Watts Up With That (WUWT) has a new post by Willis Eschenbach called the Cloud Radiative Effect (CRE). The CRE refers to the influence that clouds have on the Earth’s albedo. Clouds act to reflect incoming radiation and hence have a negative forcing.
Willis appears to have downloaded CERES data and produced a very nice figure showing the forcing due to clouds. The figure is below and if you average over the whole globe you find that the clouds produce a negative forcing of -20.6 Wm-2. I believe this was discussed in an earlier WUWT post in which they got very excited because they thought this could completely cancel any anthropogenic forcings. This was nicely rebutted by Bart Verheggen in a post called confusing the net cloud effect with a cloud feedback: Very different beasts. Basically, the figure shows the net radiative forcing of clouds (measured by CERES) relative to having no clouds at all. So, this illustrates the role that clouds play in overall greenhouse warming (making the Earth about 30oC warmer than it would be with no greenhouse warming) and does not indicate how their influence has changed since pre-industrial times.
An illustration showing the net radiative forcing due to clouds from CERES data (credit : Willis Eschenbach, WUWT).
Willis has, however, gone further and also determined the rate of change of CRE with respect to temperature. This is shown in the figure below. The values can vary quite substantially from -26 Wm-2K-1 to 36 Wm-2K-1. Willis then determines an average of -2.9 Wm-2K-1.
Rate of change of CRE with respect to temperature (credit : Willis Eschenbach)
This result may seem quite interesting because how clouds influence anthropogenic warming is quite important. They tend to increase albedo and hence reduce the overall change in forcing since pre-industrial times. According to the IPCC, however, the change in radiative forcing due to clouds since 1750 is between 0 and -1.2 Wm-2 with a best estimate of about -0.4 Wm-2 (unless I’ve read the diagram wrong – correction added : These are the wrong numbers. This is the aerosol-cloud effect which is one of the anthropogenic forcings. In fact, as Karsten points out, the likely cloud feedback is positive with a best estimate of +0.6Wm-2 and with a range of -0.2Wm-2 to 2.0Wm-2). Willis’s estimate would suggest that it should be more like -2.9 Wm-2. This is large and would actually cancel a large fraction of the anthropogenic forcings if true. Willis actually says
Finally, there’s one more oddity. This is the fact that overall, as an area-weighted average trend, for every degree the globe warms, the warming is strongly opposed by the cloud radiation effect. The action of the clouds reduces the downwelling radiation by 3 W/m2 for every degree the planet warms … in IPCC terminology, this is not only a negative feedback, but a strong negative feedback.
And the cooling effect of the clouds is even stronger in the ITCZ. There, for every degree it warms, the downwelling radiation drops by ten W/m2 or so …
I think, although I’m by no means sure, that this is the first global observational analysis of the size of the so-called “cloud feedback”. It shows that the cloud feedback is strongly negative overall, -3 W/m2 for each degree of warming. In addition, in the critical control areas such as the ITCZ, the cooling effect is much larger, 10 W/m2 or so. Finally, it shows a very strong negative cloud feedback, 20 W/m2 or more, in the area of the Southern Ocean
So, is this some amazing analysis that suggests clouds have a much bigger influence than current mainstream climate science suggests? No, I don’t think it is. I think it’s an illustration (sadly maybe) of why some basic training in science/mathematics is actually quite important (and I hope I don’t have to eat my words here :-)). Let’s see if I can explain why. To determine the average CRE, one could simply sum over the fluxes for each grid cell, Fi, multiplied by the area of each grid cell, Ai and then divide by the surface area of the Earth, A.
Willis then, I believe, calculates the gradient with respect to temperature, for each grid cell, using something like
To get his average gradient, he then simply uses (unless I misunderstand what he’s done)
This, however, isn’t the same as what is referred to in the IPCC radiative forcing estimates. This is simply an average of the gradient of the forcing with respect to temperature (in a sense, this is probably also a weather effect, rather than a climate effect). What you really want to know is how does the overall forcing change as the average global temperature changes. As far as I can tell, to get this you’d need to do something more like
In other words, you’d need to know the difference between the average cloud forcing when global surface temperatures are T and when global surface temperatures are T ’.
So, in a sense I’m writing this to illustrate two basic points. Having some basic training is important. Unless, I’ve made some silly blunder, Willis has just illustrated that without some basic scientific/mathematical training, you can easily make a silly mistake without realising. If I want to determine the overall gradient of a function, I don’t take an average of the gradients at different points along the function. You need to consider how the function has changed over the whole interval. The other point I would make is that if you think you’ve discovered something important that other really bright people have missed, consider the possibility that you’re wrong (I’ve done this myself a number of time – see yesterday’s post on climate sensitivities for example). This is especially true if what you’ve done is quite straightforward. If your calculation is right, why would those who work in the field not have done it already themselves? I’ll even accept that maybe I’m wrong about Willis’s calculation, so – if I am – feel free to let me know. But, as far as I can tell, Willis hasn’t really discovered a new way of determining the change in cloud forcing. He’s really just illustrated why one should be cautious when you think you’ve just determined something that many other bright people have missed (something I should possibly consider myself sometimes, to be honest).
Wotts Up With That Blog 
If the climate system were dominated or even strongly influenced by a significant negative cloud feeback, paleoclimate behaviour would be completely different. Most of the substantial variability – often in response to relatively small changes in forcing – would not have happened because nothing much could have happened.
The highly variable nature of paleoclimate behaviour is a very strong indication that the climate system is in fact dominated by positive feedbacks. Glacial terminations under Milankovitch forcing being an excellent example which is why I can be relied upon to trot it out every time.
Quite how Willis can be unaware that all known paleoclimate behaviour contradicts his “hypothesis” beggars belief.
Indeed, that there are other reasons why cloud feedback cannot largely cancel positive forcings is not surprising. I think I’m just am still amazed that someone can go on with all these seemingly complicated analyses and not realise that they’re basically making mistake after mistake. In this case, a fairly basic one. What’s more, noone seems to have corrected him in the comments (although maybe I shouldn’t be surprised by that).
And another thing… 😉
Wotts, you post up your thinking and working and ask commenters to check it over and look for errors. Willis slaps his stuff up and makes big claims – invariably of the same nature, concerning his belief that there are fundamental errors in the current scientific understanding. You are curious and enviably modest. Willis’s wonderings take a very different tone.
Willis is suffering from the Dunning-Kruger effect.
Thanks. It may be because I’ve got so much wrong (initially at least) in my time, that I’ve lost any sense of shame 🙂
Either I got something terribly wrong, or there is an issue with your IPCC forcing estimate. IPCC AR5 states: The sign of the net radiative feedback due to all cloud types is less certain, but likely positive. […] We estimate […] the cloud feedback from all cloud types to be +0.6 (−0.2 to +2.0) W/m2/K
Yes, I was surprised to read that the net effect of clouds was negative. I thought it was positive.
Yes, I’ve found that now. I was looking at Section 8 (radiative forcings) and had quoted the Aerosol-cloud term. I presume, however, that the reason why cloud forcing is overall positive is because of the impact they can also have on outgoing long-wavelength radiation. Presumably, the albedo affect of clouds has to always be negative though.
As I mention below, I was quoting the aerosol-cloud term in the radiative forcing diagram in Section 8 (Figure 8.17). I had assumed that maybe this was only quoting the albedo influence of clouds, but it could be that I don’t quite understand this term. Reading parts of section 7 does seem to suggest that this is what this term is (or it’s the extra albedo from aerosol seeded clouds). Anyway, whatever the actual correct number, it seems like it is nowhere near the -2.9Wm-2 quoted by Willis.
I’ve just found this in Chapter 8 – RF of the aerosol-cloud interaction (previously denoted as the cloud albedo effect), which makes me think the aerosol-cloud term I quoted in the post is what used to be called the cloud albedo (so may be the correct term for what I was discussing). This seems to be related to human emitted aerosol seeding clouds and hence changing the cloud albedo. I’m not sure, though, if there is an extra cloud albedo feedback that isn’t associated with aerosols.
Now I see the problem. One must not confuse feedback and forcing. Your number is the indirect aerosol forcing, mainly controlled by the cloud albedo effect due to enhanced aerosol concentrations (as you correctly noticed). Has nothing to do with the cloud feedback due to temperature changes at all. That’s down to SW and LW changes owing to more or less low or high clouds. In general, it is thought that low clouds get less (pos. SW feedback/neg. LW feedback) and high clouds (due to more tropical convection) get more (pos. LW feedback if I’m not wrong). The net feedback is positive (SW moderately pos and LW slightly neg.)
Thanks. I did blunder here slightly in that the numbers were so big I thought that Willis was only considering the albedo effect of the clouds, but he wasn’t. His calculation was for the net effect and so his estimate is even more wrong than I had realised (I should have realised this when I was doing some background reading. If I remember correctly the net albedo effect is -50Wm-2 while the LW effect is +30Wm-2 giving a net feedback of -20Wm-2).
To be fair – as you point out – I had also somewhat confused the forcing and feedbacks here. So, again I learn something new every day. Anyway, the main point of the post was to illustrate that you can’t determine the the change in cloud feedback, with temperature, by averaging over the temperature gradients in each CERES data cell (which I think I got right :-)).
Still confused here. Do clouds not contribute to forcing by increasing the greenhouse effect of the atmosphere? I understand that they reflect the sun as well.
To be fair, I was confused as well (probably still am about some things). I think (and maybe Karsten should be the one who answers, but I’ll have a go anyway) that if you consider the standard radiative forcing diagram that is often used (with all the bars for the different components), this only represents the direct anthropogenic forcings. This includes how anthropogenic aerosols can seed clouds, and this produces a net negative forcing. In addition to this, clouds can also have an influence simply because of the change in global temperatures (i.e., it’s not directly anthropogenic). This both acts to increase albedo (negative) and to absorb and re-radiate LW radiation (positive). This causes a net positive feedback as pointed out by Karsten in his first comment. This is also regarded as a feedback and not a forcing. In addition, for example, the increased temperature also increases atmospheric water vapour and this is also regarded (if I read the IPCC report properly) as a feedback rather than a forcing.
Maybe Karsten can clarify something though. As far as I understand it, the term forcing typically refers to those that are directly anthropogenic. The feedbacks are those that are indirect (i.e., changes in clouds or water vapour due to the changes in temperature). These feedbacks still have units of Wm-2 and so are still technically “radiative forcings” but are not directly anthropogenic.
What about orbital forcing? That’s not anthropogenic. Not yet anyway 🙂
Maybe 🙂 When people use that CO2 lagging temperature in ice core samples proves that CO2 doesn’t cause temperature rises, I typically point out that this is a consequence of orbital variations that drive temperature variations which are amplified by CO2 and that if CO2 lead temperatures in ice core samples, it would imply that rising atmospheric CO2 concentrations were influencing our orbit around the Sun :-).
Rachel, yes I see what you mean now. I may be back to being confused again, but I think what the figure actually includes are external forcings (anthropogenic and solar) and feedbacks are things like water vapour and clouds (apart from those seeded by aerosols) that respond to the change in temperature, rather than directly to the external forcing.
I’ll stick my neck out, in the spirit of this blog:
Forcings are external to the “climate system”. So solar variability (intrinsic or from orbital dynamics) is a forcing. GHGs released from geological sinks by volcanism or human activity are forcings. Aerosols from volcanism and human activity are likewise forcings.
Everything else is a feedback.
I am happy to be corrected here!
Yes, that’s how I now understand it too. It’s taken a while but I’m slowly getting there 🙂
I thought the greenhouse effect of clouds was a forcing and the albedo effect also a forcing. The net effect of the two probably being positive. The changes that occur in cloud cover due to changes in climate would then be a feedback. But because I don’t want to fall victim to Dunning-Kruger myself, I will say that I know nothing…
My comment about orbital forcing being anthropogenic one day is having heard someone mention somewhere that perhaps in the distant future, humans will be able to alter Earth’s orbit.
If we ever get any solettas up, we will be altering solar forcing. Well, we’ve already altered the radiative properties of the atmosphere, so why stop now?
I’ve got a horrible feeling that cloud forcing is changed when perturbations to the climate system alter net cloud forcing, at which point it’s also a feedback.
😉
Again, I am sticking me neck out.
wotts and BBD, you have it exactly right. There are a few things which are not easy to categorize as either forcing or feedback (e.g. sea ice albedo, which is considered a feedback on short time scales and a forcing on very long time scales), but in most cases there are no ambiguities. Feedback is a response of the climate system to a forcing, while the forcing itself can be either external (natural) or anthropogenic. Since both, feedbacks and forcings, perturb the planetary energy balance, its strength is universally expressed in W/m2. Using the effective radiative forcing/feedback at the top of atmosphere as metric, they become directly comparable.
Needless to say, Wondering Willis is dead wrong. If there is one thing you can learn from these people, it’s how not to do it. They are also reliable insofar, as the exact opposite conclusion is most likely be correct (*). In this case: Net cloud feedback is positive. There you go: You learned something 😉
(*) Argument based on countless personal debunking experiences at a time when I thought these guys might have something to contribute. Gosh, only today I know how ncomprehensibly naive I’ve been …
… or shouldn’t it better read “how incomprehensibly naive I was …”? Otherwise it would indicate that I’m still naive … if my modest grammar skills are right!?
Your grammar is fine. “I was” and “I’ve been” do not imply that you are now naive. Rather the opposite.
😉
The differences between forcing and feedback are in causality and in timing. Water vapor is strongly driven by temperatures, responds within 5-10 days, and is a feedback. Cloud changes are driven by water vapor and temperature, and are also a feedback relative to long lived (non-condensing) greenhouse gases like anthropogenically created CO2 and CH4.
In the Milankovitch cycle, by contrast, orbital forcing changes are long enough in duration that thermal solubility of ocean CO2 can respond, making CO2 a feedback in that situation – CO2 levels caused by/responding to the orbital changes. Aerosols and land use are forcings, temperature-driven cryosphere changes in albedo a feedback, etc.
Hi,
Maybe I miss something but your 2 equations about the feedback and forcing are equivalent (Ti is equal with T and Ti’ is equal with T’ in this case).
No, because T and T ’ are the average global temperatures at two different times, while Ti and Ti’ are the temperatures in each grid cell at two different times. The individual grid cell temperatures can be very different from each other and aren’t the same as T and T ’. So, the second to last equation is the average of the gradient in each grid cell, while the last equation is the gradient, with respect to average global temperature, of the average cloud forcing.
Of course, doing the calculation properly is probably even more complicated. You’d need to consider at least one years worth of data so as to get an average global temperature and an average cloud forcing (probably more than one year). Then at some later time, when surface temperatures have risen, you would need to compute an average global temperature and an average cloud forcing again. You could then compute how the cloud forcing changes with global temperature. I suspect we don’t have enough CERES data yet to do that with any accuracy (but I could be wrong).
I notice that Willis has now redone his calculation and corrected for solar radiation. I’m not quite sure how he’s done this but his average CRE per 1oC rise in temperature is now 0.7Wm-2, pretty much the same as that quoted in the IPCC document. I think, however, that this is still simply coincidental and that all he has calculated is the average of some gradient, rather than the overall gradient.
There’s a question we ask in our first-year physics workshops that might illustrate the issue. Consider a rower rowing across a 1 km wide river. They row from the first bank to the other at 2 km/hr and then back again at 1 km/hr. What’s their average speed? Most people would say 1.5 km/hr, but this is wrong. It took them half an hour to cross from the first bank to the other, and then 1 hour to get back. So, they traveled a total of 2km in 1.5 hours and hence their average speed is 1.333 km/hr. I think this is essentially what Willis has done in his calculation.
Sorry for late answer! I think in this case the exercise is to illustrate the effect for a 1 deg increase/per grid cell. As such if one applies 1 deg increase per each cell then the total average temp will also increase by 1 deg. and of course the deltas Ti’-Ti are equal with T’-T equal with 1 degree and then, the 2 equations you wrote are equivalent. This exercise is theoretic, i.e in reality of course that all Ti’-Ti is not equal with T’-T.
However, you are right that the second formula should be applied to calculate the average CRE; i.e. CRE = dF/dT = Sum(bi*dti)/Sum(dti) where for each grid cell bi = dFi/dti.
PS
In my previous comment: Of course I wanted to say that deltaTi = delta T and not that Ti=T and Ti’=T’.
Yes, you’re right that if each cell changed by the same temperature then the average would also change by the same temperature and it would be equivalent. This also illustrates another issue with the analysis because an increase of 1 degree in average temperature is not due to a change of 1 degree in every cell, so even if you were to consider a 1 degree change in each grid cell, that still wouldn’t give the correct answer.
Actually, I see what you mean. Willis’s calculation was a bit subtler than I had realised (as you noted). He has computed dF/dT for each cell and then determined the change for a 1 degree change in temperature. However, as I mention above, a 1 degree rise in average surface temperature is not due to a 1 degree rise in every cell, so it still doesn’t give the “right” answer (although it is a better approximation than I maybe gave it credit for).
Thanks to Mircea, I now realise that Willis’s calculation was a bit more sophisticated than I first realised (teach me to write a post on a Friday evening before rushing home from work). He actually calculated the gradient in each cell and then used that to determine how the forcing would change due to a 1 degree rise in temperature. Averaging this over the whole surface then gives a change due to a temperature change of 1 degree. Given that he now gets a result similar to that obtained by the IPPC is intriguing. However it still is probably not correct in that a rise in average surface temperature of 1 degree is not due to a rise (in each cell) of 1 degree, hence I think it would still need to be done the way I suggest. It would also be interesting to know how robust his gradients are. I’m surprised he can take what I imagine is quite complex data and get a nice simple gradient in each cell.
Pingback: Another Week in the Ecological Crisis, October 6, 2013 – A Few Things Ill Considered