.Guest Post by Willis Eschenbach
As usual, Dr. Judith Curry’s Week In Review – Science Edition contains interesting studies. I took a look at one entitled “Cloud feedback mechanisms and their representation in global climate models“, by Ceppi et al., hereinafter Ceppi2017. The paper looks at the changes in the radiative effects of clouds. From the paper:
The radiative impact of clouds is measured as the cloud-radiative effect (CRE), the difference between clear-sky and all-sky radiative flux at the top of atmosphere. Clouds reflect solar radiation (negative SW CRE, global-mean effect of −45 W m−2) and reduce outgoing terrestrial radiation (positive LW CRE, 27 W m−2), with an overall cooling effect estimated at −18 W m−2 (numbers from Henderson et al.[36]).
The Ceppi2017 Figure 1 shows that almost all of the models report that as the modeled surface warms, the modeled clouds change in such a way as to increase the modeled warming. On average, Figure 1 shows that for every degree C that the modeled surface warms, the modeled clouds add on another ~ 0.5 W/m2 of additional modeled forcing.
Figure 1. First figure in Cepp1017, as detailed in their caption.
Let me say that I find such a large positive cloud feedback to be very doubtful. Setting that aside for the moment, the Ceppi2017 authors have included a graphic showing the average change in the modeled cloud radiative effect from a number of models.
Figure 2. Average modeled net cloud feedback, from Ceppi2017.
I thought, hmmm … I wondered how that compared to the CERES data. Here’s a look at the same thing, net cloud feedback … except theirs is modeled and the CERES satellite data below is observational.
Figure 3. Average of 180 months of CERES data showing the relationship between changes in temperature and corresponding changes net cloud feedback. The calculations are done on a gridcell by gridcell basis, with the monthly gridcell climatology removed before the calculations.
Now, while the models kind of get it right, there are several problems with them. In the CERES data above, you can clearly see the Inter-Tropical Convergence Zone (ITCZ) as the yellow/green area above/below the equator in the Atlantic, Pacific, and Indian Oceans. In the models, it is only weakly visible in the Atlantic and is missing in the Pacific and Indian Oceans.
Next, the CERES data shows that much more of the planet has negative net feedback than the models claim. The entire southern extra-tropics shows negative cloud feedback, some of it quite strong.
Next, because theirs is an average of various models, it doesn’t capture the full variation in the net cloud feedback. In the real world, there are areas of both strong positive and strong negative feedback.
Finally, on average the CERES data shows that the net cloud feedback is negative. Now, we have to take the accuracy of that number with a grain of salt, in that we are looking at trends. Trends are a ratio, and ratios tend to distort averages. For example, the area-weighted average of the trends as shown in Figure 3 is -1.4 W/m2 per °C. A better measure is likely the area-weighted median of the trends, which is -0.5 W/m2 per °C.
Alternatively, we can look at the relationship on a global basis. Here’s a scatterplot of the monthly residual changes in CRE versus the monthly residual changes in temperature (after removing the global monthly climatology).
Figure 4. Scatterplot of the monthly global CRE and temperature data.
This gives us a third estimate of the relationship between CRE and temperature. This one is between the other two; we have estimates of the cloud feedback factor of -1.4, -1.0 ± 0.3, and -0.5 W/m2 per °C
Whichever way we estimate it, however, the CERES data shows that the net effect of clouds is negative, not positive as the models claim. The average of the models’ estimates of cloud feedback is about plus one-half of a W/m2 per °C. The CERES data, on the other hand, gives a value of about minus one W/m2 per °C.
This is a net swing on the order of ~ 1.5 W/m2 per degree C between the model estimates and the CERES data … and thus a 1.5 W/m2 reduction in the estimated climate sensitivity.
==============
Let me say in closing that I don’t think that “climate sensitivity” is a real thing. I say this because of ample evidence that the climate is a governed system, with a variety of thermoregulatory climate phenomena that work together to constrain the global temperature to a very narrow range (e.g. ± 0.3°C variation over the entire 20th Century). When such a system is in a steady state like that of the earth, the temperature is essentially decoupled from the “forcing” (the changes in downwelling radiation). And because it is decoupled, there is no such thing as “climate sensitivity”.
My best to all,
w.
My Usual Note: When you comment, please QUOTE THE EXACT WORDS YOU ARE DISCUSSING, so we can all understand the exact subject of your response.
Data: The CERES data is here.
Further Reading: Here are a few of my previous posts on the subject of the regulation of global temperature by emergent phenomena:
The Thermostat Hypothesis 2009-06-14
This was my original post on the subject, in 2009. Abstract: The Thermostat Hypothesis is that tropical clouds and thunderstorms actively regulate the temperature of the earth. This keeps the earth at an equilibrium temperature.
The Details Are In The Devil 2010-12-13
I love thought experiments. They allow us to understand complex systems that don’t fit into the laboratory. They have been an invaluable tool in the scientific inventory for centuries. Here’s my thought experiment for today. Imagine a room. In a room dirt collects, as you might imagine. In my household…
It’s Not About Feedback 2011-08-14
The current climate paradigm believed by most scientists in the field can be likened to the movement of balls on a pool table. Figure 1. Pool balls on a level table. Response is directly proportional to applied force (double the force, double the distance). There are no “preferred” positions—every position…
Emergent Climate Phenomena 2013-02-07
In a recent post, I described how the El Nino/La Nina alteration operates as a giant pump. Whenever the Pacific Ocean gets too warm across its surface, the Nino/Nina pump kicks in and removes the warm water from the Pacific, pumping it first west and thence poleward. I also wrote…
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
The cloud radiative feedback value has been carefully chosen by the IPCC and the Hansen’s of the world because if it were 0.0 or -1.0 W/m2/K, then all the catastrophic predictions of global warming theory fall by the way-side.
That is why they refuse to do what Willis has done and actually measure/calculate it.
When you put in -1.0 W/m2 cloud feedback into the calculations, the 3.0C per doubling narrative falls apart and all we get is 1.37C per doubling. Furthermore, there is little feedback on feedback impact and the earth’s atmosphere adjusts to increased CO2/GHGs very quickly – as in within two days. There might still be an ocean uptake lag but it is much smaller.
The IPCC and global warming is OVER if cloud feedback is as negative as -1.0 W/m2/K. It is time for Willis’ numbers to be published.
Bill,
Willis did not “measure/calculate” the cloud radiative feedback value. All he did was plot monthly change of cloud radiative effect CRE in W/m2 versus the monthly change in temperature, got a near random scatter plot, calculated a best fit slope of -1.0 W/m2/K, and falsely called it net cloud feedback. See my comment of May 25, 2017 at 1:29 pm. This procedure falsely assumes that only temperature change caused the cloud change. The definition of cloud feedback is the change in the top-of-atmosphere radiative flux resulting from the cloud response to a temperature change. But causality also runs the other direction. The cloud changes also cause a temperature change. Each point is a mixture of internal radiative forcing, ie cloud changes causing a temperature change, and cloud feedback, ie temperature change causing cloud changes represented by the change in the cloud radiative effect (CRE).
The Spencer paper shows that the internal radiative forcing contaminated the signal, so the slope is not the cloud feedback, but is greater than the true feedback. That is, the true cloud feedback is less than(more nagative than) -1 on this monthly time scale. But the BIGGER problem is that is only for monthly changes, and there is no way to show that this short-term feedback operates on long time scales relevant to AGW.
The sentence in the top post “Here’s a look at the same thing, net cloud feedback … except theirs is modeled and the CERES satellite data below is observational.” is misleading, as it is not the same thing at all! The modeled net cloud feedback was determined by an abrupt quadrupling of CO2 concentrations, causing a HUGE forcing, so the issue of the internal radiative forcing becomes insignificant. In the models, you can compare the time histories of the resulting warming to the resulting changes in the Earth’s radiative budget, and you can indeed extract an accurate estimate of the feedback in the models. But this tells us nothing about the real world. Spencer’s work shows that the short term cloud feedback could be as much as -6W/m/K. Spencer wrote “Unfortunately, there is no way I have found to demonstrate that this strongly negative feedback is actually occurring on the long time scales involved in anthropogenic global warming.”
It is interesting that the paper referred to, Ceppi et al 2017, shows that in the multi-model mean, almost all of the positive cloud feedback is the long wave cloud feedback of 0.42 W/m2/K, with the short wave, or cloud albedo effect, contributing a tiny 0.02 W/m2/K.
“The definition of cloud feedback is the change in the top-of-atmosphere radiative flux resulting from the cloud response to a temperature change. But causality also runs the other direction. The cloud changes also cause a temperature change.”
As there is no recorded ‘runaway’ temperatures on earth, the above mechanism is a closely coupled system keeping general temperature within narrow limits. AKA negative feedback.
“Equilibrium climate sensitivity” indeed does not exist outside of the models, almost at the level of definition. In order to calculate it, the models need to be assured of calculating all possible future climate states, or at least be assured of calculating a truly representative ensemble of climate states. That, in itself, requires a large dollop of faith. Modeled or real climates may end up sitting in a non-equilibrium state for indeterminately long periods of time, skewing results by unknown amounts.
How long should a modeler wait to see? Probably only until funding for CPU cycles runs out, but it may still give significantly different results depending on small changes in input parameters or the day of the week. Back in the real world, reality will choose only one path and there is no reason to think that events will proceed smoothly down a representative path towards a hypothesized equilibrium.
Equilibrium climate sensitivity remains a fancy way of saying “this is what my model says may happen under a certain set of conditions”. To pretend otherwise is to somehow dignify the concept as being on a par with real world measurable physical constants.
Willis, is there anyway you can explain to readers the implications of the sign change in cloud feedback?
How much does that change the global warming estimate in degrees?
My work with old climate models showed this was basically almost all of global warming in this one parameter.
Willis, thank you for the essay.
“And because it is decoupled, there is no such thing as “climate sensitivity”
Wouldn’t it be more accurate just to say the sensitivity is effectively zero?
But actually I see your point, within the very idea of ‘climate sensitivity’ is the assumption that it is not only linear, but also that it never gets saturated, buffered or approaches zero. Rather big assumption that any chemist could point out.
We really don’t know clouds
At all
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2974794
I am trying to understand figure 1. I understand one basic calculation to be
Warming in watts per meter^2 x Sensitivity in K/Wm^2 = warming in K OR
Warming in watts per meter^2 Inverse Sensitivity in Wm^2/K = warming in K
but the values on the y axis are inverse Sensitivity in Wm^2/K . what gives? Feedbacks should be a function of the function greenhouse Gas warming, in wm-2, with gain and also possibly lags, integral, and/or derivative terms, right?
Ignore my question I see now that it is w/m-2 feedback for 1 C of surface warming.
It is interesting that both the CMIP5 and Ceres data agrees that the cloud feedback is strongly negative in the 40-60 degrees south latitude band. This helps explain the great global climate impact of the opening of the Drake Passage in the Oligocene. It not only isolated Antarctica climatically, it quite possibly also changed the net cloud feedback from positive to negative.
Willis wrote: “On average, Figure 1 shows that for every degree C that the modeled surface warms, the modeled clouds add on another ~ 0.5 W/m2 of additional modeled forcing. Let me say that I find such a large positive cloud feedback to be very doubtful.”
Why? Clouds reflect an average of about 80 W/m2 to space. Why can’t this value decrease by 0.6% per K of surface warming?
If I understand correctly, the data in Figure 3 shows ratios, W/m2/K, not W/m2. So everywhere there was relatively little temperature change (say +0.02 K), even a modest change in TOA flux will produce a large change in feedback – either positive or negative. And if the temperature change were negative (say -0.02 K) the magnitude of the feedback would be just as big, but the sign would be different. This is another way of looking at the problem you discussed with median vs mean.
Figure 4 may be showing us which data points are most robust – those with the largest amount of warming and cooling. However, if you simply plotted dW vs dT, then the climate state in 2000 (a strong La Nina) and in 2015 (a weak? El Nino) could easily be dominating your data. Starting with and El Nino and finishing with a La Nina might make the results look completely different. If you do a linear fit between W and T at each location over the whole period, this problem would be minimized. However that approach doesn’t provide individual dW and dT values to plot. In theory, you have the trend at each locations X+/-Y K/15 yr for both temperature and TOA flux. If you take into account auto-correlation, the error bars are fairly huge. HADCru4 GLOBAL warming at Nick Stoke’s blog is 1.23 +/- 0.71 K/century (0.184 +/- 0.106 K) for 1/2000 to 12/2015. Local variability is far greater than global variability. If you put error bars on the points in Figure 4, the result may not be very convincing.
IIRC, there is combined TOA flux data from many satellites going back to 1979, but it isn’t as homogeneous as CERES. The global warming rate is 1.72 +/- 0.24 K/century (0.67 +/- 0.09 K). With smaller error bars, perhaps the data would be more convincing.
Nic has been discussing these recent papers, which suggest that cloud feedbacks don’t remain constant with time in models.
Gregory (2016) http://onlinelibrary.wiley.com/doi/10.1002/2016GL068406/full (paywall)
Andrews (2015) http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-14-00545.1 (Free)
Willis, I know you have said you are not too interested in the peer-review literature debate, but would a reply to Ceppi et al in the journal not open this issue up to further debate?
“Let me say in closing that I don’t think that “climate sensitivity” is a real thing. I say this because of ample evidence that the climate is a governed system, with a variety of thermoregulatory climate phenomena …” In support of this point, I note that the upward heat delivery rate implied by a one-inch-per hour rainfall event is about 16,000 watts/M^2. Even at a tenth of an inch per hour, it is 1,600 W/M^2. So what? It becomes clear that the “climate,” whether defined locally, regionally, or globally, is the composite result of a large number of small-scale heat exchange events of very high power. Any expression of forcings of single-digit W/M^2 or single-digit dec C per doubling are not really meaningful in view of how the atmosphere expresses its operation for all to see. Steam engines rule.
Willis has briefly described a ‘flywheel’ aspect of Earth’s climate system that is all too easy to take for granted. Flywheels are essential components in almost all of mans’ gadgets, from gyroscopes, through engines on the Titanic and your bracero’s lawn mower. They work well and are- here’s that word again- Essential. It would be a surprise if a 4+ billion year-old, superbly working machine didn’t have one.
RACookPE1978 May 27, 2017 at 11:32 pm
Thanks, RA. Not sure where you are getting your figures. Per CERES, downwelling TOA solar over the ocean averages 347 W/m2 (on a 24/7 basis).
On the other hand, at the surface downwelling solar over the ocean averages 186 W/m2.
This is 186 / 347 = 54% of the incoming solar making it to the surface of the ocean … far from the “5% to 20%” you claim.
Next, of the total energy you say, “that 80% of original energy is absorbed into the atmosphere, heating it directly.”
That is simply not true. A goodly percentage of downwelling solar is reflected back by the clouds, and doesn’t heat the atmosphere. Per CERES, over the ocean it’s about half and half. About 80 W/m2 is reflected by the clouds back to space, and about 79 W/m2 is absorbed in the atmosphere.
Best regards,
w.
Willis,
Averages hide a multitude of ‘sins.’ I think that this problem has to be attacked piecemeal, with whatever integration slices or aggregations make sense. Hence my harping on how specular reflection changes with angle of incidence and I suggest that frustums be the first-order sampling and modified where land covers the oceans. Further, the impact of cloud cover over land and ocean has to be taken into account. This is a devilishly complex layer-cake problem where the saying, “The Devil is in the details,” is appropriate!
I haven’t verified RA’s claims about actual values, but clearly, the flux through a cloud-free atmosphere decreases with increasing angles of incidence because the footprint of a light beam spreads out, and a longer slant-range allows for more scattering and absorption.
Willis Eschenbach
Thank you for the courtesy of your reply. I’m working 7×12 night shifts right now, and will get back to you with comments about your feedback as soon as practical.
I do note that your CERES average value of 347 watts/m^2 at TOA would require 1388 Watts/m^2 for TSI. I understand the correct TSI (average value at a nominal earth radius) is much less, only 1362 watts/m^2 TSI now the standard.
(Of course, the original 1988 TSI value of 1368 watts/m^2 was used in the original catastrophic global warming calculations by Hansen and his cronies, so why are those same 1988 calculations and constants and coefficients valid when we “apparently” already “lost” 6 watts/m^2 …. And, if after losing 6 watts/m^2 already, why are we still going to heat up the earth catastrophically due to 3.0 degrees of added warming from CO2? It would seem we would be cooling now, and facing a deficit of 3 watts/m^2 that must be replaced just to keep temperatures the same. But that is a separate question entirely. )
Thanks, Clyde. The question of 1388 W/m2 vs 1368 W/m2 is because this is just over the ocean, and there is more ocean near the tropics.
w.