Over-reaction?

There was a fun press release from the University of Bristol about using a Global Circulation Model (GCM) to study the climate of Tolkien’s Middle Earth. A number of people on Twitter (Richard Betts, Gavin Schmidt, Doug MacNeall) seemed to find it quite amusing. As far as I can tell, it was mainly a bit of fun. This kind of thing can, however, have some real value in that it can generate some public interest and can make people more aware of this topic. Generally a good thing to do.

Roger Pielke Sr., however, seemed somewhat less taken with the idea and issued a series of Tweets which finished with

It does seem to be a little ironic to claim that someone’s made a gross over-statement, by then grossly over-reacting. Also, the quote from Richard Pancost in the press release actually says

“Because climate models are based on fundamental scientific processes, they are able not only to simulate the climate of the modern Earth, but can also be easily adapted to simulate any planet, real or imagined, so long as the underlying continental positions and heights, and ocean depths are known.”

So, as far as I can tell, there is no claim from those involved in the study that GCMs are based only on fundamental physics.

However, Roger does seem to be making a potentially interesting point – how tunable are GCMs? In an earlier post I discussed a 2011 paper by James Hansen. In this paper it’s argued that the response function (which determines – I believe – the rate at which the oceans at taking up energy) is probably too slow. The reason that the models are working alright is because they’re also underestimating the aerosol forcings, and so the two effects partly cancel. This would seem to imply an element of tunability, but not necessarily complete tunability (I think I may have just made up a word).

It’s certainly correct, I believe, that GCMs only directly solve some physical equations (Navier Stokes, for example) and that much of the rest is probably parametrized at some level. This, however, is very different to being tunable. Convection, for example, is very difficult to model on a grid because it’s a very small-scale process. Unless your resolution is incredibly high, you simply can’t capture convection properly. There are, however, semi-analytic ways (mixing length theory, for example) of including convection. This may well introduce a parameter, but the parameter is likely to be constrained by actual physics or actual observations.

Of course I’m not someone who’s ever run a GCM, so maybe Roger is completely correct, or maybe I’ve misunderstood what he’s implying. He certainly appears to be suggesting that GCMs are very tunable and that the results are very dependent on some free parameters. There may be some truth to this, but I would be surprised if GCMs contain lots of parameters that are completely free. There is a big difference between having a large number of completely free parameters, and having a set of parameters that are there to simulate actual physical processes, but that are constrained by known physics or by actual observations. So I would be interested if anyone who knows more than me can comment on whether GCMs are tunable to the extent that Roger seems to be implying. Bearing in mind that Roger appears to not have a sense of humour, maybe those who comment should be a little careful about what they say.

Advertisements
This entry was posted in Climate change, Climate sensitivity, Comedy, Global warming and tagged , , , , , , , . Bookmark the permalink.

41 Responses to Over-reaction?

  1. Rachel says:

    I thought this paper, submitted to the Journal of Hobbitlore, was very good for the reason you give – that it can generate public interest – but also because it has improved my understanding of climate models. If someone asked me what a climate model was I don’t think I could have explained it before now but after reading The Climate of Middle Earth I feel as though I could say something almost sensible.

    I don’t understand Roger’s criticism so can’t comment on that.

    I also think they got the climate of the Shire right. Peter Jackson chose poorly when he picked the North Island of New Zealand. It should have been set in the South Island, somewhere near Dunedin would have been perfect.

  2. Actually, the model description in the paper (which I really should have linked to in the post) is actually quite good. It mentions sub-grid-scale processes and parametrisations (and acknowledges that this introduces uncertainties). Also, the beginning of the paper very clearly states that climate models

    are based on our best theoretical understanding of fluid motion, physics, chemistry, and biology, written in the form of equations, and then converted into a form which can be solved by a computer.

  3. MikeH says:

    I suspect they are trying to promote their MSc in ‘Climate Change Science and Policy’.

    But forget Middle Earth. The climate of New Zealand cannot be that hard. How about modelling the climate of the Seven Kingdoms in Game of Thrones. Look at that ice wall – 700 feet high and stretching 300 miles across the northern border of the Seven Kingdoms. And remarkably it is snow and ice on one side and a Mediterranean climate on the other.

  4. Rachel says:

    Andthen,
    Yes, that’s the bit I was referring to and also the paragraph immediately after that which compares climate models with weather forecast models. He’s written that paper with the general public in mind.

  5. MikeH, you may be right. That does put a slightly different slant on it 🙂

    Game of Thrones – that would be a challenge. If you could run a GCM that produced such an ouput, then I may actually agree with Roger Pielke Sr. that they’re tuneable.

  6. Rachel, yes I agree, that was also a useful explanation. I’ll put it below for others

    Climate models and models used to make weather forecasts are very similar to each other, except that climate models typically simulate longer periods of time than weather models (years to centuries as opposed to days to weeks), and therefore, due to limits on computer time and power, make predictions at a lower spatial resolution (typical scale of hun- dreds of kilometers as opposed to kilometers).

  7. Roger Jones says:

    Yairs. This is an interesting pickle and one that comes up a lot.

    The basics that get the model running is physics. And as you quite rightly imply, there are schemes that tinker with these physical processes to better reproduce observations.

    However, and this is something that RPSr conveniently overlooks, tuning can only use observations to now. It is done to get cloud processes, ice and ocean represented better. It is necessary because of scale problems, incomplete understanding of processes and incomplete observations in the main. The climate change in future projections is an emergent product of the model code under GHG forcing. The future is not tuned. This is continually and conveniently overlooked by people who want to cloud the issue.

    Something that Hasselmann and Lorenz both understood is that climate science is a boundary problem and an initial conditions problem. The former governs where the climate eventually gets to and the other how it gets there. The boundary problem is largely dealt with by radiative physics (really well understood) with feedbacks from clouds, water etc. The initial conditions problem means that internal variability governs how climate change evolves.

    RPSr is claiming that the tuning in the models violates the initial conditions problem. He has a lot invested in showing that changing land surface feedbacks due to human intervention changes the boundaries substantially. I challenge this – they cannot provide the forcing sufficient to overcome GHGs and they do not provide the cloud feedback because we would already see markedly changed behaviour. No-one who argues that non-GHG forcing can overcome GHG forcing has been able to come up with a valid model that explains past climate changes.

    These models can also be adapted to different planets because of the basic physics, but aren’t tuned to the same detail because we don’t have met stations on Venus. But they do get the basics right. So why not Middle Earth?

    There wasn’t an AWS on Mordor when I last looked, and that bastard Sauron nuked all the satellites. So I reckon that RPSr has gone all Orcy because someone is messing with his precioussss!

  8. Roger Jones, interesting comment. Thanks. Certainly in my experience noone would bat an eyelid if someone presented a model in which certain parameters had been constrained by comparing with some observations and then used that model to make some kind of prediction. Seems a fairly standard practice – kind of how science works.

    It’s interesting that you say

    I challenge this – they cannot provide the forcing sufficient to overcome GHGs and they do not provide the cloud feedback because we would already see markedly changed behaviour. No-one who argues that non-GHG forcing can overcome GHG forcing has been able to come up with a valid model that explains past climate changes.

    In the post I wrote yesterday, I was trying to argue that there are some really simple arguments (a positive energy imbalance being what I was using) as to why global warming cannot be pre-dominantly natural. The only explanation that is both physically motivated and consistent with observations is anthropogenic forcing dominating. It still surprises me that some serious people are still trying to argue for a significant natural factor. Like you, I can’t see how this can work and have yet to find a model that can explain past climate changes.

    I had an exchange on this blog with Roger Pielke Sr. a few months ago, in which he seemed to be arguing that there is no evidence, yet, for any positive feedbacks. Given that the surface temperature has risen by 0.85oC and the net anthropogenic forcings are around 2.3 Wm-2, it’s hard to see how this conclusion is correct, but – unless I was misunderstanding what he was suggesting – Roger seemed to be suggesting exactly that.

  9. I rather like the idea of predicting results based on fundamental principles. If I roll a dice, or toss a coin, I can use some pretty basic principles to predict that the dice is likely to show a face, as will the coin. However if I could take that one stage further and use the basics to predict which face I would be a happy man. We could feed in the variables and make a better prediction, but for the prediction to be accurate, all those external variations would have to be exactly the same each time, which I’m sure you will agree is a bit of a challenge. Trends are the thing that are most reliable, but if the coin lands heads up 5 times in a row I know what the trend is, but the chance of it happening again is still 50/50. Nice site by the way. And try not to make a Hobbit of interpreting middle earth climatology, it will only Ent in tears.

  10. chris says:

    There’s a bit of a definitional issue here. All computational models are potentially tunable. After all the parameters can take any value one chooses to give them.

    I’m more familiar with the sorts of computational models that won the Chemistry Nobel Prize this year, but these have similar qualities as climate models in that the values given to particular parameters are assigned according to empirical evidence or basic physics. In fact one of the basic tenets of computational modeling is that one doesn’t fiddle about with the values of the parameters in order to match “reality” in any particular run, although one might do multiple runs with variation of parameter values to test likely effects on computational outcomes, especially where the value of a parameter isn’t terribly well constrained empirically or by physics.

  11. Chris,

    Yes, I agree. That’s why I added this to the post

    He certainly appears to be suggesting that GCMs are very tunable and that the results are very dependent on some free parameters

    It really depends on what Roger means by tunable. If he just means that some parameters can be varied, then I would imagine he’s correct. On the other hand, if he’s implying that there are parameters that are completely free and are unconstrained by physics or observations, then that would seem a little extreme. I don’t know if it’s true or not, but from what yourself and others have said here, it sounds more like there are parameters that can be adjusted but that are still constrained in some sense.

    Gareth, thanks for the comment. I’m trying to think of a suitably witty, Lord of the Rings like response, but it’s escaping me at the moment 🙂

  12. Tom Curtis says:

    There are several things to be said about the climate modeling of Middle Earth, the most obvious of which is that it failed to reproduce the climate of Middle Earth. This can be seen clearly in the Forodwaith which is shown as far to warm relative to observations. The predicted broadleaf forest over the grasslands east of Mirkwood are also a concern.

    It is difficult to assess whether is a problem with model physics, or the set up. Certainly, the set up does not match geographically with Middle Earth as it is known. In particular, Tolkien said:

    ““The action of the story takes place in the North-West of ‘Middle-earth’, equivalent in latitude to the coastlands of Europe and the north shores of the Mediterranean (…) If Hobbiton and Rivendell are taken (as intended) to be about the latitude of Oxford, then Minas Tirith, 600 miles south, is at about the latitude of Florence. The Mouths of Anduin and the ancient city of Pelargir are at about the latitude of ancient Troy.”

    (Quoted from here, note, the map at that site is a load of crap)

    Being more specific, The Grey Havens, the Shire, and Rivendel are all approximately on the same latitude, about that of Oxford (52 degrees North) on modern Earth. Contrary to that, the model places the latitude of the Grey Havens at about 60 degrees North. The model does correctly place Pelargir and the Mouths of the Anduin at about 40 degrees North, however, but the consequence is that Middle Earth is expanded in scale.

    The more northerly latitude of the Shire in the model is difficult to reconcile with the the warmer than observed Forodwaith. Clearly the observed Middle Earth is colder overall than is shown in the model, a fact possibly attributable to the lower solar constant in the past, but more probably attributable to a lower CO2 content, thus explaining the reasonable temperature estimates for more southerly lands in the model. A CO2 content significantly lower than the preindustrial value used would allow colder northerly latitudes due to arctic amplification.

    The poor model geography is not restricted to just the poor scaling of Eriador (the territory west of the Misty Mountains) and Rhovanion (that east of the Misty Mountains as far as the Iron Hills). The model topography is clearly shown as including Numenor and the undying lands. That clearly identifies it as a model of the Second Age of Middle Earth. It is, however, purported to by a model of Third Age Middle Earth, as clearly indicated by its purported (see below) author, Radagast the Brown. Radagast as one of the Istari, did not arrive on Middle Earth from the undying lands until 1050 of the Third Era. Indeed, as Middle Earth was flat until the fall of Numenor, at which time the undying lands were removed from the Earth and became only accessible on elven vessels, it is impossible that the model be both of the Second Age, and hence also impossible that the model topography is correct.

    I need not go into the entirely speculative extended eastern and southern continents shown in the model topography.

    In addition to representing poor scholarship, the article is a transparent forgery. This is most clearly seen in the fact that the “Elven” and “Dwarvern” versions are merely transliterations of English into the Tengwar and Cirth scripts. Even the identification of Cirth as Dwarven is incorrect, it having been originally devised by Elves, and merely adopted for the Dwarves. Had “Radagast” being who he claimed to be, he would no doubt have composed the article in quenya, the more scholarly of the two elvish languages. He also would probably not have produced a Khuzdul version, the dwarves being secretive and not teaching the language to others.

    I am disappointed that you have ignored these clear modeling flaws, not to mention the transparent forgery to concentrate on trivia about some inconsequential “climate scientist” who appears not to have the least knowledge of Middle Earth.

  13. Tom Curtis says:

    Sorry, but I had to do it.

  14. Tom, I would recommend submitting that as an open review of the paper. I’m sure the authors would appreciate it 🙂

  15. verytallguy says:

    Tom Curtis.

    You’re not an expert on Middle Earth climate, you’re a very naughty boy.

    [sorry, felt I had to as well]

  16. Tom, are you an Elvish impersonator?

  17. Tom Curtis says:

    In order:
    AndThen, how would I do that?

    vtg, isn’t it possible I’m both?

    Gareth, I may once have been, in another life, Cutholen of the Laiquendi.

    [Third attempt at posting. Please delete prior attempts if they show up.]

  18. DrTskoul says:

    Even solution of Navier-Stokes involves some sort of “parameterizaton”. Also problems of mass transfer/heat transfer/mixing are cast as phenomenological models involving some sort of empirical models for the exchange coefficients. What most people fail to get, mostly because nobody from the scientific field tells them, is that most parameters are set using standard problems, where the solution is already known analytically or we have enough data for. After the initial setting, the parameters are kept constant. Those parameters are capturing the proper physics. Therefore these are physical models, but not ab initio problems. Similar to the computational chemistry codes that were mentioned above.

  19. DrTskoul says:

    “ab-initio models” not problems….

  20. DrTskoul, thanks. Yes, you’re right, there’s lots of “parametrization” in such models.

    What most people fail to get, mostly because nobody from the scientific field tells them, is that most parameters are set using standard problems, where the solution is already known analytically or we have enough data for.

    You may have a point here, but I think there are also lots of people who are misrepresenting (knowingly or not) the uncertainties associated with these parametrizations. It’s hard for scientists to counter some things when they then get accused of being dishonest, for example.

  21. DrTskoul says:

    True. We, from our side, we gave to always argue that these are indeed physical models. They are not just empirical/fitted equations. The Pielke’s of the world are knowingly misrepresenting, to fit it in their agenda.

    Their over-reaction is fake.

  22. Lars Karlsson says:

    Tom, the paper says it is a model of the second age (see e.g caption of Fgi 1).
    This is a serious topic, so please read the paper carefully before you comment on it.

  23. John Mashey says:

    Well, before I disappear into AGU today, I can’t resist this, given the strange combination of relevant items:

    1) strongly recommend RealClimate’s FAQ on climate models Part i and Part ii.

    2) See this comment for discussion of ways in which various disciplines tend to misunderstand climate modeling, including one where a biochemist simply didn’t believe they could work … until it was realized that he’d been exposed to some of the early protein-folding codes, which of course have to make exactly the right choice at every stage, and when they didn’t, could diverge radically from reality, i.e.. initial value problem with touchy steps, rather than boundary value problems of GCMs. This wasn’t obvious, since the number of people exposed to both computational chemistry codes and climate models is … small. 🙂

    3) As it happens, there is an amusing intersection. Among other things, I used to architect supercomputers and then visit/talk with customers/prospects about their problems, see how our systems could help them, and talk about our directions. (When expecting to spend $Ms, such customers liked to talk to Chief Scientist-types.) Hence, starting in the 1990s, I talked with folks at NCAR, GFDL, NASA, etc and they used big SGI systems for climate models.

    4) During the 1990s, I used to visit NZ every year on business, most years including a visit to Wellington to see Weta Digital, which started as a garage-shop-size effort, although by 1999 was rather larger. They of course used SGI gear for LoTR, and I helped them with planning their servers/ storage systems, which were the same SGI Origin computers used for climate modeling, albeit not quite as large. But they did have Terabytes of expensively-created data, back when a TB was still Big Data. Now, at that time I don’t think they were running GCMs, 🙂 as their models were more for those giant crowd scenes, where they pioneered interesting & respected new algorithms, like when orcs see a catapulted rock arriving.

    The last time I visited, since they’d seen me for years, they showed me half an hour of early footage about a year before FotR came out, which is rare (the movie folks are generally very secretive).
    Of course, I had to sign a (strong) Non-DIsclosure Agreement, of which I’ve signed many.
    But they had an unusual kicker, which only they could claim:
    “And if you tell anyone anything about this before it comes out, we have special halfbreed orc-lawyers to send after you.” FEAR! Regular lawyers are bad enough!

    5) Finally, just last year, Jim Salinger (well-known NZ climate scientists) had former NZ Prime Minister Helen Clark,@ Stanford to talk, very good. During Q&A I could not resist asking this que4stion:
    “I’ve been dying to ask YOU a question for 10 years. Did I really see you on the Travel Channel running around NZ with someone who was NZ “Lord of The Rings Travel Minister”? Was there really someone with that title?”
    She laughed and spent 5 minutes’ answering that one. (Yes, yes, and they’d create such jobs when needed.) LoTR of course was a huge boost for NZ tourism.

    Anyway, Middle Earth, NZ, computer models: an unexpected journey.

  24. Tom Curtis says:

    Lars Karlsson, it cannot be a model of the 2nd Age of Middle Earth, for in the 2nd Age Middle Earth was flat. The author attempts to tackle that problem, saying:

    “A decision had to be made of how to wrap the apparently flat, circular world of Mid-
    dle Earth onto the sphere required by the climate model. I chose the relatively simple solution of a straightforward direct mapping of the circular Middle Earth onto an equal-latitude/longitude grid. The missing corner regions were set to be oceans.

    However, as hardly needs saying, a flat world will not generate the same winds as those found on a globe. Therefore by that decision they cease to model the climate of 2nd Age Middle Earth.

    However, my critique does in fact contain a flaw, for I find now that their arrangement of the continents does have a basis in Tolkein’s works, albeit a map of the First Age, and non-canonical.

  25. That's MR BALL to you. says:

    So does that mean it will be impossible to ever model the climate of the Diskworld? But I though sufficiently advance science was indistinguishable from magic?

  26. Tom Curtis says:

    Mr Ball, I am sure the climate of Diskworld can be modeled as a static disc with a continuous flow of water and air over the outer edge. This may not reproduce the climate “observed” in the Diskworld novels. No doubt we will have somebody along shortly asserting that this proves the models to be unreliable 😉

  27. BBD says:

    What a wonderful thread.

  28. BBD, I was thinking the same. I’m insufficiently quick-witted to participate though 🙂

  29. Rachel says:

    It’s been keeping me entertained all day 🙂

  30. Lars Karlsson says:

    Tom Curtis, you are a flat-middle-earther!

  31. It’s difficult to imagine a model that doesn’t parametrize something. Lattice QCD is so fundamental that it might satisfy Dr. Pielke Sr, but even it has to specify a grid.

    It sounds like Dr. Pielke Sr. is waiting for a supercomputer fast enough to model the Earth’s climate baryon by baryon using lattice QCD on a grid with Planck length spacings. Given that modelling a few protons brings our fastest supercomputers to their knees, it might be a while before we can simulate Earth’s 6E24 kg of matter in this way. (Maybe Dr. Pielke Sr. will mercifully let us parametrize the interior of the Earth as a slowly decaying heat source, but then again he might insist that we also model the Sun’s 2E30 kg baryon by baryon lest we fall prey to the evils of parametrization.)

    Surely it’s a coincidence that this ridiculous burden of proof is equivalent to delaying any action on climate until God lends us His laptop.

  32. Dumb Scientist – but aren’t a lot of Lattice QCD simulations simply Monte Carlo simulations? Isn’t that the ultimate in parametrisation 🙂

  33. Tom Curtis says:

    Lars, only till the downfall of Numenor 😉

  34. Sure, if there were just one universe. But since we probably live in a multiverse, Monte Carlo simulations are just reminding us that there’s a universe exactly the same as this one but without any shrimp, or where Britney Spears teaches semiconductor physics.

  35. DrTskoul says:

    Hilarious!!! BS teaching QM.

  36. Lars Karlsson says:

    Tom Curtis, Earth was not really flat: it was only commonly believed to be flat. It was all a fraud concocted by the Eldar Protection Agency in order to discourage exploration.The surviving Númenóreans eventually exposed it.

  37. Perhaps it was a particularly quirky way of RPSr to be funny … I mean, no way on middle-Earth he could have meant that comment seriously. There just isn’t so much straw around to do that.

  38. jyyh says:

    spite Tom Curtis’ hard critique, this is a landmark study on the climate of the Middle Earth. In the defence of the veritable wizard, the few curiosities he presents about the study might be explained if there is a slight slant to right in their north arrow on their map. This might have happened on the step where they assumed a spheroid world, in map transformations there are a few difficulties with near-zero values. Additionally the standard measure of lenght has been a ‘ranga’ (a lenght of one step by an ‘average man’) which probably has been shortened because ‘in general the height of Numenoreans has decreased ever since they were forced to leave their island.’ Maybe this can be explained by the changes in diet. This could have led the author to underestimate the size of the rounded Arda. But nevertheless some of Curtis’ critics stand, and further study might be required.

  39. AnOilMan says:

    You people should listen to a real scientist when it comes to this stuff. There is nothing to fear from the rolling plumes of mount doom. It will bring nothing but benefits to all of middle earth. In fact last year was warmer than the year before so more cloud cover will only offset that effect.

    Besides the Orcs aren’t so bad once you get used to them, see;

    Seriously though, I want to see a climate model for Endor. That’s a lot of firewood, so it must rain a lot.. but if the trees pull carbon out of the atmosphere, wouldn’t that eventually kill the trees? And where does the water come from?

  40. DrTskoul says:

    Benefit myths abound. Iris effect is in the minds of old men. Don’t expect deus ex machina.

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