RealClimate logo

On Sensitivity Part II: Constraining Cloud Feedback without Cloud Observations

Filed under: — group @ 4 January 2013

Guest Commentary by Karen M. Shell, Oregon State University

Link to Part I.

Clouds are very pesky for climate scientists. Due to their high spatial and temporal variability, as well as the many processes involved in cloud droplet formation, clouds are difficult to model. Furthermore, clouds have competing effects on solar and terrestrial radiation. Increases in clouds increase reflected sunlight (a cooling effect) but also increase the greenhouse effect (a warming effect). The net effect of clouds at a given location depends the kind of clouds (stratus, cumulus etc.), their distribution in the vertical and on which radiative effect dominates.

Not only is it difficult to correctly represent clouds in climate models, but estimating how clouds and their radiative effects will change with global warming (i.e., the cloud feedback) is very difficult. Other physical feedbacks have more obvious links between temperature and the climate variable. For example, we expect and have strong evidence for the increase in water vapor in a warmer climate due to the increased saturation specific humidity, or the reduced reflection of sunlight due to the melting of snow and ice at higher temperature. However, there isn’t a simple thermodynamic relationship between temperature and cloud amount, and the complexities in the radiative impacts of clouds mean that an increase in clouds in one location may result in net heating, but would correspond to a cooling elsewhere. Thus, most of the uncertainty in the response of climate models to increases in CO2 is due to the uncertainty of the cloud feedback.

So how cool is it then that the recent paper by Fasullo and Trenberth estimates the net climate sensitivity without getting into the details of the cloud feedback then? Quite cool.

Figure 1 from Soden and Held (2006) showing ranges for each model for each of the key atmospheric feedbacks for the CMIP3 models.

The complexity of the cloud feedback means that it is difficult to observe and evaluate. Ideally, we would have accurate global observations of clouds for a number of decades to compare to the climate models. However, without that, we have to estimate cloud feedbacks using the shorter high-quality observational record, and then try to relate the short-term behavior to long-term (century-scale) cloud feedbacks. For some feedbacks, such as the water vapor and ice-albedo feedbacks, there has been some success in evaluating models directly using observations. For example, the modeled water vapor feedback in response to the eruption of Pinatubo agrees with the satellite data (Soden et al., 2002), while modeled Northern Hemisphere (NH) snow and sea-ice feedbacks underestimate the observed feedback (Flanner et al., 2008). Evaluation of cloud feedbacks has been less successful, due to the difficulty of obtaining homogenous global cloud properties from either satellite or surface-based observations. Many assumptions (such as for cloud droplet size or vertical distribution) must be made for these retrievals, and instrument and calibration errors may still be significant. Additionally, there isn’t an obvious correspondence of the short-term cloud behavior with the long-term behavior in climate models (Dessler, 2010; Masters, 2012).

Correlation of zonal mean relative humidity against sensitivity, and the definition of two key regions.

Correlation between the two mid-tropospheric regions and model sensitivities.

The Fasullo and Trenberth paper identified a relationship between the modeled seasonal change in relative humidity in the subtropical dry zones (the downwelling branch of the Hadley circulation, centered around 20-30°N and S) and the long-term feedback behavior of clouds in models. This is a very promising methodology because, if the relationship holds, we could evaluate climate models using observations of the seasonal cycle of relative humidity (which are much easier to obtain than cloud measurements). We don’t actually have to observe clouds at all! Fasullo and Trenberth use satellite data to estimate the present-day (1980-1990) May through August relative humidity and find that the CMIP3 models that best match the observations have strong moist zones in the tropical lower troposphere, strong dry zones in the subtropical upper troposphere, and high climate sensitivities. Thus, Fasullo and Trenberth conclude that the relative humidity observations are most consistent with higher climate sensitivities (around 4°C for a doubling of CO2).

One piece missing in this work is the direct link of the RH observations to the cloud feedback. Since clouds form in saturated air, the lower the RH, the lower the cloud amount (broadly speaking), and the lower the planetary albedo (since less sunlight is reflected to space). While Fasullo and Trenberth don’t specifically calculate cloud feedbacks, their figure 3 relates the climate sensitivity to changes in solar fluxes at the top-of-the-atmosphere (TOA) in the subtropics. Outside of the polar regions, clouds are really the only things which could be changing the TOA fluxes this much, but it would be nice to confirm this by comparing the northern hemisphere summer RH in the dry zones to cloud feedbacks specifically.

Another issue is that it is not clear how exactly to improve modeled RH. Subtle model details influence the emergent dynamics of the system. There’s not a single “knob” that can be tuned to influence the RH seasonal cycle, and models might not correctly capture the dependence of cloud properties on RH. Alternately, RH and clouds could be responding to something else that is controlling both processes. Even if models are capturing the response of subtropical clouds to climate change, other feedbacks (water vapor, temperature, snow and sea ice, or high-latitude cloud feedbacks) may not be related to subtropical RH or may behave differently for the seasonal cycle compared with climate change. Finally, the seasonal cycle omits cloud changes in response to CO2 directly (Gregory and Webb, 2008; Colman and Mcavaney, 2011), which will influence climate change. Nevertheless, this new simple diagnostic is an encouraging step linking observations to climate sensitivity.


  1. J.T. Fasullo, and K.E. Trenberth, "A Less Cloudy Future: The Role of Subtropical Subsidence in Climate Sensitivity", Science, vol. 338, pp. 792-794, 2012.
  2. B.J. Soden, and I.M. Held, "An Assessment of Climate Feedbacks in Coupled Ocean–Atmosphere Models", Journal of Climate, vol. 19, pp. 3354-3360, 2006.
  3. B.J. Soden, "Global Cooling After the Eruption of Mount Pinatubo: A Test of Climate Feedback by Water Vapor", Science, vol. 296, pp. 727-730, 2002.
  4. M.G. Flanner, K.M. Shell, M. Barlage, D.K. Perovich, and M.A. Tschudi, "Radiative forcing and albedo feedback from the Northern Hemisphere cryosphere between 1979 and 2008", Nature Geoscience, vol. 4, pp. 151-155, 2011.
  5. A.E. Dessler, "A Determination of the Cloud Feedback from Climate Variations over the Past Decade", Science, vol. 330, pp. 1523-1527, 2010.
  6. T. Masters, "On the determination of the global cloud feedback from satellite measurements", Earth Syst. Dynam., vol. 3, pp. 97-107, 2012.
  7. J. Gregory, and M. Webb, " Tropospheric Adjustment Induces a Cloud Component in CO 2 Forcing ", Journal of Climate, vol. 21, pp. 58-71, 2008.
  8. R.A. Colman, and B.J. McAvaney, "On tropospheric adjustment to forcing and climate feedbacks", Clim Dyn, vol. 36, pp. 1649-1658, 2011.

78 Responses to “On Sensitivity Part II: Constraining Cloud Feedback without Cloud Observations”

  1. 51
    Ed Barbar says:

    So clouds effect climate sensitivity, and no one has a clue in what direction? That’s amazing.

    Let’s all hope it’s in the positive direction.

  2. 52

    #51–“So clouds effect climate sensitivity, and no one has a clue in what direction?”

    No. There are clues. What is lacking is ‘certainty,’ or (better stated) much tighter quantification.

    “Let’s all hope it’s in the positive direction.”

    Personally, I’m hoping it’s in the negative direction (i.e., lower CS.) I’m guessing that’s what you meant?

  3. 53
    Dan H. says:

    Most scientists acknowledge that clouds have a dual-acting mechanism; blocking incoming solar radiation, and retaining terrestrial radiation. The general view is that the blockage of incoming radiation is greater, resulting in a net temperature reduction. Also, different cloud types have a larger impact; high, cirrus clouds block less solar and more terrestrial, while low, stratus block more solar and less terrestrial.
    The bigger question is, to what effect will global warming affect cloud formation? There is a scientific divide on this. On one side are those who contend that increased temperatures will lead to increased evaporation, and increased cloud formation. On the other side are those who feel the opposite. There appears to be more evidence to support the former side that warming will lead to increased cloudiness, and hence, the negative direction (lower CS). As Kevin says, there are clues, but what is lacking is certainty.

  4. 54
    Ray Ladbury says:

    Dan H.: “The general view is that the blockage of incoming radiation is greater, resulting in a net temperature reduction.”

    Citation seriously needed!

    Dan H.” “There appears to be more evidence to support the former side that warming will lead to increased cloudiness, and hence, the negative direction (lower CS).”

    Again, citation seriously fricking needed.

    Dan H., what is lacking in your post is not “certainty”, but honesty.

  5. 55

    Speaking of citations, the AR4 discussion on the topic is here:

    “The Partial Radiative Perturbation (PRP) method, that excludes clear-sky changes from the definition of cloud feedbacks, diagnoses a positive global net cloud feedback in virtually all the models (Colman, 2003a; Soden and Held, 2006). However, the cloud feedback estimates diagnosed from either the change in CRF or the PRP method are well correlated (i.e., their relative ranking is similar), and they exhibit a similar spread among GCMs.”

  6. 56
    Chris Colose says:

    Ray (54)

    There’s no doubt that clouds provide a net cooling effect of the planet, but that doesn’t mean they provide a negative feedback. It’s a confusion of the sign of a function with the sign of its derivative…

  7. 57
    David B. Benson says:

    The assumption has been a (nearly) constant relative humidity (RH) global annual average. Somewhere I glanced at a recent claim of a slight decrease in RH. I would suppose that implies a slight decrease in cloudiness?

  8. 58
    Magnus W says:

    How novel is this approach to get the sensitivity?;jsessionid=FDF33A1FB2762A4213CD031F01EF2A10.d01t04

    2000 figures make the difference
    When the researchers at CICERO and the Norwegian Computing Center applied their model and statistics to analyse temperature readings from the air and ocean for the period ending in 2000, they found that climate sensitivity to a doubling of atmospheric CO2 concentration will most likely be 3.7°C, which is somewhat higher than the IPCC prognosis.

    But the researchers were surprised when they entered temperatures and other data from the decade 2000-2010 into the model; climate sensitivity was greatly reduced to a “mere” 1.9°C.

  9. 59
  10. 60
    simon abingdon says:

    #56 Chris Colose

    “It’s the confusion of the sign of a function with the sign of its derivative…”.
    I think you’ll find that “cooling effect” and “negative feedback” are both first-order derivatives of temperature.

  11. 61
    Chris Colose says:

    #60 Simon

    The critical issue is the behavior of the function Rc(Ts) where Rc is the net cloud radiative impact on the planet (R_cloud – R_clear sky, where R is the difference in the net solar radiation and outgoing thermal radiation), and Ts is surface temperature. This function may have interesting structure and even bifurcation points that deviate significantly from linear over a wide range of climate regimes.

    The problem here is to understand its behavior over a very small range on the energy-temperature surface that is relevant for modern global warming. For example, I have plotted two different hypothetical functions dictating how clouds impact the net radiation balance of the planet. In both cases, they are acting to cool the current planet at a modern temperature of ~290 K by -20 W/m2. However, in one example, the slope of the curve is locally positive…indicating a tendency for less cooling in a warmer climate (i.e., a positive feedback), and in the other case, the slope is locally negative, indicating a tendency for further cooling (i.e., a negative feedback) in a warming climate.

    Some people make the incorrect interpretation that clouds act as a negative feedback in any climate regime that falls underneath the zero point (dashed red line). I just made these functions up randomly, and the real behavior depends on much more than temperature (humidity, wind speed, stability, etc) but it’s at least illustrative. It also points to the problem of using much different climate analogs or linear feedback analysis in climate sensitivity interpretation. This is still an under-explored problem in the subject of climate sensitivity.

  12. 62
    perwis says:

    Magnus W #59: Re the (unpublished?) Norwegian study by researchers from CICERO + Norwegian Computing Center.

    Andy Revkin has a good comment by Knutti here:

    Anything to add?

  13. 63
    sidd says:

    “The critical issue is the behavior of the function Rc(Ts) where Rc is the net cloud radiative impact on the planet (R_cloud – R_clear sky, where R is the difference in the net solar radiation and outgoing thermal radiation), and Ts is surface temperature. This function may have interesting structure and even bifurcation points that deviate significantly from linear over a wide range of climate regimes.”

    Shouldn’t Rc be a function of, if anything the tropospheric and strato temp ?
    Why in the world would you make Rc a function of T surface ? T surface as well as well as rel. humidity, and density of CCN and a host of other things, but why only T surface? I see that surface T might impact on R clear sky, but R_cloud cannot depend only on T surface.

    If you insist on graphing Rc against Tsurface, ofcourse you will see bifurcations and other strange animals because there are more gremlins on heaven and earth than are dreamt of in your philosophy.

  14. 64
    Magnus W says:

    No, I tweeted him… however not really sure what is going on… presumably the research grope will publish more papers… we do not know what they will say though…

  15. 65
    Chris Colose says:


    Yes, cloud behavior depends on those other things. The atmospheric temperature is not independent of the surface temperature, but climate sensitivity is often defined with respect to the surface boundary…even if the radiation flux depends on the vertical integral. Of course, there may very well be two or more solutions for any given Ts in the example plots above, depending on forcing agent or even the history the climate took to get to a particular point (as is the case in Snowball Earth hysteresis diagrams).

  16. 66
    Magnus W says:

    So now new study from Norway if any one missed that… still they ofc will try to publish new ones…

  17. 67
    Dr. Delos says:

    There are certainly a lot of theories and lots of complicated math to back them up. However, there are at least three enormous factors that are not known for certain that make ALL of the theories worthless in predictive value.

    1. There is no way to predict accurately the la Nina/el Nino warming/cooling effects.

    2. There is no way to accurately predict cloud behavior and I have seen many theories that ‘prove’ that clouds are positive feedback and others that ‘prove’ clouds are negative feedback. Both sides admit that cloud effects vary depending on their altitude, density etc.

    3. Although most theories now espouse that volcanoes are primarily a negative feedback, NOBODY knows when, where and how powerful the next eruption will be.

  18. 68
    Ray Ladbury says:

    Dr. Delos,
    Actually, I believe you are wrong on all three counts. You need to catch up.

  19. 69
    Dr. Delos says:

    If you can predict temperatures, flows and the magnitude of la Nina and el Nino, perhaps you can share that trick with us. As far as I can tell, nobody can even predict if there is one or the other, let alone any of the specifics.

    Clouds are even more chaotic a factor. [see above]

    If you can foresee volcanic eruptions, get yourself over to Asia, they know there will always be more volcanoes, but not when, where or how big they will be.

    After you take care of all those, the most important factor of all – climate sensitivity to a doubling of CO2. Oh sure, there are estimates and ranges, but the exact figure is absolutely necessary in order to make any prediction that are more accurate than blind guesses.

    Essentially what we have now is a set of big, complex formulas with several extremely critical factors unknown and unknowable.

    Here’s something else to think about. Wall Street has billions and billions of dollars at stake daily depending on predictions based on FAR fewer unknowns, yet they haven’t perfected the calculations necessary to buy at the bottom and sell at the top. They have had a 70+ year head start as well.

  20. 70
    Hank Roberts says:

    > the exact figure is absolutely necessary

    Delay is the deadliest form of denial.
    — C. Northcote Parkinson

  21. 71
    Dan H. says:

    Dr. Delos,
    The exact figure is probably not absolutely necessary. However, somewhere much closer than our current range of possibilities would be nice. One last piece to add to your uncertainties is how they may change in the future with higher atmospheric concentrations of carbon dioxide.

  22. 72
    Ray Ladbury says:

    Doctor Delos, Things look complicated to you because you haven’t bothered to study them for fear that it would spoil your “objectivity”. Absolute knowledge is not essential in any field of science–in fact,it is anathema to it. As to the rest of your post…well let’s just say that you aren’t wowing us with your understanding or finance either.

  23. 73
    Jim Larsen says:

    Dr Delos,

    Your analogy fails because of bounds. The stock market can fall to 0 or go to 10000000000000000, but temperatures are tightly bound to CO2 concentrations. Yep, ENSO can warp things up or down, as can volcanoes, BUT the central tendency as described by CO2 concentrations CAN NOT BE CHANGED. So yep, we get WEATHER, but CLIMATE is determined by CO2 concentrations, as driven by orbital variations and human activities.

    [Response: “Cannot” is a pretty strong statement, and almost certainly wrong in detail. Your general points is well taken though.–eric]

  24. 74

    “…climate sensitivity to a doubling of CO2. Oh sure, there are estimates and ranges, but the exact figure is absolutely necessary in order to make any prediction that are more accurate than blind guesses.”

    Uh, no. There’s a big difference between ‘ranges’ and ‘blind guesses.’

  25. 75
    Hank Roberts says:

    > Dr. Delos
    One bait, five fish.

  26. 76
    Jim Larsen says:

    73 Eric, yes, my statement was specifically wrong but generally right. Thanks for pointing that out. I should have made that clear myself. Next time I’ll do better.

  27. 77
  28. 78
    Hank Roberts says:

    another bit on what’s available for condensation nuclei.
    Found a big increase in dust this century vs. last, at one site that’s well picked:

    Studies on relationship between Asian dust outbreak and
    the stratosphere-to-troposphere transport in spring
    with coupled ice-core-meteorology

    …Yasunari TJ, PhD_Dissertation

Switch to our mobile site