Bob Carter’s cherry picking

Last year, global warming denialist Bob Carter wrote a Tech Central Station article where he claimed that satellite measurements

show little or no long-term trend of temperature change.

I emailed him to point that the satellites actually showed significant warming. He replied that this didn’t count because:

this trend is most likely produced by the single exceptionally warm 1998 El Nino year.

This year, he has written a paper where he asserts (my emphasis):

Four alternative predictions of near-future climate, based on empirical models drawn from the palaeoclimatological record, are described. Three agree that the likely trend the 21st century is one of cooling, and the fourth (based on Milankovich predictions) predicts cooling over the longer term. In keeping with the generality of these predictions, averaged global surface temperature has been falling for the last 6 years.

That is, of course, only true if you include the single exceptionally warm 1998 El Nino year.

Comments

  1. #1 TallDave
    May 17, 2005

    Bah, I hate small sample sizes. So unreliable. Why does anyone bother?

  2. #2 Scott Church
    May 18, 2005

    It seems clear that Mr. Carter understands neither the satellite record nor small trends in the upper atmosphere. The
    1998 ENSO event he refers to was preceded by one in 1981-82 that was almost as large–but of course we can’t ignore
    that one because doing so would pull down the early end of the time series and remove the cooling trend

    that
    is so precious to skeptics. And of course (why doesn’t this surprise me?), he seems to be drawing only on early versions

    of the UAH satellite retrieval (Christy et al., 1995; 1998) and ignoring the work of RSS (Mears et al., 2002; 2003;
    2003b), Fu’s team (Fu et al., 2004; Fu and Johanson, 2004; 2004b), and others. As Tim points out, he also ignores
    virtually all of the latest research on the upper-air record. For more on the satellite record and skeptic abuse of it
    (of which he seems to be typical) see my upper-air review and my paper on
    skeptic abuse of it.

    As for his “four alternative predictions of near-future climate” he offers no citations or details of any kind, but he
    seems to be referring to comments by Richard Lindzen, vociferous and highly paid climate skeptic extraordinaire, in
    regard to models discussed in the IPCC Third Assessment Report (IPCC, 2001). He doesn’t even bother to cite
    Lindzen’s remarks. Damn, this guy wants his tracks covered. Can’t say I blame him–if he actually cited any of his references someone might be tempted to check the quality of his work.

    A check of the Third Assessment Report reveals no models that indicate cooling for “near-future climate”. Even if they
    did, why would anyone use a paleo-climate model to predict near-term climate trends and assess whether short-term
    anthropogenic impacts were affecting them? Well, probably because they weren’t getting the answer they or their
    corporate benefactors wanted. Naturally, anyone truly interested in evaluating climate trends as represented by models
    would use state-of-the-art general circulation models–which, by the way, have come a long way even since the Third
    Assessment Report. Carter might have used GISS SI2000 for instance using runs incorporating deep-ocean models like
    Ocean’s B or E (Hansen et al., 2002; Sun and Hansen, 2003), both of which do an excellent job of reproducing recent past
    trends regionally, globally, and vertically based on both the satellite and radiosonde records. He might also have used
    HadCM3 (Stott et al., 2000) or DOE PCM (Meehl et al., 2003; 2003b).

    But of course, this would have wrecked the party. As for Milankovitch cycles, they are driven by changes in the earth’s
    orbit and/or changes in its rotation or nutation with respect to the ecliptic caused by gravitational perturbations from
    near encounters with asteroids or other astronomical objects. They typically involved response time constants on the
    order of several millennia. Some have become apparent only after 50,000 to 100,000 years. They have absolutely
    nothing whatsoever to do with climate change over the next few centuries, anthropogenic or otherwise.

    From what I can see, not only does Carter have little understanding of up-to-date climate change research, he appears to
    have little or no understanding of the most basic concepts in the field. And all this even before he begins his
    cherry-picking.

    REFERENCES

    Christy, J.R., R.W. Spencer, and R.T. McNider. 1995: Reducing noise in the MSU daily lower-tropospheric global
    temperature dataset. J. Climate, 8, 888-896.

    Christy, J.R., R.W. Spencer, and E.S. Lobl. 1998: Analysis of the merging procedure for the MSU daily temperature
    series. J. Climate, 11, 2016-2041.

    Fu, Q., and C.M. Johanson. 2004: Stratospheric Influences on MSU-Derived Tropospheric Temperature Trends. J. Climate,

    To be published Dec. 15, 2004.

    Fu, Q., C.M. Johanson, S.G. Warren, D.J. Seidel. 2004. Contribution of stratospheric cooling to satellite-inferred

    tropospheric temperature trends. Nature, 429, (6987), 55-58.

    Fu, Q., C.M. Johanson, S.G. Warren, D.J. Seidel. 2004b. Atmospheric science: Stratospheric cooling and the troposphere

    (reply). Nature, 432, doi:10.1038/nature03210 Brief Communications. Dec. 2, 2004. Abstract available at

    http://www.nature.com/cgi-taf/DynaPage.taf?file=/nature/journal/v432/n7017/abs/nature03210_fs.html. Accessed on Dec. 27,

    2004. Subscription required for access to the full article.

    Hansen, J., M. Sato, L. Nazarenko, R. Ruedy, A. Lacis, D. Koch, I. Tegen, T. Hall, D. Shindell, B. Santer, P. Stone, T.

    Novakov, L. Thomason, R. Wang, Y. Wang, D. Jacob, S. Hollandsworth, L. Bishop, J. Logan, A. Thompson, R. Stolarski, J.
    Lean, R. Willson, S. Levitus, J. Antonov, N. Rayner, D. Parker, and J. Christy. 2002. Climate forcings in Goddard
    Institute for Space Studies SI2000 simulations. J. Geophys. Res., 107 (D17), 10.1029/2001JD001143.

    Mears, C.A., M.C. Schabel et al. 2002: Correcting the MSU Middle Tropospheric Temperature for Diurnal Drifts.
    Proceedings of the International Geophysics and Remote Sensing Symposium, Volume III, pg. 1839-1841, 2002.

    Mears, C.A., Schabel, M.C., Wentz, F.J. 2003: A Reanalysis of the MSU Channel 2 Tropospheric Temperature Record. J.
    Climate, 16 (22), 3650-3664.

    Mears, C.A., Schabel, M.C., Wentz, F.J. 2003b: Understanding the difference between the UAH and RSS retrievals of
    satellite-based tropospheric temperature estimate. Workshop on Reconciling Vertical Temperature Trends, NCDC, Oct.
    27-29, 2003. Available online at http://cimss.ssec.wisc.edu/itwg/itsc/itsc13/session9/9_mears.pdf.

    Meehl, G. A., Washington, W. M., Wigley, T. M. L., Arblaster, J. M., Dai, A., 2003. Solar and greenhouse gas forcing and

    climate response in the twentieth century J. Clim. 16, 426-444.

    Meehl, G. A., Washington, W. M., Arblaster, J. M., 2003b. Factors affecting climate sensitivity in global coupled
    climate models. Paper presented at the American Meteorological Society 83rd Annual Meeting, Long Beach, CA.

    Stott, P.A., Tett, S.F.B., Jones, G.S., Allen, M.R., Mitchell, J.F.B., and Jenkins, G.J. 2000. External control of 20th

    century temperature by natural and anthropogenic forcings. Science 290 (5499): 2133-2137.

    Sun, S., and J.E. Hansen. 2003. Climate Simulations for 1951-2050 with a Coupled Atmosphere-Ocean Model. J. Climate,
    16, 2807-2826.

  3. #3 Louis Hissink
    May 18, 2005

    In keeping with the generality of these predictions, averaged global surface temperature has been falling for the last 6 years.
    That is, of course, only true if you include the single exceptionally warm 1998 El Nino year.

    Yes, if we include 1998 as a high year.

    Son of a gun – Tim Lambert has discovered something.

    What, though, what…..

  4. #4 Steve Bloom
    May 19, 2005

    Regarding Scott’s mention of Lindzen, my impression had been that he had not been tainted with the ExxonMobil (or similar) funding that has wrecked the credibility of so many skeptics. Is there information to the contrary?

    I’ve seen elsewhere (sorry, no link) that so far 2005 is on track to be even warmer than 1998; we shall see if this holds up.

    On the general subject of the recent warming trend of the last 15 years or so, my understanding is that what respectable climate scientists are able to say is that while this trend is consistent with the (anthropogenic) warming predicted by the models, it is not yet of sufficient length to call it climate change.

  5. #5 Scott Church
    May 19, 2005

    Steve, My info on Lindzen comes from a few places. Ross Gelbspan’s 1998 book “The Heat Is On” discusses his activities and funding. Gelbspan interviewed Lindzen, who told him that he consulted for industry as a skeptic scientific expert at a rate of $2500 per day (!), and Gelbspan also reports that he has received expense funding and underwriting for various skeptic project from Western Fuels (a coal-fired power company) and OPEC. He’s been involved at one time or another with the Cato Institute, Tech Central Station, and the Annapolis Center for Science-Based Public Policy–all industry funded ultra-conservative special interests. He was interviewed by Scientific American a few years back, and reported that he’d made some $15,000 as an industry consultant the previous year which at the rate he quoted to Gelbspan amounts to 6 days of work. You can also find more on his industry funding and connections in the ExxonSecrets fact sheet about him.

  6. #6 Louis Hissink
    May 20, 2005

    Scott,

    So,

    And how do YOU earn your living ?

  7. #7 Dano
    May 20, 2005

    An amazing amount of data goes thru Scott’s hands, some of which he channels off and does a nice job of analysis. His living sharpens his analytical skills so he can write stuff that we use, Louis, to highlight your dissembling.

    Now make yer point Louis.

    D

  8. #8 Scott Church
    May 20, 2005

    Louis, For 20 years plus I earned my living as an aerospace engineer. My MS is in Applied Physics. Today I am a Network Administrator and run a Data Center for a global e-commerce operation (Getty Images). I could easily bore you for hours with my many exploits in both industries, but I don’t discuss them here. Like mining, neither has any bearing on climate change (though my physics training puts me far closer to atmospheric sciences than geology would) and bringing them up here whenever possible would be little more than an attempt to toot my own horn in the hopes of somehow getting more credibility for my remarks than my data and conclusions merit. Anyone who believes my statements are off base is welcome to present properly researched data and analysis that says otherwise–I, and everyone else here would benefit. Beyond that, whether I’m an engineer, the Messiah, or a Baghdad camel proctologist is irrelevant. All the best.

  9. #9 Dano
    May 21, 2005

    Your quackpottery made Scott forget to say he does a nice job of analysis, Louis. Don’t ruin our amusement at your half-baked statements by making people upset.

    D

  10. #10 z
    May 23, 2005

    Well, I’m not ashamed to say that I receive a huge salary from the powerful and wealthy environmental cartel in return for nothing more than promulgating sheer lies about the innocently victimized energy companies and their noble political defenders, so that the wealthy and powerful academic climatologists can maintain their vise-like grip over all our lives.

  11. #11 David
    May 25, 2005

    Louis, if we have a year hotter than 1998 – say this year (which is a good chance) – will you dump your nonesense beliefs that global cooling might be around the corner? I guess not, but its worth asking.

    BTW Scott, are you aware of some very recent work by Mears et al. apparently to be published in GRL which has identified a significant error in the Spencer and Christy MSU datasets?

    David

    PS

    I have a vested interest in having science prevail over quakery…

  12. #12 Scott Church
    May 25, 2005

    David, I’m not aware of the new Mears et al. work but I’d love to see it! Do you know where I could get a preprint in PDF format, or at least a full abstract? There doesn’t appear to be one at RSS’s web site and so far, I haven’t come up with one on the Internet. Thanks.

  13. #13 David
    May 26, 2005

    Scott,

    am yet to see anything, but have heard bits and pieces from friends at NCAR. I will see if I can get something concrete – this will probably take some days.

    Regards,

    David

  14. #14 David
    May 27, 2005

    Scott,

    are you aware of the paper “Satellite-Derived Vertical Dependence of Tropical Tropospheric
    Temperature Trends” by Fu and Johanson which has been submitted to GRL? This deals with the issue of diurnal drift in the Spencer and Christy datasets, which appears to heavily dampen the warming trend in the MSU2lt dataset. The results are very convincing… and I understand that the MSU2lt data set is being revised to address this problem. I can email you a copy if you like?

    One will have to wait for the fully revision to occur, but one can extrapolate that the corrected warming trend should come out near 0.2C/decade – near or slightly ahead of the surface trend.

    These results are pretty exciting, as they demonstrate the dangers of overturning centuries old thermodynamic theory in favor of a short and patchy data record (as has been done by the sceptics). It is a no brainer that warming tropical sea surface temperatures must lead to a warmer tropical troposphere (as failure to do so will make the atmosphere convectively unstable, triggering convection which will warm the troposphere – this is why the lapse rate sites on the moist adiabat). It would seem faith or ignorance have led some sceptics to ignore such simple theory.

    One might hope that the sceptics now turn their attentions to where the real uncertainties in climate change are – how much will feedbacks enhance or suppress the anthropogenic greenhouse affect. One can hope, but I fear such hope is unfounded…

    Regards,

    David

  15. #15 Scott Church
    May 28, 2005

    David, I’d love to see the Fu and Johanson work. As it turns out, I hail from Seattle, and Fu and Johanson are at the University of Washington not 15 minutes from my house. I got some valuable feedback from him when preparing my own papers, and from one of his original Fu method co-authors (Dian Seidel). Shoot me a copy of the paper through my web site email and I’ll give it a read. Thanks, and all the best!

  16. #16 anon
    May 30, 2005

    There is a very simple refutation of most global warming research: the
    scientists have no way of knowing whether their models are overfitting the
    data.

    Eg, take the research described here: http://climateprediction.net

    There are so many poorly understood parameters in the model that by choosing
    different plausible settings for those parameters you can get “an Earth that
    boils or freezes, or oscillates between very hot and very cold every couple of
    years” (http://climateprediction.net/science/strategy.php)

    So what do the scientists do? Instead of being honest and admitting that one
    should not draw strong conclusions from such a poorly understood model, they
    search for parameter settings that give close to the predicted historical
    climate data for some past period, and then using those parameters they look
    at the impact of increased CO2 levels on future climate.

    The problem with this is that with so many tunable parameters (and such
    discontinuous model behaviour as a function of those parameters), you cannot
    know whether model has overfitted the historical data, and hence has little
    predictive power. Any good statistics PhD will be able to tell you that.

    This is by far the biggest problem with the climate change debate. It is just
    bad science.

    We should never forget that the scientists concerned are almost invariably
    driven by political ideology. Heed the words of the late Aaron Wildavsky,
    professor of political science at UC Berkeley:

    “Warming (and warming alone), through its primary antidote of withdrawing
    carbon from production and consumption, is capable of realizing the
    environmentalist’s dream of an egalitarian society based on the rejection of
    economic growth in favor of smaller population’s eating lower on the food
    chain, consuming a lot less, and sharing a much lower level of resources much
    more equally.”

  17. #17 sarcasto
    May 30, 2005

    Another expert who’s never studied climatology or anything related to it, yet who knows everything about everything thanks to his superior intellect and meta-analytic insight. [You can tell he’s no girrll!]
    Truly we ordinary, mere mortal climatologists are blessed that such minds as “anon”‘s have nothing better to do with their time than to educate the world via their regular (very, very regular) contributions to blog comments. And to top it all – off a quote from a political scientist! Thanks, anon!

  18. #18 Eli Rabett
    May 30, 2005

    Actually anon, the results of the climateprediction.net first experiment, using a relatively simple model were reasonably bounded http://www.climateprediction.net/science/pubs/nature_knutti_180402.pdf
    and, if anything, pointed to a higher climate sensitivity then the now four year old IPCC TAR.

    But, given that there are so many free parameters, how can the GCMs be overdetermined?

  19. #19 Tim Lambert
    May 30, 2005

    I think Mr anon doesn’t know the difference between underdetermined and overdetermined. This does not inspire confidence in his or hers claims about climate models.

  20. #20 Scott Church
    May 30, 2005

    What interests me most about global warming skeptic arguments like Mr. Anon’s is how predictable they are, and how limited in scope. “Refutations” of climate models are a case in point. Without exception, virtually every single one I’ve ever encountered,

    a) Does little more than wax eloquent about how wishy-washy and uncertain climate models are… without once ever presenting specifics about any actual model.

    b) Sooner or later ends up on a Far-Right ideological rant.

    Mainstream climate scientists encompass a wide range of philosophies, religions, and political views apart from their science. It’s telling that virtually without exception, everyone who thinks climate models are “fatally flawed” and is “unconvinced” by the “theory” of global warming is possessed of Far-Right ideologies or religions. One is reminded of the creationists and intelligent design types who insist on their “purely scientific” skepticism toward a century and a half of biology and their desire for “open-minded dialog” on origins, yet always turn out to be far-Right Bible-Belt fundamentalists.

    Once, just once before I die, I’d like to see one of these global warming skeptics actually attempt a specific and thought out examination of a real climate model–like say, HadCM3, DOE PCM, or GISS SI2000, with multiple forcing scenarios and oceanic, sea-ice, and other components–and demonstrate these fatal flaws they keep harping about. At least Mr. Anon has cited the Climateprediction.net site and some of the work there, a small (very small) step in the general direction of thoroughness… but of course, he never gets around to any specifics.

    But for once, just once, can we leave please the Right Wing ideology out of it?… If I wanted to hear a rant about those damn “environmentalists” and their anti-corporate wealth, anti-pollution, anti-Bush, un-American (or un-Australian), “anti-growth”, socialism I’d arrange to have Ann Coulter or Tim Blair put in a cell with Michael Moore and some knives and blow-torches, and kick back with a beer and enjoy the ensuing festivities. I come here to discuss science, and hopefully to learn something.

  21. #21 sarcasto
    May 30, 2005

    Another expert who’s never studied climatology or anything related to it, yet who knows everything about everything thanks to his superior intellect and meta-analytic insight. [You can tell he’s no girrll!]
    Truly we ordinary, mere mortal climatologists are blessed that such minds as “anon”‘s have nothing better to do with their time than to educate the world via their regular (very, very regular) contributions to blog comments. And to top it all – off a quote from a political scientist! Thanks, anon!

  22. #22 sarcasto
    May 30, 2005

    Sorry!

  23. #23 anon
    May 30, 2005

    Funnily enough, I do think I know what I am talking about. I suggest sarcasto, Rabbett, Church and Lambert start reading around on the theory of uniform convergence of empirical processes. A good starting text is Devroye et al “A Probabilistic Theory of Pattern Recognition”.

    The climate models are _underdetermined_ by the data, not overdetermined. Once that happens, you cannot make claims about performance of the model off the dataset on which it was tuned, without making strong assumptions about the
    complexity of the model class (eg bounded apnik-Chervonenkis
    dimension).

    For climate models it is even worse: the usual theory of empirical process convergence assumes independent and identically distributed data. But the climate is not independent (today’s weather depends on yesterday’s) and not
    even stationary, so you need a measure of the mixing time of the climate process to determine how much historical data is sufficient for the tuning process not to overfit.

    Without being a climatologist, it is hard to guess, but I’d say the climate mixing time must be at least equal to the “time constant” of the oceans (ie, how long it takes the ocean temperature to equilibriate with atmospheric temperature). Given the size of the oceans (big), that would have to be a long time. That implies you need a _lot_ more data than is currently available to tune the models reliably.

    And please note, these are general arguments that apply to any modeling scenario in which you estimate parameters based on empirical data and then extrapolate to unseen data. The theory is independent of the application.

  24. #24 anon
    May 30, 2005

    That should be “Vapnik-Chervonenkis dimension”

  25. #25 anon
    May 30, 2005

    That last post doesn’t read as well as I intended. Forget the maths for a moment – the idea is very simple: if you are tuning a complex model to match empirical data, and you wish to use that model predictively, then you have to validate the predictive power of the model somehow.

    The easiest way to do it is to use totally independent “test data sets” – independent in the sense that they were not used in _any_ way to tune the model parameters (strictly, they should be something the researchers have never even seen before). If the model predicts the independent test sets well, then you have some confidence that they’ll predict future data (but even then, lack of stationarity in climate is a problem for the statistics).

    But I don’t see any such validation going on in the results reported so far from climateprediction.net. Instead, as far as I can tell, they just tuned the parameters to find non-extreme (no ice-age or boiling) settings, then doubled the CO2 levels and drew this graph: http://climateprediction.net/science/results_cop10.php.

    From that they claim there is a much greater global warming potential than previously thought. And it gets a run in nature to boot. This is dishonest at best.

  26. #26 sarcasto
    May 30, 2005

    1. If you have an issue with something on climateprediction.net, whatever its possible relevance to Tim’s post – wouldn’t it be an idea to tackle it/them/whatever with it?

    2. Really, looking now at the “funnily enough” post above, if you have to get this stuff off your chest you might like to try getting a response somewhere like realclimate.org. Why are we talking about Vapnik-Chervonenkis or climateprediction.net in the comments to a post on Bob Carter’s delusions of competence in climatology? Do you think Bob the geologist would have a fleeting notion of what you’re on about, or doesn’t it matter to you because all science is only politics anyway (if you’re doin’ it right)? Why _did_ you quote a political scientist in your opening gambit?

  27. #27 Scott Church
    May 31, 2005

    Anon, I suggest that you actually examine a real climate model rather than indulge in vague speculations about them. Real climate models, like HadCM3 or GISS SI2000 (Hansen et al., 2002, J. Geophys. Res., 107 (D17); Sun and Hansen, 2003, J. Climate, 16) are not built from datasets alone–they’re build from known physical principles that are then checked against a wide range of independent datasets. They are also tested with countless different forcings reflecting many different scenarios, observed and not. Most use multiple components which can be tested independently of each other. GISS SI2000 has no less than 5 ocean representations, only one of which is based on actual past datasets (Ocean A). When these models are tested–“tuned”–it’s done over many, many scenarios and establishes among other things a range of climate sensitivities (see for instance IPCC, 2001, Chapter 8). It’s never as simple as throwing something together, using “a” dataset to “tune” it prior to making sweeping predictions.

    Your attention to mathematical details is admirable, and in general, considerations of how data are used vis a’ vis prediction is important. But you cannot meaningfully critique how any of this plays out in state-of-the-art general circulation models of climate if you haven’t any idea how those models actually work.

  28. #28 David
    May 31, 2005

    Prehaps anon you might consider going back to high school and learning a little physics. Climate models are primarily dynamical, with the dynamics reflecting conservations of energy, mass, water, and momentum. Their ain’t much ability to fiddle with these laws.

    It is not clear to me whether you simple don’t know what your talking about, or are an idealgoue spreading nonesense pseudo science. Either way, your clearly not a meteorologists or an oceanographer – ie an expert! Prehaps you might elighten us?

    Regards,

    David

  29. #29 anonn
    May 31, 2005

    It seems there is some kind of filter in place or I am doing something wrong or there is a bug in the software, because if I post as “anon” my posts don’t appear. Hence the change in handle.

    sarcasto, your response to my first post was to completely denigrate me. When
    I responded with more details, instead of apologizing or at the very least
    addressing my substantive point, you effectively tell me to post elsewhere.

    I did try to post this against the “bingo” thread but for some reason it
    didn’t work, so I stuck it here. And as a large fraction of this blog site is
    dedicated to dismissing the concerns of global warming skeptics, it seems as
    good a place as any to raise these issues (I just happened on this blog by
    chance from a link on John Quiggin’s blog). I have intended for some time to
    write a paper on this, but I have no time, so raising the issue in a blog
    (which AFAIK climateprediction.net do not have) seems like a good alternative.

    I closed my first post with Wildavsky’s quote because I believe there
    is a lot of truth in it.

    Scott Church: I might buy your argument except that these supposedly
    well-understood physical models (at least the models on climateprediction.net)
    have plausible parameter settings that yield “an Earth that boils or freezes,
    or oscillates between very hot and very cold every couple of years”. That
    doesn’t sound like a well-understood physical model to me.

    With that kind of unphysical variation in model behaviour, there is a real
    danger that the desired conclusion (global warming) is inadvertantly built-in
    to the model or inadvertantly built into the results through cherry-picking of
    the model parameters. One can insure against this by testing the models
    against independent validation data.

    David: I know a little physics. I have a first-class honours degree (top of
    class) in theoretical physics, and a PhD in Mathematics. I am not a
    climatologist but I know an awful lot about modeling. Why don’t you address
    my arguments instead of attacking me?

    “Climate models are primarily dynamical, with the dynamics reflecting
    conservations of energy, mass, water, and momentum. Their ain’t much ability
    to fiddle with these laws.”
    The parameters that are tuned in the models are
    not the laws of physics but unknown constants to do with all kinds of factors
    from the shape of ice droplets to coefficients of heat transfer. If these
    were derivable from first principles or measurable, there would be no need to
    tune parameters to rule out boiling or freezing the earth, or oscillating
    between the two every two years. Those sceanrios would already be excluded by
    first principles.

    Note: there is nothing wrong with tuning unknown parameters to fit a physical
    model to the data. But you cannot draw conclusions from the resulting model
    unless you have validated the model on data on which the parameters were not
    tuned, or unless there are so few parameters that you can make strong
    statistical guarantees. As far as I can tell, neither of these criteria are
    met by the existing climateprediction.net experiments.

  30. #30 Eli Rabett
    May 31, 2005

    Anon.n you don’t get it, those physical parameters CAN’T yield a boiling earth. You are confusing a throw away line in an advertisement for computer space with the actual constraints on the models. As a matter of fact one of the first things you do in an atmospheric sciences course is learn WHY a run away greenhouse effect occurred on Venus but can’t happen here. If you have access to a university library, look up, Richard Wayne’s Chemistry of Atmospheres, third edition pp 59.

    Briefly put, there are two inputs. The first is that given the flux of energy from the sun you can calculate the temperature that the Earth would have if there were no atmosphere. At that temperature the energy striking the earth from the sun and being absorbed is balanced by the energy radiated by the earth. It is about 256 K. This is also the temperature at the “top of the atmosphere” where the Earth is radiating to space.

    The second is the phase diagram of water. As the vapor pressure of water rises, you can calculate to first order the greenhouse warming from the top of the atmosphere temperature. This warming is limited because at relatively low pressures on Earth (10^-3 atm) the air becomes saturated with water vapor, liquid forms and no further greenhouse warming occurs.

    There are two ways to drive the temperature really high (boiling). The first is to have a huge increase in the solar constant, in which case we would all be toast. True you could twiddle that knob in the model, but given direct and proxy measurements of the sun’s emissions, I think even a mathematician would agree that this parameter is strongly limited. The model CAN explore what would happen if the solar constant changed by a percent or so (which is larger than the observed change).

    The second would be to increase greenhouse gas concentrations by an order of magnitude or more (depending on the greenhouse gas, of course). That too is not in the cards. We are looking at an increase of 2x from pre-industrial, which would be about 580 ppm, but for the type of increase you are requiring we would have trouble breathing before worrying about the heat. So yes, Virginia, there are limits.

  31. #31 anonnn
    May 31, 2005

    Eli, I am only quoting directly from the climatologists themselves, so if I “don’t get it”, then with all due respect, neither do they.

    From http://climateprediction.net/science/strategy.php, titled “Climate Science, Experimental Strategy”:

    Every climate model has to make a number of approximations, called parameterisations. To read more about these, click here. Basically this means that there are numbers in the model which are given a certain, fixed value, but this value is not known for sure and a range of values could be equally realistic. The experiments will investigate the effect on the modelled climate of varying the value of 20 of the most poorly understood parameters in the model – such as the relationship between the number of raindrops in a cloud and how much it actually rains (to see all 20, click here). It is possible that some combinations of parameters may replicate the past climate equally well, but produce widely different forecasts for what might happen in the future. Some combinations of parameters will not work at all, produce a completely unrealistic climate ( for example an Earth that boils or freezes, or oscillates between very hot and very cold every couple of years) and probably crash the model. It is not possible for us to tell beforehand what these combinations will be.

    I believe I have been completely faithful to this description. If I can’t believe what the climatologists themselves write, who can I believe? Besides, even if boiling is impossible in the models, it doesn’t alter the substance of my point unless the entire quote above is suspect.

    I have to say I am surprised by the level of personal attacks my comments have induced, and by how little of the substance of my arguments has been addressed. If this is representative of how global warming advocates deal with valid criticism then I have even less faith in the science.

  32. #32 James Annan
    May 31, 2005

    There is a grain of truth in anon/anonn’s comments, but s/he is certainly making more out of it than can be reasonably justified. Certainly we can have more confidence in model predictions when they have already demonstrated their ability to simulate independent data. This is already regularly done by for example paleoclimate simulations, hindcasts of the last century, the response to volcanic eruptions (generally, little or none of this data was used in model building and tuning). This is an active area of research and improvements will undoubtedly be made, but it would be quite wrong to pretend that nothing is being done at present.

    Also, even where there is limited validation, it does not make the forecast worthless, but does mean that there is a subjective element to assessing its reliability. I am not aware of any good reasons to assume that all or most of the uncertainty is on the low side, and furthermore it is quite possible to develop policies that are robust with respect to a great deal of uncertainty.

  33. #33 Scott Church
    May 31, 2005

    Anon,

    “Scott Church – I might buy your argument except that these supposedly well-understood physical models (at least the models on climateprediction.net) have plausible parameter settings that yield ‘an Earth that boils or freezes, or oscillates between very hot and very cold every couple of years'”.

    “Plausible parameter settings” are not the same thing as reality and no responsible climate modeler would interpret them as such. The ClimatePrediction.net experiment is intended to test model sensitivities to countless different parameters and inputs including those that push the models to extremes in one direction or another at the furthest limits of their confidence intervals–in no small part to help us understand the range over which models can be expected to give valid results instead of gobbledigook. These are not the same thing as realistic evaluations of those parameters and those that yield “an Earth that boils or freezes, or oscillates between very hot and very cold every couple of years” would not be used. Real state-of-the-art climate models have idiosyncracies of their own which those who work with them understand, and these people know what to accept and what to avoid when running them and interpreting results.

    I spent over a decade as an aerospace computer modeler. The models I ran were quite different from those used for climate prediction (specifically, they were finite element type models), but they too had parameters which could easily be varied in one direction or another yielding widely varying and unrealistic results. But here’s a big surprise,

    I knew this.

    And I treated them accordingly. I did not make use of inputs or parameter setting that gave wild results. I knew where my models were useful… AND where they weren’t, and I used them accordingly and checked my results against other data–yes, including data which had contributed to how they were built. These models were used to provide design loads for engine pylon structure, anti-ice systems, and even engine casing sizes, which if miscalculated will cause a commercial jet engine to explode. Guess what…

    They worked!

    And they continue to work to this day. The aircraft designed from these models include the Boeing 737-700, 800, and 900, and 777-200 series of aircraft. To date, all have logged hundreds of thousands of flight service hours worldwide without any engine explosions or major surprises in loaded behavior. No doubt you’ve flown one of these aircraft at some point in your life. I would submit to you that if your concerns about uncertainties in model parameters are valid, apart from the way their designers actually use them, by all rights you should be dead now.

    While climate models are quite different, savvy climate modelers have a similar understanding of how and how not to use their models also. The sort of tire-kicking model tests being conducted by ClimatePrediction.net are providing badly needed information about model characteristics and sensitivities. But by design they’re all over the map and aren’t representative of the sort of runs used to make trusted climate predictions. Once again, I would encourage you to examine a real model run–the kind that would be trusted to make predictions, AND… to consider exactly what those responsible for those runs are and are not concluding from them, and why. Again, I would point you to GISS SI2000 as a great place to start. See Hansen et al. (2002, J. Geophys. Res., 107, D17) or Sun and Hansen (2003, J. Climate, 16). All the best.

  34. #34 cannon
    May 31, 2005

    Scott,

    I am almost certain that the finite element models you dealt with were far
    better understood than the climate models, and had far more data to validate
    against. After all, you can build the damn thing and test it, and your
    simulations work at a much finer scale. All we have for climate models is
    limited historical data and a very very coarse grid. That makes the risk of
    overfitting the model that much greater. And hence the need for independent
    validation that much more important.

    If, as you say, we should not trust the
    climateprediction.net models to make predictions, then how do explain this graph
    and the attendant global hype that global warming could be much worse than
    predicted?

    If the consensus here is that climateprediction.net
    models are not to be relied upon for prediction, then why didn’t you say so
    up-front? I would have accepted that response immediately. If my mistake is to
    assume that their research is representative, then I, along with the rest of
    the world that read the dire new predictions, have been taken for a ride.
    But then why are other climatologists not speaking out in dissent against
    their conclusions?

    I will check your references, thanks for the pointers.

    James,

    to be confident there isn’t more uncertainty on the low side of the models, I
    would want to try really hard to tune those parameters and the model structure
    to fit the historical data and yet produce a global cooling effect with
    increased CO2. If I really could not do that and I couldn’t see an obvious
    “cook” in the models that meant such behaviour was artificially ruled out,then I would be a lot more convinced. But I would still want to see
    independent validation of the model.

    The fact that (in the words of the climateprediction.net folks themselves) the
    models can exhibit such wildly varying behaviour for plausible parameter
    settings makes me suspect that I could produce such a scenario with the
    model. This is something I would like to try when I have the time.

    If this post gets through then I am almost certain the owner of this blog is
    attempting to prevent me from posting my responses. So far, I have had to
    change my handle and email address. When neither of those worked for this
    post, I posted through a proxy and it got through.

    If I really am being censored then I won’t bother posting anymore here. Intelligent debate clearly not welcome.

  35. #35 Tim Lambert
    May 31, 2005

    anon, there is no filter that is preventing you from posting. Certainly, changing your name isn’t going to make a difference. (Though posting as “anon” is asking not to be taken seriously.)

    You are in no position to complain about people attacking you instead of your arguments when you opened with a ludicrous attack on climate scientists, accusing them of gross incompetence and of being commies who want an end to economic growth.

    Here’s a little heuristic I use: when someone shows up with an argument about how all the scientists are wrong because they didn’t consider something really obvious, it’s likely that the person is wrong and the scientists did consider it. Are you seriously arguing that the climate modellers have no clue about modeling? If so, you are going to have look at their models and get down to specifics.

  36. #36 sssarcasmo
    May 31, 2005

    So anon you feel you’ve been denigrated? Was it you who wrote the following as your opening gambit here?

    This is by far the biggest problem with the climate change debate. It is just bad science.

    We should never forget that the scientists concerned are almost invariably driven by political ideology.

    As a PhD in maths you’d be able to estimate (despite knowing nothing whatsoever of the field you were presuming to criticize) how many scientists on Earth – more than one of whom, btw, has now responded politely to you in these comments – you denigrated with your words, wouldn’t you? Why not estimate that number for us, let us know your estimate, and perhaps we’ll see whether the critic who feels he deserves my apology has the grace to make his own. I’m betting you are indeed that graceful, and I’ll also simply point out to you that the suggestion I made that you try realclimate with your polite enquiries was intended to be helpful, not dismissive of your humble scientific curiosity.

  37. #37 anon
    May 31, 2005

    Tim, it seems strange that I can only post through a proxy, but I’ll take your word that I am not being filtered.

    I did not accuse the climatologists of gross incompetence. Nor did I accuse them of having no clue avout modeling. But I do believe – at least insofar as the climateprediction.net results are concerned – that they have missed a very important issue which calls the predictive power of their results into question. And that is bad science.

    It seems the point has basically been conceded – quoting Scott: “But by design they’re [the climateprediction.net models] all over the map and aren’t representative of the sort of runs used to make trusted climate predictions”.

    We don’t see such a caveat appearing in the Nature article or on the website.

    These climate models are used to justify drastic policy measures that have serious economic impact. If there are caveats to their interpretation, they need to be exposed not hidden.

  38. #38 anon
    May 31, 2005

    sarcasmo, you did denigrate me. And you have yet to address any of your remarks directly to my substantive point. And as a whole, this site is excedingly derogatory towards global warming skeptics (eg, the bingo thread). However, on my “opening gambit”:

    This is by far the biggest problem with the climate change debate. It is just bad science.
    We should never forget that the scientists concerned are almost invariably driven by political ideology.

    I still haven’t seen anything to suggest that the problem I raised with the results on climateprediction.net are not real. Scott Church has said don’t rely on those results – look at other models and I will. So I don’t resile from my first remark: it _is_ bad science to use predictively a heavily tuned, high-dimensional model that has not been independently validated. If this approach is not representative of climate modeling as a whole, then good. But I am not yet convinced.

    As for my ad-hominem attack above: I unreservedly apologize to all those climatologists that are not driven by political ideology. However, that is a difficult number for me to estimate: every climatologist I have ever met has appeared so driven. How many right-of-center global warming advocates are there?

  39. #39 :
    May 31, 2005

    Anon, you might perhaps also explain why we should necessarily care what one website, with or without a paper published in Nature, has to say about something. It’s not as though Tim or anyone else here had breathed a word of climateprediction.net before you burst into print on the matter, is it?

    You say ” it _is_ bad science to use predictively a heavily tuned, high-dimensional model that has not been independently validated. If this approach is not representative of climate modeling as a whole, then good.” and we all agree. Hooray. You say that you’re “not convinced” – but why should we care whether you’re convinced or you’re a wombat? What is it about one website that has you so excited that you’ve apparently forgotten one of the basic rules of scientific inquiry (the one about doing your research before jumping to your august conclusions)?

  40. #40 :
    May 31, 2005

    Anon, you might perhaps also explain why we should necessarily care what one website, with or without a paper published in Nature, has to say about something. It’s not as though Tim or anyone else here had breathed a word of climateprediction.net before you burst into print on the matter, is it?

    You say ” it _is_ bad science to use predictively a heavily tuned, high-dimensional model that has not been independently validated. If this approach is not representative of climate modeling as a whole, then good.” and we all agree. Hooray. You say that you’re “not convinced” – but why should we care whether you’re convinced or you’re a wombat? What is it about one website that has you so excited that you’ve apparently forgotten one of the basic rules of scientific inquiry (the one about doing your research before jumping to your august conclusions)?

  41. #41 :
    May 31, 2005

    Anon, you might perhaps also explain why we should necessarily care what one website, with or without a paper published in Nature, has to say about something. It’s not as though Tim or anyone else here had breathed a word of climateprediction.net before you burst into print on the matter, is it?

    You say ” it _is_ bad science to use predictively a heavily tuned, high-dimensional model that has not been independently validated. If this approach is not representative of climate modeling as a whole, then good.” and we all agree. Hooray. You say that you’re “not convinced” – but why should we care whether you’re convinced or you’re a wombat? What is it about one website that has you so excited that you’ve apparently forgotten one of the basic rules of scientific inquiry (the one about doing your research before jumping to your august conclusions)?

  42. #42 :
    May 31, 2005

    Anon, you might perhaps also explain why we should necessarily care what one website, with or without a paper published in Nature, has to say about something. It’s not as though Tim or anyone else here had breathed a word of climateprediction.net before you burst into print on the matter, is it?

    You say ” it _is_ bad science to use predictively a heavily tuned, high-dimensional model that has not been independently validated. If this approach is not representative of climate modeling as a whole, then good.” and we all agree. Hooray. You say that you’re “not convinced” – but why should we care whether you’re convinced or you’re a wombat? What is it about one website that has you so excited that you’ve apparently forgotten one of the basic rules of scientific inquiry (the one about doing your research before jumping to your august conclusions)?

  43. #43 sssarcasmo
    May 31, 2005

    Ahem, that were me …

  44. #44 Ian Gould
    May 31, 2005

    <H<ow many right-of-center global warming advocates are there?>>

    Well we can start with Kyoto signatories John Major, Helmut Kohl and George Bush Senior. Then there’s US Senator John McCain, co-author of the McCain-Leibermann Climate Responsibility Act. Oh and Governor George Pataki of New York state…

    Climate change only became a partisan political issue when George W. Bush decided to dump the treaty his father co-authored and smear his critics with accusations of political bias.

  45. #45 James Annan
    May 31, 2005

    I have to wonder if anon has actually read the paper in question, as they explicitly disclaim any predictive ability in their results.

    we cannot provide an objective probability density function for simulated climate sensitivity

    I would, however, accept that the some of the press coverage was somewhat misleading.

  46. #46 anon
    May 31, 2005

    “What is it about one website that has you so excited that you’ve apparently forgotten one of the basic rules of scientific inquiry (the one about doing your research before jumping to your august conclusions)?

    Well that’s a case of the pot calling the kettle black if ever I heard one.

    Describing the results from climateprediction.net as “one website” makes it sound as insignificant as this blog. It is a lot more significant than that. It has already produced one Nature paper, co-authored by academics from prestigious institutions (Oxford among them). Their results were reported with much fanfare and scaremongering around the world when they first appeared a few months ago (I first read about them on Slashdot). This is no ordinary “website”.

    I would have thought serious flaws in their methodology would matter to scientists in the field, particularly if those flaws are not representative of the field as a whole.

  47. #47 anon
    May 31, 2005

    I have read the paper. Since we’re quoting:


    Six of these model versions show a significant cooling tendency in the doubled-CO2 phase. This cooling is also due to known limitations with the use of a simplified ocean (see Supplementary Information) so these simulations are excluded from the remaining analysis of sensitivity.

    I’m willing to accept that as an outsider I may be misinterpreting, but that sounds awfully like cherry-picking to me. What about the samples with unnaturally high heating tendency due to perhaps unknown limitations in the model?

  48. #48 John C
    May 31, 2005

    Cannon / Anon:

    Back at 16:06:50 you said:

    But then why are other climatologists not speaking out in dissent against their conclusions?

    I think the answer is that you have framed the question wrong. The correct question would be what do other climate scientists think of the study. And the answer would be that a number believe that the work should be taken in context of what was done with the tools.

    For example, the most informative blog on climate issues is – arguably – RealClimate. They produced this in regards to the paper under discussion. I think it shows that a number of climatologists did try to present the results in context.

    Regards,
    John

  49. #49 sssthingo
    May 31, 2005

    Amusingly I think, it was “pot kettle black” or “throwing stones in glasshouses” that first sprang to mind when anon claimed something said to him had been rude, said thing having been said to him just after he’d dropped his shorts and poked out his tongue at an entire field of scientific research. Heh, it’s just like usenet around here sometimes :)

  50. #50 Tim Lambert
    May 31, 2005

    anon says “every climatologist I have ever met has appeared so driven”. I’m guessing that the number of climatologists anon has met is zero.

  51. #51 Chris O'Neill
    May 31, 2005

    On Bob Carter generally and also Ian Plimer, I get the feeling from their writing e.g. http://troppoarmadillo.ubersportingpundit.com/archives/003140.html that they don’t actually care if global warming is artificially caused. Plimer says “Global warming commenced again at 500 BC, there was an excess of food and great empires such as the Ashoka, Ch’hin and the Romans grew. Contemporary records and Roman clothing shows that conditions were some 5C warmer than today.

    Plimer and Carter like to warn about the next ice age and I agree that an ice age would be a real problem for humanity.

    So their view about artificially caused global warming seems non-genuine to me when you consider that they don’t actually think that global warming is a problem anyway.

  52. #52 Eli Rabett
    May 31, 2005

    Another two things that are perhaps non obvious are that while GCMs are the top of the climate system food chain, and have parameterizations, they sit on more complex but limited experiments and systems models (ocean, carbon cycle, radiative, etc). These do indeed set strong limits on the range of parameters that can be used in GCMs. Using these models you can tune on one area and test on another. This bounds the range of the parameter space in GCMs, meeting Cannon’s original objection. Frankly, given what I know about these smaller systems there is no way in hell that a modern GCM or even an old one dimensional model could produce a boiling ocean. That is a cooked red herring and what you get when you deal with folk who have no common sense or are trying to fake it to satisfy their primal urges.

    Second, one of the important reaons for the climateprediction.net experiment was to see if there were any singularities within the possible range of parameter space and it appears that there are not, although the work is going forward. That there were a few choices that produced cooling or large warming is not unexpected, but has to be evaluated in the context of how likely the set of parameters which produced these results were. I believe that point was discussed in detail in the paper Canon read.

  53. #53 Scott Church
    June 1, 2005

    Cannon/Anon,
    “If, as you say, we should not trust the climateprediction.net models to make predictions, then how do explain this graph and the attendant global hype that global warming could be much worse than predicted? If the consensus here is that climateprediction.net models are not to be relied upon for prediction, then why didn’t you say so up-front? I would have accepted that response immediately. If my mistake is to assume that their research is representative, then I, along with the rest of the world that read the dire new predictions, have been taken for a ride.”

    I did say so up front, but you may have misunderstood me–my apologies if I was unclear. It’s not that the ClimatePrediction.net models aren’t reliable for prediction, it’s that the specific runs being generated by the project aren’t–at least, the majority aren’t. The problem here is that you’re talking about apples and oranges. The link you referred to is to an ensemble of runs covering a bewildering array of parameter inputs, some of which will be quite representative of real climate and useful for prediction, but many, many of which absolutely will not be. You simply cannot treat that graph, and the entire unfiltered collection of ClimatePrediction.net runs as representative of what savvy climate modelers would get by running these models for all they’re worth in a real prediction exercise.

    The analogy is a little crude, but the ClimatePrediction.net experiments are a sort of crash-dummy test. They are not, and never were intended to be predictive. Their objective is to test model limits and sensitivities within the range of parameter inputs that might be conceivably used for one reason or another in model studies. This is like Ferrari, say, putting a Testarossa through rigorous test drive sessions with all sorts of extremes–flips, curb jumps, spinouts, wall bashing sessions, and more in an attempt to find out its limits prior to marketing. These tests are absolutely necessary to establish the vehicle’s performance and handling characteristics, and its safety, but no one believes they’re representative of what it can do in the hands of an experienced driver at the race track. It’s also worth pointing out that the label “crash-dummy test” is not intended in any way to be derogatory. Without these test we would not know much about the sensitivities of climate models and their limits any more than Ferrari would know that the Testarossa’s limits are. It’s just that we have to be clear about the differences between crash-dummy tests and the race track and not confuse the two.

    As for “hype” about warming being possibly worse than expected, these claims are not, and never were based on models alone. A great deal of research has been done in this area. For instance, you might start with the NAS’s examination of some questions related to potentially sudden climate change in Abrupt Climate Change: Inevitable Surprises. There is also much about the possibility of climate change being worse than expected in the IPCC Year 2001 Report. And of course, there is much research since then as well. You would do well to consult some of this, and investigate some real predictive climate model studies instead of dwelling exclusively on crash-dummy tests…. before labelling them as “hype”.

    “I am almost certain that the finite element models you dealt with were far better understood than the climate models, and had far more data to validate against. After all, you can build the damn thing and test it, and your simulations work at a much finer scale. All we have for climate models is limited historical data and a very very coarse grid. That makes the risk of overfitting the model that much greater. And hence the need for independent validation that much more important.”

    Actually, they weren’t all that well understood and I could not test them. It is true that they were considerably simpler than most climate models and required far fewer parametrized inputs. But beyond that the data against which I was able to test them was quite sparse compared with what I was after. Mostly, they were dynamic models that involved parametrization of damping, broadband forcing, non-linear material responses and other things. The grid and parametrization scales were no better than their climate model counterparts for what I was measuring, and I had very limited data regarding any of the variables I had to parametrize. All were built from known physical principles and what I did know regarding the parametrizations and inputs, and were checked wherever I could do so against multiple data sources. These models were intended to reproduce the response of structural components to a nearly infinite range of aircraft engine vibrations during 20 years of rain, shine, calm, and stormy flight service–there was no way I could get well characterized data for any of this, even in principle. But the response of my models, and the resulting design loads, were sensitive to all of it. So I had to rely on my own best judgment and what data I did have. Along the way I learned what I could and could not trust my models for…. Sound familiar?

    By contrast, climate models have surface and upper-air data on countless different inputs over time scales that are plenty long enough to test model characteristics. There are over 900 stations providing anywhere from 50 to 100 years of climate data for the surface alone, and that’s just for starters. Despite the many known problems and gaps, I would have killed to have had datasets this good!

    But I didn’t.

    So I made do with the data I did have and my own judgment, and in over 10 years, I haven’t seen so much as one single criticism of any kind directed at climate models by global warming skeptics that could not have been directed at my own models with equal force.

    But the flight service record of the resulting aircraft speaks for itself.

    If I have understood your arguments properly and done them justice, the problem here is that you’re trying to pass off crash-dummy tests as representative of prediction exercises without a clear understanding either of the differences between the two, or the art of climate modeling as practiced by experienced modelers. Which is why I keep pointing back to real predictive exercises like the ones cited in my last post. I believe you would benefit greatly from a trip to the race track to watch real drivers handle the runs. All the best.

  54. #54 anon
    June 1, 2005

    Scott, I just don’t buy your physical aircraft component simulation analogy. The physics is way better understood and the mechanisms far less complex than the climate. In most cases we don’t even need simulation – we have simple newtonian calculations that tell us if a bracket is stiff enough. Houses stay up without simulation – but that doesn’t tell me anything about climate modeling.

    900 stations with 50 to 100 years of climate data for the surface of the earth is pretty small, given the scale of the problem. We know the earth’s climate varies massively of its own accord, without any forcing by humans.

    All I am asking for is that models be validated against data on which they were not tuned. In my opinion, that’s pretty simple, and quite reasonable.

    If you have not read the paper, please do: http://www.climateprediction.net/science/pubs/nature_first_results.pdf.
    And then read some of the attendant hype, eg http://www.newscientist.com/article.ns?id=dn6934, and maybe watch Bob Carr’s efforts from lateline (I think it was) shortly afterwards, where he put up maps of Australia with 50C temperatures. And then tell me honestly that I am misrepresenting the science as presented to the public.

  55. #55 Dano
    June 1, 2005

    The PR campaign has a new wrinkle – make up climate data and see how the GCMs handle it. Excellent. I’m _sure_ there’s someone out there who’s made fake forecasts with fake data. Do you think they’ll give up the source code?

    [Snark apology]

    Fortunately, we can fall back on an old wrinkle to balm our consciences: We know the earth’s climate varies massively of its own accord, without any forcing by humans.

    Ahhh…the old non-sequiturs sound so sweet…

    D

  56. #56 James Annan
    June 1, 2005

    anon, you apparently no longer dispute that the cpdn results do not claim any predictive capability (roughly speaking, they were presenting possibilities, not probabilities). And by picking on one paper, you ignore the large body of validation that I mentioned earlier. Here is not the best place to discuss the limitations of the cpdn paper in detail (I suggest sci.environment), but certainly, I do not expect to see claims for climate sensitivity of >10C take up much space in the next IPCC report.

  57. #57 Scott Church
    June 1, 2005

    Anon,

    Once again, you’re completely missing the point. Yes, the basic physics behind aircraft models is well understood. I’ll even grant you that it’s better understood than the basic physics behind climate dynamics. But “basic physics” is not the issue here–it’s the specific application of that physics where knowledge of inputs is incomplete and the difference proper judgment by experienced modelers makes. Whether the physics of aircraft models is “way better understood” or not makes absolutely no difference of any kind whatsoever. Even if I understand the physical principles of structural response and “the stiffness of a bracket” better than God himself,

    a) I do not understand how that bracket and the model it’s part of are being forced by a jet engine that’s ripping along at several hundred miles per hour generating 30,000 lbs of thrust while turbomachinery whirling at tens of thousands of RMS’s are shaking it violently as it flies through storms, wind shear, or hard landings for 20 odd years.

    b) I do not understand how those divinely understood physical principles are going to manifest themselves in each of the countless specific structures where I do not have enough data to completely characterize all joints, components, boundary conditions, etc.

    c) The models I construct under these circumstances will be useful for some things and not for others, and my own judgment is crucial for getting the most out of them–regardless of how many uncertainties there are my parametrizations and forcings, and regardless of how rediculous the results will be if I run them with all sorts of wild inputs and no proper modeling judgment.

    Yes, climate models are far more complex than the structural models I built in the aerospace industry, and they they’re also fundamentally different than mine. But in both cases,

    a) The models are representing systems that are far more complex than they are.

    b) Limited data is available as to how their separate components interact with each other.

    c) Knowledge of how they are being forced is limited.

    d) How the art and craft of modeling is practiced by savvy modelers in the building and running of the model make literally all the difference in the world as to how effectively a) through c) are handled, how the results are or are not used, and how much can be learned from them.

    e) a) though d) are true regardless of how many wild, unrealistic scenarios can be derived from them using any combinations of inputs a vivid imagination can dream up, even if some of them, taken in isolation, are possible.

    f) The exact same criticisms about uncertainties, assumptions, and lack of data can be leveled against both types of models.

    But as I’ve already pointed out, these models work. They work because the men and women who build them understand how to deal with points a) though c). Most importantly, they know the difference between a proper handling of these points and model crash-dummy tests. And yes, they do in fact test their models against data that is independent of that used to build and/or calibrate them.

    I can only repeat these points so many times, and then I simply have to let them stand whether they land on listening ears or not. I appreciate your concern for the importance of independent data, and you are right to say that models need to be tested against data other than that used to calibrate them. But until you get off of this ClimatePrediction.net crash-dummy test bandwagon you’re never going to learn anything about how climate models are used in real predictive exercises by those who understand them. Nor will you learn anything about the ways that they are in fact, tested against real independent data (as you wisely point out they must be). You cannot effectively critique that which you have no experience with and do not understand, and no amount of preaching about uncertainties and variability of results are going to change that–any more than the many similar criticisms I received from some of my aerospace colleagues changed the effectiveness of my model runs, or shortened the service lives of the aircraft built from them. All the best.

  58. #58 anon
    June 1, 2005

    Scott, I don’t need to repeat myself anymore either. An argument by analogy is only as good as the analogy itself. Until you explain precisely how your aerospace models are comparable with climate models in terms of complexity of parameterization, sensitivity to those parameters, predictive power, etc, I will be about as convinced as if you said “I built a house based on a model and it didn’t fall down – therefore the predictions from the climate models must be good”.

    You were the one who created the term “crash test dummy” to describe the climateprediction.net results, not they. If you read the website, the paper and read the media commentary, it sounds a helluva lot more serious than that. Eg on New Scientist: http://www.newscientist.com/article.ns?id=dn6934

    Try these headlines:

    “Soaring global warming ‘can’t be ruled out'”
    – actually, that one is pretty empty – can’t rule out that the moon has a core of French brie either.

    “When you see large areas of the northern hemisphere at 11C above pre-industrial levels, you think this is quite scary,” says Stainforth.

    “Geological data shows the Earth’s climate has been much warmer in the past. Temperatures were around 6C higher during the Cretaceous period, for example, when dinosaurs roamed the Earth. But Bob Spicer, an expert in the palaeoclimate at the Open University in Milton Keynes, UK, says there is no evidence that temperatures have ever been as high as in some of the climateprediction.net simulations.”

    This is very inflammatory stuff. There is definitely a predictive element to the statements. But perhaps the biggest red flag for me comes from the filtering of results described in the paper:


    Some iterations of the models showed the climate cooling after an injection of CO2, but these were discarded after close examination because the temperature fall resulted from an unrealistic physical mechanism, says Stainforth. In these scenarios, cold water welling up in the tropics could not be carried away by ocean currents because these were missing from the models.
    There are no obvious problems with the high temperature models, he says.

    This may be totally valid, but without independent validation of the models, it is hugely risky. I am a scientist. I know how hard scientist’s try to prove their hypotheses. Without in any way impugning their integrity, I would be very surprised if the scientists concerned did not work a lot harder at finding an explanation for the anomolous cooling than they did for the anomolous heating. It is only human. It is precisely this kind of thing that you need independent validation sets for: they keep you honest.

    I think a fair conclusion to this thread would be:

    1) Any predictive claims from the climateprediction.net results should be taken with a pinch of salt;

    2) The presentation of the climateprediction.net results to the public did not include appropriate caveats on their interpretation.

    3) Complex tuned models should be validated against independent data.

    For me the question remains: given the consensus here that climateprediction.net models and results are not representative of climate science in general, are the representative models and predictions better founded? I guess that’s up to me to sort out for myself.

    A final word on one of Scott’s criticisms:


    You cannot effectively critique that which you have no experience with and do not understand, and no amount of preaching about uncertainties and variability of results are going to change that.

    I completely disagree. There are general principles of modeling that apply independently of the model concerned. There is a huge statistical literature devoted to it. As someone who has a lot of experience applying those general principles in practice across a number of different fields, what I read on climateprediction.net raised serious questions for me. And from this thread it would appear that my concerns were not unfounded.

  59. #59 Dano
    June 1, 2005

    Ooh! Looky! Conflation! One outlying study is conflated up to the whole enterprise!

    And headlines are scientific evidence!

    Whoo-ee, fun.

    D

  60. #60 Eli Rabett
    June 2, 2005

    A.non might want to take a look at the first figure (about a third of the way down the page) at http://www.earthinstitute.columbia.edu/news/2004/story12-13-04b.html which shows the prediction of a relatively unsophisticated 1988 model projected forward to 2020. The predictions, based on three different emission scenerios, has done pretty well compared to the middle scenerio which is the one closest to what has happened in the intervening 17 years, (red line).

    Certainly this shows a case where data used to validate the model was not used to tune the model. I know that there is a major effort to validate models involving large groups. I have heard that models are tuned on one set of data and then validated against hindcasts involving other data which certainly also meets a.non’s objections. Finally another method is to tune the model against global data and then downsize it to smaller regions at finer resolution to validate. In short, I think to make the claims that a.non is making have any validity, he would have to know a lot lot more about how models are validated.

  61. #61 Scott Church
    June 2, 2005

    Which goes right back to what I’ve been saying all along. Models are in fact tuned, and then validated, against multiple independent datasets, and predictions based on them are not generated randomly. This is why it’s a vain and ill-informed exercise to trot out endless editorial speeches about model uncertainties and/or wild outcomes from model runs, even from scientific forums, before one has actually investigated any real model-based predictive studies. I have yet to see even one single instance of any global warming critic who has actually examined a genuine model based study of future climate trends such as those I’ve cited above, and demonstrated that properly used, they cannot provide meaningful forecasts. The best any of these critics can come up with is generalized rants about “uncertainties” or “rediculous” results defended with examples from studies that were not intended for such–like Anon’s.

  62. #62 Dano
    June 2, 2005

    I usu. point septics here, Scott. Nary a peep after that, except that some issue dying paroxysms.

    D

  63. #63 Scott Church
    June 2, 2005

    Wow, great link Dano! Thanks! I’m going to have to put that one at my site. BTW, how are you liking Buckley? Do you get up my way very often these days? I miss the coffee runs!

  64. #64 Ian Gould
    June 2, 2005

    To return briefly to the question of right of centre advocates of te need to address global warming – adding Arnold Schwartzenegger’s name to the list. http://www.terradaily.com/2005/050602003701.hj03axfv.html

  65. #65 David
    June 3, 2005

    ……”Climate Change Scientists Robert Carter” will be a guest speaker at a Victorican Farmers Federation regional forum next week – he will discuss is “climate change cool science or hot air”. See: http://www.vff.org.au/index.php?id=70255 .

  66. #66 James
    June 22, 2005

    I would be most interested to know what someone with appropriate expertise has to say about the evidence in
    http://www.john-daly.com/artifact.htm

    (which is I am told “a layman’s summary”. However “the reference list at the
    back contains the details of the refereed papers on which the
    work rests”.)

    The significance of the CO2 absorption factor being overstated by 80 times in the parameter used for climate models is obvious. I am not familiar enough with the terminology or the science to have confidence in my assessment of the research quoted.

The site is currently under maintenance and will be back shortly. New comments have been disabled during this time, please check back soon.