hypothesis at all, we have definitely outdone him 🙂

    Perhaps we can devote some of the climate science research budget to Knuckle cracking research, as Robert thinks it important! It may actually be a better use of climate science money:)

    Roy Weiler

  26. Lucia, have you heard of the three economists in the lifeboats, surround by canned food they are unable to open.
    The first starts;
    “Assume we have a can opener”

  27. Andrew:
    My second reference is actually to cracking knuckles not causing arthritis. I am in FL also, Miami, if you do not mind me asking, where are you?

    Roy Weiler

  28. Lucia,

    Arrogant, aggressive, insulting, poorly informed, and obnoxious… humm… I don’t think your conclusion of “really, really, really bad bedside manner” is much of a leap.

  30. Lucia,

    Back to the slide at the top–I did not actually think that was a real model of anything, just something he made up to illustrate his point. I agree that if it is a real model, it’s pretty funny, but I’m not 100% sure of that!

  32. Julio (Comment #81869)-topography in models can often be very unrealistic. They either have to model the land forms as waves, or average heights within a gridcell. This is part of the reason why just about everyone agrees that regional/local climate “projection” is much less reliable than global predictions

  33. Andrew_FL,

    Modeling the land forms as waves isn’t inherently wrong. 😉

    However, modeling them as a ‘roughness’ probably is in many cases.

  34. expect robert to run away.

    I was dipping into “Hippocrates’ Shadow” and came across the astonishing fact that medical science does not know why, when you crack your knuckles, it makes a sound. Really. No one knows where that sound comes from, and many people have tried, without success, to figure it out.
    #####
    Huh? why is that astonishing? all of the competing theories would be pretty damn hard if not impossible to test. You’re left with modeling things like cavitation. Lucia? is cavitation an easy problem? Its pretty damn ordinary that theories that cant be tested ( hey joe lets make fingers to test our theory ) would go unresolved. especially if they are unimportant. Nope nothing astonishing there whatsoever.

    “You might think that a science that can’t figure out why cracking your knuckles makes a sound would have nothing useful to contribute to human civilization. But you would be wrong. ”

    huh? In other words, there are stupid people who make illogical conclusions? I think you prove that to us every time you come here robert

    “The moral being that while humility and critical reassessment are very good things, there is nevertheless a big difference between limited understanding and total ignorance.”

    How is that the moral? Nobody doubts that there is a big difference between limited understanding and total ignorance. Anybody here? anybody here think there is no difference between limited understanding and total ignorance? And you know the moral of this story is that 2+2 does not equal 5. It follows, right?

  36. Out of curiosity, does anyone know what Smith’s slide actually means? What are the height units? What is his pointy bit? Is it a peak? Average height? What do models actually do? Are they really blockwise? I know some use terrain-following coordinates.

    I see a lack of scepticism here.

  37. Apologies for pursuing the OT, but Robert do you have a link for your HS piece on knuckles? Like other commentators I was under the impression that the established explanation is cavitation of the sinovial fluid. I’d be interested to know why HS discount it. Thanks.

    Nick – Re: the slide – I wondered if English is not Leonard’s first langauge?

    “Sciences knows more than we can Model”? Might be a typo., maybe he meant “Science” or perhaps he meant “Scientists”?

    Without the context of the whole slide pack and talk the slide really doesn’t tell me anything. If he were presenting to a non technical/modelling aware audience and he was backing it up with commentary similar to David H’s comments 81842/4 then it would make a useful supporting example of the point “Output of models needs to be interpreted with due regard to all their limitations”. He seems simply to be saying that correct algorithms can be applied to unrepresentative data.

    FWIW I don’t see the step to “a lack of scepticism” – or does volume of a cow really = 4/3’pi’r^3? 🙂

  39. Double cool – Anthony mans up to it:

    [While the -21wm2 and ~1.2 W/m2 values are correct, the comparison is wrong, and it is my mistake. The values are Top of Atmosphere and Surface, which aren’t the same. This prompts a new rule for me, I shall not publish any posts after midnight again (other than something scheduled previously during the day), because clearly I was too tired to recognize this mistake. I’ll add that I have emailed Dr. Allan regarding the question of feedback on hisfigure 7, and have not received a response. – Anthony]

  40. Curious #81881
    FWIW I don’t see the step to “a lack of scepticism”
    The lack of scepticism consists of proceeding to talk about spherical cows without trying to find these things out. We don’t know what sort of cow it is.

  44. Lucia? is cavitation an easy problem?

    Depends. Some cavitation problems are very easy. That you should check a piping network to see if water will obviously cavitate under the planned operation is discussed in introductory fluids classes and the issue might appear on a test. (Or not. Usually not because while checking is easy, they solution to the rest of the problem pretty much just “stops”.)

    Its pretty damn ordinary that theories that cant be tested ( hey joe lets make fingers to test our theory ) would go unresolved. especially if they are unimportant. Nope nothing astonishing there whatsoever.

    Yep. I’m surprised the guys who did the study with the mockup of the fingers did the study.

    Nick

    The lack of scepticism consists of proceeding to talk about spherical cows without trying to find these things out.

    Huh? Maybe you should go read what skepticism actually is at wikipedia.

    Laughing at the idea illustrated by a graphic is not the same as assuming that the graphic is meant as a literal representation of every possible thing. Moreover, if you read comments prior to yours, you’ll see the notion that the graphic which made me laugh might not be literally true was discussed by comment the 3rd comment, which I posted and which uses phrases like “assuming the illustration … is correct, this would seem ”. The notion that the cartoon might notcomport with reality was emphasized in italics to make it stand out! Then, this was discussed again with Julio. Moreover, how variable height surfaces might really be treated in models has also been discussed by Oliver and others.

    So, if you are suggesting there has been some mis-diagnosis that the cartoon might not represent literal truth, you either haven’t read the conversation or you haven’t managed to register what was said.

    As for your question asking for quantification: Smith didn’t include units which suggests it’s a cartoon (possibly intended to make people laugh. If so, it succeeded.) The pointy bit is likely a mountain peak– as would be consistent with the caption “missing mountain ranges”. If you want to know what models actually do, maybe you should ask Leonard.

  45. Nick

    Well, why are we talking of spherical cows at all here?

    We are discussing it because the cartoon showed a funny idea that made me laugh. It’s the sort of idea that is somewhat analogous to ‘spherical cow’ jokes.

    I don’t know why you are bent out of shape about this.

  46. Sorry Nick – I was after a “straight” answer!

    However your question prompted me to look at Leonard’s source presentation and IMO it looks reasonable stuff. With your knowledge of models and modelling do you think he is being unreasonable?

  47. lucia (Comment #81851) September 21st, 2011 at 5:05 pm

    Driving east and west through the Cascades, it’s easy to see the dry side and the wet side…. then it would be surprising if they could predict snowfall in mountains, retreats of glaciers, water resource amounts from mountain rain etc.

    They can’t.

    As I live in the foothills of the Cascades the computer generated weather forecasts(accuweater, weather.com) are pretty useless at timescales beyond one or two hours and sometimes they are even useless at current conditions. X weather system might be traveling at Y MPH until it pushes up against the Cascades in which case it stops traveling altogether and dumps. So frequently the ‘rain in the AM with afternoon clearing’ ends up being total nonsense.

    I’ll quote a line from out local weatherman’s report this morning
    Computer guidance suggests the best chance of rain will be from Seattle north. That’s not to say that some rain won’t fall to the south of Seattle, it’s just less likely.

    I’m South of Seattle and the clouds are bunching up against the Cascades and getting pretty dark. I’d put the chance of rain somewhere around 90%. The meteorologist that forecast todays weather has only lived here 2 years. I don’t think he’s figured out the ‘computer’ doesn’t know about the Olympic and Cascade mountain ranges yet, even though he’s learned to ‘hedge’ his forecasts substantially.

  48. nick,

    The entire slide presentation is linked at Judith’s. It’s pretty clear what the problem is if you look at his slides.

    Tim Palmer ( you know him) kind of turned me on to Smith.

    Anyway you can see the slide and the outputs from the GCM.

    where does the water go? In the real world the water doesnt go through the mountain. So the spatial distribution of precipitation is a function of the mountain. In the modelling world that spatial feature is missing and you get lousy representation of precepitation the spatial distribution of . So think about south america.

  49. I am assuming that the 0 – 360 along the x-axis represents longitude. In which case, why is the elevation at 0 different than the elevation at 360?

  50. Lucia,

    This point from viewgraph 14 caught my eye:

    “Big Surprises arise when something our models cannot mimic turns out to have important implications for us.”

    I found this interesting because of a few plots I generated of interannual ocean temperature variances which the models have difficulty mimicing:

    http://sites.google.com/site/climateadj/ocean_variance

    I’m not sure if this difficulty is important or not, but maybe it is?

  51. Mosher (#81898) correctly provides the context for the slide — the models ignore significant local/regional features (like the Andes) in pursuit of homogenized global averages.

    I am persuaded that the most probable nature of the Big Surprise is that if there is significant climate change it is more likely to be highly regionalized–some areas experiencing big change (good or bad) most others little or nothing.

    I recall something in the climategate emails where Phil Jones is dismissive of the supposedly odd preoccupation with regional climate factors and issues (like land use) by Roger Pielke Sr. But I think that for purposes of predicting and mitigating the Big Surprise, Pielke Sr’s perspective will be a hell of lot more useful than the Climategate clique.

  52. John

    I am assuming that the 0 – 360 along the x-axis represents longitude. In which case, why is the elevation at 0 different than the elevation at 360?

    Two not-mutually exclusive possibilities come to mind:
    1) three is a step discontinuity in the elevation in the cartoon model at 0/360. Similar step discontinuities appear elsewhere.
    2) it really is a cartoon intended to convey an idea. Not-withstanding Nick Stokes’s highly literal questions , the illustration which has all the hallmarks of being a cartoon is not intended to be a quantitative description of any particular model. The cartoonist didn’t take any particular care to make sure the cartoon contained no apparent inconsistencies.

  53. Lucia

    “The cartoonist didn’t take any particular care to make sure the cartoon contained no apparent inconsistencies.”

    That would be very ironic in a cartoon meant to emphasise the need for checking for internal inconsistencies.

  55. Oliver (Comment #81873)-It is when you get repeating sets of Andes extending out over the Pacific Ocean.

    I’m trying to find the interview with David Legates explaining this, not succeeding…

  56. Re: Andrew_FL (Comment #81913)
    September 22nd, 2011 at 3:09 pm

    Oliver (Comment #81873)-It is when you get repeating sets of Andes extending out over the Pacific Ocean.

    What, that bothers you? 😛

  58. The thing that made me laugh when I first saw the graphic was the caption in the upper right hand corner: “Schematic of Missing Mountain Range”.

    Is that anything like the “missing heat” and if we find one will the other be hanging out with it?

  59. boballab (Comment #81918)-It’s a bit inverted. We know exactly where the mountain range is in the real world, but in the model it isn’t there. The missing heat, well I wouldn’t say it doesn’t exist in the real world, only that in models we have a certain amount of heat, and we know where it is supposed to be…but finding all that theoretical heat building up in the real world? Well…

  60. Thanks Andrew_FL – re: fig 13 – good reference. Any idea if they fixed that problem? And if so, what did it do to the model output? (I’ve had a quick look around the R30 site but there are some broken links so I thought it quicker to ask!)

  61. Just read the Q and A section after the presentation and think I might have answered my own question. Seems that they are unlikely to have fixed it as it looks to be an inherent property of the R30 model. But I’d welcome something from someone who knows.

  62. On yesterday’s episode of Big Bang Theory, Leonard gave a talk which he started with a joke about a spherical cow. (I watch too much TV.) Lead author Sheldon Cooper disliked everything about the talk except the joke and got into a fistfight with Leonard after the talk. Penny stared blankly at the joke.

Comments are closed.