What About the Clouds, Andy?

February 21st, 2009 by Roy W. Spencer, Ph. D.

(edited 4:30 p.m. CST Feb. 22 for the discussion of Dessler et al.-diagnosed water vapor feedback value of 2.0 W m-2 K-1)

I’ve been receiving a lot of questions lately about Andrew Dessler’s water vapor feedback paper which supports the positive water vapor feedback exhibited by the IPCC climate models. Dessler and co-authors used AIRS temperature and humidity sounding retrievals from the Aqua satellite during 2003-2008 to compute how the specific humidity changed with warming.

By way of comment, I suppose I could say the same thing I hear whenever I find negative feedback in satellite data: “So what? Feedbacks on such short time scales might have nothing to do with feedbacks associated with long-term warming.”

But I’m not going to do that (insert smiley). Instead, let’s review what Andy has found…and what he hasn’t found.

There are two main categories of radiative feedback involved in climate variability and climate change, and Dessler et al. have addressed a portion of one of them.

These major categories of feedback are the same as the two components of Earth’s radiative energy balance: (1) absorbed/reflected solar (shortwave, SW) radiation, and (2) emitted infrared (longwave, LW) radiation. The LW feedbacks are mainly due to water vapor and high clouds, while the SW feedbacks are mainly due to low clouds.

Longwave Feedback

In response to a warming or cooling influence, the LW feedback is mostly controlled by changes in water vapor and high clouds…Dessler et al. addressed the water vapor part, and got indications of positive water vapor feedback (specific humidity increasing with warming).

And guess what? Using the CERES radiation budget data from Aqua during 2002 through 2007 I get about the same result as they did. In fact, I got an infrared feedback parameter right in the middle of the range of all of the IPCC models. (Note that CERES includes the effect of both cirrus clouds and water vapor, so at face value this would suggest the Infrared Iris effect was not operating during this time…but see “Cause or Effect?” below for another interpretation.)

The Rest of the Story: Shortwave Feedback

The other half of the feedback story which Dessler et al did not address is the reflected solar component. This feedback is mostly controlled by changes in low cloud cover with warming. The IPCC admits that feedbacks associated with low clouds are the most uncertain of all feedbacks, with positive or negative feedback possible…although most, if not all, IPCC models currently have positive SW feedbacks.

But I found from the CERES data a strongly negative SW feedback during 2002-2007. When added to the LW feedback, this resulted in a total (SW+LW) feedback that is strongly negative.

Is my work published? No…at least not yet…although I have tried. Apparently it disagrees too much with the IPCC party line to be readily acceptable. My finding of negative SW feedback of around 5 W m-2 K-1 from real radiation budget data (the CERES instrument on Aqua) is apparently inadmissible as evidence.

In contrast, Dessler et al.’s finding of positive LW feedback of 2 W m-2 K-1 inferred from the AIRS instrument is admissible.

But whether my SW feedback work is published or not misses the main point. Unless you know both LW and SW feedbacks, you don’t know the sensitivity of the climate system, and so you don’t know how much global warming there will be in the future.

The modelers would probably even claim that everyone already knows water vapor feedback is positive, and so it didn’t need any further observational verification.

Spencer’s Mea Culpa?

While I have believed for years that water vapor feedback might be negative, I will admit the latest evidence is looking more and more like the real story could on the reflected solar side instead. The radiation budget guys have been trying to tell me all along that it was the SW feedback that was the most uncertain…maybe they are right. Of course, it could be that long-term feedbacks are opposite of the short term ones, like others have tried to tell me when I find negative feedback (insert second smiley)….

Cause or Effect?

But, as we HAVE published in Journal of Climate, there is an issue regarding feedbacks that could throw all of our satellite diagnoses of feedback into a cocked hat anyway. That is the issue of causation.

The issue is related to something that Forster & Taylor (2006 J. Climate) and Forster & Gregory (also 2006 J. Climate) have previously demonstrated: In order to estimate radiative feedbacks, you must first remove any sources of time-varying radiative forcing from the data. No one has ever bothered to do this for the time-varying radiative forcing due to natural cloud variations in the satellite data. It appears to be the largest source of decorrelation in both the satellite data and the IPCC model output.

Such variations can even be proved to exist…they produce spiral patterns when you plot running averages of temperature versus radiative flux. We have even found those patterns in all 18 IPCC models we have analyzed. It is the only possible explanation for those patterns…I challenge anyone to find an alternative explanation.

And here’s what happens if you don’t remove the effects of time-varying radiative forcing from the data before feedback diagnosis: It decorrelates the relationship between total (LW+SW) radiative imbalance versus temperature. This is because the temperature change lags the forcing…90 degrees out of phase for harmonic forcing…which is what causes the spiral patterns. This will cause the diagnosed total feedback parameter to be biased toward zero (which would be a borderline unstable climate system) — even if the true feedback is negative!

In fact, we showed that if the forcing is 100% radiative (e.g. from natural cloud fluctuations) the error in the diagnosis of the total feedback will be 100%! In other words, you can not measure feedback in response to an unknown amount of time-varying radiative forcing. At least not until someone invents a new way.

The Simple Version

If this sounds too technical, it can also be explained in terms of causation: When researchers see what looks like positive cloud feedback with warming…how do they know that the warming wasn’t the result of the clouds (forcing), rather than the other way around (feedback)? Time-varying radiative forcing due to cloud fluctuations completely obscures the evidence of feedback, giving the illusion of a sensitive climate system

Takin’ it To The Street

Unfortunately, our J. of Climate article has been greeted with deafening silence. Apparently, everyone is too busy burning up computer cycles to see how much virtual global warming they can create in their models. I’m sorry for sounding so cynical, but given the importance of this issue policy-wise, you would think that someone besides me would be working on it.

So…my book describing these issues is almost finished…it’s due at the publisher by March 2. Since even the public understands “cause versus effect”, I decided to put it all down in as simple terms as possible.

Maybe there will be some physicists or engineers out there who understand what I’m talking about.


Forster, P. M., and J. M. Gregory (2006), The climate sensitivity and its components diagnosed from Earth Radiation Budget data, J. Climate, 19, 39-52.

Forster, P.M., and K.E. Taylor (2006), Climate forcings and climate sensitivities diagnosed from coupled climate model integrations, J. Climate, 19, 6181-6194.

Spencer, R.W., and W.D. Braswell (2008), Potential biases in cloud feedback diagnosis: A simple model demonstration, J. Climate, November 1.

Comments are closed.