I’ve Looked at Clouds from Both Sides Now -and Before

October 8th, 2011 by Roy W. Spencer, Ph. D.

…sometimes, the most powerful evidence is right in front of your face…..

I never dreamed that anyone would dispute the claim that cloud changes can cause “cloud radiative forcing” of the climate system, in addition to their role as responding to surface temperature changes (“cloud radiative feedback”). (NOTE: “Cloud radiative forcing” traditionally has multiple meanings. Caveat emptor.)

But that’s exactly what has happened. Andy Dessler’s 2010 and 2011 papers have claimed, both implicitly and explicitly, that in the context of climate, with very few exceptions, cloud changes must be the result of temperature change only.

Shortly after we became aware of Andy’s latest paper, which finally appeared in GRL on October 1, I realized the most obvious and most powerful evidence of the existence of cloud radiative forcing was staring us in the face. We had actually alluded to this in our previous papers, but there are so many ways to approach the issue that it’s easy to get sidetracked by details, and forget about the Big Picture.

Well, the following graph is the Big Picture. It shows the 3-month variations in CERES-measured global radiative energy balance (which Dessler agrees is made up of forcing and feedback), and it also shows an estimate of the radiative feedback alone using HadCRUT3 global temperature anomalies, assuming a feedback parameter (λ) of 2 Watts per sq. meter per deg (click for full-size version):

What this graph shows is very simple, but also very powerful: The radiative variations CERES measures look nothing like what the radiative feedback should look like. You can put in any feedback parameter you want (the IPCC models range from 0.91 to 1.87…I think it could be more like 3 to 6 in the real climate system), and you will come to the same conclusion.

And if CERES is measuring something very different from radiative feedback, it must — by definition — be radiative forcing (for the detail-oriented folks, forcing = Net + feedback…where Net is very close to the negative of [LW+SW]).

The above chart makes it clear that radiative feedback is only a small portion of what CERES measures. There is no way around this conclusion.

Now, our 3 previous papers on this subject have dealt with trying to understand the extent to which this large radiative forcing signal (or whatever you want to call it) corrupts the diagnosis of feedback. That such radiative forcing exists seemed to me to be beyond dispute. Apparently, it wasn’t. Dessler (2011) tries to make the case that the radiative variations measured by CERES are not enough energy to change the temperature of the ocean mixed layer…but that is a separate issue; the issue addressed by our previous 3 papers is the extent to which radiative forcing masks radiative feedback. [For those interested, over the same period of record (April 2000 through June 2010) the standard deviation of the Levitus-observed 3-month changes in temperature with time of the upper 200 meters of the global oceans corresponds to 2.5 Watts per sq. meter]

I just wanted to put this evidence out there for people to see and understand in advance. It will be indeed part of our response to Dessler 2011, but Danny Braswell and I have so many things to say about that paper, it’s going to take time to address all of the ways in which (we think) Dessler is wrong, misused our model, and misrepresented our position.

33 Responses to “I’ve Looked at Clouds from Both Sides Now -and Before”

Toggle Trackbacks

  1. mdmill says:

    Well, Dr Dessler is apparently (implicitly) conceding that if a large radiative “source” is independent of surface temperature and IS a significant component of surface temp change, then the validity of his results are in question.
    (You can score a point if you wish!- he would probably reply that he never disputed that particular point theoretically)
    But certainly he is arguing that such a large source does not and cannot exist, according to his argument(I have not read the final paper yet.); and therefore it need not be accounted for. And whether or not this is a separate point, it is certainly the key point now. We are all ultimately interested in the true situation, regardless of past disagreements .
    So please be clear Dr Spencer; Are you disputing that key assertion and those SPECIFIC arguments and data having read his published version? and why? Are you arguments similar to your recent blog on the subject?
    Thanks…Sincerely M D Mill

    • You are correct….Dessler concedes the theoretical possibility that cloud radiative forcing *could* obscure cloud feedback, but then goes on to explain why he thinks cloud radiative forcing is too weak to do so.

      What I have posted here is consistent with my previous posts.

      • …oh, and if you read his new paper, he seems to have a typo early in the paper…he claims we do NOT believe cloud radiative forcing causes temperature change. Of course, the opposite is the case. I have no idea how such a statement could have made it through 2 peer reviews.

      • mdmill says:

        It occurs to me that Dr Dessler is implicitly assuming that the outward power flux at the top of atmosphere equals the radiative plus non radiative outward flux at the ocean surface; ie that there are no significant energy storage fluctuations in the atmosphere (over 3 month periods).
        …But might not energy storage fluctuations in the atmosphere, especially vapor-latent energy fluctuations, over 3 month periods, be significant enough to invalidate this assumption and his results? Also the EFFECTIVE mix depth over 3 month periods, may be much less than he is assuming…as little as 30 meters(ie the “mix layer” does not act like a single lumped element in smaller time frames- I know you have already suggested this)… I have not worked out the details of this, but these are some lines of questioning submitted for your consideration.

        I have derived a simple equation that computes the surface temp time derivative. It involves both the absolute ocean current speed(very important) and the ocean SURFACE power flux time derivative(radiative plus non-radiative), an effective mix depth, and the SPACIAL GRADIENT OF TEMP in the direction of flow . The velocity term incorporates the component of temp changes that occur due to ocean currents transferring energy around the ocean[note: if current speed and flux derivatives are both zero then the temp time derivative must be zero also…ie no change in temp]. All of these terms are available from measurements(ie NCEP/UCAR- satellite reanalysis project data, float drift records, etc.)which allows a calculation of the effective mix depth.
        I picked a spot in the pacific ocean where the current seemed relatively constant in direction(westward) and magnitude(.12M/sec)at 12S, 130W, and using available data I calculated an EFFECTIVE mix depth of 30m. This equation then indicated that a change of surface power flux of 0.4W/m^2 would result in a change in the temperature time derivative of 0.1C/year surface temperature change(comparable(?) to observed non-seasonal variations). This is certainly much different than Dr. Dessler concludes is his paper.
        This is only a rough preliminary result in one spot in the ocean; but I conclude that atmospheric conditions(including evaporation rates) can CAUSE significant ocean temp changes, as opposed to ONLY “heat transport by the ocean” as Dr. Dessler writes…that’s my opinion for now, for what it’s worth.
        I hope these thoughts are of interest.

        Sincerely, M D Mill

  2. Kim says:

    Following this thread.

  3. Dallas says:

    Actually Dr. Spencer, the cloud relationship is probably due to conductivity change at the oceans surface air interface, the coefficient of heat transfer are hugely different, 1000/1 is my guess, so only a same change in conductivity can create a much larger impact. Neat relationship, I have yet to prove.

  4. Joletaxi says:

    Hello Mr.Spencer

    I think we have a long way to go before we understand the mechanism of cloud formation.
    Just yesterday, I watched the thick layer clouds under the plane, and something that had never realized before, I’m surprised that they remained confined to a certain altitude?
    After all, the atmosphere is constantly changing and is driven by bottom-up, turbulence. Why the cloud remains confined to a certain altitude?
    On the other hand, I see sometimes mentioned a correlation between solar activity and seismic and volcanic on earth.
    Have you considered this? Strong volcanic activity inevitably has an influence on the temperatures on earth, probably also on cloud cover?
    In any case, the pattern more simplistic water vapor causes more clouds seems unrealistic.
    And deny the overwhelming influence of clouds on the temperature is acting in bad faith.
    Are you aware off a study on the correlation volcanism, solar activity.


  5. Stephen Wilde says:

    Looking at the big picture since the 1950s with two major shifts at around 1975 and 2000 I think the primary cause of changes in the amount of solar energy getting into the oceans is variability in the behaviour of the global surface air pressure systems under combined solar and oceanic influences.

    That makes the jet streams either:

    i) Move towards the poles and/or become more zonal for reduced global cloudiness.

    ii) Move towards the equator and/or become more meridional for increased global cloudiness.

    That is how the system input varies to the extent observed.

    The potential feedbacks are effectively overwhelmed by those changes in system forcing.

  6. Thanks Dr. Spencer. You have made very clear to me that although clouds are a response to surface temperature, they are more of a cause of surface temperature.

  7. P. Solar says:

    Indeed. Perhaps the logical complement to this graph would be the same CERES data compared to the rate of change of temperature.

    One can see it to some extent in the present graph. Since lambda is a constant , the blue line is effectively a representation of surface temp.

    What can be noted is that (with some lesser deviations) positive energy balance gives a positive SLOPE and vice versa. This is to be expected since radiative forcing produces not a temperature change but a RATE of temperature change.

    Plotting dT/dt in an accompanying graph would probably demonstrate that there is a much closer correlation with the rate of change of temp (ie radiative forcing) than with temperature itself (which relates to the feedback).

  8. Buck Smith says:

    I am not clear if cloud radiative forcing is due to longwave IR radiated form clouds or clouds blocking and unblocking incoming solar radiation or the sume of those two. Does it include water state changes back and forth form vapor, droplet, ice in the clouds?

    Thanks for any help on this.

  9. Buck Smith says:

    sorry for the typos, should read

    I am not clear if cloud radiative forcing is due to longwave IR radiated from clouds or clouds blocking and unblocking incoming solar radiation or the sum of those two. Does it include water state changes back and forth from vapor, droplet, ice in the clouds?

    Thanks for any help on this

  10. The problem with Dessler ,and crew is they ignore past data as well as present data.

    They won’t accept any data that goes counter to their (incorrect) thoughts.

    They don’t seem to believe in all the temperature changes (sometimes abrupt ) that have happened throughout earth’s history.

    Mann’s famous hockey BS is a great example.

  11. Christopher Game says:

    Dr Spencer writes: “NOTE: “Cloud radiative forcing” traditionally has multiple meanings. Caveat emptor.”

    Surely this is not fair dealing. It is the duty of the writer to make himself clear, not the duty of the reader to unscramble the writer’s ambiguities or idiosyncratic terminology. Christopher Game

  12. Christopher Game says:

    Dr Spencer continues to let the IPCC dictate the terms of the debate, to the loss of science.

    Implicitly accepting the vile IPCC’s ridiculous “forcings and feedbacks” formalism, Dr Spencer writes as if he thought that there is a one-and-only true definition of the distinction between forcings and feedbacks. Properly, the distinction has to depend on the model.

    Cloud forcing and feedback from land-sea surface temperature cannot be properly analyzed in terms of the the IPCC’s ridiculous “forcings and feedbacks” formalism, because that formalism has only one dynamically determined internal state variable, when it is desired to distinguish effects which need to be expressed by two dynamically determined internal state variables, temperature and cloud extent. The IPCC formalism ridiculously puts all feedback into a single term which is merely a timeless mathematical function of the land-sea surface temperature and does not distinguish dynamic differences between for example water vapour and cloud effects. That makes it a first-order formalism, when a second-order model is neeed for the present purposes.

    While it is perhaps true that every cloud has its silver lining, it is also true that you can’t make a silk purse out of a sow’s ear. Christopher Game

  13. If clouds cause negative feedback, we would expect to see more visible spectrum radiation reflected from the earth following warm periods, less following cold periods.

    The data I would expect to see is the convolution over past temperatures that best predicts present radiation from the earth as a function of past temperatures over the preceding days and months.

    It would be possible to read off feedback from the convolution.

  14. MikeN says:

    How does the Pacific thermostat theory relate?
    I heard Mann say that he thought even more than clouds, the modeling of ENSO is the biggest uncertainty.

  15. P. Solar says:

    Dessler (and others) do an ols regression on data with significant noise/error in the independent variable. It is a mathematical result that the ols estimator WILL be less than the true “slope”. How much less depends on the amount of noise. The data has more noise than signal. The regression attenuation will be very significant.

    Since the ols value is less than the neutral 3.3 W/m2/K of the plank response he comes to the erroneous conclusion that he has demonstrated positive feedback.

    If you want to dismiss the “positive feedback” results of Dessler, F&G et al this is the easiest and most incontrovertible route. Since the results of this flawed method are the only empirical justification for positive feedback in models and that is the only reason the models produce alarming warming this seems to me to be the crux of the whole AWG battle.

    While it would be good to be able to prove a “correct” result this is not necessary to refute the claimed result of a positive feedback.

  16. Steve says:


    Has Dessler told the clouds there not meant to causing radiative forcing?

  17. Lewis Guignard says:

    On a different note: what is the long curve on the temperature anomaly graph based on? I can’t seem to find that. Thank you.

  18. Pehr Bjornbom says:

    Dr. Spencer, the changes in radiative forcing are also obvious from the phase plane plots introduced by you and Dr. Braswell.

    If the radiative flux is R(t) = N(t) – ??T(t) and we are looking at two times with the same temperature anomaly we obtain

    R(t2) – R(t1) = N(t2) – N(t1) – ?(?T(t1) – ?T(t2)) = N(t2) – N(t1)

    since ?T(t1) = ?T(t2).

    Thus the change in radiative flux between those two times is a pure change in radiative forcing.

    Look for example at figure 4a in Spencer and Braswell (2010) in JGR. The difference in radiative flux between Dec 07 and Sep 08 is about 1.2 W/m2. Since the temperature anomaly is the same at both times this shows a pure change in radiative forcing amounting to 1.2 W/m2.

  19. KR says:

    Dr. Spencer – If clouds change in response to temperature, they are _feedbacks_, if they change for some reason unrelated to temperature, they are _forcings_. I am most certainly not aware of multiple definitions outside of attempts to redefine the issue for some point of view.

    You have presented no evidence, no physical model, supporting cloud radiative changes separate from temperature. Water vapor levels respond extremely quickly to temperature, convection responds to temperature – without a temperature independent physical mechanism for cloud changes _they are a feedback_.

    Secondly, I’ll note that monthly or even yearly changes to cloudiness are far too short a time frame to correlate to climate changes – the general estimate is 30 years, I would prefer >40 based on the leveling off of the temperature anomaly standard deviation, and a short time period is completely insufficient time to separate internal climate variability from long term trends.

    I’ll note in passing that cosmic rays do not appear to correlate to temperature changes – at best all estimates of GCR’s, solar magnetic fields, and other esoteric changes on cloudiness are, at best, quite weak effects.

  20. JohnB says:

    KR, a physical mechanism in one word – “Currents”.

    If you accept that ocean temps effect cloudiness then warm and cold currents effect cloudiness. From a big picture perspective, the rise in sea levels since the Glacial Maximum must have resulted in widespread changes in ocean currents. This is obvious.

    In fine, the sea floor is not static and must be constantly changing, most of these changes are small and slow but others (like when an island is born) are fast and make significant changes to currents in a short space of time. However the principle remains the same.

    The effect of ocean currents on clouds can only be static if the currents are static. The currents can only be static if the sea levels and sea floor remain static. Since these things do not remain static, then currents must change over time and therefore their effect on clouds must change.

    The effect must be there, but how large or small it is and how to quantify it I have no idea. Regardless, it is something that will change clouds independently of temperature and therefore answers your question re “forcing”.

  21. C S Mill says:

    Although I’m something of a time-expired cloud physicist (I left the Manchester APRG in 1995) I find it unbelievable that people try to argue that any changes in the net thermal effects of the total cloud cover of the Earth are entirely driven by temperature changes – especially when considering periods of the order 50 years as we must when looking at climatic changes.

    My particular interest was in cloud microphysics and I have to admit that our understanding of the microphysical processes in clouds remains woefully short of a quantitative, predictive science. You mention cloud condensation nucleus concentrations as just one of the factors that can have an influence, and with this I, and many other would concur. The effect of relatively minor changes in CCN on the net thermal effect of clouds may well be profound (see for example Latham, J., 1990: Control of global warming? Nature 347. 339-340 ). By influencing the initial cloud droplet spectra they can subsequently influence the cloud lifetime and the optical properties of the cloud during that time. Given the diverse processes – some of them biological – involved in CCN production and transport it seems highly unlikely that temperature change would be the only driver for changes in CCN over long timescales.

    Unfortunately, far from having a reliable global picture of CCN concentrations going back 50 years we have, at best, a few dubious spot-measurements of CCN (my own experience of CCN counters, as with some other cloud physics instruments, is that their output depends as much on the serial number on the back as it does on the air sample going in at the front).


    Earth’s climatic system is in a delicate balance, yet it is not easy to change the balance in that the climate always wants to try to revert back to the way it is, and yet we know from past history, that this is not always the case.

  23. Dallas says:

    Doc, When I compare the mid-trop to strat it seems to be pretty interesting.

  24. Jos Hagelaars says:

    Dr. Spencer,
    On WUWT I put forward a question, it is one of many and probably overlooked. Therefore I repeat the question on your own blog:
    It looks to me that the graph you show here is a combination of the graphs 1A and 1C in Dessler’s 2010 paper, in your graph delta-Ts is just multiplied by the constant lambda. Dessler says about this delta-Ts: “Also plotted is an ENSO index (23), and the close association between that and delta-Ts verifies that ENSO is the primary source of variations in delta-Ts.” and
    “Obviously, the correlation between delta-Rcloud and delta Ts is weak (r2 = 2%), meaning that factors other than Ts are important in regulating delta-Rcloud.”
    It seems to me that you’ve reached the same conclusion as Dessler, in that there is a low correlation between the radiative forcing at the TOA and the delta-Ts. Am I correct ?”

  25. KR says:

    JohnB – “Currents”

    That’s an excellent point, JohnB, currents do affect heat transfer, distribution of relative humidity, etc. Aerosols are also implicated, as in the ‘global dimming’ seen due to aerosol effects on cloud seeding/cover.

    But the paleo data, including that covering times when (for example) the continents were in different locations – every study I know of shows similar global climate sensitivities (on the order of 3C/doubling of CO2, or more properly 0.7-0.8C/W/m^2 forcing change), despite radically different ocean currents.

    I see ocean currents as affecting regional heat distribution much more than global mean temperatures, for example the Gulf Stream keeping Northern Europe much warmer than it would be otherwise.

    But I quite frankly do not know of any strong evidence to suspect currents as a significant global forcing.

    • Troy_CA says:


      Within the context of these papers (the back and forth between Spencer and Dessler), the monthly and yearly variations of the cloud forcing _IS_ the topic of importance. SB08 was critiquing the method of Forster and Gregory 06, which determines climate sensitivity using seasonal and yearly (and later in Murphy et al 09, monthly) measurements of TOA flux and temperature changes. The TOA flux observed will be a combination of forcing and feedback, but the basic FG06 method will assume that there is no cloud forcing in the signal, so the leftover TOA flux can be considered feedback. SB08 and SB11 point out that IF there is forcing in the measured TOA flux that is counted towards the feedback, this will lead to a confounding of the signal and a significant underestimate of the feedback (and hence an overestimate of climate sensitivity, its inverse).

      So, whether clouds in the long-term can induce climate change is irrelevant to this particular issue, where the focus is indeed on whether clouds can be a “forcing” on the seasonal/yearly scale used for the regressions — that is, can they vary for reasons quite apart from temperature? As Jos Hagelaars points out above, Dessler 2010 essentially agrees that they can: “Obviously, the correlation between DRcloud and DTs is weak (r^2=2%), meaning that factors other than Ts are important in regulating DRcloud.”

      To your original comment, then, yes, clouds can act as a forcing on the timescales relevant to these analyses. The next question is whether they can do so on a magnitude that would lead to a significant underestimate of the feedback, or whether they have only a minor effect. Dessler 2011 mistakenly used ARGO down to 700 m (rather than the 50-100 m mixed layer) to support his notion that the OHC fluxes were much greater than any possible cloud forcing, and compounded this by failing to look at the _yearly_ average fluctations or uncertainty data, which suggested that much of what he considered to be “ocean forcing” was actually noise.

  26. JohnB says:

    Thanks KR, it’s an idea I’ve been kicking around. The closing of the Panama gap would have radically changed current patterns and should have radically changed the climate. So why didn’t it? The answer is deceptively simple, forcings aren’t all that important.

    The most common paleo temp reconstruction is Scotese and is often graphed with the CO2 from Berner to give the graph found here;

    When we compare the temps to forcings, there is zero correlation. It doesn’t matter that the Sun warmed, or that the CO2 was way higher or that the continents went walkabout all over the globe changing the currents.

    Forcings don’t matter, the feedbacks do.

    It would seem that the climate “equation” (which I think of as similar to the Drake equation except with interwoven parameters) is preprogrammed to produce one of two answers. So long as the forcings stay within certain limits it doesn’t matter how much they change as the feedbacks will automatically correct for forcing changes. It also means that within those limits, feedbacks are always nett negative.

  27. Stephen Wilde says:

    “So long as the forcings stay within certain limits it doesn’t matter how much they change as the feedbacks will automatically correct for forcing changes. It also means that within those limits, feedbacks are always nett negative”


    And the mechanism is the speed and/or size of the water cycle via changes in the surface pressure distribution and the sizes, position and intensities of the permanent climate zones.

    The surface pressure distribution changes so much as a result of solar and oceanic variability that the CO2 effect is not discernible.

  28. KR says:

    JohnB – “When we compare the temps to forcings, there is zero correlation”

    I would have to completely disagree. That particular graph is quite interesting, but since CO2 is not the _only_ forcing, it’s a bit limited. And misleading.

    The sun was rather dimmer in the past – as per any G class star, it brightens over it’s life as the internal fuel mix changes. If you combine the lower solar forcing with the higher CO2 forcing in the distant past, it _does_ correlate with temperatures, as per the “faint young sun paradox”: http://en.wikipedia.org/wiki/Faint_young_Sun_paradox

    Climate feedbacks have kept the Earth somewhere between frozen and boiling, which is why historically CO2 (as a feedback) has combined with insolation to keep the Earth within a certain range.

    Just don’t forget the _response time_ for those CO2 feedbacks (such as ocean and mineral sequestration). They are really slow on a human timescale, and may not save us from our own actions. Our emissions are changing things hugely faster than geologic feedbacks of the past.

  29. Pinkie Catalan says: