The theory of anthropogenic CO2 forcing is that outbound IR radiation is plugged up a little more in the atmosphere, causing a steeper lapse rate and warmer surface. The issue of inbound NIR water vapor absorption is that, if inbound NIR solar radiation is plugged up a little more in the atmosphere, then you would presumably have the opposite effect.

If NIR water vapor absorption parameters were under-estimated (as appears to have been the case in GCMs), it seems to me that this would explain some GCM defects, such as a too negative tropopause temperature. The amounts are non-trivial as they exceed (in wm-2) the impact of 2xCO2. It also seems to me that you would have an over-estimate of water vapor feedback effects. Some of these questions were passed very indirectly to Alan Arking, who argued that such errors do not matter. Here are some of his comments.

I only have part of the correspondence, which I’m reproducing here, because the matter is so topical. . I can’t see anything confidential in this correspondence. NIR water vapor absorption is not a topic that I pretend to any authority. There are a couple of paragraphs in IPCC 2AR, but remarkably (as far as I can tell) almost nothing in IPCC 1AR or IPCC TAR about infrared aspects of anthropogenic CO2 forcing. 

Fred and Al,

here are some former thoughts on the negative forcing of H2O, CO2, and CH4 (yes, also CO2 and CH4) from a communication with Steve McIntyre. I still think that this is a reasonable starting point for an in-depth analysis (which is not my professional task and for which I do not have the time nor the spectroscopic data base). I am still convinced that negative forcings, caused by NIR absorption, are underestimated. This can simply be seen by graphical integration of some NIR bands - without any numerical calculations - and a "guesstimate" of NIR band broadening, as may be assumed from general knowledge in IR spectroscopy. Note that the negative forcing starts already with atmospheric absorption bands of 4.5 µm and below. The concentration dependence should indeed be logarithmic.

best regards
Hartwig

Dec. 28, 2004

Fred:

I am on an overseas trip at the moment, without easy access to my computer files. Hence, I cannot  vouch for the quantitative aspects of your statement.  Let me state that the water vapor feedback in the solar spectrum is negative, in contrast with the terrestrial spectrum, which is positive. 

Approximately 25% of the incident  solar radiation is absorbed in the atmosphere, primarily by water vapor.  The effect is NOT linear with respect to water vapor amount.  I have plots (and a least squares fit) showing a logarithmic relationship, and I can send copies when I get back next week, or you can find them two papers I authored in 1999, one in Geophys Res Lett, the other in J. Climate. 

Water vapor absorption is included in climate model calculations, although most models show less absortion than observed (approx  20%; ECMWF model is a notable exception,showing slightly more than observed). 

With best regards,
Al Arking

Dec. 29, 2004

Fred:
I will be back Jan. 2, and will call that day or the next. 

The answer to Q. 3 is yes; the negative feedback is included in almost all climate models.  The radiation codes might not agree quantitatively (with each other or with observations) but solar absorption is included. I just realized that the papers I mentioned can be downloaded from http://aa.eps.jhu.edu.  When I get back, I can provide you a more up to date analysis. 

Al Arking

Jan. 3, 2005

Fred:

I am back from travel, and can now respond in more detail and (I hope) with greater clarity.  There are problems with the modeling of water vapor absorption, which I will describe briefly below with illustrations from a forthcoming paper.  But the effect on WV
feedback is small.  The reason is that most of the solar radiation that is not absorbed by the atmosphere is absorbed by the surface.  So, changing the amount of solar radiation absorbed by the atmosphere has only a small effect on the total column absorption (i.e., net absorption at the top of the atmosphere) and, consequently, only a small effect on the WV feedback as usually defined (i.e., the response of sfc temperature to a change in net flux at the top of the atmosphere).    

However, that does not mean that correcting the radiative transfer scheme in climate models has no effect.  On the contrary, altering the partitioning of absorbed energy between the atmosphere and surface affects the stability of the atmosphere, convection, and how much of the flow of energy from equator to pole is carried by the atmosphere versus the oceans.  It also affects the hydrological cycle, because transferring solar absorption from the surface to the atmosphere causes less evaporation and, thereby, less clouds and precipitation.   

While practically all climate models include atmospheric absorption by water vapor, the magnitudes and sensitivity to water vapor vary greatly from model to model.  As an example, Fig. 1 compares the zonal, annual mean absorption in several models used in climate research, showing wide variations amongst the models in the overall magnitude and latitudinal dependence of atmospheric absorption. 

The observed relationship between atmospheric absorption and water vapor is shown in the upper plot of Fig. 2  for a mid-latitude site (36.6 N, 97.5 W) during the fall season.  Water vapor has, by far, the largest systematic effect on mean daily absorption.  It increases steadily with column WV amount, and fits a logarithmic expression quite well. 

The lower plot of Fig. 2   shows the effects of clouds.  The clouds do not absorb solar radiation, but because they reflect radiation back to space, they can have a large effect on total column absorption and on the partitioning of the absorption between surface and atmosphere,   depending on cloud thickness and height.  

Radiative transfer models based on theory, as opposed to those that are empirically tuned, tend to underestimate atmospheric absorption, as illustrated in Fig. 3  , which compares state-of-the-art model calculations with the clear-sky observations in Fig. 2  .

Let me know if you want more info, or references to published papers. 

With best regards,
Al Arking

Jan. 12

Fred: 

Sorry for the delayed response.  It seems that my last e-mail as tagged with a rarely used address that is not forwardable, and it is only by chance, and after more a week's delay, that I found your recent msg sent to that address.  Please use my regular address.

Before responding directly to your comments, let me first show Fig. 4  (attached ), in which the lower right graph is identical to Fig. 3 in my previous msg (showing atmospheric absorption vs WV, but showing only the Chou model calculations).  The upper right panel shows albedo at the top of the atmosphere.  As you can see, there is a bias between model and observations, but substantially smaller than the absorption bias.  More important, the TOA albedo for either model or observations is practically
independent of WV.  Hence, the WV-albedo feedback is extremely small, and correcting the bias in the model will not change it significantly.   

It might appear puzzling that increased atmospheric absorption (due to increased WV) has little effect on TOA albedo, but one must note that as water vapor increases (causing albedo to decrease), aerosols are also increasing (causing albedo to increase), with the two effects apparently canceling.  (The correlation coefficient between WV and aerosols over the 13 days of collected data is 0.74.)

Now, in response to your comments, note that the data in Fig. 4 are for clear days only.  On days that include cloud cover, Fig. 2  (in previous msg) shows that clouds have a substantial effect on atmospheric absorption---sometimes increasing it, sometimes decreasing it---but the effect is not systematic, because cloud droplets do not absorb (they only scatter) and, because clouds can be at any height, both  the sign and magnitude vary (high clouds --> less absorption, low clouds --> more). 

Concerning dependence on air mass, I have lots of figures, but I think it is more meaningful to look at averages over the course of a day; otherwise, there is one more variable to consider.  In Figs 2-4, all plotted points are a daily average. 

With respect to CASE #1, it is not an energy balance problem when the net energy at TOA does not change.  Although the discrepancy between observations and model calculations is quite large, for the reasons given above and in my previous msg, the feedback effect depends on the hydrological cycle far more than on the radiation.  Since we are not sure we can calculate accurately enough the change in the climate system when one changes the partitioning of energy between atmosphere and surface, we should be less sure about calculating the SENSITIVITY of the climate system to such changes.    

CASE #2 is the real issue: What is the relationship between WV and clouds, and the corresponding relationship between WV and TOA albedo?  The models might very well be inadequate in that regard, just as they may be inadequate with respect to the temperature-WV relationship.  So, the problem is not one of calculating the radiation correctly, so much as calculating the changes in WV and clouds in response to changes in radiation (caused by GHGs, aerosols, and solar luminosity).  Relatively speaking, the accuracy of the radiation codes are a minor issue, compared to the hydrological calculations. 

I hope you agree with this assessment.
 
Al Arking

Jan. 21, 2005

Fred:

I now realize that some of the explanations I wrote you are not correct, any may very well have been confusing.  But the figures I sent and the analysis behind them ARE correct.  Regretfully, my written explanations were written hurriedly, and I was not paying attention to the info content of the figures.  So, let me now correct my written comments. 

What is missing in most RT codes, including the ones based on line-by-line calculations (the closest to theory) is some absorption in the atmosphere that my work shows is related to WV.  (A number of years ago, some well-known scientists were claiming it was associated with clouds, and referred to it as "anomalous cloud absorption, but I do not think anyone believes it anymore.)  Most GCMs show that bias.  In some GCMs the bias is as much as 30 W/m^2 in mean daily atmospheric absorption out of ~80 W/m^2.  However, the GCMs have little or no bias in TOTAL absorption (i.e., atmosphere plus surface), which is ~240 W/m^2.  And it is total absorption that is behind the WV feedback associated with the net radiative flux at TOA (to which one usually refers in discussions of WV feedback). 

Of course, there are many other feedbacks in which WV plays a role---e.g., indirect radiative effects through the effect of WV on clouds and aerosols, and non-radiative effects through the effect of WV on dynamics and the hydrological cycle.  

It is interesting how GCMs with highly biased radiation codes can get the total absorption (which is 1 - albedo, when expressed in units relative to the incoming solar flux) correct.  Simply, there are too many knobs that can be turned---e.g., albedo of the surface and clouds, cloud amount, etc.  In essence, if the albedo is correct, then the simple WV feedback is correct.  However, the OTHER feedbacks involving WV may have large errors, and this is what I was showing in Fig 6 of my previous msg, where there was a large change in the SH +LH at the surface.  (One gets substantially the same results with a GCM, as shown by Fei Liu, my former student, in his Ph.D. dissertation.)

By the way, the question of whether the overall feedback is positive or negative depends on what other processes are in the model.  In my 1-d radiative-convective model, the feedback is POSITIVE, because total absorption increases with WV, but the feedback is very small, because most of the solar energy not absorbed by the atmosphere is absorbed by the sfc.  Without any absorption, the change in sfc temp due to doubling CO2 is 1.96 deg; with the standard code, 2.01 deg; enhanced RT code, 2.11 deg. 

I hope this clarifies things. 

Al Arking

March 19, 2005

Dear Hartwig,

When I re-read this material and Arking's comments, I am more convinced than before that there is an important and salient point here. I find Arking's answer very unconvincing. Could you re-transmit the figures from Arking as they were not included in this email.

It seems to me that Arking's answer can be easily re-butted. He says that, if the NIR is not absorbed in the atmosphere, it is absorbed at the surface. Thus, it doesn;t affect total forcing at the top of the atmosphere.

However, the whole theory of anthropogenic CO2 forcing is that outbound absorption in the atmosphere is plugged up a little more, causing a steeper lapse rate and warmer surface. If inbound absorption is plugged up a little more, it would surely completely offset this. Obviously both you and I agree on this.

Again re-reading the earlier materials, the failure of the GCMs to correctly model atmospheric absorption, the admitted uncertainty of the NIR water vapor parameterization and the likelihood that the models would be improved by increasing water vapor NIR absorption - points to the possibility that there is a very real problem here as we surmise.

.... Everyone thinks that these GCMs and infrared parameterizations are a lot more certain than they really are. Belmiloud's comments are really quite scathing on the infrared parameters.

It would be a pretty interesting project to work through the NIR issues.

Regards, Steve

March 22, 2005

Steve,

...  It seems to me that neither Fred Singer nor Arking understood the argument, which below you outline correctly. (Formerly I also tried to show that the NIR CO2 absorption might include a substantial negative forcing in the NIR). Unfortunately some time ago I found it close to impossible to get reliable absorption data to further quantify this subject, too many variables, temperature and pressure dependant. Something else, which might back the NIR aspect: the strongest warming of the last decades or century is observed in the driest world regions (regions which are cold and have an extreme "continental", i.e. dry climate). This might be a further indication of an unknown negative WV feedback. The warming in more humid areas is simply not there. Corresponding figure with most continental climate (highest thermal amplitude) attached.

regards,  Hartwig