Constraining model ECS

To follow up on Michael’s post I thought I might just highlight this recent paper by Fasullo, Sanderson & Trenberth called Recent Progress in Constraining Climate Sensitivity With Model Ensembles.

It’s hard to summarise, as it’s essentially a summary itself. I think it’s open access, so you can read it yourselves if you wish. It mainly discusses different methods for constraining climate sensitivity using model ensembles. The two main methods are Perturbed Physics Ensembles (PPE) and Multi Model Ensembles (MME). A PPE is when you use a single GCM, but perturb the physics to see how that influences the resulting ECS. You can then weight the result on the basis of climatological skill. An MME is when you use multiple models and then try to identify emergent constraints (EC) that can then be used to constrain the resulting ECS.

Credit : Fasullo, Sanderson & Trenberth (2015)

Credit : Fasullo, Sanderson & Trenberth (2015)

As I said above, it’s hard to really summarise this paper as it’s a summary itself. However, the top figure on the right shows an example of an EC – the strength of mixing in the lower troposphere over warm tropical oceans – that can be used to constrainn the ECS in an MME. They lower figure illustrates other possible emergent constraints and

suggest[s] an underestimation of ECS by models due to cryospheric and cloud feedbacks. While no single EC study should be regarded as definitive, the collective guidance of this literature broadly fails to support the hypothesis that model error is responsible for the divergence between GCMs and estimates of ECS based on simple models and the instrumental record …… Rather, it suggests the opposite, with the evidence showing that model error has more likely resulted in ECS underestimation.

The one thing that struck me about this work is possibly related to the point Michael was making in the previous post. Why has noone taken a GCM and shown that ECS could be much lower than we currently think? It seems that – in a sense – the PPE and MMEs are essentially trying to do something like this. What they seem to find, however, is that if you constrain your results according to climatological skill, it is difficult to get a result that is inconsistent with the IPCC range. Most of the results seem to put ECS at around 3oC – or slightly higher – with a range of around 1oC.

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

291 Responses to Constraining model ECS

  1. “Why has noone taken a GCM and shown that ECS could be much lower than we currently think?”

    I believe the lowest GCM in the IPCC “fold” is 2.1C

    It would be quite an experiment to actually look for settings which lead to lower numbers for ECS.

    i

  2. David Young says:

    From the summary and conclusions: Especially the 3rd paragraph.

    ” The challenge of constraining ECS in part stems from the difficulties entailed in validating the broad range of processes involved, the often vague linkages tying mean state biases to trends, and the unclear connections between internal and forced variability. Observational uncertainties and challenges in comparing simulated and observed fields and especially clouds are also fundamental. The community has access to two types of flawed ensembles on which to test hypotheses: PPEs which have potentially very large sample sizes but can only explore behavior within a single and imperfect model framework, and MMEs, which are systematically diverse but insufficiently sampled to make any robust statements based on correlation alone. Furthermore, analyses of PPEs and MMEs have remained largely independent, although it seems increasingly clear that any comprehensive assessment of uncertainty in ECS will need to take account of both parametric and systematic uncertainties in a single framework.
    Nonetheless, in recent years, a diversity of approaches has been proposed and tested as constraints on both individual feedbacks and ECS generally, and these can be viewed as serving a range of purposes. Firstly, when designed appropriately, they offer an approach for benchmarking model fidelity and provide associated insights to guide model development priorities through highlighting biases that are both physically and statistically linked to targeted model predictands. Moreover, when correctly interpreted, they offer qualitative guidance on the potential implications of model error on ECS and provide a basis for either weighting or screening the models in PPEs and MMEs. However, as evidenced by recent assessments of the snow cover feedback, such constraints, no matter how apparently direct and physically plausible, may in some instances be uncertain.
    Moreover, recognizing the structural and statistical limitations of PPEs and MMEs, it remains doubtful that either can offer a strong and useful reformulation of the probability distribution describing ECS based on existing model archives. Moreover, given that these archives already push the limits of the available computational infrastructure, providing a framework for fully addressing both structural and parametric uncertainties in models may lie beyond available capabilities for the foreseeable future.
    These drawbacks do not render PPE or ECs useless, however. Recent results, summarized in Fig. 1b, generally suggest an underestimation of ECS by models due to cryospheric and cloud feedbacks. While no single EC study should be regarded as definitive, the collective guidance of this literature broadly fails to support the hypothesis that model error is responsible for the divergence between GCMs and estimates of ECS based on simple models and the instrumental record (e.g., [21, 25]). Rather, it suggests the opposite, with the evidence showing that model error has more likely resulted in ECS underestimation. The EC literature therefore redirects the focus of the effort to reconcile instrumental and GCM estimates onto the untested base assumptions and data sensitivities of alternative approaches.”

    This approach is being used in fluid dynamics too and it has fundamental limitations. Eddy viscosity models are far simpler than GCM sub grid models but they make the Bousinesq assumption that the eddy viscosity is isotropic, an assumption well known to be false in some cases. However, to test that you would use a Reynolds’ stress model and there are too many parameters to really constrain with high quality data. The result is that varying the constants in the model has strong limits in terms of evaluating the uncertainty. The other thing is that even simple eddy viscosity models have a very large number of nonlinear algebraic terms. Varying the forms of all these terms quickly turns into a dart throwing exercise.

    This paper seems to me to raise more questions than it answers and the task ahead is rather daunting.

  3. This paper seems to me to raise more questions than it answers and the task ahead is rather daunting.

    Sure, and it may well be daunting. However, it still appears that what this paper is highlighting is that it’s hard to constrain ECS within the range 2oC – 4oC. There seems little evidence to support the idea that it’s likely below 2oC, or much above 4oC.

  4. Overall, I think it might be difficult to find ‘the emergent constraint’ for ECS. Usually, the more successful ones that are based on physical insight will just provide constraint on a particular aspect, say low-latitude low-level cloud feedbacks. Further, there is no guarantee that reality lies within the ensemble of models.

    “Why has noone taken a GCM and shown that ECS could be much lower than we currently think?”

    We tried to lower ECS in this http://www.nature.com/ngeo/journal/v8/n5/full/ngeo2414.html paper. I wouldn’t rule out that it would be possible to lower ECS further, however, we have spent quite a bit of effort lately to try to find a way with fairly little success. Perhaps starting from a less sensitive base than the MPI model would be easier.

  5. Thorsten,
    Thanks for the comment. What you say in the first part of your comment is something I had wondered, assuming I understand what you’re getting at. If you focus on a single EC, then that might tell you how that EC influences ECS, but doesn’t tell you that that model’s ECS value is actually a good representation of reality.

    I wrote a post about your paper with Bjorn.

  6. Thorsten Mauritsen says:

    Yes, I realise now that I was being too implicit. I meant that many factors will influence ECS and so a single aspect or process constraint will not necessarily be very relevant. Also, I realise that your post had mostly understanding ensembles of models in mind, though ECs are often leveraged to say where reality might be.

    Oh, and thanks for the post on our paper, very well done!

  7. Thorsten,
    Thanks for clarifying.

  8. BBD says:

    Never mind the models. How are we going to tune palaeoclimate behaviour so it fits with low sensitivity?

    Now that’s the real challenge.

  9. David Young says:

    Thorsten, Thank you for commenting here. I do find it surprising that it is seemingly difficult to construct a low ECS GCM, but I’m not expert on all the sub grid models used. I suspect what you mean is its difficult to construct a “credible” GCM with low ECS. I could just remove the radiation GHG absorption model and I would get a model with 0 ECS, No?

    It would be helpful to me if you could summarize what you did in the paper you referenced.

    Best Regards

  10. David,
    There is a basic summary here. It may not be as good a Thorsten could do, but it’s a start.

  11. Pingback: Science's best friends - Ocasapiens - Blog - Repubblica.it

  12. David Young says:

    I looked at your post, ATTP, and it was informative even though I probably missed something important. I believe that they were trying with cloud feedbacks to get below 2C for ECS and couldn’t if they included all the feedback models. In a broader modeling validation context, that doesn’t really say too much for me. To really examine this, you would have to try to change the fundamental cloud assumptions, the model resolution, the algebraic relationships, etc.

    Just as an example, say I have a Navier-Stokes code with an eddy viscosity model. There are many constants and algebraic relationships to tune. So, I could make thousands of runs varying these assumptions. However, no matter what I did, there would still be the Bousenesq assumption, viz., that the eddy viscosity is isotropic. If the flow is a challenging (and most separated flows do not obey this assumption) one, I will never succeed in getting the Reynolds’ stresses to match the data. You might of course get lucky and try a flow for which the Reynolds’ stresses make little difference no matter how wrong they may be. That would be an example of selection bias.

    So, I am quite skeptical about these kinds of exercises. In a context where there are probably hundreds if not thousands of algebraic relations and assumptions, it is a very big undertaking, and perhaps undoable.

    With a simple model, there is a chance of understanding things in a reasonable amount of computer time.

  13. > Just as an example, say I have a Navier-Stokes code with an eddy viscosity model.

    Just out of curiosity, have you ever chosen another example, DY?

  14. Kevin O'Neill says:

    David Young – as BBD asks, how are we going to tune paleo data so that the fast-feedback sensitivty is lower lower than 3C +/- 0.5C ? As Hansen writes, 800,000 years of paleo data nails the fast-feedback sensitivity.

    Dithering about models seems rather tangential to what we know before we initialize the first run of one of today’s GCMs. Navier-Stokes code with an eddy viscosity model notwithstanding.

  15. David Young says:

    Kevin, I actually agree that paleoclimate is important. My non-expert opinion is that we desperately need better data so we can estimate for example the delta T for the LGM better than within a factor of 2. But I’m no expert and don’t have the time to delve into it. I just have observed that the field is tremendously controversial. Given this state of affairs, I can’t really justify having any opinions worth your paying attention to, not that that would happen anyway.

  16. dhogaza says:

    David Young:

    “If the flow is a challenging (and most separated flows do not obey this assumption) one, I will never succeed in getting the Reynolds’ stresses to match the data. ”

    This happens with solid objects like airfoils in high-velocity airstreams.

    What, exactly, is the relevance to climate? Does a 20 knot onshore flow over a forested slope lead to flow separation and … oh, wait, this is so subgrid that detailed modeling of the flow over said slope isn’t of course even attempted by GCMs.

    Give us a climate-scale example where flow separation actually happens and matters.

    It has been suggested before that problems with modeling airflows over airfoil-scale physical objects might not have relevance at the global climate scale, and you’ve never responded adequately.

    So – be specific with your example, please.

  17. David Young says:

    There is separation all over the place in climate such as downstream of mountains. The vortex street downstream of a massive separation is dynamically the same as the mid latitude vortex patterns we call weather. Shear layers are all over the place too, called fronts in weather. Convection is also an ill posed problem that is just as challenging if not more so than separation.

    But the more fundamental challenge is turbulence itself. As you know from your last airplane flight, the atmosphere is extremely turbulent. GCM’s use the same eddy viscosity models for the planetary boundary ayer as used in CFD. I’ve been quite interested in finding out how they account for clear air turbulence, but haven’t been able to get a response. I suspect, this mystery is treated as an afterthought in GCM’s since there are so many other complex effects to deal with.

    If this doesn’t satisfy you, Nick Stokes has said many times that “its all computational fluid dynamics” anyway. This is of course true and obvious with a moments thought. Of course aerodynamic flows are nowhere near as complex as the atmosphere and eddy viscosity models are at least comprehensible by a single individual. I doubt that’s true for GCM’s. They have large teams with specialists in different sub grid models.

    Nick has also said he doesn’t think that CFD numerical methods are that relevant to GCM’s. I disagree with him on that.

    But the take away is that the atmosphere is vastly more challenging than most other flows of practical importance.

    Do you know anything about the use of eddy viscosity in GCM’s? I would like to know the details of what is done. Clear air turbulence can be huge in effect, being of equal magnitude to the “average” flow velocities and the properties of such turbulent air are dramatically different those of non turbulent air.

  18. BBD says:

    Yes, but Dave, you’d said all this before and it remains deeply unclear where you are going with your argument.

    Are you suggesting that modeled estimates of ECS are too high to the extent that there are policy implications?

    Because if so, I return you to palaeoclimate behaviour, which contrary to your claim above, is well enough understood to extract a value range for ECS which leaves no policy wiggle room at all.

    That being the case, it’s hard to see what point there is to your fixation on models. They are not the primary source of knowledge on sensitivity.

  19. dikranmarsupial says:

    David Young wrote “There is separation all over the place in climate such as downstream of mountains.”

    I think the key point is whether such phenomena are of sufficiently large scale to be captured in the GCM directly, given the grid sizes currently feasible, or whether they are too small to be captured. If the latter is the case, then these effects are normally dealt with using downscaling methods (https://en.wikipedia.org/wiki/Downscaling).

  20. verytallguy says:

    DY

    I would also claim that these turbulent midlatitude eddies are in fact easier to simulate than the turbulence in a pipe or wind tunnel in a laboratory. This claim is based on the fact the atmospheric flow on these scales is quasi-two-dimensional. The flow is not actually 2D — the horizontal flow in the upper troposphere is very different from the flow in the lower troposphere for example — but unlike familiar 3D turbulence that cascades energy very rapidly from large to small scales, the atmosphere shares the feature of turbulence in 2D flows in which the energy at large horizontal scales stays on large scales, the natural movement in fact being to even larger scales. In the atmosphere, energy is removed from these large scales where the flow rubs against the surface, transferring energy to the 3D turbulence in the planetary boundary layer and then to scales at which viscous dissipation acts. Because there is a large separation in scale between the large-scale eddies and the little eddies in the boundary layer, this loss of energy can be modeled reasonably well with guidance from detailed observations of boundary layer turbulence. While both numerical weather prediction and climate simulations are difficult, if not for this key distinction in the way that energy moves between scales in 2D and 3D they would be far more difficult if not totally impractical.

    http://www.gfdl.noaa.gov/blog/isaac-held/2015/06/07/60-the-quality-of-the-large-scale-flow-simulated-in-gcms/

    Engaging there might help your appreciation of how these issues are perceived and dealt with by the community.

  21. bill shockley says:

    BBD,

    Apparently, and surprisingly, some eminent climate scientists do not accept the paleo derivation of ECS, Michael Mann among them. I’m not a student of the IPCC reports, but it seems possible they use model-generated ECS numbers as the basis for their analyses, which would make their conclusions less certain and expressed over a broader range of possible outcomes. Tamino covered the Mann paper yesterday on his blog.

  22. bill,
    In what sense is that Michael Mann not accepting the paleo derivation of ECS? His article is really just pointing out that if we get to 700ppm CO2eq, that there is a non-negligible chance of something like 6oC of equilibrium warming. Such a large chance of such extreme warming (i.e., much larger than if the distribution were Normal) means that we have a fat tail and that there is a non-negligible chance of a really extreme outcome. I don’t think paleo estimates rules this out.

  23. bill shockley says:

    ATTP,

    But how is Mann deriving this range of probabilities (never mind the shape of the distribution, for now)?
    If Hansen were asked this question, he would say the range of possibilities is 2.5C – 3.5C and the bell curve would be so constrained, rather than the almost 10C range in the Mann graphic. Unless Mann is talking long-term effect, i.e., ESS.

  24. dhogaza says:

    dikranmarsupial:

    “I think the key point is whether such phenomena are of sufficiently large scale to be captured in the GCM directly, given the grid sizes currently feasible”

    Yep. There are other points, too, but this is a good place to start.

    Note that DY didn’t actually answer my question …

  25. bill,
    It’s certainly long-term (i.e., run to equilibrium with a concentration of 700ppm CO2eq) since the peak is at 3C. I’m not 100% sure of the origin of the graph, though.

  26. dhogaza says:

    This is almost funny …

    Earlier, I wrote …

    “What, exactly, is the relevance to climate? Does a 20 knot onshore flow over a forested slope lead to flow separation and … oh, wait, this is so subgrid that detailed modeling of the flow over said slope isn’t of course even attempted by GCMs.”

    David Young responded:

    “There is separation all over the place in climate such as downstream of mountains.”

    Yes, of course. It’s what I said. You ignored the important part of what I said, though.

  27. bill shockley says:

    ATTP,

    Thanks. I’ll have to give it more thought. But… I can’t see Hansen ever considering 0.5C within the range of possibilities—ECS or ESS.

  28. bill,
    Oh, are you talking about the left-hand side of the figure, or the right-hand side?

  29. bill shockley says:

    ATTP, as you pointed out, the highest probability in the curve is just below 3C, so the intention has to be ECS. If it was ESS, then the peak probability would be near 6C. Also, Mann mentions in the article that slow feedbacks increase risks (the context of the paper) later on in the century. So, I think this is a model-derived curve and doesn’t pay attention to what we’ve learned from paleo studies.

    My mention of 0.5C is in reference to the vanishingly low, extreme left end of the curve.

  30. BBD says:

    If the highest probability is ~3C ECS (Mann) then how is this not congruent with palaeoclimate-derived estimates of ECS?

  31. bill shockley says:

    BBD, because the range of probabilities in the curve include values outside Hansen’s range of certainty.

    Mann’s curve: 0.5 – 10C
    Hansen’s range of certainty: 2.5 – 3.5C

  32. bill shockley says:

    I paraphased Mann as saying:
    Also, Mann mentions in the article that slow feedbacks increase risks (the context of the paper) later on in the century

    To avoid ambiguity, I was referring to this:
    With additional warming comes the increased likelihood that we exceed certain “tipping points”, like the melting of large parts of the Greenland and Antarctic ice sheet and the associated massive rise in sea level that would produce.

    This means that as we pass 2C, 3C, 4C, the risks of passing tipping points increases. Hansen thinks that if we arrive at 2C this century, heating in the pipeline + slow feedbacks take us irreversibly to 3 or 4C.

  33. bill shockley says:

    Oh, and sea level rise. 🙂

  34. BBD says:

    I’ve been under the impression for some time now that there has been decreasing confidence in and emphasis on the fat tailed distribution. In fact I thought James Annan dispatched it in about 2006 😉

  35. bill,

    Hansen thinks that if we arrive at 2C this century, heating in the pipeline + slow feedbacks take us irreversibly to 3 or 4C.

    Yes, I suspect that that is true. If we can’t find a way to reduce atmospheric concentrations once we’ve reached 2C, then 3C probably becomes almost unavoidable and – given slow feedbacks – 4C probably becomes unavoidable in the long-term.

  36. BBD says:

    Sorry, bill, but it took me a moment to find this post on James Annan’s blog discussing his 2006 paper.

  37. bill shockley says:

    Thanks, everybody, for agreeing.

    As Jennifer Francis would say, “we’re in a pickle”. 😦

  38. BBD,
    Thanks, I had seen that before.

  39. BBD says:

    Oh yes, we’re in trouble all right. In a sense this is why I find arguments about fat tailed distribution a bit annoying. Yes, there is a 10% chance that the situation may even be worse than ~3C / 2xCO2 but since this is itself a fairly nightmarish prospect under BAU, why run the risk of being labelled alarmist (incorrectly, of course, as Mann points out, but still labelled all the same)?

    Contrarians will seize on literally anything they can twist to make it look as though ‘warmists’ are also ‘alarmists’. Don’t hand them the nightstick and then kneel down is my advice.

  40. bill shockley says:

    BBD, thanks for the Annan link. Makes good sense. Commons sense, even. He beat Hansen by a few years.

    Good points about the “moderate”/more certain truths being bad enough. Same point you made about the Antarctic glaciers being past their tipping points.

    There is an overabundance of evidence that we need to do literally, everything we can, starting now. It will take a revolt of the type that happened during the Great Depression, when FDR taxed the rich and employed the millions of unemployed, and is already showing signs of beginning…. here, and especially in Europe.

  41. bill shockley says:

    Makes you wonder sometimes if these guys read each other’s papers??

  42. dhogaza says:

    Bill Shockley:

    “Makes you wonder sometimes if these guys read each other’s papers??”

    Yes, they do. They also aren’t shy of expressing skepticism of papers. True scientific skepticism, you know, the kind that ultimately deepens our understanding of the natural world.

  43. bill shockley says:

    dhogaza,

    Got examples? Not that I doubt that what you describe happens, but I’m wondering how frequently, and how significant a part it plays in scientific progress. From my limited experience, it’s thinking outside the box rather than thinking reactively that leads to progress in science. Jennifer Francis spending a year, sailing the arctic in a small boat, looking, thinking about things. James Hansen, reflecting that Earth’s history is where to look for answers, and following his own advice.

    But that is not really the point I was making. Why does M. Mann appear unaware of Hansen and Annan’s sound proofs that there is not so big a range-risk in the ECS number? I’ve seen this type of apparent ignorance in more than one place, by distinguished scientists who should know better (actually, I have one other example LOL). Mann is not challenging the old, but rather, ignoring the new.

  44. The figure in Michael Mann’s HuffPo articles comes from Wagner & Weitzmann’s book – Climate Shock. I think what they’ve done is take this statement in the AR5 SPM

    Equilibrium climate sensitivity is likely in the range 1.5°C to 4.5°C (high confidence), extremely unlikely less than 1°C (high confidence), and very unlikely greater than 6°C (medium confidence).

    and tried to assign actual probabilities. Remember that according to IPCC speak likely means 66% – 100% chance; very unlikely means 0 – 10% chance, extremely unlikely means 0% – 5%. Because there is a bigger probability of more than 6C, than there is of less than 1C, they’ve produced a skewed distribution with a fat tail. I tend to agree with people here that such a fat tailed distribution for ECS seems unlikely. However, I’m less convinced that given slow feedbacks, the possibility of carbon cycle feedbacks starting to operate, etc. that we should be ignoring the possibility of a black swan.

  45. bill shockley says:

    ATTP,

    I’m pretty sure I’ve seen that figure before, too. I might even be able to come up with it. Probably somewhere in the IPCC literature.

    Regarding the Black Swan, you’re preaching to the quire. It just struck me how out of sync Mann is with the thinking of the #1 climate change scientist. IMO. Seems like Hansen gets shrugged off a lot. There is extreme risk, but Mann’s argument is not accurate. Hansen argues that the big uncertainty risk is in the unknown radiative forcing of aerosols.

    My other example is Kevin Anderson, who thinks we need to mitigate at a rate of 8-10% per year. Apparently he is unaware of Hansen’s view that through forest and soil management we can draw down in excess of 100 GtC over the course of the century (the earlier the better), and that greatly lowers the necessary mitigation rate (he recommends 6%/year). If the world finds a way to coordinate their efforts well enough to mitigate emissions at a high rate, then we should also be able to coordinate the soil and forest management efforts. Drawing down carbon also allows us to lower the existing CO2 burden, possibly to 350ppm this century.

  46. bill,

    Apparently he is unaware of Hansen’s view that through forest and soil management we can draw down in excess of 100 GtC over the course of the century (the earlier the better), and that greatly lowers the necessary mitigation rate (he recommends 6%/year).

    I hadn’t heard this before. Do you have some kind of link? It’s an interesting idea.

  47. bill shockley says:

    Google image search all leads back to your Climate Shock source:

    Still feel like I’ve seen it somewhere else.

  48. bill shockley says:

    ATTP,

    Excerpt from a youtube lecture:
    CO2 drawdown:
    We pointed out in this paper that to do that, if we make an assumption that if we improve our agricultural and forestry practices enough to restore some of the carbon in the forests and in the soil by as much as a hundred gigatons of carbon—which we thought was an ambitious target—then you would need to reduce CO2 emissions 6% per year if you want to stabilize the planet’s energy balance this century. And that seems a little difficult, practically speaking. But one positive thing is that… the system is taking up approximately half of our CO2 emissions, so we’re burning enough fossil fuels to increase CO2 by about 5ppm—almost 5ppm per year— but it’s only going up 2 point some ppm per year. The other half is going into the ocean and into the soil or biosphere. And, in fact, that number is not understood very well. Because the models had said, 25 years ago, that it was 40 per cent disappearing, and they said, well these sinks are filling up and so it’s acutally going to decrease to 30% or 20%. Well, it hasn’t decreased—instead it’s increased to 50%, even though the emissions have gone up, so the size of the sink has really increased, a lot, for reasons that aren’t fully understood. So I think the potential for getting more stored in the soil and the biosphere is maybe more than a hundred gigatons, so we wouldn’t have to reduce as fast as 6% per year, but we need to understand that better.

    The paper he refers to is is Assessing Dangerous Climate Change

  49. Bill,
    Thanks, very interesting.

  50. bill shockley says:

    ATTP,

    Thrilled to be of service!

    If you search your blog for one of my first comments, the same reference is there, with the discussion about how the carbon sink is increasing in strength. And there is a response by someone who has the relevant Hansen paper that speculates on the mechanism whereby the sink is increasing. If you can’t find it easily, I’ll probably be able to do it. Maybe search for “sink”. It should not be very far back.

  51. bill shockley says:

    Here it is. It was Johnny On the Spot:

  52. mt says:

    Thanks for this discussion; I learned quite a few things. Not much to add, except to reiterate that I think a new code base and a proper career path for coder/scientists would help a lot.

  53. David Young says:

    VTG, Thanks for mentioning this Held post which I have reread and found interesting. You caused me to think about it again. He was unable to tell me anything about the clear air turbulence modeling. I suspect he runs the models but is not a developer. We exchanged some email about it but i didn’t get many details. 🙂

    His fundamental point seems to be that because the gradients in the vertical direction are smaller than those in the other directions, the flows are more 2D than many other fully turbulent flow. I understand that point and like a lot of engineering intuitive explanations, it is partially true. I do question though how much of a difference there really is and here is my reasoning :

    1. Fully developed 3D turbulence is a prominent feature of the atmosphere. These small eddies must come from the energy cascade from larger scales. How rapidly the energy decays would be a subject for testing using balloon measurements of wind and turbulence perhaps. In a lot of weather station data i’ve seen, the turbulent fluctuations are around 100% if the mean velocity, far larger than in aerodynamic flows. However, that’s in the boundary layer. At altitude, we know from flight testing that the story is about the same. That’s a lot of turbulence.
    2. The primary mechanism of mid latitude “weather” which is what Held is talking about is vortex dynamics. This can be a very challenging problem because the shear layers break up into smaller and smaller vortex cores. The exact rate of that breakup is critical to have any chance of determining the strength of even the large vortex cores. It’s also a problem in 2D incidently where its still very difficult to get it reasonably right. It’s exactly the same problem as predicting the dangerous wake effects of an aircraft on following aircraft., a problem where modeling leaves something to be desired.
    3. Tropical convection is also a very important weather and climate mechanism and Held has an old post on modeling this that shows extreme sensitivity to the size of the computational domain. Of course it’s a famous ill posed problem. His intuition says nothing about this. I am suspicious that models don’t match the data very well in the tropics because something may be missing there.

    Dikran: Downscaling seems to be something else entirely. Basically, a way of taking a very coarse grid solution and using it as a boundary condition for a small scale simulation or statistically correlating features of the course grid solution with local subgrid properties. That’s OK, even though there are some uncontrolled errors in this version of downscaling. A more modern method is domain decomposition that CFD usually uses where there are no errors in solving the discrete system. DD requires iteration between the coarse and fine grid scales and is probably far too expensive for weather modelers to use effectively. There are so many sources of errors, this one is probably a second order one.

    The main question here is how do the small scales affect the large scales? The effect is in some cases large, for example a mountain range or turbulence generally which changes the bulk viscosity of the fluid at a coarse grain level. This is where climate modeling is exactly like turbulence modeling, which it uses of course. You try to define the effect of the small scales by changing the large scale course grain properties of the solution. That’s hard enough for eddy viscosity models which have strong limits. It’s even harder with things like clouds or convection.

    Of course, that’s the answer to Dhog’s question too. Separation was just an example. Bousenesq’s assumption fails to hold in lots of nominally attached flows too. I think Dhog saw the reference here a while ago to the Aeronautical Journal survey paper on turbulence where some of those were discussed. He could reread that if he is interested in the details. But the broader point about subgrid model validation is really a general point independent of my example here. It’s just an obvious and widely known example where there is a vast literature to fall back on. All sub grid models rely on algebraic or differential approximations that make technically unjustified assumptions. In any real validation, one must test varying those assumptions also and that’s a real challenge.

  54. dikranmarsupial says:

    David Young wrote “Dikran: Downscaling seems to be something else entirely. ” I think you are missing the point. Downscaling models are used in climatology precisely because GCMs are too expensive to run at the sort of scales required to incorporate the effects of local topography (e.g. mountains). Using a nested high resolution regional model is only one way of performing downscaling, statistical downscaling is another (which I have worked on in the past), which is not affected by these problems (assuming the linkage between *modelled* large scale circulation and local climate are reasonably stable).

    “DD requires iteration between the coarse and fine grid scales and is probably far too expensive for weather modelers to use effectively.”

    In which case, it will be *way* to expensive for climate modellers to use at all.

    So far you have provided no evidence AFAICS that the sort of problems you raise are relevant to the sort of grid scales on which GCMs currently operate, just the assertion that they do. If you provide this evidence, then it will help us to appreciate the point you are trying to make.

    It is my understanding that climate models are used in generating the various reanalysis products; I would have thought that if there were substantial issues in modelling large scale atmospheric circulation, this would have been exposed by these projects where weather model outputs and observations are assimilated.

  55. BBD says:

    @ bill

    Johnny on the Spot? I barely turn up for my own breakfast these days 😉

    Note that Hansen is very far from sanguine about the *temporary* increase in drawdown, which he does, after all, refer to as a Faustian bargain.

  56. verytallguy says:

    DY,

    I think if you want relevant input to your points you really need to publish in the relevant literature. I certainly lack the expertise to help you.

    A paper demonstrating that these issues are both relevant and not currently recognised in the climate modelling literature would move your agenda far faster than any number of blog comments.

    Good luck!

  57. bill shockley says:

    BBD,

    LOL. Breakfast is about the only place I turn up… once a day… all day!

    I agree, the growing sink feels fragile at an intuitive level. He is still saying 6%/year, even though in that video he says a slower rate of mitigation might be possible. He also thinks 6% pushes practical limits.

    We need to leave as little as possible to luck, which does not favor those late to the show.

  58. dpy6629 says:

    Dikran,

    There is actually some strong evidence provided by Paul Williams about 5 years ago where he improved Weather forecasting skill in one model by improving the overly dissipative and very old and outdated leapfrog scheme. There was a Newton institute video on this that is good. That’s very strong evidence.

    The problem I raised is that downscaling contains uncontrolled errors in the interactions of he course and finer scales. Other errors may be larger however.

    The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right. That’s said to be due to the dogma of the attractor. I think Tol is right that there is no rigorous basis for that.

  59. dikranmarsupial says:

    dpy6629 thanks, I’ll see if I can find the video. However I am not sure that constitutes *strong* evidence as the needs of weather forecasting are not the same as for climate prediction (even if they do use similar models). For climate prediction the important thing is for the statistical distribution of weather, given the forcings, to be correct, rather than that the detailed reproduction of individual weather events is correct.

    I don’t see how statistical downscaling in particular suffers from the problems you seem to suggest. However, the point I was making is that climatologists don’t expect GCMs to capture the effects of topology (e.g. mountain ranges), and deal with such issues in other ways.

    “The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”

    Because the limited forecast horizon for the weather prediction is limited by the sensitivity to the initial conditions, which would be the case even if the model physics were perfect and the resolution infinite. The initial conditions however have almost no effect on the 1000 year averages. Climate modelling is not long term weather forecasting, the thing of real interest is the forced response of the climate, which is essentially what is left after all the “weather noise” is averaged out.

  60. dikranmarsupial says:

    The example I normally use for why the limited forecast horizon of weather models doesn’t mean that climate models are unreliable is the double pendulum. Consider a double pendulum (https://en.wikipedia.org/wiki/Double_pendulum), made of iron, with an electromagnet placed to one side. The double pendulum is chaotic, i.e. its dynamics are deterministic, but highly sensitive to initial conditions. This means even if we make a perfect computer model of the double pendulum, its exact path will deviate from the path followed by the real pendulum, unless the initial conditions are know with complete precision. We can however use the model to predict the statistical distribution of the position of the system without any problem at all, we just need to runs lots of simulations in parallel and look at the distribution of states that we observe.

    Now if we increase the current in the electromagnet, the statistical distribution of pendulum states will change as the pendulum is attracted to the electromagnet, and our computer simulation, if it includes a model of the electromagnet will be able to predict the change in the distribution, even though it can’t predict the exact position of the pendulum with any accuracy for very long after the start of the simulation.

    Hopefully the analogy to climate modelling is fairly obvious. The chaotic oscillation of the double pendulum represents the weather, the statistical distribution of pendulum states represents the climate (i.e. the statistical properties of the weather) and the electromagnet represents climate forcings (GHG, volcanic, solar etc.). The change in the statistical distribution corresponds to the forced response of the climate, which is the thing that the climate modellers are primarily interested in (for the purposes of informing policy).

    Of course a climate model is much more complicated, but hopefully this illustrates why the common argument that we can’t predict weather more than a few days in advanced, so how can we predict climate 100 years in advance, is incorrect. Weather prediction depends heavily on initial conditions, centennial scale climate predictions has no real dependence on initial conditions.

  61. dpy6629 says:

    Dikran, I know all the simple analogies. I have nicknamed this the dogma of the attractor and I believe there is no foundation I have found for it despite looking very hard because of the implications for CFD.

    I also have counterexamples where the initial conditions do change very strongly the long time behavior. See AIAA journal Aug 2014, Venkat et al . These multiple solutions were found by small changes in time stepping or preconditioning strategies. These solutions are sometimes close to each other and sometimes totally different. This paper is the first report on this because we are the first ones to use powerful modern nimerical methods and try to really get grid converged answers. However now these these have been found in other codes too. Why only now? Selection bias is a powerful thing that people often do without really realizing it.

    My conclusion is that the dogma of the attractor must be false in general. I think the logical assumption is that these things exist in other applications of CFD such as GCMs. The fact that “every time I run the model I get a reasonable climate” seems like a absence of evidence is not evidence of absence thing.

    Sorry about posting under a different name. My different devices use different accounts .

  62. dikranmarsupial says:

    dpy6629 (don’t worry, the username didn’t cause any confusion)

    O.K., so you would agree that there are chaotic systems that have predictable statistical behaviour (double pendulum) and there are others that don’t (doi: 10.2514/1.J052676 ?). What is your evidence that the climate is one of the latter?

    One line of evidence that it is one of the former is provided by paleoclimate data (I suspect there are others). The shifts between glacial and interglacial periods over the last 800,000 years or so bring the climate back to pretty similar states between transitions (“tipping points” if you like). This would be hard to explain if the statistical behaviour of the climate were unpredictable.

    Are you really arguing that a difference in initial conditions large enough to disrupt a short term weather forecast would also be large enough to cause a non-negligible change in the Earths mean climate over the next 1000 years? That seems rather doubtful.

  63. dpy6629 says:

    OK, but there is evidence I think that changes like Panama forming can change currents enough to have big changes. However, I don’t trust my or anyone’s intuition about climate’s sensitivity to initial conditions.

    What I can prove is that for Navier Stokes (RANS) there are a plethora of these multiple solutions despite 20 prior years of “expert” belief that that would not be the case.
    The most compelling ones and the most alarming (for RANS faithful) are the ones hat are close together because it might mean our past code validation could be wrong and an example of selection bias. But we don’t know enough yet to say that either. It is alarming though if you want to use this stuff where public safety is at stake.

    And we would have never found this without modern numerical methods and a brand new code.

  64. What is your evidence that the climate is one of the latter?

    A good question and David’s response is nowhere near an actual answer to this. What irritates my about David’s basic view is that (as I’ve said many times) he’s using knowledge of the technical details to imply things about our climate that virtually all experts regard as not true. We understand the basic energy flows in the climate system (globally at least) and we know the heat capacities of the different parts of the system. This means that we really do not expect large, chaotically-driven variations because it would almost require violating conservation of energy.

  65. dpy6629 says:

    ATTP, I’m not irritated however RANS conserves mass momentum and energy and there are a surprising plethora of multiple steady states. Some are close together. Intuition about NS or climate is just that, and it’s not really scientific. I would ask you before getting too irritated to read the paper. AIAA Journal, August 2014, Venkata et al. Now we are finding this in older codes too. I am quite confident the paper is right because the authors really hoped it was not but finally ran out of excuses. But maybe you can find a problem for which I would be grateful

  66. David,

    I’m not irritated

    I don’t care if you are or aren’t. I am, though. You’re doing what you always do which is highight highly technical details without illustrating that you have any understanding of the basics. Show some understanding of the basics and I might take you more seriously. While you keep doing what you’re doing now, I will treat you with the disdain you deserve.

  67. dpy6629 says:

    Just to be clear. I am not saying we can’t predict the final state in some cases, just that there are some quite surprising counterexamples. More importantly, to know when you can predict and when you can’t requires better methods and codes, a petty non controversial point.

  68. dpy6629 says:

    Ok, so you can’t say that I’m wrong or even read the paper proving I might be right. General intuition about climate is not an argument unless I missed something. As Nick Stokes says it’s all CFD anyway. I don’t understand what you want me to do. It’s a referred paper that shows 20 years of expert opinion ( or more accurately expert hope) was wrong at some level. And it’s an interesting result in its own right. Don’t worry I am not an author so you can pretend it’s some one less irritating to you if that will help.

  69. David,

    I don’t understand what you want me to do.

    I would like some sense that you understand the basics of climate science. Albedo, fluxes, insolation, Planck functions, heat capacities,….. It’s not that hard. Focusing on details, like non-linearities, without showing you understand the basics, just means that talking with you is likely to be a waste of time. Also, what paper? You don’t seem to have provided any kind of link.

  70. dpy6629 says:

    I referenced it on a previous thread and offered to send it to you privately and you declined saying you could find it. It’s AIAA Journal, August 2014, Venkatakrishnan, Kamenetskiy, Johnson et al. It follows a paper by me which you will ignore I suspect even though it’s good too😀
    Your library will have it or you can pay a small fee online at aiaa.org or you could swallow your pride and email me for a free copy. You have my email

  71. @dpy6629 O.K., so you have no evidence that the long term statistical behaviour of the climate is unpredictable.

    “However, I don’t trust my or anyone’s intuition about climate’s sensitivity to initial conditions.”

    I provided more than intuition, I provided evidence, in the form of what we see in paleoclimate data. This isn’t absence of evidence, this is evidence of stable regularities in the climate system and its response to forcings.

    “…there is evidence I think that changes like Panama forming can change currents enough to have big changes.”

    Well yes, but that would be regarded as a change in the forcings, rather than chaotic internal variability, so it doesn’t really support your contention as far as I can see. The closure of the Panama gap was also a huge change in the climate system (which includes the oceans), so of course it had a huge effect. The fact that climate models, given suitable adjustments to position of landmasses and forcings, give plausible simulations of the climate of the past (e.g. IIRC uplift of Appalachians) again is evidence that while weather may be chaotic, the climate probably isn’t.

    At the moment, you have provided some evidence of an issue in fluid dynamics models, and provided a suggestion of some evidence that weather prediction can be improved if this issue is addressed to some extent. However that does not imply that it is something that will make a big difference in climate modelling. Say fixing this issue doubled the prediction horizon for weather forcasting, would that make a substantial difference to climate modelling? I very much doubt that it would. I have been trying to help you make your case; I don’t accept something without evidence either (although I don’t totally disregard intuition or experience either, hence I am taking your argument seriously), so if you want to make progress, you need to provide evidence, or at least explain why paleoclimate results do not suggest that climate behaves in a largely predictable manner.

  72. David.
    It’s a paper about multiple solutions for the Navier Stokes equations. Fine. That paper does not, however, indicate that we would expect large variations in our climate, from some kind of long-term average, simply because of the chaotic nature of the system. That’s the problem. That’s why I think you need to show that you at least understand these basics, because it seems to me that you’re invoking complexity to sow doubt, without even trying to understand whether or not such complexity can do what you seem to be suggesting.

  73. Dikran’s last paragraph in the above comments hits the nail on the head.

  74. David wrote: ” I am not saying we can’t predict the final state in some cases, just that there are some quite surprising counterexamples”

    Climate modelling is not really about predicting the final state, it is about determining the forced response of the climate and assessing the variability around the forced response caused by the chaotic internal climate variability (e.g. ENSO).

  75. dpy6629 says:

    What I see here, Dikran and attp is some hints of the forced vs internal variability dogma which is a very mushy thing. The point here is that energy can be conserved and the forcing be the same and there be many final states each of which may be chaotic itself. This seems obvious in retrospect.

    Sometimes these states are far apart and sometimes close together which is the more alarming case. You say that expert intuition says this doesn’t happen. Aside from intuition what’s the evidence. Schmidt said people had looked for this and not found it In the models. It was never found in NS either until we built a new code with real powerful methods. However, Schmidt indicated experts thought it a possibility.

    I have to stop now and maybe this is not the forum to continue anyway. I gave you some evidence that seems to contradict your intuition. Think about it from the point of view of simple math. The more nonlinear a polynomial equation is the more likely it is to have multiple solutions.

  76. David wrote “What I see here, Dikran and attp is some hints of the forced vs internal variability dogma which is a very mushy thing.”

    No, actually these terms are quite well defined and understood. The double pendulum example I gave earlier is one in which there is very obviously a forced response to the electromagnet and a chaotic variability around that forced response. Dismissing this a “dogma” and “a very mushy thing” is unhelpful, and is not a satisfactory response to the gap in your argument that I have been politely trying to help you to address.

    ” The point here is that energy can be conserved and the forcing be the same and there be many final states each of which may be chaotic itself. ”

    As I have pointed out, climate modelling does not aim to predict final states, it aims to estimate the forced response of the climate. The fact that you view this as a “very mushy thing” may be merely an indication that you don’t really understand this distinction, and that a better understanding would help you make our point more convincingly.

    ” I gave you some evidence that seems to contradict your intuition.”

    Actually, no. I am willing to provisionally accept there is an issue with the fluid dynamics that may well improve the forecast horizon for weather prediction (I suspect ATTP may agree that far). However you have provided no evidence that this affects the long term statistical behaviour of the climate (which is what matters for climate modelling), and have provided no argument against the evidence from paleoclimate that I gave that indicates that long term climate is predictable.

  77. David Young says:

    Dikran, I guess as you say I don’t understand the distinction between final states and responses to changes in forcings. Seems like climate models aim to predict the final state resulting from changes in forcing among other things. That state can be time dependent and have chaos of course. Is this distinction really well defined? Maybe you can help me on that. In CFD, there is a very old dogma, viz., that we can’t predict absolute forces, but we can predict increments resulting from changes. This sounds a little like what you are perhaps getting at? Asking, not saying definitely. We have a new paper showing this is not really true, but it is true for a very narrow range of situations if you spend 100’s of man years tuning your methods.

    Please try to help me on this one.

    I also don’t understand what your paleoclimate evidence has to do with this. If indeed states are close to each other, we would have little chance of seeing that in the very coarse grain paleoclimate data. Are you assuming that the only multiple states must be far apart? I don’t think that is really likely for all of them and there are many of them.

    Another factor to consider is that the model itself may indeed have different behavior than the system being modeled and I think all modelers who are realistic would admit that this is obviously true even in course measures like regional climate.

    My main point is the same as Tobis’, namely that its time to abandon existing models and build at least one modern one. The money saved can be spent on better data, which for the paleoclimate seems to be to be desperately needed. This idea the paleoclimate data which is so controversial and uncertain should tell us much about a nonlinear chaotic simulation and its veracity is a little farfetched. I think the simulations are pretty weak at predicting ice ages anyway as I recall and that’s OK of course, but there is huge room for improvement. Maybe I didn’t read that part carefully enough though.

  78. Kevin O'Neill says:

    DY writes: ” This idea the paleoclimate data which is so controversial and uncertain …”

    What is controversial about paleo data and what is uncertain? We’re talking pretty much about the last 800,000 years. You keep coming back to this, but I know not what controversy you speak of.

  79. David Young says:

    Dikran, After some more thought while walking let me try to say it right in your language.

    We can’t really compute absolute temperature or even really the regional climate in its final state. But we can compute small changes in temperature due to small changes in forcing. Is that it more or less?

    I would respond that the way we operationally do that is to run the model with the baseline forcing. You then run the model with the perturbed forcing and then take the difference in the two “end states.” I’m not sure if “final states” has some other meaning for you.

    In addition, one would want to make sure that the two runs were equally “converged” whatever that means. And our work shows that’s a very important issue.

    Am I on the right track?
    DY

  80. David,

    forced vs internal variability dogma

    Why do you keep using words like dogma? It is quite possible to do a fairly basic calculation as to what would be required to see a large internally-driven variation away from the equilibrium point. It’s possible to do a basic calculation as to how we should respond to changes in forcing. The reason that people regard it as a system that responds to changes external forcing plus internally-driven variability around the forced response caused by the chaotic internal climate variability, is because that appears to be the best description of the system based on our understanding of the underlying physics. If you’re going to keep using words like dogma I’m going to keep treating what you say with disdain.

    You also keep mis-representing what Michael Tobis is saying, and you don’t appear to have even attempted to show that you understands the basics of climate science.

  81. David Young says:

    ATTP, My god, have you read the paper? Do you disagree with my characterization of the standard climate science statement? I am bending over backwards to understand what the dogma is so I can then show you the paper where we refute it. I know facts and data are a little challenging. I am not Tol and you are not Winston Churchill. Really.

  82. David,
    Do you not understand why describing a scientific position as a dogma is rather insulting?

    Do you disagree with my characterization of the standard climate science statement?

    If you mean the one with dogma in it, then yes.

    I am bending over backwards to understand what the dogma is

    It’s not a dogma. If you really want to understand this then you should ignore GCMs for a while and try to simply understand the basics. What is a forcing? What are feedbacks? What role do the oceans play? What are the heat capacities of the different parts of the system? What can cause us to warm/cool? What role do volcanoes play? What about the Sun? ……

    At the moment you simply seem to have shown is that you’re familiar, maybe an expert, on the details of CFD. You need – IMO – to also try and understand the basics of climate science so that you can put climate models into the right context. Simply pointing out that the system is non-linear and chaotic is not sufficient to show that this will be relevant to what climate models are trying to determine. Most experts think it isn’t.

  83. mt says:

    The gap between CFD people and climate modelers can be instructive. I advise more patience toward David Young in particular. He seems sincere and smart, though to be sure I’ve been disappointed often enough by people’s stubbornness.

    Climate models are structured as particularly messy CFD models, but the problems of climatology are very different from the problems of mechanical or chemical engineering. It’s good to explain how, but hard.

    One of my core points is that rigorous disciplines underestimate each other habitually. That’s why neither software engineering nor computational science seem able to make a dent in climatological methodology.

    Anyway I don’t think CFD as usually practiced has a lot to directly offer GCMs, though there are some process studies where a closer collaboration would be very helpful. On the other hand, some of what they do is enormously impressive in its core applications.

    Who knows. I wonder about my own deeply unimpressed assessments of economics. Maybe I’m missing something. But that’s another matter…

  84. BBD says:

    It’s just daft. I had the same discussion with another ‘chaos and unpredictablity’ peddler a while back and they too were unable to address the remarkably predictable way in which interglacials have played out over the last ~800ka. Same palaeoclimate agnosia guff too.

    It’s junk rhetoric and it’s boring.

  85. BBD says:

    mt

    I advise more patience toward David Young in particular. He seems sincere and smart, though to be sure I’ve been disappointed often enough by people’s stubbornness.

    Having had several interactions with DY in the past I am going to have to disagree with you in essence. He is the problem, not everybody else.

  86. MT,

    I advise more patience toward David Young in particular. He seems sincere and smart, though to be sure I’ve been disappointed often enough by people’s stubbornness.

    Okay, it’s your post 🙂

    I also think DY is smart. In a sense that’s what I find frustrating. Smart people who haven’t bothered to famaliarise themselves sufficiently with a topic and think that their deep understanding of an aspect of the topic allows them to draw broad conclusions.

  87. DY writes: “Am I on the right track?”

    Sadly, no, I’ll see if I can help explain.

    “We can’t really compute absolute temperature or even really the regional climate in its final state. But we can compute small changes in temperature due to small changes in forcing. Is that it more or less?”

    If you by “small changes in the forcing” you mean changes in forcings small enough not to produce a tipping point, then yes, the expected response of the climate is predictable (although we may have different definitions of a “small change in temperature”, 2 degrees change in GMSTs is not a small change). However, you are still talking about final states, which is missing the point. Individual states of the climate system at any point in time are chaotic and unpredictable. We are instead trying to establish the change in the statistical distribution of possible states that result from a change of forcings.

    “I would respond that the way we operationally do that is to run the model with the baseline forcing. You then run the model with the perturbed forcing and then take the difference in the two “end states.” I’m not sure if “final states” has some other meaning for you. ”

    Well I think this is the problem, the way in which climate modellers use modellers is not quite the same was as the way in which they are used in your field. If you run a climate model multiple times with the same forcings, they don’t all end up in exactly the same state (because of internal climate variability). You can view the individual model runs using tools such as climateexplorer and you will find this is the case. They do all however show broadly the same tendencies for temperatures to rise in response to increasing forcings, and that is what climatologists are interested in, the specific final states are not that interesting AFACIS. The other thing of interest is the spread around this mean behaviour, which represents (some of) the uncertainty of the projection.

    “In addition, one would want to make sure that the two runs were equally “converged” whatever that means. And our work shows that’s a very important issue.”

    As we don’t expect the simulations to end in exactly the same state, the idea of convergence doesn’t really crop up, AFAICS, so it isn’t a very important issue for climate modelling. As I understand it, the models are often run for a while as a “burn in” period, but I think that is to allow them to “forget” the initial conditions.

    Now I am being patient, as MT suggests, but frankly if someone from outside your field told you that some central idea in your work was dogma, when you used it to try and explain to them why you thought they had misunderstood you, it is likely that you would interpret that as rudeness and/or that they weren’t taking what you were saying seriously. So if you want people to be patient with you, then it would be a good idea to avoid calling other scientists views “dogma”.

  88. DY, consider the double pendulum example again, this time we can have an ensemble of double pendulums, each with their own electromagnet (providing indentical forcings). As we can’t set them all off in the identical initial conditions of the real double pendulum we are trying to model, the best we can do is set them off in similar initial conditions (perhaps some random jitter to cover the range of plausible initialisations). Now when we set the simulations running and slowly increase the current in the electromagnets to match the observed forcings, the states of the pendulum in the different model runs will quickly loose any coherence they had at the start (as it is a chaotic system). This means the final states of the simulated pendulums will all be different (note, as I have said, final states are not what really matters). However if we compare the statistical distribution of those final states with that of a control simulation where the forcings are constant, we will find there is a small shift in the distribution, which is an estimate of the forced response.

    A particularly important point here is that the ensemble mean (i.e. the mean state over the ensemble) is not directly a prediction of the position of the real pendulum at any particular point in time. This is because it is just an estimate of the “bias” in the pendulums movement caused by the forcing and it has its own chaotic movement superimposed on that bias. Similarly the ensemble mean of a GCM ensemble is not directly a prediction of the Earth’s future climate, but just of the expected effect of the change in forcings, over which there will be the effects of internal climate variability (expected to lie within the spread of states seen in the model runs).

    So, in the case of the double pendulums, do you agree that the final states of individual model runs are not expected to converge to the same thing and that this is not important in estimating the expected [mean] effect of the electromagnet?

  89. mt says:

    I didn’t think it was an intentional insult the way DY used “dogma”. I don’t fully agree with marsupial’s pitch here either, though some of the points are good.

    But both DY and marsupial seem to agree that the purpose of CGCMs is to establish the global mean temperature trajectory or the sensitivity, I contest this.

    That’s a side effect. It isn’t really what the builders of the models find interesting.

    It certainly is a key product of climatology as an applied science. But the models are not built to achieve that. They are built to represent and advance the science itself.

    That a sensitivity range inevitably emerges from the models is an interesting and informative fact in itself. But you don’t build a CGCM to determine GMST or its trend. If you do, you’ll get nothing of value.

    You build a CGCM to embody a complex collection of simple theories in such a way as to produce a system closely analogous to the real world. There are many advantages to such a tool. One is that you can do experiments on it – better to do them on a digital world than the real one.

    The impetus to do 21st century scenarios follows.

    But if you create GCMs with the idea of projecting the future or indeed with any idea other than embodying a complex collection of simple theories to create a realistic simulation, you are barking up a very wrong tree.

    Insomniacally yours, mt.

  90. mt says:

    DY “My main point is the same as Tobis’, namely that its time to abandon existing models and build at least one modern one. ” No, I said nothing about abandoning extant models. Though I imagine success in the sort of endeavor I propose will hasten the long overdue disappearance of Fortran from active use, I don’t expect this to occur quickly.

    “The money saved…” How is any money to be saved by adding yet another codebase to the collection? My ideas are definitely not about penny pinching. I think climate science is too important to be done half-heartedly.

  91. Andrew Dodds says:

    Regarding the ‘close end states’ problem..

    Bringing this to the real world, if these states were within the range of internal variability and common pertubations – say +-0.2K on a global scale – then the climate would shift between them more or less at random, and for all intents and purposes they would simply be part of internal variation. There may even be an example of this in the PDO.

    So if we have a variety of stable end states for a given starting state, they must be further apart than normal variability, otherwise they are effectively the same state.

  92. MT it isn’t really a pitch, I am just trying to explain to DY what he needs to do to show me that the CFD issues he is interested in has a substantial bearing on climate prediction rather than weather forecasting.

    I think I made the distinction earlier that the forced response is what is of primary importance for the scientists in helping to inform the debate on what we should do about climate change. Of course their real interests are in understanding the world, but I wanted to keep the discussion focussed on the key issue where DY seems not to appreciate the operation of climate models. I completely agree with MT on the fundamental aim of building a GCM. I am also not against the idea of building the issues DY mentions into the next generation of the models if it means they then reflect reality more closely. However that doesn’t mean that it will result in a substantial change in what the models say about climate, or indeed that they are not other areas where work will produce more important improvements in the models. The best way would be for DY (as an expert on the topic) to work with modellers to improve the models.

    According to Wikipedia “Dogma is a principle or set of principles laid down by an authority as incontrovertibly true”, which does not apply here as I am capable of explaining (with the double pendulum example) what is meant by forced response without reference to any authority, and I am quite happy to discuss its validity. Calling it dogma is rude as it is implying that this is not a reasoned understanding of the science, but merely an unquestioned belief passed down by some authority, which is deeply anti-scientific.

    Most of the time when scientists use the word dogma to describe an aspect of their own field, they are using it ironically to indicate that they know it isn’t completely valid, for instance the “central dogma of molecular biology”. These days when my biologist friends say dogma it seems to be pronounced with the “”. Now I am comfortable talking to biologists about the central dogma of molecular biology, as I also pronounce it with the “” and they will get it that I understand its limitations. However, I would not call someone else’s science dogma, especially if it was an argument they were using to question my views. Fortunately I am generally fairly placid by nature and don’t respond, however if DY wants to make progress in what has become a rather confrontational debate, it would be better for him to avoid needlessly insulting the people he is talking to, as a hostile response is perfectly natural and expected.

  93. MT,
    What you say probably illustrates why I find these kind of discussions frustrating. In a fundamental scientific environment you build model to try and understand physical systems. Either you’re using them to try and understand some observations, or to try and understand how a system might evolve. As you say, you typically build them on the foundation of relatively simple theories. Sometimes the models themselves can remain quite simple, in which case you’re probing a subset of the physics that can influence the system, or they can become very complex, in which case you’re trying to represent the system as fully as you think you can. That doesn’t really change that the goal is to understand and probe the system.

  94. mt says:

    A whole essay on stubborn wrongness has emerged in my head.I hope I get to it. I find it easier to write these things mentally than to actually, you know, write them.

    The moral of the story, though, is that people who mean well and are reasonably true-sense skeptical still find it hard to adequately challenge their own ideas. Generally there is some ego involvement.

    My experiences with DY are too brief to know whether this is his case, but I gather there is already some frustration hereabouts with him.

    But the other side of the coin is that it may be the fault of people trying to explain the contrary view to him.

    In this case, I agree, that improved CFD is of very minor importance in directly improving GCMs. And I would explicitly include global weather models here. (Again, I do see a role for better numerical methods in some important subfields of climate, though.)

    Making the case is rather subtle though. In a case like this it helps to take some time to understand the intellectual world the other person comes from. If you say it’s not like this, it’s like that, the other person may be as frustrated by your misunderstanding of his “this” as you are by his misunderstanding of your “that”.

    In the present case, I am open to that interpretation, but again, I’ve been sorely disappointed in the past.

    The alternative hypothesis, which apparently some here hold, is that DY is not attending to the arguments against his position.

    Two specific occasions come to mind, where I was arguing with an intelligent person (2 different people on 2 different subjects at 2 different occasions) who was unambiguously wrong on a matter, and felt I could convince them of it with a tight logical argument. It turns out, though, that success in a direct contradiction of an argument is extraordinarily rare, far more so than one would expect, even when the opponent obviously has the intellectual capacity and background to understand the argument. In both cases, the conversation was public on the internet, so a certain amount of ego threat may have been involved.

    But DY is not, that I have seen, as unambiguously wrong as those two others. I don’t have an unambiguous proof that higher order accuracy in the PDEs is irrelevant, for instance, though that is the first thing CFD people think when they look at our hulking models.

    As usual with scientists (and I exclude economists) we are not clueless about our own domain. It turns out that high order accuracy is not an issue in GCMs or even weather models for reasons I won’t attempt to explain just now on a sleepless night. If it were, we would have done it by now.

    But it’s not a trivial or uninteresting question.

    It’s interesting what questions get asked of climatology by what disciplines. Usually there is something to learn in constructing a good answer, and something else ot learn in trying to get someone else to understand your argument.

    Of course, eventually there is a point of futility.

    So the question of how much y’all and he have been talking past each other and how much DY is being stubborn is unsettled in my mind. If ATTP doesn’t mind another go around, I’m happy to take it on and suggest to DY that he start from tabula rasa stating his position so I can try to understand it.

  95. If ATTP doesn’t mind another go around

    Not at all, I think there are some very interesting issues here. I just find it hard to hide my frustration at times.

  96. MT wote “But both DY and marsupial seem to agree that the purpose of CGCMs is to establish the global mean temperature trajectory or the sensitivity, I contest this.”

    Actually, I don’t agree with this either! Even if you are only interested in informing policy in the climate debate, if you just want to estimate the GMST trajectory or climate sensitivity, then much more simple physical models will give pretty much the same sort of answers AFAICS. From a policy perspective, the thing GCMs give is a handle on the consequences for regional climate that you can’t get without modelling the general circulation (atmospheric and oceanic) AFAICS. The discussion largely considered GMST, but that doesn’t mean it is the only thing that is important.

  97. BBD says:

    mt

    You might find this occasion when DY attempts to ‘educate’ Gavin Schmidt interesting background as it sets the tone for much that has followed since.

  98. MT wrote “In a case like this it helps to take some time to understand the intellectual world the other person comes from”

    Yes, one way of doing this that seems sensible to me is to consider a simpler system (for instance a double pendulum) and find out their understanding of that. If DY were willing to discuss the (un-)importance of final states in that examples and the forced response as a change in the distribution of states, then that might happen. Unfortunately, people often seem reluctant to discuss simple systems/thought experiments (I have others ;o).

    I agree about the point regarding backgrounds causing communications difficulties. I come from a computational statistics background and Monte-Carlo methods and inference-by-simulation seem pretty natural concepts. This means some aspects of climate modelling seem pretty natural to me, others rather less so.

  99. mt says:

    BBD, thanks. It’s an interesting exchange at RC. I don’t think DY came off that badly. I also think Gavin missed an opportunity to clarify at #70.

    I’m getting the sense that DY lacked a clear understanding of what the output of a climate model is; that is, it is not clear what “climate” means to him. Some of what I read here is similar. Maybe we should start there.

    There are subtleties there, too. There’s a fairly facile definition of “climate”, but I think it leaves one in a position where It is not easy to define “climate change” in a mathematically sound way.

  100. MT,
    Feel free to correct me, but it seems that a great deal of this is related to confusion about terminology, or what climate models are being used to do. If we were wanting to make precise predictions about weather events, then the results would be very dependent on the initial values and would blow up (or diverge from reality) after some time interval. However, that’s not what climate models are for. They’re being used to understand trends and climates (suitably averaged weather). Similarly, I think some misunderstand what people mean when they refer to boundaries. Formally, the boundary values are the values on the boundary of the grid. However, I think that in this context people often use it to mean what bounds the systems. In other words, insolation, albedo, atmospheric composition.

    Firstly, maybe you disagree with the above, but if it does have some merit, then understanding both the terminology and the uses would probably help. I am, however, probably being woefully naive.

  101. I don’t know if the definition of climate as being the statistical properties (distribution) of the weather system (all fields, not just GMSTs) is the fairly facile one (as a statistician, of sorts, it seems reasonable/useful to me). I’d be happy to hear deeper or more useful ones.

    A nice “definition” I’ve often seen is “climate is what you expect, weather is what you get”, which seems a good start.

  102. bill shockley says:

    dikranmarsupial said:
    the thing GCMs give is a handle on the consequences for regional climate that you can’t get without modelling the general circulation (atmospheric and oceanic) AFAICS.

    Seems like you could at least get a handle on these things from paleo studies. I have no idea what is in the literature.

  103. Paul S says:

    dikranmarsupial,

    …if you just want to estimate the GMST trajectory or climate sensitivity, then much more simple physical models will give pretty much the same sort of answers AFAICS.

    Yes, though mostly simple models are only considered legitimate to the extent that they agree with GCMs.

  104. mt says:

    “the thing GCMs give is a handle on the consequences for regional climate”

    Well, no, I don’t think so, not at present. Not to say that they shouldn’t or couldn’t.

    I hate to be channeling RPSr., but current models don’t agree very well inter-model and so must be considered unreliable at regional scales. In scenario runs they may give a flavor of what to expect at large but not quite planetary scales, e.g., drying subtropics, extreme warming in the Arctic, both of which we are already seeing. Not much detail.

    In particular, as I understand it, models severely underrepresent severe events. So they don’t tell us much about the trajectory of something that practically we care very much about.

    One of the questions I raise is whether it isn’t worth the public’s while to try to greatly improve this situation. But it’s a risk – the extent to which that is possible is unclear.

    I would also say that modeling is only half the story. While the initial condition of the atmosphere doesn’t matter at all, the initial condition of the ocean does, a great deal, and at present it is underspecified. Fixing this would be an expensive and relatively unglamorous pursuit.

    The main thing GCMs of the present vintage give is confidence and progress in the science.

    It is my belief that we don’t know to what extent we can do better in terms of operational prediction. If we can’t have better models, fans of geoengineering really ought to drop it.

  105. mt says:

    The trouble with defining climate as the statistical properties of weather is that it presumes that there is such a distribution – for practical purposes that means a stationary process. So how do you define climate change? Remember you only get a single realization from a changing distribution.

  106. BBD says:

    mt

    Impressive clarity for someone who hasn’t slept a great deal. Thanks.

  107. David Young says:

    Dikran, I do think the problem here is my use of the term “final state.” What I mean is the “shape” of the attractor. Calling it the statistics of the attractor is also fine i guess. Rephrasing then, the ultimate statistics it seems to me can depend on initial conditions. The paper I referenced is about steady state models, but in fact those realizations are just averages of chaotic velocity fields anyway. In addition to initial conditions, the statistics can very definitely depend on numerical methods, time step used, etc. etc. Paul Williams showed an example. In any case, we are talking here about computational models of fluid dynamics. We have no idea if these multiple solutions are all realizable in experiments or not. Its a very interesting question I think. In the past people didn’t particularly look for these things.

    The use of the D word is a device to provoke thought I picked up from some very smart mathematicians. I can try to stop using it here. There are a lot of things in science, especially at the level ATTP wants me to “understand” climate that are just glosses or intuitive explanations that aren’t really even quantitative or very meaningful in my opinion. One such example is the idea that we can compute increments for example but not absolutes. In this case there is some truth to it that can be quantified in very narrow ranges of applications, but some strong limits too. The problem here is that these glosses can be dangerous because their advocates often are pretty assertive and over confident.

    I don’t think there is even any reason to believe there can’t be many attractors, in fact I recall from graduate school that there can be an arbitrarily large number of fixed points for example. That’s pretty well established in testing for example for real flows. What is perhaps more surprising is that these attractors can be somewhat close in statistics. I was surprised a little by that myself but the result I think is right.

    So that’s I think the disagreement as clearly as I can state it. The existence of these multiple solutions was very surprising to RANS experts and practitioners because their whole cultural belief system, career goals, and plans were based on this belief in unique and grid converged solutions and the idea that RANS would get the right answer if we just ran the code right or used he right methods. They looked very hard for alternative explanations for years before running out of excuses. That’s not a criticism, even though some of us were not surprised. 🙂

  108. David Young says:

    MT, Thanks for the comments. I’m busy right now, but will read them in detail and respond soon. In fact, I have some questions about things like modeling clear air turbulence in weather models that I really want to learn about. Perhaps you can help me.

  109. Willard says:

    > The use of the D word is a device to provoke thought I picked up from some very smart mathematicians.

    You might also like to pick up the L letter, a powerful device by smart physicists to provoke thought:

    CONDITIONS

    You will make sure:

    – that my clothes and laundry are kept in good order;
    – that I will receive my three meals regularly in my room;
    – that my bedroom and study are kept neat, and especially that my desk is left for my use only.

    You will renounce all personal relations with me insofar as they are not completely necessary for social reasons. Specifically, You will forego:

    – my sitting at home with you;
    – my going out or travelling with you.

    You will obey the following points in your relations with me:

    – you will not expect any intimacy from me, nor will you reproach me in any way;
    – you will stop talking to me if I request it;
    – you will leave my bedroom or study immediately without protest if I request it.

    You will undertake not to belittle me in front of our children, either through words or behavior.

    http://www.listsofnote.com/2012/04/einsteins-demands.html

  110. MT I don’t think our views on the uses of model are that different, by “a handle on” I didn’t mean “a good idea of”, and by regional I meant the next step down from global (continental scale?).

    “The trouble with defining climate as the statistical properties of weather is that it presumes that there is such a distribution – for practical purposes that means a stationary process. So how do you define climate change? ”

    I’m not sure I agree there. The actual climate is the distribution of possible weather behaviour for the conditions. We may only be able to work with a stationary approximation, but that doesn’t mean a true distribution does not exist. This however seems a rather fine point, rather than being “fairly facile”.

    “Remember you only get a single realization from a changing distribution.”

    That limits what you can infer about the underlying distribution, but doesn’t imply that it doesn’t exist or is stationary.

  111. DY wrote: “The use of the D word is a device to provoke thought I picked up from some very smart mathematicians.”

    I very much doubt that smart mathematicians would use such a word to describe an idea that can be clearly explained without resorting to authority and which nobody minds you questioning (provided you engage with the responses).

    “I can try to stop using it here.”

    I would suggest you stop using it everywhere, it gives the impression of arrogance and of not taking the opposing argument seriously, which I assume you did not intend. This is especially true for electronic forms of communication, where tone is often difficult to convey reliably.

    Now given that you have not answered my question, I think I will leave the discussion to MT as he is (a) much more of an expert on the models than I am and (b) appears to have less an distance in terms of terminology. I look forward to reading the discussion (and finding a paper or two from Williams to read). Thank you for the discussion.

  112. Howard says:

    Is it even feasible to improve GCMs to the point where they will better inform mitigation strategies in the time-frame required to do any good? From what I read, the answer is no. Therefore, this discussion is strictly academic.

    I do agree with MT about dropping geoengineering (except CCS) from the menu. We are already doing multiple geoengineering projects via industrial energy, transport, manufacturing, agriculture, urbanization, etc. Waste minimization makes the most sense.

  113. David Young says:

    MT, When I started at RC years ago, I was really much more a numerical algorithm specialist and not very knowledgable about things like turbulence and fluid dynamics. I’ve been trying to learn though.

    You mention this problem of defining the “final state” for a system that is not stationary. In CFD, we usually assume a stationary flow to justify Reynolds averaging and turbulence modeling. It may be wrong in CFD too, I haven’t looked into it in detail. But using that assumption people had assumed for a long time that we could define the “final state” clearly, namely, a grid converged, residual converged, numerical solution. Some of these new results cast that cultural belief into doubt I think. GCM’s I believe also use eddy viscosity models. 🙂

    There was an interesting paper about 5 years ago showing for a low Reynolds’ number airfoil problem something similar to Paul Williams result for the Lorentz system, viz., that using standard numerical methods could give very deceptive results with the behavior strongly dependent on the details of the numerical methods. I can find it again if you are interested. Here I am talking about the long time behavior and average forces for example.

    BTW, The point about better numerical methods is one we have been trying to infuse into CFD for 35 years. The situation is similar to the climate modeling world perhaps. There are 100s of codes, some of the commercial ones are literally a “giant pile of FORTRAN” and very messy legacy code. There is a massive industry of engineers who “run” the codes and they often do good work. But there is overconfidence among those code runners and often a cultural belief that if the code is run right, the answer will be useful. The real truth is that selection bias is a very powerful thing and the literature is in my view very misleading about how accurate the methods and codes really are. I always had my suspicions about CFD based on lack of rigor and seemingly over confident leaders of the field, but in the last 15 years we have started to produce the papers to prove the case and some progress is evident.

  114. DY,

    There are a lot of things in science, especially at the level ATTP wants me to “understand” climate that are just glosses or intuitive explanations that aren’t really even quantitative or very meaningful in my opinion.

    Given that you haven’t even tried to show that you understand the basics, the idea that you can dismiss what I suggested as “just glosses” or “intuitive explanations” is galling. I’m asking you to illustrate that you understand the basic physics. These basics are very important. If – as seems clear – you do not understand them, and have no interest in understanding them, then you’re just spreading doubt.

    Rephrasing then, the ultimate statistics it seems to me can depend on initial conditions.

    Show this. People who work in this field do not think that this is true. The basic physics (which you seem to not understand and seem unwilling to even begin to understand) suggests this isn’t true. If – as seems the case – you simply do not understand the basics, then you’re doing exactly as I’ve suggested you’re doing, which is take some technical detail in a related field and think that you can simply apply this to a field in which you have not even a basic understanding. Only MT’s request that we show some patience is stopping me from saying what I really think of this.

  115. Joshua says:

    ==> “Calling it dogma is rude as it is implying that this is not a reasoned understanding of the science, but merely an unquestioned belief passed down by some authority, which is deeply anti-scientific.”

    Repeated for emphasis.

    Consider David’s description of alternative views of the science as dogma in the following context:

    ““David Young | June 8, 2014 at 5:01 pm |

    What I found is that its impossible to really discuss any possible problems with climate science or even Mann [at ATTP]”

    So he writes comments at a blog where he believes it isn’t possible to discuss any possible problems with climate science, and describes alternate views (to his own), as dogma.

    Looks to me like a self-reinforcing logic.

  116. BBD says:

    ATTP says (of DY; emphasis mine):

    If – as seems clear – you do not understand them, and have no interest in understanding them, then you’re just spreading doubt.

    Bingo!

    And *that’s* what hacks me off.

    The great big fat subtext here is that DY is peddling an unsubstantiated belief that sensitivity has been substantially overestimated. That’s the real game, but it is played by proxy. By the endless riding around of hobby-horses in tight little circles.

  117. Willard says:

    > Is it even feasible to improve GCMs to the point where they will better inform mitigation strategies in the time-frame required to do any good? From what I read, the answer is no. Therefore, this discussion is strictly academic.

    You’re right to recall that in the end the bottleneck lies elsewhere, Howard. However, it’s worse than that. Wondering if the attractors we find are not the artifact of the models seems to rely on a view from which we could arbitrate just about anything. This is what appealing to “ultimate statistics” does to me, at the very least. Ironically, this view from nowhere oftentimes ensnares itself into what is called dogmatism:

    Supposing that the dogmatist assents to something, say p, on the basis of a reason, say q, and gives r as his reason for q, etc., how should the Pyrrhonian react in order to avoid the snares of dogmatism?

    http://plato.stanford.edu/entries/skepticism/

    “What if I told you there were many attractors?” or “What if I told you I had the ultimate statistics which would cast doubt upon our current implementations of GCMs?” are not that interesting, from an academic point of view.

  118. dpy6629 says:

    MT, Your point about GCMs being a tool to advance scientific understanding is a good one. On that score, the post about the Trenberth paper is a good one. This task is daunting enough with turbulence. For GCMs and climate its kind of beyond a grand challenge. I’m not on a good device to say more — I hate Mobil phone keyboards. I am very interesting in continuing however. I am going to ignore the peanut gallery if you are ok with that.

  119. David,

    I am going to ignore the peanut gallery if you are ok with that.

    With apologies to MT, I’m not okay with that.

  120. dhogaza says:

    DY has yet to provide any reason to believe that the small-scale CFD issues have any relevance to the climate-scale simulations done by GCMs. Above, he introduced another red-herring regarding timescale. “But what about the emergence of Panama?”, he asks, “wouldn’t that have a large effect?”. Well … from a policy point of view our interest in GCMs is the modeling of the climate system over the next several decades. Changes on geological timescales – such as the effects of plate tectonics – can be ignored on decadal timescales.

    DY has yet to provide a single reason why the issues he raises are relevant to the purpose for which GCMs are built.

    Even at the small scale of modeling airfoils and airplanes, he overstates his case:

    DY: “But there is overconfidence among those code runners and often a cultural belief that if the code is run right, the answer will be useful.”

    The Boeings and Airbuses of the world have found them to be extremely useful, and I doubt anything DY has to say will convince them that their findings are simply a result of “cultural belief” with no empirical evidence based on experience to back up those findings.

    And describing the engineers who are users of such models as simply being “code runners” is just another example of DY’s unwarranted and arrogant assumption of superiority over professionals in a field that is not his own.

  121. BBD says:

    Classic DY. Point out that he is denying palaeoclimate behaviour that contradicts his shtick and that what this is really about is peddling unfeasibly low sensitivity and he tries to pretend that you are the peanut gallery.

    Not very smart, DY.

  122. Joshua says:

    David –

    ==> “I am going to ignore the peanut gallery if you are ok with that.”

    You have said that you consider it impossible to discuss the problems with computer science here. So then why areyou here?

    It certainly does seem logical to conclude that you’re only here to advance an agenda, or (rather successfully) annoy people or get a rise out of them, or something like that.

  123. Kevin O'Neill says:

    DY writes: “I don’t trust my or anyone’s intuition about climate’s sensitivity to initial conditions.”

    Intuition? 800ky of paleo data.

    On a daily basis that’s 292 million initializations of the climate. At what point can we rule out any significant effect on the real world?

    David, when asked about paleo data, what it tells us, you’re simply crickets. You claim the paleo data is controversial, but when asked what is controversial about it, crickets again.

    Throw away all the GCMs ever made and and all their results and we’d still come to the same basic conclusions. They inform us about the details – they don’t dictate our basic understanding.

  124. mt says:

    turbulence cannot be modeled in a GCM, where the grid cells are order 10 km square by some tens of meters thick, It must be parameterized.

    I learned a little about that parameterization some decades ago but cannot call much of it to mind. I am sure a little diligent searching will turn up the relevant information. Usually source code comments link to the relevant research. I don’t think it’s first order important to this discussion.

    If DY wants to continue this conversation on my blog without the baggage of my knowing any of his past history here or elsewhere I am amenable. But Willard tells me all the cool kids are hanging around ATTP now.

  125. dhogaza says:

    mt:

    “turbulence cannot be modeled in a GCM, where the grid cells are order 10 km square by some tens of meters thick, It must be parameterized.”

    A lot larger than that, I believe. But, yes, that’s why a few of us have been hammering on the vastly sub-grid nature of DY’s complaints. We know this. I don’t know why DY doesn’t.

    Yes, people have been telling DY this for a year? years? and every time he pops into a new thread he starts at ground zero – climate modelers aren’t aware of the problems with numerical methods used in CFD and therefore … well.

    Perhaps this history will help you understand the frustration and the belief that DY is not posting in good faith that you see here.

  126. dhogaza says:

    mt:

    “If DY wants to continue this conversation on my blog without the baggage of my knowing any of his past history here or elsewhere I am amenable.”

    DY tends not to converse, but rather to assert.

  127. dhogaza says:

    DY:

    This is relevant how? You still refuse to answer relevant questions.

  128. DY,
    No idea what your point is. “Someone else does it, so it’s okay if I do it?”.

    MT,

    If DY wants to continue this conversation on my blog without the baggage of my knowing any of his past history here or elsewhere I am amenable.

    This might be better, but if DY can actually engage without sounding like an arrogant git and without dismissing perfectly reasonable comments, then here is fine. I’m just not convinced that he can do this, and I’m not convinced that I have the patience to let him carry on as he is.

  129. Andrew Dodds says:

    I see that as part of the peanut gallery I must be far too plebeian for you to deign to respond to, but nevertheless I would note that the case of Paxil appears to be that of a large, commercial company twisting evidence to make it’s product look safe and effective.

    In the climate ‘debate’, there are some large, commercial companies – some of the largest in the world, which depend for their continued existence on minimal action being taken on global warming. But I’m sure that they would not be trying to twist evidence to make global warming look uncertain or trivial. Perish the thought.

  130. David Young says:

    I tried to make a long comment yesterday and it seems to have been deleted. i’m not sure why. Perhaps breaking it up will help since there is some important content.

  131. David Young says:

    Just to provide context to some of the statements being made here about CFD that are not well informed, I quote from LeDoux et al, AIAA Journal, July 2015. There is also a follow on paper providing even further cautions about CFD increments even for attached flows. I apologize for the LaTex isms. I haven’t mastered the internet font thing yet. This is a consensus of a large number of authors and not just my view. Some of the authors are indeed “code runners.”

    “Three model problems associated with aerodynamic drag minimizations are studied. These test cases have been proposed by the aerodynamic design optimization discussion group, and include an inviscid NACA0012 non-lifting airfoil, a viscous RAE2822 lifting airfoil, and a viscous lifting wing based on the NASA Common Research Model. Various optimization methods are utilized, including MDOPT, TRANAIR, SYN83, and SYN107. The resulting designed and associated baseline geometries are cross-analyzed by several CFD codes, including OVERFLOW, TRANAIR, GGNS and FLO82. Pathological issues are unveiled in both of the {\em simple} airfoil model problems. Designed geometries for the inviscid symmetric test case exhibit strong tendencies to permit non-symmetric flow solutions. Designed airfoils for the viscous lifting
    case also support non-unique solutions and hysteresis loops at or near the design point in RANS and IBL simulations. These results provide further evidence that single-point aerodynamic optimization is often ill-posed. In extreme cases, it can yield designs with very undesirable aerodynamic characteristics, at least as analyzed by RANS and IBL methods, occurring at off- and even on-design conditions. We use these examples to document further the multiple and
    “pseudo-” solution phenomena for steady state RANS. This provides evidence that even in practical engineering settings, numerical methods to assess stability and uniqueness of steady state solutions and/or predict the bifurcations of these solutions have value. The single point wing design problem likewise is ill-posed in the spanwise direction. Multi-point design with potentially a large number of points or the inclusion of inequality constraints can regularize the problem.

    Over the past four decades computational fluid dynamics (CFD) has matured to the point that accurate incremental aerodynamic performance analyses are now possible for complete aircraft configurations, provided that the flow in the viscous shear layers remains predominantly attached
    to the geometry surfaces. Fortunately, this is usually the case for well-designed aircraft at their intended cruise flight conditions. However, fundamental problems remain for predicting separated flows, corner flows, transition, and even for modeling nominally attached flows as documented in this paper. These difficulties are simply reflections of the fact that the Navier-Stokes equations are an example of a nonlinear dynamical system, and as such are ill posed, either as an initial value problem (as discovered by Lorenz over 50 years ago \cite{Lorenz}), or as a boundary value problem. Thus, many standard numerical strategies commonly developed for and used in the elliptic boundary value problem domain and commonly used in CFD are problematic and can provide inconsistent results.”

    Airbus’ view is not very different because the facts and science don’t magically change on the other side of the Atlantic. The bottom line is that the literature in fact gives a skewed view of the accuracy, stability, and replicatibility of CFD. And of course, the CFD companies are in the business of selling their wares, enough said about that. This is not different than in other fields of science, where there is a growing consensus that there is a serious problem. This is not sowing doubt, it is simple fact and the scientific method at work.

  132. David Young says:

    MT, What is your view about how to determine when a climate simulation has reached some kind of point where it is meaningful for comparison with another simulation with different inputs, parameter settings, gridding, etc.? Maybe its not really rigorous, but what is your view? It seems to me that this is the first question to answer.

    As I said, in most CFD simulations we can define that rigorously, because they are steady state simulations, even though in practice this rigorous definition is usually very far from being satisfied. We are starting to see more and more colorful time accurate simulations using Detached Eddy Simulation. There are some challenges here, for example its not possible to really define grid convergence. And then there is the issue of when the statistics are meaningful. For steady state runs we have proof that the usual criteria for convergence can be very misleading. But we don’t have much of a clue for time accurate simulations. Any thoughts would be most welcome.

  133. mt says:

    1) You take observational data and extract means, variances, and seasonal patterns 2) You spin up a climate model and extract similar statistics 3) You set up a metric, formally or informally, for the distance between these statistics 4) You determine whether your model produces a model climate that isn’t too embarrassing to publish, i.e., works about as well as the state of the art 5) You publish papers describing your model’s baseline climate and the points of agreement and disagreement with observations. 6a) You start using it to do science and/or 6b) You do CMIP runs with it for IPCC.

    Does this process produce something useful? Yes.

    Is it prone to overtuning? Less than you’d think, but maybe.

    Does it produce code that is brittle and overcomplicated? That’s my complaint.

    Is it rigorous? Well I can imagine you claiming it isn’t. But I can’t imagine doing much better at present.

    Does it reach a steady state? That depends on whether you are speaking of an AGCM or a coupled atmosphere/ocean CGCM or ECM. The atmosphere has a short memory, which allows chaos to dominate the exact instatntaneous state, but that in fact makes a number of other issues addressable. That is, given stable boundary conditions, atmospheric models empirically produce a stable model climate.

    In a coupled system, it’s more complicated, and some compromises must be made. On the whole the ocean predictability limit is much longer than the usual time scales of interest (notwithstanding eddy fields near coastal currents) while the atmosphere’s is shorter. So in practice we just go ahead and do it and it works well enough to be informative.

    I am sure this doesn’t address the question in the terms you prefer. I would love to be more mathematically rigorous, but I think I understand what the models are doing well enough to justify their use.

    The issues just aren’t CFD-flavored at all.

    I think from your point of view I’d claim that the models are highly dissipative and highly forced, so the convergence issues you seem so interested in aren’t very relevant. Energy is being pumped into and out of the system very rapidly compared to the fluid time constants.

    At the same time, many of the parameterized phenomena are poorly specified. We use these models to try to figure out what the system is. We do not actually have a handle on it in an engineering sense. We don’t have a “right answer”, and the faster we force the real climate the less effective observations can be in specifying our target.

    Further, (caveat – I may be wrong here as there may have been recent progress) some parameterizations are scale-specific; the usual CFD ideas of convergence in the high-resolution limit don’t even apply. If your model resolves convective clouds, you need to change your convective cloud parameterization for instance; the submodel you come up with to represent cumulus mass and energy exchanges at 100 km may be invalid at 1 km.

  134. David Young says:

    MT, Thanks for this detail. I will review it very carefully.

    A couple of points jump out at me. I think but am not sure that aerodynamic flows have more energy density pumped in than the atmosphere. That’s just based on average velocities. However, we do find that minimizing dissipation does really improve accuracy often dramatically. This was critical for the code where we first found multiple solutions, subsequently verified in many codes. That’s my reasoning for why dissipation is worth working on in GCM’s and Paul Williams work seems to verify that. But I could be wrong.

    And that is where understanding perhaps the long time behavior of the attractor(s) and bifurcations might be important. In CFD right now, we just run the code in time accurate mode and have little guidance. In all honesty, the results are often pretty bad, but then the data is pretty noisy too. But we want small increments, so this level of noise really makes things not very useful except for very qualititative work. But that doesn’t stop the salesmen from showing results that match data perfectly. 🙂

    Yes, the sub grid models seem very challenging to me, especially after reading Lacis’ descriptions. The fact that he has devoted an entire career to just one is telling. How to validate those seems to me a very daunting task and I can’t even think of how you would do it given computer limitations. Turbulence gives us fits already.

    Lots of things to understand better.

  135. dhogaza says:

    DY:

    ““Three model problems associated with aerodynamic drag minimizations are studied….”

    Airplanes again. The planet is not an airplane, and can not, is not, and never will be modeled on the grid scale of an airplane.

    This is such a simple point, why do you continue to bring it up and ignore those who make the point that it’s a different problem space on a different scale?

  136. dhogaza says:

    MT:

    “Further, (caveat – I may be wrong here as there may have been recent progress) some parameterizations are scale-specific; the usual CFD ideas of convergence in the high-resolution limit don’t even apply. If your model resolves convective clouds, you need to change your convective cloud parameterization for instance; the submodel you come up with to represent cumulus mass and energy exchanges at 100 km may be invalid at 1 km.”

    Amen. Scaling. DY either doesn’t understand scale or does and pretends it doesn’t matter.

  137. dhogaza says:

    DY:

    “Lots of things to understand better.”

    Let it be written that DY admits he is not God. Let us remember. Let us not place bets on whether or not he admits he’s not God elsewhere, where he is intent on “proving” that GCMs are worthless.

  138. This may be a key point

    Energy is being pumped into and out of the system very rapidly compared to the fluid time constants.

    Let’s see if I can elaborate a bit on this (MT can correct me if he thinks I’m going wrong somewhere). Let’s imagine that we think it’s possible that non-linearities could lead to a dramatic change in how much we think we might warm in the coming decades. Let’s imagine that some small-scale non-linear effect can lead to dramatic cooling in the Arctic, increases in Arctic sea ice and ice sheets, etc. The problem is that the incoming solar insolation is well-defined and predictable. How do you sustain a region being much cooler than we’d expect from basic energy balance for long enough to lead to significant changes in ice coverage. The heat capacity of the surface/atmosphere is not very large. It should warm up relatively quickly. Similarly if you want to sustain a warmer than expected region.

    Now, you can have internally-driven warming/cooling that can persist for a reasonable length of time (years). This can be associated with ENSO events, but also with internally-driven radiative perturbations (clouds/water vapour). The problem here, though, is that the feedback response is – on average – smaller than the Planck response. Although you might violate this in some region for some time, you would not expect large internally-driven perturbations to persist for a long time (decades) because the system will have a tendency to return to its state of quasi-equilibrium, defined by the incoming solar insolation, the albedo, and the typical composition of the atmosphere.

  139. BBD says:

    Although you might violate this in some region for some time, you would not expect large internally-driven perturbations to persist for a long time (decades) because the system will have a tendency to return to its state of quasi-equilibrium, defined by the incoming solar insolation, the albedo, and the typical composition of the atmosphere.

    That is my understanding too. Climate isn’t self-propelling. It is reactive on decadal and longer timescales.

  140. I should probably have added that the influence of the atmosphere (greenhouse effect) is largely determined by the composition/concentration of the persistent greenhouse gases (CO2 mainly) not by the preciptable greenhouse gases.

  141. mt says:

    Regarding the chaos / predictability question, some very nice comments especially from ATTP. I will add my 2 cents as well. In the end I will diverge from what seems to be the consensus here.

    Here’s the statistician’s view as I understand it: weather is the state, climate is the distribution of which the weather state is a sample.

    Weather is highly chaotic. It is possible for weather to be chaotic even in a fixed distribution – indeed most simple examples of chaos (Lorenz system e.g.) are part of a fixed distribution. Far from “climate” bifurcating in such systems, it doesn’t change at all. Indeed that is the normal, simplest case.

    Worse than chaotic is inergodic -systems whose detailed history (weather) determines the final distribution (climate)

    Climate models’ weather is empirically ergodic; i.e., they do not have that problem. There is no sign that different distributions of weather emerge from minimally perturbed sets of runs. I imagine that this observation could be proved for somewhat simplified models.

    Weather, the instantaneous state, diverges quickly in perturbation experiments, per Lorenz, but the resulting **distribution** of states is the same.

    Is the real world ergodic? Well, since we have only one realization, it may not be a well-formed question. I tend to think that chaos, predictability, ergodicity, are properties of models, not of real systems.

    But when we talk about the weather/climate distinction in this way we are implicitly saying their is a meaningful distribution of weather. Given the physics of the underlying system, it’s hard to imagine an important bifurcation in the distribution as ATTP points out. That atmospheric models are ergodic under quasi-steady forcing is unsurprising and reassuring.

    But I think that once the deep ocean notices the huge anthro forcing, all of these arguments get pretty fuzzy and possibly moot. We don’t care if the abyss is ergodic. The distribution may be stationary, but we will perceive the deep ocean “weather” as climate in a practical sense even if it is only weather in the “climate is the distribution and weather is the state” sense.

    It’s little consolation if the distribution of the unforced system is stable, if the forcing is rapid enough and the system is slow enough. As a weakly stratified system, the ocean already has shown its capacity to operate in surprising ways. The mathematical ergodicity may not protect us from some unexpected new dynamics involving interactions between the surface and the abyss,

    Indeed, to the extent that the surface is not warming as fast as we expected because of heat export to middle depths, the ocean is surprising us already.

  142. mt says:

    Regarding DY, to me it’s peculiar. I’ve only been aware of DY for a short time. But I do not see him insisting on the climate chaos-in-the-formal-sense point everyone is rebutting. I in fact don’t see him grinding any particular axe, but honestly attempting to fit climate modeling into a fairly rich conceptual framework that he has which is certainly not entirely unrelated.

    Does he have something of value to add?

    Well, I hung around the CFD folks at U of Chicago occasionally to see if I could learn something. I did learn that our problem is very different from the usual engineering ones that engage the CFD community. But I also learned that they have domain specific computer languages related to their interests which are extremely powerful abstractions, while we are still mucking around (the analogy of shoveling filth manually is pretty visceral for me in this regard) in human-written Fortran, and I was intensely jealous of their capacities.

    Is DY grinding some sort of naysayer axe? Maybe but so far I haven’t seen it.

    Can we do better in climate? I believe I managed my career well enough to have some idea how to do that, but I didn’t manage it well enough that anyone will fund me to do it. I certainly don’t have the energy or will power or chops to do it alone on a volunteer basis. But it bugs me. I’m trying to convince somebody else to start a climate model from a clean slate, preferably with Python and PETSc for the heavy lifting.

  143. Without wanting to pile-on to DY too much, my issue is not that he doesn’t have something to add; I’m sure he does. It’s more than he seems to be overly willing to be simply dismissive of other aspects that are relevant and worth understanding. That leaves me with the view that even though he has something to add (and given that he publishes relevant papers, presumably is adding something) the reluctance to recognise the broader picture and the context in which climate models are used makes me less convinced that it will end up being something worthwhile. I also, FWIW, have problems with people who appeal to their own authority in the way that DY does. Call it a failing, if you will, but I have it nonetheless.

  144. dikranmarsupial says:

    “Here’s the statistician’s view as I understand it: weather is the state, climate is the distribution of which the weather state is a sample.”

    I’d say climate is the statistical properties of the weather, which could include its dynamic behavior. Viewing climate as the distribution of weather states may be useful for some applications, but it doesn’t need to stop there if more is required and a more complex model is justified. For example a Markov model can represent the statistical distribution of some particular set of strings of symbols. The model comprises of both the emission probabilities for each symbol for each state of the model as well as the probabilities that govern the transitions from one state to another. Change the transition probabilities and we change the strings that are well explained by the model. I’m quite interested in weather typing, so the Markov model idea is not completely unapposite.

    It also seems to me that the true definition of climate includes the distribution of weather states that we could have observed but didn’t (i.e. climate is the true generative model of the weather, which as I understand it is basically what a GCM tries to do via Monte Carlo simulation).

  145. mt says:

    Marsup, no argument – by “distribution” I meant any generalized properties of weather trajectories. Maybe it was the wrong word. At least I should have defined it.

    The Markov suggestion applies to climate models. A digital climate model really literally is a (gigantic, messy) Markov system. The time-continuous system we are modeling probably isn’t, though, really.

  146. dikranmarsupial says:

    “But I do not see him insisting on the climate chaos-in-the-formal-sense point everyone is rebutting”

    I don’t think that is what everyone is rebutting; the key point for me is whether there is evidence to suggest that the problems he raises apply to the scales on which GCMs operate, rather than being dealt with via parameterisations or via downscaling etc (however I may not have communicated that clearly). The answer appears to be “no, so far”, but David would have done more good for his case by actually saying so.

    The idea that we can’t predict weather more than a few days in advance so how can we trust climate projections 100 years into the future is though a very common climate skeptic myth, and did suggest a fundamental issue in David’s understanding of climate modelling (hence discussion of double pendulums to pick something with less terminology/baggage, again it is a pity that the analogy was summarily dismissed).

  147. dikranmarsupial says:

    MT “Maybe it was the wrong word.” always the problem with inter-disciplinary discussions! I think we are actually in good agreement about most things (especially importance of oceans).

  148. mt says:

    Communicating across disciplinary boundaries is hard.

    You, ATTP, are a master – I don’t know of anyone not formally trained in climate who has grasped climate better. But you shouldn’t expect comparably complete and rapid success from everyone who takes an honest interest.

    My working hypothesis remains that DY is honestly interested, and should be treated with respect.

    It’s clear he doesn’t jump the disciplinary gap very well. Much of what he says is hard for us (well, me anyway) to understand, and clearly he misses some of what we are saying. But I don’t see that as a reason to stop trying. I have had plenty of fruitless discussions in a quarter century on the internet. This one doesn’t have that flavor for me, at least as yet.

  149. My working hypothesis remains that DY is honestly interested, and should be treated with respect.

    I will endeavour to do better 🙂

  150. mt says:

    One though I hold is that atmosphere components of CGCMs are good enough and ocean components are not. If there are ways to control mixing due to model artifacts it would constitute a real advance.

  151. dikranmarsupial says:

    I should point out though that DY has not treated us with respect; calling an idea “dogma” (even to provoke thought) and summarily dismissing analogies as having already been seen is not in any way respectful.

  152. David Young says:

    I do agree that this is one very strong frustration of these discussion here and elsewhere is that people seem to assume that one is arguing that the “chaos means nothing is right” idea (which to those who hold it is simply a gloss on a far more complex reality). Why does that happen? Perhaps because the idea is easy to rebut and people have their favorite examples to prove it wrong. However, there are real complex questions here about long time behavior of nonlinear systems, MT understands this. One reason for assuming that the “opponent” is saying “chaos means nothing is right” is I think very superficial knowledge. That was my position 15 years ago and it takes a big effort to move beyond that level of knowledge.

    I would argue that the extent to which climate is just a mid latitude Rossby wave machine that dissipates energy to turbulence, its not very interesting. I think this characterization is unlikely because vortex dynamics is not generally that predictable, but I don’t know for sure. The interesting parts are surely the nonlinear parts such as ice sheets, ice ages, ocean dynamics, etc. An ice age is an example of NO change in total forcing but a small change in the distribution of forcing causing a large change. That would be an illustration of an ill-posed problem, no? I also doubt that the maximum extent of the ice sheets is predictable at all. Certainly its a very daunting modeling challenge.

    MT, I searched for proofs on the ergodicy of Navier-Stokes a few years ago and came up empty. This is an example of some theory that would be really helpful. Maybe its too hard to solve, but it would really help. There is some mathematics on the dimension of the attractor. It is bounded by the Reynolds’ number which for practical flows is huge. I don’t think this helps much except to say that we can’t rule out very complex attractors with very long time scales.

    One reason for skepticism however is that in CFD, there really was a cultural belief that steady state RANS was well posed for a very long time. We have now proven that to be wrong. But as I say, there were lots of glosses out there that “explained” it was well posed. And there is still a very strong cultural preference for this.

    You touched on something else that is a big factor and that is selection bias. You said that if your climate result “is not too embarrassing’ you publish it. In CFD, that’s why the literature is unreliable. Everyone wants to make their code look good, so you virtually never see sensitivity of the results to parameters studies and you never see bad results, unless of course, they have found and fixed the problem they had earlier. A result that disagrees with the data is simply chalked up to “bad grid ding, unconverged results, turbulence modeling errors, etc.” and never published. There is a whole host of excuses and they are I think sincerely believed. It’s not a consciously dishonest attitude.

    In the climate debate, there is a strong tendency to deny that such biases affect the scientific literature. That’s another frustration here. Such denials are I think becoming a very strongiy minority position and the evidence of a problem is really very convincing.

    BTW, I had another comment deleted last night. Any idea why?

  153. DY,

    I do agree that this is one very strong frustration of these discussion here and elsewhere is that people seem to assume that one is arguing that the “chaos means nothing is right” idea (which to those who hold it is simply a gloss on a far more complex reality). Why does that happen?

    It’s because it is what it seems that you’re primarily trying to say. If it isn’t what you’re saying, then maybe you need to try harder to recognise the relevance of these non-linearities.

    BTW, I had another comment deleted last night. Any idea why?

    Because I just decided to delete it. Okay?

  154. JCH says:

    Yes, the sub grid models seem very challenging to me, especially after reading Lacis’ descriptions. The fact that he has devoted an entire career to just one is telling. …

    How is this not a horrendous swipe at Lacis? How is it telling? What is he adding here?

  155. BBD says:

    An ice age is an example of NO change in total forcing but a small change in the distribution of forcing causing a large change.

    A very large change in seasonal and spatial forcing, if you are referring to insolation change at eg. 65N latitude at peak orbital (obliquity) forcing.

    In the climate debate, there is a strong tendency to deny that such biases affect the scientific literature.

    NB, *mt*, NB…

  156. Willard says:

    This sounds polite enough:

    However, there are real complex questions here about long time behavior of nonlinear systems, MT understands this.

    It does not sound that respectful, however.

  157. dikranmarsupial says:

    “An ice age is an example of NO change in total forcing but a small change in the distribution of forcing causing a large change. That would be an illustration of an ill-posed problem, no?”

    I’d say it was an example of a “tipping point”. I think most people would agree that tipping points in that sense exist (e.g. methane calthrates releases), but that doesn’t mean that anthropogenic climate change is necessarily going to provoke one. Note the small change in orbital forcing is amplified by carbon cycle feedback, which seems only tenuously related to the soft of CFD issues being discussed, AFAICS.

    However, this is getting away from the original point of why the planet comes back to such similar conditions during intergacials if the climate system is inherently capricious.

  158. In a sense the irony in what DY is saying is that if any of the rest of us made the arguments he’s making we be accused of alarmism. It is possible that something extremely surprising could happen and that we could pass some kind of tipping point. However, with some exceptions maybe, these are mostly regarded as unlikely in the coming century, or we have low confidence in whether or not such a tipping point could actually occur given what we expect to happen in the coming century. Most evidence suggests that the response to changes in forcing of the magnitude that we might expect in the coming decades is likely to be approximately linear. It might not be, I guess, but – if not – the outcome could be very unexpected and potentially not very good.

  159. JCH says:

    Well, is not Hansen discussing passing a tipping point, possibly in this century? AMOC goes dizzy. Europe chills. Freshwater pulses. Snow starts falling on the Southern Ocean instead of Antarctica.

  160. dikranmarsupial says:

    David Young wrote: “One reason for assuming that the “opponent” is saying “chaos means nothing is right” is I think very superficial knowledge.”

    Again, not very respectful.

    David Young wrote “The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”

    You see this argument from climate skeptics quite frequently, and “chaos means nothing is right [in GCM projections]” is generally exactly what they mean.

  161. Willard says:

    BTW, for those who are interested in visuals:

    http://demonstrations.wolfram.com/FiveModeTruncationOfTheNavierStokesEquations/

    The caption reads:

    From a version of the three-dimensional Navier–Stokes equations for an incompressible fluid with periodic boundary conditions, a particular five-mode truncation was derived in [1]. The resulting set of nonlinear ordinary differential equations allows only a finite number of Fourier modes and behaves as a system with five degrees of freedom, thereby resembling the behavior of the Lorenz attractor.

    Also note that the Navier–Stokes existence and smoothness problem is a tough one:

    https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_existence_and_smoothness

    When DY uses Navier-Stokes as an example, please beware that he’s using one of the Millenium problems.

  162. dikranmarsupial says:

    If the work of Williams to which David refers is that described in this paper

    http://www.met.reading.ac.uk/~williams/publications/2010MWR3530.pdf

    I don’t think it supports David’s apparent argument very strongly. I’ve only skimmed the paper and it really isn’t my area (so caveat lector), but it appears that the method “only” improves the forecast window for weather prediction by about a day. This seems like a big win for weather forecasting, but I don’t see why that implies a fundamental issue for climate modelling. After all, if you would ask:

    “The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”

    how would the answer differ if the question was

    “… just why would a weather model that becomes inaccurate after a few days [and one day more] get the 1000 year averages right.”

    It more or less confirms that it is initial conditions that limit the forecast horizon, rather than CFD issues.

    The other point is that the paper suggests that the monthly climatologies are not significantly changed by the improved method, which kind of suggests that climate (in a statistical sense) is not overly sensitive to the CFD issues.

    However, that is just my non-expert reading of it, I’d be happy(ish) to find out I am wrong!

  163. dpy6629 says:

    Another part that seems really interesting is tropical convection, another very difficult problem. Held has a post showing it is computationally very sensitive. It is important to climate through the h2o feedback and the current macro theory for the tropics doesn’t seem to agree with the data very well. Please let’s not rehash the hot spot wars though. Can’t say more now due to darn mobile keyboard

  164. dpy6629 says:

    Dik ran, the parts out climate is not in the paper probably. It’s i n the Newton isn’t. Video. I have seenthing in the CDd lit on this that I will post this pm. Damn mobile keyboard 😀😡

  165. dikranmarsupial says:

    David, do you agree that the monthly climatologies not being significantly affected suggests that the CFD issue does not greatly affect climate prediction? If not, please explain why.

    Please can you explain how does being able to extend the weather prediction horizon by one day alter the answer to your question “… just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”? As you say that you are not arguing that “chaos means nothing is right”, I am interested to understand what you did mean by this question.

  166. mt says:

    1) I have no idea as yet where DY is on the alarmism / complacency spectrum insofar as policy is concerned : as far as I have seen he is limiting himself to pure science, 2) I can imagine being upset at “dogma” but I wasn’t especially. To suggest that climatology is somehow immune to groupthink is implausible to me. Nobody objects when Hansen makes a similar point. 3) We are endeavoring to have technical discussions about models as grownups with some skill who think modeling is worth pursuing. This is orthogonal to climateball, and I respect anyone taking a position of rigorous policy neutrality, as Willard did for some years before finally joining our camp. I think we need some people in that neutral role.

    Has David expressed any opinion whatsoever on climate policy?

  167. mt says:

    Willard, climate flows are very closely related to Navier-Stokes; were it not for convective processes both atmosphere and ocean would be NS problems.

  168. mt says:

    “An ice age is an example of NO change in total forcing but a small change in the distribution of forcing causing a large change. That would be an illustration of an ill-posed problem, no?”

    actually Milankovic cycles imply quite a large change in distribution of forcing if that matters.

  169. Willard says:

    When did I join any camp, Dr. Doom?

    Here’s an interesting concern that bears on climate policy:

    The first prerequisite to being effective in the public sphere is to be fully honest and to correct errors and problems. That is the normal course of science and it has harmed the implementation of mitigation policy. It is all compounded by the appeals to deep dark conspiracy theories about the fossil fuel industry and their supposed paid mouthpieces.

    https://andthentheresphysics.wordpress.com/2013/10/26/engaging-with-skeptics/#comment-6344

    Notice the signature.

  170. Willard says:

    Wait. There’s more:

    I have tried to steer clear here from touching hot button issues and getting into the apologetics of climate science with your clientele, some of whom I know are on the same wavelength as St. Thomas.

    https://andthentheresphysics.wordpress.com/2013/11/03/judith-currys-stadium-wave/#comment-7045

  171. Willard says:

    Sometimes, the dogwhistling is less subtle:

    I agree that the planet is going to continue to warm. The crucial question is how much. For that the IPCC seems to rely on GCM’s while admitting in AR5 that in fact many of them are too sensitive to greenhouse forcing. That’s a crucial question that transcends the political squabbling and the apologetics. We need to get this right and that should be the top priority.

    https://andthentheresphysics.wordpress.com/2013/11/03/judith-currys-stadium-wave/#comment-7047

    No wonder DY signed “Lukewarmer” in one of his first comment here.

  172. Eli Rabett says:

    Given that just about every large subsonic wind tunnel has been closed to save money, Eli suspects that CFD for subsonic aircraft is a solved problem as far as the manufacturers are concerned

  173. dikranmarsupial says:

    MT ” I can imagine being upset at “dogma” but I wasn’t especially.”

    I am not upset by “dogma”, the problem is that it is condescending, which is usually an indication that someone isn’t taking the other side of the discussion seriously and is adopting a teacher-pupil attitude rather than trying to have “[a] technical discussion[s] about models as grownups with some skill who think modeling is worth pursuing”.

    “To suggest that climatology is somehow immune to groupthink is implausible to me. Nobody objects when Hansen makes a similar point.”

    I don’t think anyone is implying that climatology is immune to groupthink, and I think the discussion here shows that this group is open to the idea that DY seems to promulgate, but just want evidence that it affects the sort of scales on which climate models operate. Responding with this sort of behaviour is far from convincing.

    Basically if you want to be treated with respect (and I agree with MT’s call to treat DY with respect), it isn’t a one-way street and you need to be respectful yourself. This is not complicated, it is just the golden rule.

  174. Eli Rabett says:

    MT

    My working hypothesis remains that DY is honestly interested, and should be treated with respect.

    You’ve been wrong before

  175. Willard says:

    > [C]limate flows are very closely related to Navier-Stokes; were it not for convective processes both atmosphere and ocean would be NS problems.

    Perhaps I ought to have quoted this tidbit from thy Wiki page I cited above:

    The Navier–Stokes existence and smoothness problem concerns the mathematical properties of solutions to the Navier–Stokes equations, one of the pillars of fluid mechanics (such as with turbulence). These equations describe the motion of a fluid (that is, a liquid or a gas) in space. Solutions to the Navier–Stokes equations are used in many practical applications. However, theoretical understanding of the solutions to these equations is incomplete. In particular, solutions of the Navier–Stokes equations often include turbulence, which remains one of the greatest unsolved problems in physics, despite its immense importance in science and engineering.

    https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_existence_and_smoothness

    There are more than 30 occurences of “turbul” on this page.

    ***

    Sometimes, that there’s an open problem is just not such a big deal.

  176. dpy6629 says:

    Eli, you have said this about wind tunnels before and you are completely wrong about it. I posted here the consensus opinion of a large group of engineers. Read and enjoy.

  177. mt says:

    Thanks W, duly noted on your neutrality. I merely suspect you of being my friend and ally; it’s just a suspicion which I cannot prove.

    Also duly noted on David’s points.

    “I agree that the planet is going to continue to warm. The crucial question is how much.”

    On how finely we need to discern ECS, there are a couple of reasons that I could argue that it doesn’t matter. One is explained here: http://planet3.org/2013/03/08/why-equilibrium-sensitivity-is-not-policy-relevant/

    Another is that what we really care about is the sensitivity of impacts to forcing; while impacts were theoretical we used global mean temperature as a proxy. Now that they are starting to appear the ECS is more of a theoretical concern.

    There is no longer any room for informed doubt in any sustainable ethical system on the evidence that net emissions of CO2 must go to or below zero as quickly as is feasible.

    Most people are yet to be convinced of this. This is hardly surprising.

    I find it odd that climate experts do not find resistance from other scientific sectors to this astonishing policy conclusion entirely expected.

    Verheggen et al and other studies show that the closer a scientist is to the core of climatology the more likely they are to be convinced that we have a big policy problem, But this can be turned on its head. The further you are from climatology, the less likely. This stands to reason.

    The sorts of argument that carry the day from a political activist’s portfolio are irrelevant, and dealing with individual concerns from educated people is a daunting and hard-to-scale process. Responding to such engagement as if we were facing another Morano sock puppet is the opposite of helpful, though.

    Counterclaims that climatology is without flaws or blemishes are risible. It is one thing, given the fierce attacks on our good faith, to ignore over-the-top criticism. It is another to take offense at the slightest hint of criticism and respond with excesses of enthusiasm.

    I have been restrained in my public criticisms of the climate science community because of the ethical constraints of our circumstances and the overwhelming evidence of our peril. It’s a pity I have ethical standards – I’d be quite valuable to the bad guys if I really made a habit of airing my criticisms and gripes. But in a generic sense I am willing to say that being interested in climate does not in itself confer sainthood. The problems endemic to science as a whole are not absent in climate studies.

  178. mt says:

    Turbulence “closure” is a big deal at the practical nuts and bolts level of weather and climate modeling both in GCMs and on cloud-resolving models. If you don’t bother, you don’t have a working model. It’s not a big deal for us that it’s mathematically unsolved – even if it were solved we couldn’t afford to represent it perfectly.

    But it certainly matters how we patch something in to make up for the fact that we can’t resolve it.

  179. dpy6629 says:

    JCH, I respect Andy and have said so. Sub grid modeling issues are tremendously tough and may be unsolvable. Turbulence modeling is an example.

  180. Dikran,
    Your comment largely sums up my view.

    MT,

    But in a generic sense I am willing to say that being interested in climate does not in itself confer sainthood. The problems endemic to science as a whole are not absent in climate studies.

    Absolutely. I might frame it slightly different to how you’ve done so. I see no evidence to suggest that problems in climate science are any different to the kinds of problems you find in the other sciences. Scientists/academics are human. They’re not perfect. They have biases, goals, desires,…. However, there’s no evidence that such human frailties prevent us from gaining understanding of physics systems. Expecting scientists/academics to live up to some kind ideal is naive and unrealistic. That isn’t to say that that things couldn’t be better, that we couldn’t do things differently, or that we should excuse bad behaviour. It’s simply that there no evidence that I’ve seen that climate science is somehow special in terms of these kinds of issues.

  181. dpy6629 says:

    The bad news is that the turbulence model details can matter a lot even at steady state. This is not controversial but not often highlighted. We have a new paper in work taking a very small first step.

    Turbulence models are most all eddy viscosity models meaning they adds a viscosity to account for the unresolved turbulent eddies. This viscosity is computed using very nonlinear PDEs because the physics is very nonlinear.

    It would be a miracle if the model was not too dissipative and everyone knows they are. The hope is it doesn’t matter too much despite well know cases where it does matter a lot.

    There are perhaps hundreds of different such models.

  182. dpy6629,

    The bad news is that the turbulence model details can matter a lot even at steady state.

    In what sense do they matter a lot? Please answer this question, because simply pointing out that the turbulence model matters tells us virtually nothing. As MT has pointed out

    Turbulence “closure” is a big deal at the practical nuts and bolts level of weather and climate modeling both in GCMs and on cloud-resolving models. If you don’t bother, you don’t have a working model. It’s not a big deal for us that it’s mathematically unsolved – even if it were solved we couldn’t afford to represent it perfectly.

    Are you agreeing with this, or saying something more? If more, what are you actually saying and why is it relevant?

  183. > [I]t certainly matters how we patch something in to make up for the fact that we can’t resolve it.

    Of course it does, and this is why it’s so easy to exploit this in the lukewarm argument that remains implicit in DY’s concern. The overall trick is to say that there are problems in climate science, and to talk about them over and over. If you have an open problem that is one of the toughest problems in mathematics, so much the better. One can even call it an example and induce from that examplar case of the implementation of Navier-Stokes equations that there are many problems in climate science to solve.

    Climate science has many problems to solve. Climate science has many problems to solve. Climate science has many problems to solve. That’s all you need to say to make the audience to hear before we use its output for policy decisions, or something along these lines.

    ***

    More generally, such ClimateBall move is possible with any incomplete expression, like “it certainly matters.” To say that something matters simpliciter is asking for trouble. It certainly matters, but for what exactly? The first thing an hockey coach will hammer and hammer over again to his rookies is: “finish your check”!

    ***

    Less generally, there’s a problem with DY’s stance:

    [DY0] I agree that the planet is going to continue to warm. The crucial question is how much.

    [DY1] The more nonlinear a polynomial equation is the more likely it is to have multiple solutions.

    [DY2] [W]e would have never found this without modern numerical methods and a brand new code.

    [DY3] It is alarming though if you want to use this stuff where public safety is at stake.

    DY0 and DY3 are two faces of the same lukewarm coin. DY0 and DY1 create a Dutch book to justify one’s insatisfactions against modelling. DY0 and DY2 shovel policy after we RETHINK ALL THE THINGS. DY2 and DY3 shovel policy after we RECODE ALL THE THINGS. You sure get the drift. Now, what about the relationship between DY1 and DY2 or DY1 and DY3?

    Etc ad lib. And don’t start me with the presuppositions behind these claims, their correctness, or even their relevance in the grand scheme of things.

    ***

    While I can fathom that defensiveness and aggressiveness is uncool and doesn’t help, it’s also important to understand why the lukewarm playbook triggers such reaction.

  184. mt says:

    “It’s simply that there no evidence that I’ve seen that climate science is somehow special in terms of these kinds of issues.” Absolutely agree; to the contrary the motivations toward frank scientific fraud are much greater in other fields. But there is a circle-the-wagons thing happening given all the attacks, Almost any criticism from outside is treated as unfair.

    But consider my position – I think almost all macroeconomics applied to time scales longer than a couple of years is broken. And it is exactly that sub discipline that is driving climate policy, so this failure I perceive is very consequential. But on the other hand, I can’t pretend to know a lot about it, and some of my critiques will be off base and even wrong. Still the whole business seems to me based on faulty assumptions.

    If I am to question whether IAMs are a useful tool for climate economics I have to give permission to people to also question GCMs for climate physics. I think it’s important for a field to be able to handle challenges from outside.

    If you take the right attitude and your interlocutor does too, there is plenty of stuff to be learned, The occasional arch sarcasm is a price worth paying. If you have a thin skin you shouldn’t play climateball at all.

    Eli is right; my patience lasts longer than most and it gets abused on occasion. BUt I think it’s the right approach to public conversations about climate.

    I’ve been exposed, in my peculiar life, to a lot of cultures. I’ve learned that some things that some people find obvious others find dubious, still others find impossible, and yet others find inconceivable and unheard-of. I’ve learned that almost everybody is stubbornly wrong about something, with the possible exception of myself. The odds of course are that I too am stubbornly wrong about something, but obviously I cannot imagine what that is.

  185. dpy6629 says:

    Attp, this is a question. For Nick Stokes. He gave me a reference and it was incomprehensible to me but seemed to have very old origins. Turbulence has come a long way since the leapfrog scheme. It should be pretty easy to try a modern eddy viscosity model. This is where the “pile of very old Fortran” problem comes in. I agree with mt that we need a new modern model so it becomes possible to do this kind of thing.

    The other thing to do is test the sensitivity of the result to changes in the model, a big task.

    In the 1970’s algebraic models were the only ones used because of computer limits. These old models are
    Very bad compared to pde based modern ones.

    MT, Has anyone done it?

  186. DY,
    You haven’t even tried to answer my question. Can you at least try? Do you not realise that in science simply pointing out that a model can’t do something is not relevant, unless you can also show that what it can’t do is related to what it is trying to do. All models are wrong, but some are useful. Pointing out the former, while ignoring the latter, is a waste of time.

  187. ” Almost any criticism from outside is treated as unfair.”

    Has that happened on this thread?

    “If you have a thin skin you shouldn’t play climateball at all.”

    However, it would be nice if we could discuss climate science without it being climateball. Call me hopelessly naïve, but I had thought that was what we were doing? ;o)

  188. dpy6629 says:

    W, WTF? Bertrand Russell would eviscerate you.

  189. MT,

    But consider my position – I think almost all macroeconomics applied to time scales longer than a couple of years is broken. …..

    If I am to question whether IAMs are a useful tool for climate economics I have to give permission to people to also question GCMs for climate physics. I think it’s important for a field to be able to handle challenges from outside.

    A very good point. I’m maybe biased towards being reluctant to be overly critical of methods/techniques outside my own area, because my experience tells me that in most cases the outside critic is missing something crucial. However, being able to deal with genuine/honest challenges from outside is important.

    If you take the right attitude and your interlocutor does too, there is plenty of stuff to be learned, The occasional arch sarcasm is a price worth paying. If you have a thin skin you shouldn’t play climateball at all.

    True, but occasions where both parties have the right attitude is rare. However, you’re right that if you have a too thin skin, you shouldn’t play ClimateBallTM. I do find this unfortunate, though.

    Eli is right; my patience lasts longer than most and it gets abused on occasion. BUt I think it’s the right approach to public conversations about climate.

    Your approach is commendable and I wish I could emulate it more than I do. I kind of think that it’s the right approach, but I have real trouble not giving as good as I get.

  190. BBD says:

    mt

    If you read closely, you will see that DY Is fudding over models to advance a low sensitivity argument by stealth.

    Watch, and see 😉

  191. David,

    W, WTF? Bertrand Russell would eviscerate you.

    Who is this aimed at? If me, please explain, and explain properly. I’m really doing my best to take MT’s advice on board to treat you with respect. I am, however, really struggling and my patience is running extremely thin. Give me an opportunity to simply ban you and I will take it with glee. MT seems to think that discussing this with you is worthwhile. I’m largely convinced that it is not. Convince me otherwise, or consider going away.

  192. “However, it would be nice if we could discuss climate science without it being climateball.”

    In stats, the Bayesian/frequentist divide is often pretty partisan, but in my experience, generally the two sides are able to discuss the issues without it inevitably degrading into “fundamentaldefinitionofprobabilityball”. I wonder why climate science can’t be discussed online without it being necessarily climateball; I’m pretty sure the answer has nothing to do with the science.

  193. I’m the “W,” AT.

    Care to show me how he would, DY?

    Give me your best shot.

  194. Ahh, I see. I thought he was doing a double W in WTF.

  195. Nah, AT – DY only had an affected response after seeing that his:

    [W]e need a new modern model so it becomes possible to do this kind of thing.

    The other thing to do is test the sensitivity of the result to changes in the model, a big task.

    has been more or less anticipated.

    ***

    Let’s stand aside his doubling down the “I agree with mt” (i.e. I’m with the cool guy, na-na-na) when the last time he did so it was for the purpose of putting words into MT’s mouth. That’s between him and MT.

  196. mt says:

    The worse the models are, the uglier the risk spectrum is, and the more urgently overdue climate policy is, because uncertainty is not your friend. So if David makes that argument, he is plainly wrong, but so far whatever I’ve understood of his has not been plainly wrong, so it would surprise me and disappoint me.

    Apologies to the legal staff at the ClimateBall League for brand abuse; of course I meant ClimateBall Brand Internet Flame War (TM).

  197. MT ““The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”

    seems plain wrong to me, the fact that weather forecasts have a short horizon does not imply that 1000 year averages are unreliable. If you had a weather model with perfect physics and infinite spatial and temporal resolution it would still have a short forecast horizon due to the lack of information about the initial conditions. I could be wrong, but it seems a non-sequitur to me.

  198. mt says:

    David, please summarize your current position and email it to me at mtobis at gmail or point to it. If you’d like to discuss it elsewhere I will consider posting it on my blog. Then we can leave ATTP alone and start from a blank slate. I do not promise never to get bored and frustrated. But I’m not there yet.

  199. MT,
    Sorry, my attempt to emulate your patience didn’t last very long at all.

  200. > The worse the models are, the uglier the risk spectrum is, and the more urgently overdue climate policy is, because uncertainty is not your friend.

    The lukewarm argument goes the other way around — we need prettier models because there’s too much uncertainty; prettier models should reduce CS; if we can’t get prettier models, so much the worse for climate policy; in any case, (non-libertarian) policies are not overdue.

  201. David Young says:

    W is Willard, one of whose heroes is Bertrand Russell. He and I have discussed this at length before. I think the theory of ClimateBall is something Russell would regard with as little respect as he showed to the Romantics. W will understand. 🙂

  202. David Young says:

    ATTP, I thought I answered your question to the best of my knowledge. I said I didn’t know for sure whether turbulence models make a huge difference for climate. I and most experts however I think would be surprised if it didn’t. Especially when there is lots of turbulence (as in the atmosphere) and also relatively laminar flow, I think most turbulence experts would say they would be surprised if an algebraic model worked very well.

  203. David Young says:

    Dikran, I can’t do a detailed comment on Krakos and Darmofal now. You can however look it up in the AIAA Journal I think about 5 years ago. Use your imagination and look at the NACA airfoil results. It’s Navier-Stokes at low Reynolds number so a turbulence model is not used. The result is that there is a fixed point and with various pseudo time methods you can get any answer within a factor of 2.0. These pseudo-time methods are used in ALL RANS CFD codes with a couple of exceptions (such as ours).

    Not an exact analogy to climate but it is the same flavor as Williams Lorentz attractor illustration. The other piece of evidence here is the rigorous attractor dimension upper bound I gave you. That’s proven in Temam’s excellent book on Navier-Stokes. I don’t know what the real dimension is, but if its large, we can’t rule out some rather non-intuitive long time behavior. I don’t trust my intuition about these things as it has been very untrustworthy in the past. Yours may be better, but I am skeptical. I need to look at Willard’s reference, You can tell me perhaps what the dimension of 5 means. Periodic boundary conditions is pretty limited perhaps.

    BTW, lets drop the “lack of respect” thing. I don’t care if you respect me and you for purposes of scientific arguments (hopefully) don’t care if I respect you. 🙂 Since I don’t know you, any other position is just really emotional projection.

  204. Joshua says:

    mt –

    ==> “Almost any criticism from outside is treated as unfair.”

    Dikran asked if it has happened in this thread. I’m curious if it’s happened to you more generally.

  205. David Young says:

    History of Western Philosophy, W.

    “Revolt of solitary instincts against social bonds is the key to the philosophy, politics, and the sentiments, not only of the romantic movement, but of its progeny down to the present day. Philosophy, under the influence of German idealism, became solipsistic, and self-development was proclaimed as the fundamental principle of ethics. As regards sentiment there has to be a distasteful compromise between the search for isolation and the necessities of passion and economics. …The comforts of civilized life are not obtainable by a hermit, and a man who wishes to write books or produce works of art, must submit to the ministrations of others if he is to survive while he does his work. In order to feel solitary, he must be able to prevent those who serve him from impinging on his ego which is best accomplished if they are slaves…. Not only passionate love, but every friendly relation to others, is only possible, to this way of feeling, in so far as the others can be regarded as a projection of one’s own self.. Hence and emphasis on race… By encouraging a new lawless Ego, it made social cooperation impossible and left its disciples faced with the alternative of anarchy or despotism. Egoism at first made men expect from others a parental tenderness, but when they discovered with indignation, that others had their own Ego, the disappointed desire for tenderness turned to hatred and violence. ”

    I would say that is metaphysical absolute contempt and shows a complete and deep “lack of respect” for the Romantics.

  206. dhogaza says:

    ATTP:

    “Sorry, my attempt to emulate your patience didn’t last very long at all.”

    Remember. MT does not have your medium-length history with DY, which in the beginning was marked with patience on your part. So when summed over your history with DY, you are far ahead of MT, and it will take months for MT to get there.

    MT might consider the significance of this.

    MT might also consider skimming some of the older threads here, and places elsewhere where DY has been far less polite, and consider whether DY’s concern about airfoils and his expertise in certain limited aspects of the limitations of CFD really makes him smarter than those working on GCMs. Who DY has, in the past (though not quite so explicilty in this thread), characterised as being unskilled, lacking knowledge, not keeping up with the relevant science (CFD on small scale items like his most recent call-out, “Use your imagination and look at the NACA airfoil results”), who have essentially written models that suck.

    All in a FUD campaign, as others have mentioned above, to convince us that the sucky models overestimate CO2 sensitivity.

    No, you’re not going to find this in this thread, MT. Why do you imagine that you would? Why to you imagine that everyone else here’s exposure to DY is similarly restricted to this thread?

  207. > I think the theory of ClimateBall is something Russell would regard with as little respect as he showed to the Romantics.

    Your unverifiable counterfactual is duly noted, DY.

    However, your “but Russell” wasn’t supposed to be in response to any theory of ClimateBall, but to the tension in the many things you peddled on this thread. Let me remind you:

    [DY0] I agree that the planet is going to continue to warm. The crucial question is how much.

    [DY1] The more nonlinear a polynomial equation is the more likely it is to have multiple solutions.

    [DY2] [W]e would have never found this without modern numerical methods and a brand new code.

    [DY3] It is alarming though if you want to use this stuff where public safety is at stake.

    DY0 and DY3 are two faces of the same lukewarm coin. DY0 and DY1 create a Dutch book to justify one’s insatisfactions against modelling. DY0 and DY2 shovel policy after we RETHINK ALL THE THINGS. DY2 and DY3 shovel policy after we RECODE ALL THE THINGS. You sure get the drift. Now, what about the relationship between DY1 and DY2 or DY1 and DY3?

    https://andthentheresphysics.wordpress.com/2015/09/12/constraining-model-ecs/#comment-63202

    Do you know many romantics who were using arguments like this one, and how would Russell regard the nirvana fallacy you’re peddling?

    Many thanks!

  208. dhogaza says:

    DY:

    “History of Western Philosophy, W.”

    DY, smartest person in the room, and none of us should forget it. Not even MT, apparently annointed by DY as being the second smartest person in the room.

  209. mt says:

    Amezcua Kalnay & Williams quite explicitly supports Dikram’s reading against David’s.

    These excerpts are from the Conclusions section.

    “The first question asked in the present paper is: are
    there any statistically significant changes in the monthly
    climatology of the SPEEDY model caused by the
    upgrade in the numerical integration scheme

    we found that
    there is no significant evidence to reject the null hypothesis
    of identical climatologies. In other words, for each
    month, the climatology generated by integrating with the
    RA filter is the same as the one obtained with the RAW
    filter”

    but

    “In general, an improvement… can be attributed to the use of the RAW filter, and
    the improvement is larger for medium-term forecasts
    with lead times of 72, 120, and 144 h. ”

    That is, the weather prediction skill improved but the model climate did not change in any way that matters, and this was apparently by design.

  210. dhogaza says:

    MT:

    “Amezcua Kalnay & Williams quite explicitly supports Dikram’s reading against David’s.”

    Of course. I’m sure Dikram will appreciate your taking the time to check just in case he was wrong, of course. Because DY is so often right.

  211. David Young says:

    Willard, You and I had this discussion a long time ago. Russell’s early sophomoric work is not a reflection of the mature philosopher or the mathematical logician. I do think climate ball theory is not very interesting because it basically assumes a level of dishonesty and deep layers of meaning that seem a construct. Ocams razor would demand the simpler explanation, namely, that people usually say what they mean or at least try. Climateball is really a theology of the climate blogosphere, finding small inconsistencies, imaginary differences, and a focus on minutiae and single phrases or sentences often taken out of context. Your habit of resurrecting ancient blog history is also unhelpful. We all perhaps deserve the benefit of the doubt.

    No disrespect intended. 🙂

  212. David Young says:

    MT, The Newton Institute lecture does give an example that is not in the paper.

  213. Willard says:

    Vintage 2013:

    There are two possibilities:

    1. The dogma of the attractor: You always get sucked in to the attractor in the long term so short term errors don’t matter. This is just so far as I can tell wishful thinking. There is no mathematical or theoretical justification, merely, the observation that climate models don’t blow up. This is the official ex cathedral dogma of the team and the IPCC. As I say, there is really no support for it.

    2. Clmate models have so much spurious dissipation that all error modes (and also the signal you want to track) are dissipated away leaving a simulation with forced and wrong artificial stability.

    I actually favor the second theory because it is supported by 50 years of experience in numerical PDE simulations and we know that the models have a lot of spatial dissipation added explicitly. I just finished verifying this for the latest NCAR model which by the way still uses the leapfrog time marching scheme with the filter shown by Paul Williams to be far too dissipative.

    You need something better to have a chance with people with actual mathematical background in PDE’s.

    http://judithcurry.com/2013/06/24/how-should-we-interpret-an-ensemble-of-models-part-i-weather-models/#comment-336712

    The emphasized sentence does not square very well with Bertrand’s 4th commandment: When you meet with opposition, even if it should be from your husband or your children, endeavour to overcome it by argument and not by authority, for a victory dependent upon authority is unreal and illusory.

    ***

    Vintage 2011:

    Fred, I must say that unless you watch the Paul Williams presentation at Isaac Newton Institute I am quite concerned about YOUR being open to clearly documented issues. Williams is very well documented and devastating. He shows example after example from both simple model problems and from the literature itself of instances where model outputs are sensitive to numreical methods and time stepping. Watching this again angers me that people haven’t been more careful and it makes me question whether you are being honest about the facts and data. The facts and data are pretty damning. My central concerns are based on convincing evidence, Williams results and Andy Licas’ posts here.

    http://judithcurry.com/2011/10/18/does-the-aliasing-beast-feed-the-uncertainty-monster/#comment-125195

    What does Bertrand say about ad homs, DY?

  214. Willard says:

    Between 2013 and 2011, there was no pause:

    Andy, I don’t think this point about initial value problem vs. boundary value problem is correct. Technically, both weather and climate are the same initial value problem. The difference is that weather modeling generally assumes that certain gross forcings are constant and in climate you can’t make this assumption. But the boundary value problem statement is a gloss. The real statement you are grasping for I think is that climate is the statistics of the attractor. But this is a statement that in fact implies nothing absent other strong assumptions that are certainly false such as that the attradtor is of relatively low dimension and is strong enough to over come errors in time integration. Paul Williams has shown how in simple situations this assumption has no basis.

    Reminiscing all this should start with Andy Lacis’ comment. An excerpt:

    Weather modeling/prediction and climate modeling/prediction are two very different problems in physics. Weather modeling is an “initial value” problem, while climate modeling is a “boundary value” problem. Weather may be thought as being similar to keeping track of a busload of people headed for the airport. They are all in close proximity at the initial starting point, but in a few of days they will be dispersed all across the globe. Climate, on the other hand, is more akin to a group of people scattered around the world who are all coming to attend some convention. At the end of their trip, they will all be at the same location, no matter where they might have started out from. (Weather is a diverging phenomena starting out from initial conditions, while climate is converging toward some equilibrium point defined by radiative forcings.)

    Good atmospheric dynamics and thermodynamics (and high spatial resolution), and good specification of initial wind, pressure, humidity, and temperature conditions, are the key factors for good weather modeling. Radiative heating and cooling, and the conservation of energy, are of little significance or concern.

    http://judithcurry.com/2012/03/06/ams-members-surveyed-on-global-warming/#comment-182528

    ***

    Speaking of Williams’s lecture, here are the slides:

    http://www.newton.ac.uk/files/seminar/20101208153016101-152652.pdf

  215. Willard says:

    A bedtime story:

    Another excellent point concerns computer simulations. I view this reliance on computer simulations as a potential death knell of science. Simulations of complex simulations are usually badly wrong. The science of analyzing the errors in simulations is a field of utmost importance. The fundamental problem here is asymptotics. The information / cost ratio for almost all computational simulations is terrible, effectively there will never be a computer big enough to achieve meaningful accuracy without a revolution in methods. Higher order finite element methods are a possibility.

    http://judithcurry.com/2012/03/06/ams-members-surveyed-on-global-warming/#comment-182528

  216. DY,

    said I didn’t know for sure whether turbulence models make a huge difference for climate. I and most experts however I think would be surprised if it didn’t.

    Not only does it not answer my question, I also do not think it is true; unless what you really mean is very different to what it seems that you mean. If by experts you mean experts who do climate modelling then I do not think that they think accurate turbulence modelling would make a huge difference if the goal is to try and understand how our climate responds to external perturbations, and if we are interested in scales of the size of continents/large countries. You should probably read MT and Dikran’s comments again.

  217. DY,
    I also think this speaks volumes

    BTW, lets drop the “lack of respect” thing. I don’t care if you respect me and you for purposes of scientific arguments (hopefully) don’t care if I respect you. 🙂 Since I don’t know you, any other position is just really emotional projection.

    It would, of course, be stupid to expect people to actually respect others. They either do or they don’t. However, someone who can’t even bother to treat others with respect during a discussion, is someone who probably isn’t really interested in dialogue.

  218. Willard,
    Your link goes to a very good comment by Andy Lacis, that DY should probably read. I can’t see where your quote comes from, though.

  219. I asked DY
    “David, do you agree that the monthly climatologies [in William’s paper] not being significantly affected suggests that the CFD issue does not greatly affect climate prediction? If not, please explain why.”

    Davids reply:

    “Dikran, I can’t do a detailed comment on Krakos and Darmofal now. You can however look it up in the AIAA Journal I think about 5 years ago. Use your imagination and look at the NACA airfoil results. It’s Navier-Stokes at low Reynolds number so a turbulence model is not used. The result is that there is a fixed point and with various pseudo time methods you can get any answer within a factor of 2.0. These pseudo-time methods are used in ALL RANS CFD codes with a couple of exceptions (such as ours).

    Not an exact analogy to climate but it is the same flavor as Williams Lorentz attractor illustration. The other piece of evidence here is the rigorous attractor dimension upper bound I gave you. That’s proven in Temam’s excellent book on Navier-Stokes. I don’t know what the real dimension is, but if its large, we can’t rule out some rather non-intuitive long time behavior. I don’t trust my intuition about these things as it has been very untrustworthy in the past. Yours may be better, but I am skeptical. I need to look at Willard’s reference, You can tell me perhaps what the dimension of 5 means. Periodic boundary conditions is pretty limited perhaps.”

    This is evasion, pure and simple. I find a paper (ironically with David’s help) that provides evidence that he is wrong, and David refuses to discuss that evidence and instead merely restates his position yet again. Raising Krakos and Darmofal is just a rhetorical device to try and shift attention away from the inconvenient Williams paper and create work for me. That is deeply shabby behaviour in a scientific discussion.

    David also wrote

    “BTW, lets drop the “lack of respect” thing. I don’t care if you respect me and you for purposes of scientific arguments (hopefully) don’t care if I respect you. 🙂 Since I don’t know you, any other position is just really emotional projection.”

    How ironic. For the purpose of scientific arguments, the really vital thing is to address evidence that runs counter to your argument, which David has done repeatedly in this discussion. Respect is important in scientific discussion, if you don’t take your “opponents” view seriously, you will become resistant to being corrected. If you are unable to be corrected when you are clearly wrong, for instance

    “The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.”

    whist at the same accuse others of “dogma”, you will only end up making yourself look silly. Being respectful is for your own scientific good, not mine. Of course you can’t be corrected on that either.

  220. Correction: Should have been “For the purpose of scientific arguments, the really vital thing is to address evidence that runs counter to your argument, which David has failed to do repeatedly in this discussion. “

  221. On September 18, 2015 at 9:21 pm, ATTP said,

    “DY,

    You haven’t even tried to answer my question. Can you at least try? Do you not realise that in science simply pointing out that a model can’t do something is not relevant, unless you can also show that what it can’t do is related to what it is trying to do. All models are wrong, but some are useful. Pointing out the former, while ignoring the latter, is a waste of time.”

    David Young said in reply on September 19, 2015 at 1:00 am,

    “ATTP, I thought I answered your question to the best of my knowledge. I said I didn’t know for sure whether turbulence models make a huge difference for climate. I and most experts however I think would be surprised if it didn’t.”

    What the average global surface temperature – or better yet, the heat content of the climate system (including the oceans) – will do over the next one or two centuries is mostly what this politicized global warming “debate” is all about. And on this issue, this whole discussion on turbulence seems to me to be irrelevant. That is:

    As I’ve pointed out many times before, the heat content of the climate system changes if and only if the incoming heat flux (related to a combination of solar irradiance and albedo) changes or the outgoing heat flux (related to the greenhouse gas effect) changes. And since with anywhere close to business as usual greenhouse gas emissions we can expect the latter to continue to change as it has up to now, trapping more and more heat, and since there does not seem to be any meaningful possibility that albedo will or even can change much (with solar irradiance assumed to not change in any meaningful way) based on what I pointed out before on albedo (see my comment here
    https://andthentheresphysics.wordpress.com/2015/09/09/guest-post-some-thoughts-on-scientific-software-in-general-and-climate-modeling-in-particular/#comment-62826
    September 15, 2015 at 12:04 am under the post “Guest post: Some thoughts on Climate Modeling” for my most recent comment on this point on albedo), it should seem pretty clear what’s going to happen over the next one or two centuries regardless of what one would want to say on turbulence – that is, we have much reason to believe that it’s going to get hotter and hotter in the same positively accelerated fashion it has over the past one and half centuries. (Follow the links I give starting in my comment above to see the positive acceleration in question.)

    I believe ATTP essentially addressed and all this in a more elaborate way on September 18, 2015 at 6:32 am and September 18, 2015 at 8:06 am. These two are what DY replied to above. ATTP said,

    “Let’s imagine that we think it’s possible that non-linearities could lead to a dramatic change in how much we think we might warm in the coming decades. Let’s imagine that some small-scale non-linear effect can lead to dramatic cooling in the Arctic, increases in Arctic sea ice and ice sheets, etc. The problem is that the incoming solar insolation is well-defined and predictable. How do you sustain a region being much cooler than we’d expect from basic energy balance for long enough to lead to significant changes in ice coverage. The heat capacity of the surface/atmosphere is not very large. It should warm up relatively quickly. Similarly if you want to sustain a warmer than expected region.

    Now, you can have internally-driven warming/cooling that can persist for a reasonable length of time (years). This can be associated with ENSO events, but also with internally-driven radiative perturbations (clouds/water vapour). The problem here, though, is that the feedback response is – on average – smaller than the Planck response. Although you might violate this in some region for some time, you would not expect large internally-driven perturbations to persist for a long time (decades) because the system will have a tendency to return to its state of quasi-equilibrium, defined by the incoming solar insolation, the albedo, and the typical composition of the atmosphere.

    I should probably have added that the influence of the atmosphere (greenhouse effect) is largely determined by the composition/concentration of the persistent greenhouse gases (CO2 mainly) not by the preciptable greenhouse gases.”

    DY seems to not get what ATTP is saying. That is, it seems that DY does not take the point I made above that ATTP said in a more elaborate way – it’s a heat input / heat output question as laid out above, and turbulence does not really change that (see my note above on global warming following a positively accelerated curve over the last one and half centuries – follow the links for more to the graph showing this). He seems to argue that for all we know, by “turbulence” (by “chaos”), even with business as usual emissions, we still really have no meaningful idea what will happen over the next one or two centuries – we can’t really say whether there will be an ever increasing heat accumulation in the climate system or we can’t really say whether we will see an ever hotter average global surface temperature – we could even see another ice age – even with an ever increasing heat accumulation in the climate system. I don’t see how either of these is cogent – it seems to me that each would have to violate some established physics. (People like ATTP who have PhDs in physics can say whether there might be such violations and what they might be. Note: His above seems to already have gone there somewhat.)

  222. Magma says:

    I’ll repeat the point I made in the Tol nonsensus thread about Richard Tol… many intelligent commenters are collectively spending far more time and effort refuting David Young than he is spending baiting you. If you’re writing for the benefit of other readers here, fine. In my opinion both Young and Tol are engaged in a slightly more subtle form of t‍rolling.

  223. David Young says:

    I can see we are reverting to ClimateBall mode. This gets to the abysmally partisan nature of the climate wars and those engaged in it.

    Dikran, I gave you some evidence, from Paul Williams himself (the video) and from Krakos and Darmofal that numerical methods can change the long time behavior of dynamical systems dramatically. These systems are not climate, but are model problems that are not unrelated to climate. It is really a stretch to say “we lowered one source of excessive dissipation in a weather model and it didn’t change the climate” so the climate must be stable to changes in numerical methods. The change was very valuable for weather modeling however. This tells me that there are probably other improvements that should be considered.

    1. It’s pretty obvious that working on numerical methods can have a big effect on GCM weather results.
    2. Leapfrog is a very old scheme and has been well known since the 1970’s to have problems. That it even is still used is a strong illustration of MT and my point. BTW, the RAW filter leapfrog is still probably easily improvable using backward differentiation methods. But then there is the pile of FORTRAN problem.
    3. A clean sheet GCM could be immensely valuable as a test bed for methods if nothing else.

    I don’t think it is going to help to argue here about how stable climate simulations are to changes in methods, sub grid models, etc. The paper highlighted in this post showed some results showing that climate effects are very easy to generate by such changes. This paper however just scratches the surface. There needs to be a much larger effort here even though real verification and validation may be beyond reach. Starting with turbulence models would be a logical place to go. Everyone in the field of fluid dynamics (and its a huge and old field) knows they make a difference, because turbulence is a large effect on the resolved scales of a GCM simulation.

  224. DY,

    I can see we are reverting to ClimateBall mode. This gets to the abysmally partisan nature of the climate wars and those engaged in it.

    Do you at least recognise your own contribution, or are you suggesting that somehow your behaviour has been above board and that the problems is with others only?

    That it even is still used is a strong illustration of MT and my point.

    What is MT and my point? I agree with virtually everything MT has said. I think I also understand what you’re saying. I don’t think what you’re saying and what MT is saying are sufficiently similar for you to suggest that somehow you’re saying the same thing. Of course, this is probably simply a ClimateBallTM move, which is rather ironic given the quote of yours I’ve highighted above.

  225. David Young says:

    Willard, I love you and thanks for posting the link to Williams slides.

    Dikran might look at the quote from Pfeffer et al.

    “These results give evidence that climate simulations are sensitive not only to physical parameterizations of subgrid-scale processes but also to the numerical methodology employed.”

    “In the weather and climate prediction community, when thinking in terms of model predictability, there is a tendency to associate model error with the physical parameterizations. In this paper, it is shown that time truncation error can be a substantial part of the total forecast error.” Teixeira et al, 2007.

    Pretty clear that what every fluid dynamicist knows from graduate school is also true for climate and weather modeling.

  226. DY,
    Care to comment on this

    Amezcua Kalnay & Williams quite explicitly supports Dikram’s reading against David’s.

    I think Willard would probably call this peddling

    Pretty clear that what every fluid dynamicist knows from graduate school is also true for climate and weather modeling.

    Every physicist knows that energy is conserved. I think most fluid dynamicists do too. Given my discussions with you, I’m not clear that all do, though.

  227. David Young says:

    ATTP, I know that when I am perceived as part of the opposite team, my behavior will be microscopically examined for the slightest implication that offends someone’s sense of propriety. That’s a silly game to play since its impossible to meet these very partisan standards without muzzling yourself.

    I would just note that Andy Lacis’ online persona is very aggressive and blunt, making me look like a piker. I like Andy and can look past his personality to the substance. I would hope at least some others here besides MT could do the same.

  228. David wrote “Dikran, I gave you some evidence, from Paul Williams himself”

    Yes, and I looked into Paul Williams work and found that he says the approach has no significant effect on monthly climatologies. That *is* climate and you are *still* evading the issue. This is not climateball, at least on my part; I have taken your argument seriously and investigated your sources and come back with questions that you are refusing to engage with.

    You also made an error when you wrote ““The other issue is just why would a weather model that becomes inaccurate after a few days get the 1000 year averages right.” and you are apparently unwilling to discuss that either.

    The conclusion that I have arrived at is that the sort of numeric issues that David mentions are real, and GCMs are improved by their inclusion, that they are already known about in the climate modelling community, and they are already working on it (Paul Williams for example), as they are working on many other issues. However, David provides no evidence whatsoever that they have a substantial effect on climate modelling, and refuses to discuss evidence from William’s paper that suggests that they don’t. So I am not hostile to what David has to say, but I am not going to accept it on his authority and I am going to ask questions. If he won’t answer them, he shouldn’t be surprised that he fails to convince.

  229. DY,

    I know that when I am perceived as part of the opposite team, my behavior will be microscopically examined for the slightest implication that offends someone’s sense of propriety. That’s a silly game to play since its impossible to meet these very partisan standards without muzzling yourself.

    Hmmm, that is not what I’m suggesting at all. I’m suggesting that you seem completely incapable of recognising how your own behaviour influences how these discussions progress. I’m not trying to examine your behaviour, I’m suggesting that maybe you should examine your behabiour.

    Also you quoting this is quite annoying

    “These results give evidence that climate simulations are sensitive not only to physical parameterizations of subgrid-scale processes but also to the numerical methodology employed.”

    I doubt anyone here would argue that the physical parametrizations of the subgrid-procresses aren’t important or that the numerical methodology isn’t important. That you would seem to be suggesting otherwise rather confirms my view that you have no great interest in actual dialogue.

  230. David Young says:

    MT and I both believe that new and modern GCM’s would be a big contribution. MT seems to feel this more from a software point of view and my reason is more related to using modern low dissipation methods and after our discussion here turbulence models.

  231. DY,

    MT and I both believe that new and modern GCM’s would be a big contribution.

    You seem to be suggesting that you two are the only two here who think that. If so, that is utterly bizarre. Why would you think any such thing? Are you just trolling now?

  232. David Young says:

    [Mod : Not good enough. Explain it properly yourself!]

  233. David Young says:

    [Mod : Pathetic answer. Care to try again?]

  234. David Young says:

    Williams shows some examples from the Lorentz attractor and for GCM’s that changing the time step will change the climate of the system. There are also some quotes from older papers saying that numerical methods do affect the climate. It is a little surprising that we are arguing about something so obvious.

  235. DY,
    I don’t think we are arguing about this. It’s a little surprising that you think we are.

  236. BTW as a computer scientist, I think it would be a good idea to invest in software engineering to support climate modelling; as a statistician I think it would be a good idea to invest in statisticians to support climate research. However we also need investment in collecting data, in developing new ideas, in evaluating existing ones against the observations… The list is long and the funding pot is pretty small (no matter what some would say – grants are not at all easy to get).

  237. David Young says:

    Dikran seems to be arguing about it. This is intended for him.

  238. David Young says:

    ATTP, I am not sure what your point here really is, can you restate it. I at least understand Dikran’s point and I think Williams provides some conclusive evidence about it. Climate is sensitive to virtually all the choices made in GCM modeling and numerical methods and turbulence model are two such choices.

  239. No, I don’t think he is. Maybe read his comments a little more carefully.

  240. David, Williams also shows that the monthly climatologies are essentially unaffected, and you are still ignoring that fact.

    It appears that David only wants to talk about the elements of William’s presentation that supports his view and not the elements of his work that argues against it. However, my patience has now run out.

  241. It appears that David only wants to talk about the elements of William’s presentation that supports his view and not the elements of his work that argues against it.

    That would, essentially, be my point.

    However, my patience has now run out.

    Likewise.

  242. David Young says:

    Dikran, So you acknowledge then that numerical methods, sub grid modeling choices, including turbulence models can and do influence the long term behavior of these models? Just because one statistical measure doesn’t change doesn’t tell me much.

    You just previously said: “However, David provides no evidence whatsoever that they have a substantial effect on climate modelling, and refuses to discuss evidence from William’s paper that suggests that they don’t.” I provided some evidence they make a difference to the long time behavior and that this is well known.

  243. DY,
    I have a suspicion that you’re rather misrepresenting what Dikran was saying. Care to actually read the comments again, think about them for a while and maybe respond again? If you choose not to, that would also be fine.

  244. “I provided some evidence they make a difference to the long time behavior and that this is well known.”

    David is still ignoring the monthly climatologies.

    “Just because one statistical measure doesn’t change doesn’t tell me much. ”

    It is the statistical measure designed to test the effect on the modelling of climate (the hint is in the name). If you think that doesn’t tell you much, that seems a bit, err… counterintuitive. If it were me, I would assume that the fact the monthly climatologies don’t change means that perhaps I had overestimated the significance of the other evidence. Self-scepticism is a useful guard against the conformation bias to which we are all susceptible (and I am no exception).

    As I said, my patience really has run out.

  245. Willard says:

    > I can see we are reverting to ClimateBall mode.

    You never really left it, DY. From your first comment:

    This approach is being used in fluid dynamics too and it has fundamental limitations. […] Varying the forms of all these terms quickly turns into a dart throwing exercise.

    This paper seems to me to raise more questions than it answers and the task ahead is rather daunting.

    https://andthentheresphysics.wordpress.com/2015/09/12/constraining-model-ecs/#comment-62695

    An editorial that could lead to a formal argument (“fundamental limitations”) turns into a sideswipe (“dart throwing exercise”). A conclusion with mere armwaving (“more questions than in answers”) and innuendo (“rather daunting”). Pure ClimateBall, where nothing of interest gets said and all that gets peddled is judgmental crap to dismiss the results from a whole field.

    ***

    Interestingly, there’s already one tension I underlined earlier. In fact, DY’s argument looks like a Dutch book to me:

    (DB1) There are fundamental limitations to modeling.
    (DB2) There are unanswered questions.
    (DB3) The task to answer them is daunting.

    Considering DB1, concerns about DB2 and DB3 can always be raised. In other words, if there are fundamental limitations to the modeling task, why expect it needs to surpass them before we consider it answers questions?

  246. David Young says:

    More from Williams slides (thanks to W).

    “Different low-cloud climate feedbacks in current climate models are due to complex interactions between physical parameterizations and numerical artifacts” Cheedela et al, 2010

    There are also some nice plots from Zhao & Zhong 2009 comparing leapfrog and Adams Bashforth. It makes a difference to the zonal averages.

    Losing patience hopefully means you at least thought about the point raised.

  247. DY,
    Quoting things that we all would have agreed with before this discussion started as if they support what you were saying at the beginning (when they don’t) is why people are losing patience.

    Losing patience hopefully means you at least thought about the point raised.

    Have you tried thinking about the points that are raised. Doesn’t seem like it. It just seems like you’re intent on trolling.

  248. David Young says:

    Willard, Everything you quote is I think true. You fail to quote the bits from previous posts where I say what my positive agenda would be. Better data, a clean sheet GCM, improved theory and simple models particularly for convection.

    I still think you should read your hero’s History of Western Philosophy. Might give you some perspective on game playing vs. real hard hitting discussion.

  249. Willard says:

    > Your link goes to a very good comment by Andy Lacis, that DY should probably read. I can’t see where your quote comes from, though.

    DY might have read it, since my quote after “between 2013 and 2011” is from a comment he made in response to Andy’s comment. Here’s the link, with another tidbit:

    Dynamics are critical to the feedbacks which are critical to sensitivity estimates. The dynamics are precisely the “weather.” My sense based on 30 years of modeling is that you’ve got a long way to go to make the “boundary value problem” gloss mean something. Assertions about the “average dynamics” require more than just hand waving in my opinion.

    http://judithcurry.com/2012/03/06/ams-members-surveyed-on-global-warming/#comment-182552

    The last sentence shows another piece of evidence of DY’s ClimateBall mode.

    ***

    The whole discussion is rather interesting, with ChrisC, PekkaP, FredM, Captain Kangaroo (otherwise known as Chief), JimD et al. At least from a ClimateBall point of view.

  250. David wrote “Losing patience hopefully means you at least thought about the point raised.”

    Nice try at winding me up, but it won’t goad me into returning to the discussion.

  251. Eli Rabett says:

    If you believe in entropy turbulence doesn’t make a damn bit of difference in climate modeling.

  252. Willard says:

    > Everything you quote is I think true.

    I can grant you the truth of your claim and my argument still applies, DY. Fundamental limitations implies that some questions won’t get answered. Which means that requiring that they get answered cannot be felicitous. Either accept the limitations or reject the whole entreprise as futile dart throwing. Both choices defeats your stance, for which there’s even a meme:

    Source: http://highexpectationsasianfather.tumblr.com/post/30418622608

    The obvious response to your appeal to perfection is that models are good enough for what we ask of them.

    ***

    When your strategy is unloseable, you stop doing science. There are many unloseable strategies in ClimateBall. Your strategy is unloseable. That’s why you can peddle it for years. That’s also probably why Gavin threw the towel years ago.

    Now, do you understand why I’m saying that you’re doing looks first and foremost like ClimateBall? If you don’t believe me, I could very well list all your appeals to authority you made on this thread alone.

    Incidentally, you will note that nothing in what I said requires that I attribute any motive at all to you. This refutes your conception of what you call “ClimateBall theory.” That theory, if we can call eristic a theory, can very well accomodate Russell’s decalogue:

    https://www.brainpickings.org/2012/05/02/a-liberal-decalogue-bertrand-russell/

  253. David Young says:

    Yes, Willard, there are fundamental limits. The models may be “good enough” for some things. I’ve been told they are not “good enough” for regional climate for example or even for global average temperature anomaly apparently. My point is simply that we need to bend every effort to get better and the first step is to honestly face the problems, issues, and limitations.

    We have been doing exactly this in CFD and its making a difference. We are also improving codes and methods. It is not a game.

    I use the climate blogs to learn about these issues and the science. Climate is so politicized that someone like MT who is openminded is a true diamond in the rough.

  254. DY,

    or even for global average temperature anomaly apparently.

    And who told you this? Just someone? A name, or maybe a link?

  255. David Young says:

    Pesky Rabbit, The presence of turbulence has an effect on the resolved scales that is often very large, equivalent to a large increase in local viscosity, well that’s the 1st order effect anyway.

  256. David Young says:

    ATTP, I’ve said this before. It’s in AR5. They lowered the likely range for global temp anomaly I think for 2030 below the model mean. But maybe they are wrong. The IPCC isn’t perfect.

  257. DY,
    Now you’re trolling. That’s pathetic. Please stop.

  258. Joshua says:

    willard –

    ==> “The whole discussion is rather interesting, with ChrisC, PekkaP, FredM, Captain Kangaroo (otherwise known as Chief), JimD et al. At least from a ClimateBall point of view.”

    It really was a quite interesting read – not only from a ClimateBall perspective.

  259. Joshua says:

    Anders –

    ==> ” That you would seem to be suggesting otherwise rather confirms my view that you have no great interest in actual dialogue.”

    Geebus. I posted a couple of times, David’s comment that he thinks it’s impossible for him to discuss problems with climate science here.

    Why did you think he is here? To discuss something that he thinks it is impossible for him to discuss (problems with climate science?)

    MT’s opinion notwithstanding, it seems pretty obvious that he’s here to push an agenda (I asked him for clarification but he hasn’t done so). Particularly amusing, IMO, is his comment that respect is irrelevant (which I agree with, actually), given that it seems that his ideological jihad is largely prompted from him not being treated with respect at RealClimate.

  260. Joshua,
    Yes, yes, I know.

    Why did you think he is here? To discuss something that he thinks it is impossible for him to discuss (problems with climate science?)

    Well, he has at least illustrated that it is impossible, even if not for the reasons he would claim.

  261. Eli Rabett says:

    The resolved scale is the globe. Turbulence dissipates.

  262. David Young says:

    [Mod : Sorry, I’m not that interested in this continuing at all. If you’re genuinely interested in this topic, I would recommend taking MT up on his suggestion that you contact him directly, and find a way to discuss it with him further. I think it is commendable that MT has the patience that he has, and I wish I had a similar level of patience. I, however, clearly do not.]

  263. Joshua says:

    ==>” even if not for the reasons he would claim.”

    David’s comment that it isn’t possible for him to discuss problems with climate science here was interesting in that there wasn’t any apparent recognition of how is own actions might play a role.

    Is it impossible for MT to discuss problems with climate science here?

    If so, then why is it possible for him where it isn’t for David? Because MT’s one of the tribe and David isn’t? Because of something about David’s approach?

    It also may be telling in that David’s comment suggests that he’s only interested in discussing the problems with climate science. In and of itself, that wouldn’t seem particularly meaningful. But given how the discussion played out, where others keep asking him, essentially, to contextualize the problems that he’s focusing on within the larger framework of climate modeling, and as far as I can tell he keeps refusing to do that (and further, implying that asking him to do so is merely “dogma” or an outcome of “Schmidtian ideas about [ones] own credentials or intelligence” to borrow from the other thread,), then maybe there’s more behind his statement that he can’t discuss the problems with climate science here.

  264. Joshua says:

    Sorry –

    I posted that last comment after you did your most recent moderation of David’s comment. While I’d like you to leave my last comment up for David to read (one can always hope, can’t one?), it may be somewhat unfair for that comment to stay up if David isn’t allowed to respond.

    I’d like you to allow him to respond if he sees fit, but you’re not going to allow him to do so, maybe you should delete my last comment?

  265. Joshua,
    Okay, he’s welcome to respond. I don’t have a copy of his earlier comment, though, so can’t retract the moderation.

  266. Joshua says:

    Thanks.

  267. > The models may be “good enough” for some things.

    Like preserving monthly climatologies’ properties even when benchmarked using Paul’s numerical methods, perhaps?

    DM mentioned “climatologies” on the September 18, 2015 at 5:06 pm and asked DY to acknowledge it at 6:23 pm. Then on September 19, 2015 at 4:00 am, presumably independently, MT pointed it out.

    Then DM mentioned it again on September 19, 2015 at 8:48 am, 3:34, 4:02 pm, and 4:21 pm.

    Failing to acknowledge it is not good enough anymore.

    ***

    > I’ve been told they are not “good enough” for regional climate for example or even for global average temperature anomaly apparently.

    Senior’s squirrel is duly acknowledged. The relevance of this fact (which I believe is true) in the grand scheme of things is far from being obvious.

  268. David Young says:

    Josh,

    Here’s what i originally posted. Since Willard posted the link, it seemed appropriate to the discussions here about style, tone, and respect. Particularly 5,6,7, and 8 are relevant.

    It originally appeared in the December 16, 1951, issue of The New York Times Magazine, at the end of the article “The best answer to fanaticism: Liberalism.” by Russell.

    Perhaps the essence of the Liberal outlook could be summed up in a new decalogue, not intended to replace the old one but only to supplement it. The Ten Commandments that, as a teacher, I should wish to promulgate, might be set forth as follows:

    1. Do not feel absolutely certain of anything.
    2. Do not think it worth while to proceed by concealing evidence, for the evidence is sure to come to light.
    3. Never try to discourage thinking for you are sure to succeed.
    4. When you meet with opposition, even if it should be from your husband or your children, endeavor to overcome it by argument and not by authority, for a victory dependent upon authority is unreal and illusory.
    5. Have no respect for the authority of others, for there are always contrary authorities to be found.
    6. Do not use power to suppress opinions you think pernicious, for if you do the opinions will suppress you.
    7. Do not fear to be eccentric in opinion, for every opinion now accepted was once eccentric.
    8. Find more pleasure in intelligent dissent than in passive agreement, for, if you value intelligence as you should, the former implies a deeper agreement than the latter.
    9. Be scrupulously truthful, even if the truth is inconvenient, for it is more inconvenient when you try to conceal it.
    10. Do not feel envious of the happiness of those who live in a fool’s paradise, for only a fool will think that it is happiness.

  269. David Young says:

    I acknowledge that the Williams paper asserts that this one small change to the time stepping did not significantly affect this statistical measure. It is a very small change and as MT said it made no difference in this measure perhaps “by design.” There is lots of other evidence and a lot of “physics” understanding of turbulence that is also worth considering. To assert that “climate is different” is itself a very big and ugly squirrel already pretty much shot by Nick Stokes and other experts. This squirrel violates by the way rule 5 of Russel’s decalogue.

  270. DY,
    Seriously, please do as MT suggested and contact him directly if you’re interested in taking this further. I’m really no longer interested. I don’t think Dikran is either. I see no benefit in talking with you further about this.

  271. Joshua says:

    David –

    Thanks, but I don’t see how that addresses the questions that I was asking you. (As an aside, that list seems fine to me – although as a teacher I can think of many other things that could also be on a ten commandments list that aren’t represented.

  272. David Young says:

    I didn’t see a question there Josh. I already told why I bother with this. It is because of the opportunity to learn from people who might disagree. I even learn a lot from Willard’s links which are actually a real contribution.

    The larger context you mentioned is just that these issues can make a difference, they are in principle easy to change, modulo the pile of Fortran issue. I’m not sure if its the biggest issue, I do know it has turned out to be far bigger in CFD than “physical” understanding led people to believe. And as Stokes says “its really all CFD.”

  273. DY,
    The next time you claim someone has said something to support your view, you had better provide a link. In almost all cases in which I’m aware of what the other person has said, it hasn’t been what you claimed they had said.

  274. Joshua says:

    David –

    ==> “It is because of the opportunity to learn from people who might disagree.”

    I have to say, although I don’t know your intent, that looks to me like you’re ducking the questions I was asking you to answer – because although you’ve said that “you’ve already told why you bother” you haven’t spoken to the quote of your comment from Judith’s that I keep highlighting.

    And saying that you don’t “see a question,” is, it seems to me, being unconstructively literal.

    Now I can’t follow the technical discussions very far at all, but I can follow them far enough to see a repeated pattern where people ask you questions and at least say (and with validity to the extent that I can tell), that you haven’t addressed those questions. As such, while I do appreciate the response, I don’t think it was on topic. I could repeat the questions that I was addressing (in a general sense), but I don’t want to get caught in dark hole of blogospheric talking past each other blather.

    If you’re interested in good faith exchange, go back and read the comments in this thread that I made w/r/t your engagement here. It won’t take long. If after doing that, you still don’t see which “questions” I’m asking you to address, so be it. I have to think that if you read with purpose, you would. If you do find something there to address, I would be interested. No point in just repeating that you’ve either not found a question or already answered a question that you didn’t find.

  275. As ATTP suggest, I am not really interested in dialogue with DY anymore, as he clearly isn’t taking the evidence against his position seriously, however this seems to me a good example of confirmation bias:

    “I acknowledge that the Williams paper asserts that this one small change to the time stepping did not significantly affect this statistical measure.”

    DY shouldn’t have to be asked the question half a dozen times before acknowledging this (including having this pointed out by Willard), and when DY does acknowledge it, you will note that he does his level best to minimise the acknowledgement, for instance by not mentioning that the “statistical measure” is “monthly climatology”, i.e. a direct test of the effect on climate.

    “It is a very small change and as MT said it made no difference in this measure perhaps “by design.””

    Again, more minimisation. It may be a very small change in the code base, but it has a very substantial effect on the modelling of weather, effectively adding an extra day to the prediction horizon. Given the progress of weather modelling over the last fifty years, that seems like quite a lot to me. This shows that the effects DY talks about are important for weather forecasting but not climate. Which does not support David view at all.

    ” There is lots of other evidence and a lot of “physics” understanding of turbulence that is also worth considering”

    Yes, for weather forecasting. There is very little evidence that it has an effect on climate, and DY’s constant evasion has done a very good job of demonstrating that.

    So DY has seen the work of Williams and only accepts the importance of the limited evidence that it provides to support his position (the slides suggests there is some in the talk) but steadfastly ignores the (IMO stronger) evidence against it, and when forced to acknowledge it does his best to belittle it as much as possible,

    I wonder if Bertrand Russell would have approved.

    DY as ATTP suggests, do take MT up on his offer, and leave a link here to the discussion if it goes ahead. I won’t take part, but I would be interested to see if MT can help you make your case any better than you have done here. This really is my last post. Hopefully DY might give some thought about why he has been first so evasive and then dismissive of this line of evidence.

  276. BBD says:

    mt

    See?

    😉

  277. David Young says:

    [Mod : Whether you choose to take the advice to talk with MT further, or not, is up to you. Discussing this further here, however, wasn’t one of the options.]

  278. David Young says:

    Josh, You quote something out of context and ask me to explain it. I explain it and they you say it doesn’t answer your question which you can’t seem to state clearly. Have you forgotten the question or are you perhaps not owning your own engagement issues?

  279. David Young says:

    [Mod : Playing the ref is boring.]

  280. Eli Rabett says:

    Dikran, an awful lot of the large subsonic wind tunnels have been closed in the last 30 years and not many are left. Of those that are left many have special features such as the ability to cool the tunnel to study icing and similar stuff. Even in 2000,

    Use of CFD in initial design is today the norm rather than the exception with wind tunnels being used for fine tuning and to test at the edges of the CFD design limits

  281. Willard says:

    > This squirrel violates by the way rule 5 of Russel’s decalogue.

    Russell’s decalogue fails to mention squirrels. It’s all about truthfulness. If you want a rule about relevance, you can look at Grice’s maxims, or better yet:

    4. Relevance rule

    A party may defend a standpoint only by advancing argumentation relating to that standpoint.

    https://en.wikipedia.org/wiki/Pragma-dialectics

    That I’m using a gamesmanship terminology doesn’t imply that I don’t mean business, DY. Your “it’s not a game” amounts to a double ad hom. It raises your profile and demean mine. Not that I can’t respond in kind: for instance, I could reply that a “hard hitting discussion” implies one addresses an argument, and you failed so far to address the ones I offered you.

    It’s hard to play the ball and the man at the same time. Start with keeping your eye on the ball. With some practice, you’ll learn to make physical plays while playing the ball. You’re just not ready yet, and are using cheap ad homs to brag and to whine. This contrasts with the Very Serious schtick you’re trying to sell.

    ***

    That Williams’ favorite procedure does not change monthly climatologies is far from being irrelevant to the idea that Williams’ favorite procedure would improve the quality of climatological outputs so much that we’d have but little choice but to ditch all the actual models as pretty much useless. Paraphrasing, of course. I can look for your a number of ways you formulated this concern if you prefer.

    In any case, thank you for your concern.

  282. Joshua says:

    David –

    OK. One more try. I suspect there’s no point, but what the hey?

    I meant what I said. IMO, you could read what I wrote and come up with a more substantive response on point – if you were doing so in good faith.

    I’ve already elaborated here:

    https://andthentheresphysics.wordpress.com/2015/09/12/constraining-model-ecs/#comment-63289

    I think that the questions posed by what you wrote over at Judith’s, characterizing your views on participating in discussions here, are quite obvious. Saying that the comment was out of context, IMO, is a duck. Why are you participating in discussions on a blog, at least ostensibly discussing problems with climate science, when you said elsewhere that you found it impossible to discuss problems with climate science here? What does it mean that you’re doing something that you said was impossible? It isn’t a logical course of action.

    Does it mean that the ostensible reason that you’re here is not the actual reason that you’re here (and, for example, the main reason that your here is to grind an axe you have about being treated disrespectfully over at Real Climate, or perhaps to push a political agenda that you have associated with the issue of climate change)? Or perhaps because you’ve changed your mind from when you wrote what you did over at Climate Etc.?

    You have responded on the topic a couple of times now, alternating between saying that you couldn’t figure out what I was asking you to address (which I have to say I find implausible) and answer questions that I find it hard to believe that you plausibly felt were actually my questions.

    I’d like to take the convo further, but your inability to come up with a more substantive response is a kind of litmus test, IMO, of how open you are to engaging in the kind of introspection that would enable further good faith exchange. As such, I’m fine with just leaving it right here. Just know that when I read technical exchanges such as those that I’ve read in this thread, and I can’t follow the technical aspects, I look for other signs to help me parse what I might be able to get useful out of observing the discussion.

    When I see people claiming that what they’re asking you to do is engage in good faith exchange and that you are repeatedly failing to pick up that ball, when they say that deeply embedded with in technical questions, I never know exactly where to go with that. I can have a gut feeling about it, but maybe I’m just not technically capable enough to know that you actually are addressing their questions in good faith.

    So when something similar plays out where I feel that you are being willfully obtuse about something that, frankly, IMO, is plainly obvious – then I use that as a bit of a tool to help me evaluate what was going on in the more technical convo. It’s not a foolproof method as certainly how you engage with me doesn’t necessarily generalize to how you’re engaging with others. And of course, also, there’s obviously also two sides to our exchange and of course, I have my inherent biases that I bring to evaluating your exchanges with others.

    So if you want to take up the ball (not the Climateball) with me, have at it. I’m actually interested. That’s why I’m asking you for information. But if you don’t see a way to have a meaningful response, I’m cool with that. I’m sure that you would think that it’s legitimately because I’m full of shit or a raving lunatic or just a “troll” or something like that. I”m cool with that also. But know that as teacher, you will have failed to engage a student who is telling you that he’s interested in learning, and in particular, learning more (a tough job given my limited skills) about a complicated topic. I would like to know the answer to questions about how the problems you describe, based on your expertise, play into the larger context of modeling. I’m inherently skeptical of modeling to a pretty significant degree. But if I see you participating in that discussion in ways that don’t make sense to me, and then further you can’t explain to me why I’m wrong that your manner of participation is illogical within a context of good faith exchange to share perspectives and understand better, then it’s an opportunity lost.

  283. Joshua says:

    And really, David, if you don’t see your way clear to writing a response that you honestly think that I would think is more substantive and on point, don’t bother writing anything at all. In the very least, I’m guessing you’ll save BBD from the possibility of heartburn from reading another of my monologues.

  284. David Young says:

    Josh, We all change our minds from time to time and or succumb to hope that things can change. What really happened is that ATTP and I had a much more civil discussion at Fuller’s site that made me think he might change his previous methods. And indeed things did change for a while. This post is an interesting one that shows I think the huge challenges of really validating GCM’s particularly for climate modeling. Weather modeling is another story. That’s why I commented and then things reverted to Climateball and you know the rest of the story. The strategy of deleting on topic and technically important comments started. This is just so shabby. You read the rest of the story.

  285. DY,

    What really happened is that ATTP and I had a much more civil discussion at Fuller’s site that made me think he might change his previous methods.

    Unfortunately, my general view has not changed since that discussion at Fuller’s. If anything it has been reinforced.

    That’s why I commented and then things reverted to Climateball and you know the rest of the story.

    IMO, you were the prime reason for things reverting to ClimateBall. That you appear unable – or unwilling to see this – is why my view has not changed.

    The strategy of deleting on topic and technically important comments started. This is just so shabby. You read the rest of the story.

    If you want to write and moderate your own blog, feel free. You want to comment here, you get to play by my rules. If you don’t like that, don’t comment here. I would like to say that I don’t care either way, but that wouldn’t be true.

  286. BBD says:

    Joshua

    In the very least, I’m guessing you’ll save BBD from the possibility of heartburn from reading another of my monologues.

    Not at all. I found the last one rather soothing, to be honest 🙂

  287. Willard says:

    DY found an kindred spirit:

    I just started reading the section on numerical models and it is a jewel. The most interesting point is that those who “run the models” have higher confidence in the output model results than those who actually build the models. This struck a strong chord with me because it is so true in my field. We actually are trying to address this bias and there are some new papers coming out by big names (not me) about it. But I do have a paper on it in the works trying to do a rigorous job of evaluating uncertainty. Even for problems considered trivial, its much larger than the literature would indicate.

    Of course in climate science, the problem is very bad. Political partisans and activist but ignorant scientists feel a strong need to defend the models despite deep ignorance. It’s a bad situation.

    http://judithcurry.com/2015/09/20/new-book-doubt-and-certainty-in-climate-science/#comment-732128

    The jewel comes from Alan Longhurst, a lone vigilant who reviewed a family of research fields and, with the help of Mr. T, saves scientific INTEGRITY ™ with a Galileo-like objectivity.

    ***

    Alan seems to appreciate Bob Carter’s Taxing Air:

    Congratulations on the book, it’s a winner.

    http://www.taxingair.com/complete-reviews.html

    An objective analysis that seems to share Dr. Art Raiche, Professor David Bellamy, Nick Minchin, Former Senator and Minister for Finance, Lubos Motl, Matt Ridley, Max Rheese, Emeritus Professor Don Aitkin, James Delingpole, Paul Monk, FHB5, and many otters.

    DY might note that I could not post the last part of that comment at Judy’s.

    ADD. It finally got through.

  288. Willard,
    I rest my case. I think DY has also found himself a new home. I have every confidence that he will be made to feel welcome.

  289. Willard says:

    DY’s quest for SOUND SCIENCE ™ reached a new high:

    Willard likes to pretend he is Bertrand Russell’s squirrel. But he really is just a wild and crazy one running about with no sense of direction. For him, pointless but detailed and superficial analyses are the nuts he lives off during the winter.

    Remember Russell’s Decalogue and try harder.

    http://judithcurry.com/2015/09/20/new-book-doubt-and-certainty-in-climate-science/#comment-732511

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