Climate Ball – To Play or Not To Play

mt has an article up at And Then There’s Physics arguing that climate science should engage with the red team process and try to make lemonade out of those lemons.

Climate disruption is a problem, and there are those who seek to obscure that fact for selfish reasons. So they malign the science, even though the core of it is a triumph that should be celebrated.

I’m not under any misapprehension that the impetus for the “red team” isn’t very much under the influence of this malice. The revulsion of the climate community (“we’ve been red-teaming each other for decades!”) is therefore understandable.

I’d like to advocate that we resist this visceral response and embrace this as an opportunity.


  1. I'll comment here since AtTTP has banned me. I think MTs proposal is reasonable. I just wanted to mention why there is reason to question GCMs for climate. One line of evidence is the tropospheric hot spot. This is predicted by the lapse rate theory, which is the basis of "physical climatology". The problem is that the data shows a much weaker temperature gradient than GCMs predict. That would be a good issue to focus on since it indicates that GCMs have a problem with their conversion models. It's an I'll posed problem and Held has a good post on extreme sensitivity to computational domain size.

  2. My understanding is that new evidence has the infamous "tropospheric hot spot" problem as going away under improved treatments of data, and that all along the evidence has been weak and tentative.

    If you understood my article, you didn't really respond to my suggestion.

    I have no idea what "a problem with their conversion models" means.

    If Held has an article you find relevant you ought to provide a link.

  3. Michael, this is discussed by nick Lewis here

    Helds post is here

    There are plenty of a priori reasons to doubt GCM results in the tropics. Convection is an ill posed problem and climate simulations are done on ridiculously coarse grids.

    Held also has a recent post on GCM tuning and there is a big article a year ago or so by modelers describing why its not "settled science".

    I also think that CFD experts generally are quite aware of the biases of code "users." Users tend to be overconfident in model results.

    Regardless of your views, this would be a good subject for a red team look.

  4. Modeling is the way to get at this.

    Some CFD people do sneer at the climate models. CFD people tend to look at very different sorts of problems than climate modeling does, and sometimes I've seen them baffled at the whole idea. Climate models are messy beasts and in some ways they'll never be the most elegant of models (not to say they can't be improved upon) but they do, remarkably, work.

    So someone who is utterly convinced the sensitivity is so low we don;t have to do anything about it should build one that works at least comparably well and isn't very sensitive to CO2, or they should move on.

    Thanks for the links.

  5. Well, Michael, aside from GSAT, what are GCMs skillful at? Regional climate is poorly predicted. As I recall Hargreaves et al s conclusion was there was "potential" for skill at predicting GSAT, not exactly a strong endorsement.

    As I recall there is a Russian GCM with a low ECS that is quite skillful for GSAT.

  6. GSAT?

    I am sick and tired of people not understanding what a triumph climate models are. They represent the details of the earth's climate from first principles. The goerning equations and boundary conditions are put in; weather systems come out. They do the right thing - the dry zones, the wet zones, the lapse rate, the ITCZ, innumerable details. Often when models and data disagree, it turns out that the data are wrong.

    This is an extraordinary success of science. The idea that "the models are crap" is itself crap.

    Can models predict changes in climate in detail? No, that's even more demanding. But any GCM that performs well on objective tests of climate does have a high enough sensitivity to worry about, as far as I know.

    There are models that aren't GCMs, but they don't have emergent climate phenomenology. They can be useful, but that isn't the point here.

    What Russian GCM do you refer to? I'm not just taking your word for it.

  7. Ok Michael if it will make you happy I agree that GCMs are amazing tools for weather forecasting. Many advances in CFD originated in these models. There is of course irreducible uncertainty as well. Over the last 40 years however, CFD has outpaced most of the older GCMs. The problem here is the specific skill or lack thereof. My reading indicates that the modelers themselves are quite aware of the issues but that those "running" the codes are less aware of them. Those doing the "defense" of the models can be remarkably unaware.

  8. You can find the reference to what I called the"Russian model" in one of Steve Moshers comments on ATTP.

    Nic Lewis also in the reference I gave earlier has some points about how sensitive models are to cloud and convection sub grid models. This suggests to me that it should be easy to get a lower than 2.1 ECS version with plausible values. Nick's entire write up is worthwhile. The conclusion is that AOGCMs don't provide scientifically valid evidence about ECS.

  9. Michael, My argument is that's current GCMs could be easily tuned for low ECS using the cloud and convection models whose accuracy is very questionable anyway. There is also the eddy viscositymodel of the boundary layer. Everyone knows one can get quite different separation patterns and even Bl health by such tuning. This is all often not easy to discern from this literature. The ocean models also have these models.

    Have you read the recent big GCM tuning article or Helds latest? These are actually good signs that the GCM modelers themselves are going to do better than turbulence modelers in the future.

  10. Leaving aside the stuff I don;t understand about BI health etc. I'd like to focus on "current GCMs could be easily tuned for low ECS"

    This goes exactly to my point.

    If GCMs are as useless as people (including you) posit, my position is that this would be the case, while if they embody real skill, it would not.

    So if it is so easy, please do it! Really. I'd love to have some basis to change my mind about all this.

  11. Michael, You should read the Nic Lewis writeup I referenced earlier. He cites a number of examples where ECS was reduced in some cases dramatically by different (and perhaps better, perhaps worse) cloud and convection models. I don't know what is in people's mind, but if GCMers are like CFDers, they will naturally rule out any parameters that seem to give an "unreasonable" result. They I believe look at lots of outputs to make these decisions.

    I do doubt however if one could easily get a total feedback that is negative so I would think it might be difficult to go below 1C. Mosher showed a reference that proves its possible to get quite close to 2C. Mosher asserts that that model matches real data very well.

    I don't think GCM's are useless as weather models as I said above. I do doubt their utility for long term climate simulations. I also do not believe they are real "scientific" evidence concerning the real ECS of the system. They are giant machines with hundreds of control knobs. They do relatively sell on Rosby waves and seemingly quit poorly on convection.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.