2011-02-28 22:38:36Interesting paper about water vapor feedback
Ari Jokimäki


I found an interesting paper about water vapor feedback. Hallegatte et al. (2006) say in their abstract that (emphasis mine):

"In the model here, it is found that, even though the water vapor static gain proves consistent with results from GCMs, it turns out to be negative for time scales below 4 yr and positive only for longer time scales. These results suggest two conclusions: (i) that the water vapor feedback may be fully active only in response to long-lived perturbations; and (ii) that the water vapor feedback could reduce the natural variability due to tropospheric temperature perturbations over short time scales, while enhancing it over longer time scales. This second conclusion would be consistent with studies investigating the influence of air–sea coupling on variability on different time scales."

Their Figure 4 shows the story. It seems to me that this puts the studies of Lindzen and Spencer in a new context. Strange that I haven't seen this issue discussed before.

2011-03-01 00:23:33



Do you understand how it works? Why would there be a short-term negative but a long-term positive?

2011-03-01 01:32:13
Ari Jokimäki


I don't understand it very well, Neal, but here's how they describe it - pages 8-9 of the linked PDF:

"The fast pole corresponds to the lowering of latent heat flux due to rainfall decrease, which comes from the rising temperature (corresponding to a decrease in RH). This mechanism constitutes one path of the WV feedback: any transient trajectory showing an increase in atmospheric absolute humidity requires an imbalance between precipitation and evaporation, and hence necessitates an increase in atmospheric latent energy content compared with the equilibrium state. In consequence, the WV feedback process should involve a rapid atmospheric cooling, as formalized in our model, with a time response of about a few weeks. This negative feedback can be illustrated by considering a doubling CO2 experiment that would lead to a 3 K surface temperature increase: for constant RH, absolute humidity would increase by about 10%. This would correspond roughly to an addition of 4 kg m-2 of water vapor in the atmosphere, and hence to a latent energy loss of 107 J m-2. Because the involved LW flux changes are about 3 W m-2, one month of the total additional flux would be necessary to collect that energy."


"Our characteristic time of 3 to 7 yr is very long compared to the 6 months found with B98. Considering the very crude radiative model used in B98 (a gray gas model), we take this as a confirmation that the long response time of the positive feedback gain is linked with the radiative interactions between ocean and atmosphere."

2011-03-01 04:00:04
Julian Brimelow

Ari, great find.  I was not aware of this paper.

One of the critiques of Lindzen and spencer's work is that they have been looking at too short a period of time. This work supports the notion that one needs to look at decadal-long datasets (or at least 10 years) to appropriately capture the full effect of the WV feedback.

It would be very interesting to know what Murphy, Trenberth and  Dessler think about this paper. 

Neal, if I am interepreting their graph right (Fig. 4), in their model the total feed back only becomes weakly positive after about 2-3 years, with the greatest signal form about 10 years.  So yes, it seems that at short-time scales the feedback may be negative.  I have not read Dessler's recent paper in a while, but I recall his data suggesting something similar.


2011-03-01 04:26:19


But is this the response to a "pulse" of water vapor, or to the onset to a gradual increase? The first case would hardly be applicable, since every year there is a new pulse. And in the second case, well, it's already been 150 years, so why would we worry about 3 years?

2011-03-01 04:53:19interesting
Dana Nuccitelli

From 5 years ago, too.  Seems like this is a big hit to Lindzen and Spencer's work in arguing for low sensitivity.

2011-03-01 06:00:49
Ari Jokimäki


There's also a 2004-version of this study. There are some differences but overall seems to be an early version of the same thing. I have to wonder why this seems to have gone largely unnoticed. Google Scholar lists 8 citations for the newer paper and only 1 citation for the older one. I found this by reading the Bony et al. (2006) review on climate feedback processes (where Hallegatte is one of the authors, by the way).

It would be interesting to check how this compares with Lin et al. (2010), but I seem to have misplaced my copy of the full text. I'll see if I can find it tomorrow.

Another interesting thing would be to see how the available observational estimates and other model results fit to the curve in Hallegatte et al. (2006) Figure 4.