Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models

By Nic Lewis

An important paper, Wang et al.[1], on the relationships between cloud feedback, climate sensitivity (ECS) and aerosol-cloud interaction in the latest generation of global climate models (CMIP6) has just been published. The key conclusion of the paper is:

The seeming consistency of global-mean temperature evolution between more positive cloud feedback (high ECS) models and observations requires a strong aerosol indirect cooling effect that leads to an interhemispheric temperature evolution that is inconsistent with observations.

The new open-access paper – the senior author of which, Gabriel Vecchi, is a well known professor of geosciences at Princeton University – provides further evidence that high ECS CMIP6 models do not realistically simulate the behaviour of the Earth’s real climate system.


The spread in estimated ECS has increased further in CMIP6 models; it is 1.8–5.5 K as compared with 2.0–4.7 K in the previous, CMIP5, generation. Far from the science being settled in relation to the response of the climate system to increasing greenhouse gas concentrations, insofar as it is represented by climate models uncertainties appear to be growing.

The main cause of the overall upwards shift in ECS in CMIP6 from CMIP5 – the mean ECS increased by about 0.5 K in addition to the 0.8 K increase in the highest ECS – is more positive cloud feedback, particularly over the Southern Ocean, in many models.[2] The stronger positive cloud feedback arises both from cloud cover and albedo decreasing more with surface temperature, driven by changes in the physical representation of clouds in many CMIP6 models.[3]

However, a growing body of evidence indicates that the high ECS CMIP6 models are unrealistic. For instance, paleoclimate proxy data are not compatible with the high ECS CMIP6 models.[4] The new paper provides further evidence, here from the historical period.

As the authors say in the paper’s Abstract:

We show that models with a more positive cloud feedback also have a stronger cooling effect from aerosol-cloud interactions. These two effects offset each other during the historical period when both aerosols and greenhouse gases increase, allowing either more positive or neutral cloud feedback models to reproduce the observed global-mean temperature change. Since anthropogenic aerosols primarily concentrate in the Northern Hemisphere, strong aerosol-cloud interaction models produce an interhemispheric asymmetric warming. We show that the observed warming asymmetry during the mid to late 20th century is more consistent with low ECS (weak aerosol indirect effect) models.

It was noticed back in 2007 that, in then current models, the change over the 20th century in total forcing was strongly correlated with that in aerosol forcing, and that their ECS was reciprocally related to their total forcing change.[5] That is what one would expect if modellers were tuning their models to match the observed change in global mean surface temperature (GMST). In CMIP5 models, aerosol forcing was more evidently correlated with GMST rise[6] than with ECS. That may well be because fewer modelling groups tuned their models to match the historical rise in GMST. Also, despite it being a solid, very basic part of climate science, there was huge variation in forcing from CO2 and other greenhouse gases in CMIP5 models.

Aerosol forcing, cloud feedback and ECS in CMIP6 models

To help bring out the relationship between cloud feedback and the other variables in CMIP6, Wang et al. separate out from the 30 models they analysed the 9 models with the most positive cloud feedback (range +0.78 to +1.18 Wm−2K−1: T9 models) and the 9 with the least positive cloud feedback (range −0.20 to +0.30 Wm−2K−1: B9 models).[7]

In CMIP6 there is again a strong correlation between aerosol forcing, in particular that related to cloud changes, cloud feedback, and ECS, as Figure S4 of the new paper (reproduced as Figure 1 below) shows. Figure S4 is the same as Figure 1 of the paper except that it only includes one model from each of the 20 institutions, which makes it clearer. B9 models are at the left in each panel of Figure 1, while T9 models are at the right. Due to only one model per institution being included, only six of the B9 models and six of the T9 models are shown.

Panel a of Figure 1 shows that ECS is strongly positively correlated with cloud feedback across CMIP6 models (r2=0.74, or 0.69 if all models from each institution are included).

Panel b of Figure 1 shows a quite strong negatively correlation between cloud feedback and aerosol-mediated cloud radiative response ΔRcaer (which includes both the local aerosol indirect forcing effect on clouds and nonlocal changes in clouds that result from aerosol-induced changes in the large-scale circulation). That is, the T9 models (those with the most positive cloud feedback) tend to have a more negative ΔRcaer than the B9 models.

Somewhat surprisingly, the T9 models have a less negative clear-sky direct aerosol forcing than the B9 models, but the much larger differences between indirect cloud-related aerosol forcing in B9 and T9 models dominate.

The filled circles in Figure 1 represent the top 50% of models graded by the consistency of their simulations with observed historical GMST warming (1990–2014 mean minus 1880–1909 mean); the bottom 50% are shown by open circles. The correlations are stronger when only the 50% of models most consistent with observed GMST warming: the r2 is 0.81 not 0.59 between cloud feedback and ΔRcaer, and 0.84 between ECS and cloud feedback. Of the ten models whose historical simulations most are consistent with observed GMST warming, four have positive ΔRcaer, and four have weakly negative ΔRcaer (~ –0.5 Wm−2).

The correlation between ECS and ΔRcaer, which is fairly weak (r2 =0.26) when consistency with historical warming is not tested, is strong (r2 = 0.78) across the 50% of models most consistent with observed GMST warming.[8]

Figure 1. Cloud feedbacks, ECS, and aerosol-mediated cloud radiative responses (ΔRcaer) in CMIP6 models; one model per institution. Reproduced from Figure S4 of Wang et al. B9 models are shown by the leftmost circles in each panel of Figure 1, while T9 models are shown by the rightmost circles, only six of each being included due to the restriction to one model per institution.

The paper points out that, since anthropogenic aerosol emissions are very largely in the northern hemisphere, models with stronger aerosol-mediated cloud radiative responses should show greater interhemispheric asymmetry in total cloud radiative response (cloud forcing) and hence in warming.

The authors say:

Due to the larger cooling effect of the aerosol-cloud interaction, T9 models simulate slightly colder surface temperature anomalies during the mid to late 20th century compared to the B9 models (Figure 4a), even though the T9 models have a more positive cloud feedback and a higher ECS. While this difference between the B9 and T9 models’ surface temperature anomaly is small when globally averaged (and only few scattered years are significantly different—indicated by the gray shading), the hemispheric asymmetry of the historical aerosol forcing induces substantial differences in the interhemispheric warming asymmetry (Figure 4b).

Figure 2, which reproduces Figure 4 in the paper, illustrates these points.

Figure 2. Modeled and observed surface temperature change. A reproduction of Figure 4 in Wang et al. Annual time-series of (a) the global-mean surface temperature anomaly and (b) the interhemispheric contrast of surface temperature anomaly. The black line is from GISTEMP and is rebased to match the model ensemble mean of the 1951–1980 period value. Each thin gray line represents a single ensemble from one model. The red and blue lines indicate the model ensemble mean of the T9 and B9 models, respectively. The gray shadings indicate the years that the difference between T9 and B9 are significant (t-test, p < 0.05).

The authors go on to say:

The observed interhemispheric warming asymmetry over the 20th century is more consistent with the models with weaker cloud feedback and aerosol indirect effect (B9) than those with more positive cloud feedback and aerosol indirect effect (T9).


In the paper’s discussion section the authors conclude:

Both more positive (high ECS) and less positive (low ECS) cloud feedback models are able to simulate the observed global-mean temperature record, but T9 models do it through a combination of strong warming from GHGs and strong cooling from aerosols, while B9 models do it with moderate warming from GHGs and modest cooling from aerosols. Because historical aerosol forcing has been larger in the Northern Hemisphere, the strong aerosol-cloud interactions cooling effect in T9 models produces a distinctive historical interhemispheric surface temperature evolution (red line in Figure 4b), which is inconsistent with that in observations over 1950–2000 (black line in Figure 4b). These results support the recent findings that the CMIP6 models more faithfully capture the observed evolution of surface anomalies across a range of quantities over 1980–2014 tend to have lower 21st century projected warming.

Only five out of the thirty CMIP6 models analysed are B9 models (those which are consistent with the evolving interhemispheric surface temperature contrast) that are also in the top 15 models for consistency with historical GMST warming.[9] Their range of estimated ECS is 1.81–2.86 K, with a mean of 2.5 K, far lower than the mean of 3.7 K for all the models. Those institutions that have developed CMIP6 models with ECS values comfortably above 4.5 are increasingly looking as if they may have taken a wrong turn somewhere. Somewhat surprisingly, they include several highly regarded modelling centres, such as NCAR and the UK Met Office.

Nicholas Lewis                                                                                      5 March 2021

[1] Wang, C., Soden, B. J., Yang, W., & Vecchi, G. A. (2021). Compensation between cloud feedback and aerosol-cloud interaction in CMIP6 models. Geophysical Research Letters, 48, e2020GL091024.

[2] Flynn CM, Mauritsen T. On the climate sensitivity and historical warming evolution in recent coupled model ensembles. Atmospheric Chemistry and Physics. 2020 Jul 6;20(13):7829-42.

[3] Zelinka MD, Myers TA, McCoy DT, Po-Chedley S, Caldwell PM, Ceppi P, Klein SA, Taylor KE. Causes of higher climate sensitivity in CMIP6 models. Geophysical Research Letters. 2020 Jan 16;47(1):e2019GL085782.

[4] Zhu J, Poulsen CJ, Otto-Bliesner BL. High climate sensitivity in CMIP6 model not supported by paleoclimate. Nature Climate Change. 2020 May;10(5):378-9.

[5] Kiehl JT. Twentieth century climate model response and climate sensitivity. Geophysical Research Letters. 2007 Nov;34(22).

[6] Rotstayn LD, Collier MA, Shindell DT, Boucher O. Why does aerosol forcing control historical global-mean surface temperature change in CMIP5 models?. Journal of Climate. 2015 Sep 1;28(17):6608-25.

[7] Their results are not sensitive to the number of models chosen for this purpose.

[8] Correlations are very similar when the model set is not restricted to one model per institution.

[9] CAMS-CSM1-0, FGOALS-g3, INM-CM4-8, MIROC-ES2L and MPI-ESM1-2-HR (MPI-ESM1-2-LR replaces MPI-ESM1-2-HR if no restriction to one model per institution is applied; the two model versions have almost identical estimated ECS).

Originally posted here, where a pdf copy is also available

45 responses to “Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models

  1. Thanks for this essay. I also read the study and made a comparison of the NH-SH (=interhemispheric difference -IHD) between the observations and the CMIP5 mean with the result: also the former model family CMIP5 suffered from the unreliable projection of the IHD. After carefully blending the tas (land) /tos data of the CMIP5 mean for both hemispheres this gives a similiar picture as it is shown in Fig. 2b:
    It points to an unrealistic strong response of the models to the aersol forcing, to balance out a a too high ECS in the historical period to match at least the observed GMST.
    This leaves us alone with the question: How useful are current models for the projections of the future GMST-deveopment when they match the present with the wrong reason??

    • Well it’s looking like these models are not skillful at predicting patterns of warming or patterns of much else. That’s not surprising given that the numerical truncation errors are orders of magnitude bigger than the small changes in energy flows that they attempt to model. This is absolutely not news to anyone in this field. What is still shocking is how assiduously and dishonestly these models have been defended by both scientists and the science “communicator” community. It’s another shocking example of institutional corruption.

    • Nic, I much appreciate your bringing a paper’s subject matter that is dear to my interests to the attention of Climate Etc. The paper, which I have not read at this point, evidently helps better explain what I have found with ΔRaer being the critical factor in putting higher sensitivity models more in line with the observed GMST for both CMIP5 and CMIP6 models.

      I have replied under frankclimate’s post because I agree with his points made in the last two sentences of his post and in addition that what we see as compensation for higher ECS in the historical period in CMIP6 models can also be seen in CMIP5 models.

      Critically needed for the historical (observed) period are modeled and observed data for aerosol forcing and ocean heat uptake (ΔRaer) with smaller confidence intervals. The CMIP5 and CMIP6 models respond in a consistent fashion over the historical and various scenario future periods to GHG forcing. It is the variation from model to model of ΔRaer and to a lesser extent ΔN that muddies the waters of model validation by way of GMST emulation. With tighter confidence intervals on ΔRaer and ΔN validation using the observed GMST would become more precise.

  2. Pingback: Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models – Climate-

  3. Rob Starkey

    When the models margin of error is so large how can they be realistically useful for government policymakers?
    An ecs in the 2’s means CO2 is no big deal, but esc over 4 means there could be ancillary issues.

  4. Iain Climie

    Very interesting but I’ll be boring and trot out my usual comment that there are win-win actions (e.g. reducing waste, combining conservation with careful use, increased food storage, silviculture, integrated methods, methane-reducing feed additives which can also boost growth, restoring fish stocks etc) which make sense regardless of the nature, extent, cause and direction of climate change, which address the risks of a volcanic winter (e.g. following the Tambora eruption of 1815) or if a major food crop collapsed. The repeated attempts to be right when we get it right seem perverse or maybe I’m just getting old (63 on March 8th) and grumpy.

    • Well I do think we should try to get it right or as right as its possible to be with ill-posed models. No regrets is good of course and vastly better than the current near insanity for “renewables” that rely on polluting technologies such as solar panels and batteries.

    • “Makes sense” makes no sense. For most things, if their being done makes economic sense, someone will do them.
      Some things that do not make economic sense may make sense for other reasons if their cost does not exceed their benefit. That calculation is often very difficult, even when politics and confirmation bias are not involved, and both almost always are regarding climate calculations.
      So other than vague, nebulous generalities what, if any, actual, reliable calculations support any of those proposals?

  5. I guess I simply don’t know how this field of chaotic CFD can progress given the current structure of institutional science. I doubt that current modeling technology can be pushed much further without fundamental mathematical advances with regard to dynamical systems and chaos. The funding incentives and the cultural norms of modern science are actively preventing people even trying to do this fundamental work. Most of the money is spent on ever more costly simulations and peddling myths such as that if we had a powerful enough computer these models would suddenly be skillful.

    I’ll boil this all down to a single simple and obvious observation. When numerical error is larger than the “signal” you are trying to find, any skill will be due to tuning causing cancellation of large errors that brings some outputs into alignment with the data. Skill on others measures is doubtful in this situation. That’s why models get average temperature roughly right. It’s the tuning to TOA radiation balance. But the patterns of change have no expectation of skill much less accuracty.

    • “That’s why models get average temperature roughly right.” is not correct. Models have average global temperatures varying by 2 to 3 C. Their “physics” differences are hidden by Gavin Schmidt’s anomalies. The model worlds’ physics are not the same as our world’s.

      • Yes I should have said temperature anomaly.

      • But the UN IPCC models only roughly agree among themselves on anomalies during the late-20th Century ‘tuning period.’ Earlier model hindcasts and future model projections of globally-averaged temperatures among the various models are all over the place. It is not “just the physics;” it is the political bias of the modelers. The modeler’s own statements that they “adjust parameters to get an ECS that seems about right” indicate that the models are a political statement, not a scientific statement.

      • I have to disagree. The models are a scientific statement in the loose sense of that term. It’s just that they are not good at patterns at all.

        If TOA radiation balance is correct due to tuning and ocean heat uptake is roughly right as it seems to be, then global average temperature anomaly will also be OK. I think that’s just due to conservation of energy.

  6. Iain, humankind is constantly performing “win-win” actions through the free-market capitalistic mechanisms that move us forward. Governments never provide for “win-win” solutions for complex problems. Ideology forces simple solutions for complex problems. Additionally, with ECS demonstratively below 2C/CO2 doubling, there is no serious threat from global warming.

    • Iain Climie

      Hi Dave, I have slightly mixed views on that. The problems that free markets face with food gluts or with the general concept of having enough can be disturbing. I recall my first economics lesson noting that there is no obvious upper limit on human wants despite the wit who (many years later) said “You can’t have everything, where would you put it?”.

      Having said that, planned economies are generally far, far worse; you might be interested in Frank Dikotter’s “Mao’s Great Famine” which documents the hellish effects of political dogma applied in China between 1958 and 1962. 10s of millions of people died in peacetime because (as you noted) the ideologues knew they were right. Local party officials couldn’t be seen to report under-production so exaggerated crop yields so, taking the results at face value, more food was used to feed those in cities or sold for export to fund economic development leaving those who had actually produced it short.

      • aporiac1960

        “The problems that free markets face with food gluts or with the general concept of having enough can be disturbing.”

        Not nearly so disturbing as mass starvation.

        The thing I find most disturbing is the almost complete detachment from reality of most people in advanced Western nations arising from having their basic needs met, and in most cases, greatly surpassed. The reduced sense of meaning such a condition engenders leads them to desperately attempt to find anything (whether is bears any discernible connection to their lives or not) that in any respect falls short of some ideal that never has existed, never will exist, and never could exist (because it is a mere construction of their own bored and foolish minds), so they can then go on to moan about it and turn it into a cause. And so they are full of talk of “problems,” despite not actually having any (except themselves, of course).

      • IAIN, I have no idea as to your point. Food gluts? Your “… the general concept of having enough can be disturbing.” is nonsensical. What are you trying to get at?

      • aporiac1960: yes, very interesting psychological observation.

      • Bill Fabrizio

        aporiac1960 … a well sketched social psychological profile. Now add search for meaning begets savior, savior begets movement, movement begets tyranny (or, if you prefer, inquisition).

  7. Reblogged this on Climate Collections.

  8. Richard Greene

    They are climate computer games, not models of the climate on this planet.

    Computers “predict” whatever their owners want to predict — they represent personal opinions of the future climate — their output is not real data.

    We have lived with global warming for the past 45 years. It has been mild and pleasant. Most affecting colder areas of the Northern Hemisphere. Mainly during the six coldest months of the year, and mainly at night.

    Actual PAST global warming was good news, but predicted FUTURE global warming is a climate emergency?

    I’d like an explanation of why that would happen, not just an assertion that it will happen … that I have been hearing every year for the past 50 years!

    When the average climate computer game predicts double the actual warming, and has not become more accurate over the past 30 years, we have politics, not real science.

    Your choice:
    (1) Actual experience with pleasant global warming since the 1970s, vs.
    (2) Climate computer game predictions of a horrible future climate emergency?

    My choice is (1) — reality.

    Climate computer games are nothing more than props used to spin a fairy tale of a coming climate crisis. And I am too old to believe fairy tales,

    • Richard, it has been warming over the past 300 years, but cooling over the past 6 millennia. The late 20th Century “blip” is lost in the noise.
      Fearmongering pays big dividends in academia and crony capitalism.

      • thecliffclavenoffinance

        It HAS been warming since the late 1600s, but people alive today have actual experience living with some, or all, of the global warming trend since the 1970s. It was nothing like the “existential crisis” warming always predicted for the future. Usually coming in ten years. … And in ten years, the imagined “crisis” will still be coming in ten years. And twenty years from now, the alleged climate crisis will still be coming in 10 years. (scaremongers prefer 10 years — they were VERY upset when John Kerry said 9 years).

        It warmed a lot from 20,000 years ago too.
        Obviously caused by all those pesky “snow machines burning FOSSIL FUELS that Canadians used to get around when Canada was entirely covered with an ice glacier. Or so I’ve heard.

    • Agree. There is no climate change emergency — just a climate education emergency. Since the woke crowd won’t read the actual data, maybe a video can help …

  9. Victor Ovid Adams

    To wit: WW II Pacific Fleet Admiral William F (Bull) Halsey when told by his chief meteorologist that long(er) term weather forecasts in the West Pacific theater are useless: “I know they are useless but I need them for planning purposes”. I take the liberty to add three more quotes from him for those who’d want to take a short brake from the business of preventing the coming Climate Armageddon.
    Hit hard, hit fast, hit often.
    There aren’t any great men. There are just great challenges that ordinary men like you and me are forced by circumstances to meet.
    I never trust a fighting man who doesn’t smoke or drink.

  10. “…. increasingly looking as if they may have taken a wrong turn somewhere.”

    By the end of this decade that will be a phrase that will have been increasingly used….. about a lot of things.

    I read several studies (some new, some new to me) last night that at a minimum increased uncertainty about projected SLR. Will save for appropriate post.

  11. <>

    The was Ken Arrow and the European Theatre.,he%20and%20his%20team%20could%20stop%20publishing%20them.
    Ken Arrow was a weather forecaster during World War II. But Arrow—who went on to win the Nobel Prize in economics—realized after a while that his team’s weather forecasts of European battlefields weren’t very good. He asked his superiors if he and his team could stop publishing them.

    The reply came: “The Commanding General is well aware that the forecasts are no good. However, he needs them for planning purposes.”

  12. I concluded some years ago that the climate models were mathematically unfit for purpose. The reason is very fundamental.
    The highest equator grid resolution in CMIP5 was about 110km IIRC. Typical was 280km. To avoid parameterization, the grid size needs to be less than 4km like regional weather models, still only reliable out a few days.
    NCAR says that the CFL constraint on numerical solutions to partial differentials means that halving grid size requires about 10x the computational effort. That means at least 7 more Orders of Magnitude computational effort is needed to avoid parameterization.

    So parameterization is unavoidable and will be for a very long time. The parameter tuning to best hindcast inescapably drags in the attribution problem. The warming from ~1920-1945 is indistinguishable (both visually and statistically) from the warming ~1975-2000. Yet even AR4 WG1 SPM Fig 4 said the former period was not anthropogenic induced—not enough rise in CO2. So models are tuned to the later period having some unknown natural component whilst presumed all Anthropogenic forcings. Which is why they all run hot.
    Nic’s post just shows how CMIP6 higher cloud feedback is tuned by higher aerosols to best hindcast, which then results in large hemispheric errors due to hemispheric aerosol asymmetry. It is still a fools errand.

  13. Renewables are useful for fossil fuel (coal, natural gas, oil) deposits saving for the future generations to come.
    There will be people on the planet in 10 millennia, in hundred millennia…
    I believe there will be humans on Earth in a hundred million years, and in a billion years.
    Shouldn’t we think of the future generations well being?

    Let’s save fossil fuel. Once burnt it never can be restored. It had taken billions of years to accumulate in the underground deposits.
    Fossil carbon will be in great demand very soon… So let’s not wasting it carelessly.

    • Hopefully a safer bet are things like genetically modified trees that take carbon out of the air much faster and give us a renewable source of hydrocarbons. I just read Freeman Dyson arguing that this is likely assuming science continues to progress.

  14. Ulric Lyons

    Weaker solar wind states since 1995, causing a warm AMO phase, driving a reduction in low cloud cover, could easily look like amplified CO2 warming, if one disregards the AMO.

  15. I have been a hydrodynamic modeler all my professional life. Clouds can be modelled at fine scale. A planetary scale with finite element methods would need millions more times computing power. I’m sure we will get there. All this to say that there is very little hope for numerical modelling accuracy.

    SO4 is a cloud condensation molecule. It is released naturally from oceans and is important in the vast regions of marine boundary layer stratocumulus. These clouds are blown hither and thither transporting moisture around the globe.

    Low level cloud cover in the subtropics and tropics varies with sea surface temperature. Easily noticeable in satellite data. Models are being trained as we speak. it becomes significant for multi-decadal climate variability in ocean upwelling regions.

    “Marine stratocumulus cloud decks forming over dark, subtropical oceans are regarded as the reflectors of the atmosphere.1 The decks of low clouds 1000s of km in scale reflect back to space a significant portion of the direct solar radiation and therefore dramatically increase the local albedo of areas otherwise characterized by dark oceans below.2,3 This cloud system has been shown to have two stable states: open and closed cells. Closed cell cloud systems have high cloud fraction and are usually shallower, while open cells have low cloud fraction and form thicker clouds mostly over the convective cell walls and therefore have a smaller domain average albedo.4–6 Closed cells tend to be associated with the eastern part of the subtropical oceans, forming over cold water (upwelling areas) and within a low, stable atmospheric marine boundary layer (MBL), while open cells tend to form over warmer water with a deeper MBL. Nevertheless, both states can coexist for a wide range of environmental conditions.5,7”

  16. Pingback: Weekly Climate and Energy News Roundup #445 – Watts Up With That?

  17. Thanks, Nic, as usual most interesting.

    You say:

    The spread in estimated ECS has increased further in CMIP6 models; it is 1.8–5.5 K as compared with 2.0–4.7 K in the previous, CMIP5, generation

    As I’ve said before, this 40-year-plus inability to narrow the uncertainty of “climate sensitivity”, the central and most important number of the current climate paradigm, is clear evidence that the underlying theory that changes in global temperature are a linear function of changes in forcing is simply wrong.


  18. Thanks, Willis. You could be right, however in all or almost all the GCMs global temperature changes are pretty much a linear function of forcing, at least over a GMST range from somewhat below preindustrial to 23 C or so.

    The main cause of the wide, and widening, spread in GCM’s ECS values appears to be their difficulty in properly parameterizing and/or directly simulating cloud (particularly low cloud) behaviour.

    • Nic, if it were just GCMs that are not reducing uncertainty in the ECS, you’d be right. But the same is true of ECS estimates from theory and reviews, observations, paleoclimate, and climatology.

      Taken as a whole, to me this increasing ECS uncertainty despite spending millions of computer cycles, person-hours, and dollars on the question over the last forty years is compelling evidence of something very wrong with the underlying theory.

      Again, thanks for a most interesting post.


    • I would argue that the underlying theory is probably OK since its really the Navier-Stokes equations which have been vetted very thoroughly over 100 years. However, we understand very little about these equations in terms of long term predictability and computability. There is at least one attractor generally but its properties are mostly unknown. If the attractor is weak (if its not very attractive) then computability is probably not going to be viable. We just need to get people to start addressing the deep theoretical and mathematical issues and take some deep career risks. Right now, it’s quite depressing actually. Virtually all the money is spent on building additional codes that implement the same old methods and running those codes more and more frequently. The result is just confusion and the promotion of the selection biases which are already deeply entrenched in the field of CFD.

      One can usefully consider much simpler problems than the atmosphere and oceans such as an airplane wing with a uniform onset flow free of turbulence and precipitation. Even this problem seems to be showing up the same high uncertainty and variability in results in the high Reynold’s case when separated flow is present. The same is true when there are large scale vortical structures present (this is the case in the atmosphere for example). Basically, the field has been shafted by its own marketing success. Everyone oversells their codes and results to keep funding alive. So those who fund the field have come to believe CFD is a solved problem. That is a very pernicious cultural narrative because it’s patently false. We as a field need to stop the dysfunctional and harmful arms race of “correct” results and positive findings and start showing the high variability of even simple simulations to parameter choices, model constants, etc. There are hundreds of these choices for even a simple RANS simulation.

      I have a long post prepared with references that I’m trying to get released right now. Suffice it to say knowing what I know now after 40 years of work, it is deeply dishonest for climate scientists to portray GCM’s as anything more than costly tuning machines for turning billions of dollars into skill only on those outputs used in tuning or closely related to them. Thus since TOA radiation balance is tuned to match data and if ocean heat uptake is about right as it seems to be, then by conservation of energy, global average temperature anomaly will be roughly right. But of course the pattern of warming (since its not used in tuning) will have no expectation of skill.

      The problem here is that numerical truncation errors are orders of magnitude greater than the small changes in energy flows that we care about. And the simple problems I mentioned above show that the Palmer and Stevens project to use massive computing power to resolve smaller and smaller eddies is not a guarantee of anything. This method for the simple cases I mentioned is still in its infancy. These methods seem to be selection bias machines. You have thousands of parameters. You run the code until you get a convincing result for one case and then give a sigh of relief and publish, failing to provide the hundreds of failed attempts that are where the real scientific information lies. It’s a field in need of deep reform. Senior researchers need to stand up and tell the truth. Some do so (mostly) such as Slingo and Palmer generally, even though their work does leave the impression that we can definitely predict the climate of the attractor, which is at the moment a speculative idea.

      People generally recognize that medicine is a field with deep structural issues. Medicine as actually practiced in the US is not evidence based and I believe most of the money spent makes little difference. As I’ve gotten older and tried to improve my health, I have been depressed by how little modern medicine can do to prevent or reverse heart disease. Basically statins help only a very little. You can lower your blood pressure and get in excellent aerobic condition, but that’s about it. Virtually all men will eventually have arteriosclerosis. My brother has struggled with his physician staff for decades to get them to restrict themselves to treatments that have strong evidence of efficacy. You can never succeed, only maintain nominal control over the problem by constant vigilance. What I’ve realized over the years is that CFD is no better.

  19. Pingback: Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models – Watts Up With That?

  20. Pingback: Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models |

  21. Pingback: Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models – Climate-

  22. Alongside the important question of how long you need to average properties to consider a change in climate is another question :- how long is a generation of climate science?
    The Wang et al paper is an affirmation of many of the findings of Kiehl et al 2007. The last 13 years should have seen the intelligent use of Kiehl’s information and insights into the production of a class of more credible models. Instead, the more pessimistic models become more and more divergent from observational data and more defiant of Kiehl’s findings.
    I could deprecate the Wang et al paper and perhaps accuse it of “Kiehl-hauling”, except that it is an obviously necessary and timely reminder to the climate-modeling community.

    • Steve Fitzpatrick

      Kiehl-hauling ….. got to love it. ;-)

      But more seriously, the continuing and ever growing divergence of most climate models from measured reality is simply a reflection of climate ‘science’ not being anything like what most people would consider to be science. There appears no motivation for modelers to change their models to generate results consistent with measured reality. It is more like a religion, not a science.

  23. Pingback: Compensation between cloud feedback + ECS and aerosol-cloud forcing in CMIP6 models