A Discussion of Water Vapor Feedback in Climate Change

"Anthropogenic greenhouse forcing and strong water vapor feedback increase temperature in Europe" by Rolf Philipona et al. (GRL, 2005, subscription required for full text), which has attracted a certain amount of media attention. The overall goal of the paper is to understand, from a physical standpoint, why European temperatures have been increasing three times faster than the Northern Hemisphere average. It focuses on the changes between 1995 and 2002, over which time good surface radiation budget observations are available. The paper reports some results on the role of large scale circulation changes (which they conclude are minor) but I'll concentrate on the results relating to water vapor.

The most interesting result may be summarized as follows. Measurements from a network of six Alpine surface budget stations indicate that the primary radiative forcing driving the increase in surface temperature is an increase of downward clear sky infrared from the atmosphere to the surface. The annual average increase in this term is nearly 4 Watts per square meter between 1995 and 2002. Net cloud effects are relatively less important. Moreover, the increase in downward clear sky infrared is correlated with an increase in atmospheric temperature, and also an increase in the water vapor content of the surface layer of the atmosphere. Using a simple radiation model, the authors conclude that about a third of the increase in downwelling infrared is due to the increase in atmospheric temperature, and the rest is due primarily to an increase in the water vapor content of the low level atmosphere. This happens because water vapor is a greenhouse gas, so increasing the water vapor content makes air act more like a perfect blackbody emitter, if the air is not already opaque to infrared. In this case, increasing water vapor content will make the air a better absorber and emitter, even if its temperature doesn't change. From this result we learn that: (a) observations confirm the expected increase of low level water vapor content with temperature , and (b) the increase in water vapor accounts for the bulk of the increase in downward radiation heating the surface.

The authors then subtract off the part of the downward infrared radiation increase attributable to temperature and water vapor increase, and thus estimate the part due directly (as opposed to via feedbacks) to the increase in anthropogenic greenhouse gases such as CO2. They estimate this to be about one third of a Watt per square meter. This is not in bad agreement with estimates from detailed radiation models run by the authors, which say that the change in surface radiation due to the 12ppm CO2 increase between 1995 and 2002 should be about one fourth of a Watt per square meter. It is striking that the changes in the Earth's surface radiation budget due to anthropogenic greenhouse gases are so profound that they can be directly observed on a regional scale, over such a short time period. So far, so good. Physics seems to be working as it should, and climate scientists seem to be basing their understanding of climate change on rock-solid physical principles. The authors do not fall into the trap of assuming that water vapor is the root cause of the observed warming. They understand fully well that water vapor acts as a feedback to amplify forcing due to CO2 increase, and make this clear in their paper. This paper does not, however, deal directly with the problem of whether European warming can be attributed to CO2 increase. It only shows that, whatever mechanism is causing the warming of the atmosphere in this region, the surface warming is being amplified by low level water vapor feedbacks.

The accuracy of the media coverage of Phillipona et al. is decidedly mixed. The BBC got the scientific story straight (warming due to water vapor amplifying anthropogenic effects, everything working as it should, no worries about the physics, mate.), but their otherwise sound article was published with the unfortunate sub-header "Water vapour rather than carbon dioxide in the atmosphere is the main reason why Europe's climate is warming, according to a new study." This gives the casual reader the erroneous impression that the study concludes CO2 is unimportant. It feeds the old, discredited skeptics' notion that the water vapor greenhouse effect is so dominant that there's no need to be concerned about CO2. National Geographic is a little breathless: " The latest villain on global warming's most-wanted list is all wet—and a little surprising. Water vapor, experts say, is the culprit behind Europe's rapidly rising temperatures." However, they get the basic scientific story straight, quoting Philipona as saying "It is an experiment that clearly shows which factors are driving the higher temperatures. It is not the clouds, not the sun, not the aerosols. It is the increased greenhouse gases and the strong water vapor impact." UPI is probably the worst of the bunch. They state "Swiss scientists say Europe's recent rapid temperature increase is likely due to an unexpected greenhouse gas: water vapor." Unexpected? If they were readers of RealClimate, they'd know better.

All of this was relatively harmless, but all the coverage missed the boat in the same way. Press reports failed to note that the water vapor feedback discussed in Philipona et al. is not the same water vapor feedback usually discussed in connection with global warming. It is instead a surface water vapor feedback which adds additional surface warming on top of the usual things we talk about. The effect is already incorporated in the climate models used in IPCC forecasts, but the new observational study will be useful as a reality-check.

Phillipona et al. analyzed trends in the energy budget of the Earth's surface. While this is definitely an aspect of climate change, it comes as a surprise to many that the surface energy budget plays a decidedly secondary role in climate change compared to the top-of-atmosphere energy budget. The fact is, that even if the diligent Swiss authors of this paper had found that increasing CO2 contributed nothing to the changes in the surface budget, this would have in no way contradicted our understanding of the way anthropogenic greenhouse gases influence climate. For the most part, surface temperature changes are determined by perturbations to the top-of-atmosphere budget, and the surface budget is just dragged along, accomodating itself to whatever changes in surface temperature are demanded in order to be able to satisfy the top of atmosphere budget. It is impossible to understand the greenhouse effect without thoroughly understanding this point. Even the authors of Phillipona et al. seem to be a little fuzzy on this matter. They seem to think they are looking at the same water vapor feedback discussed in various review articles on the subject (e.g. Held and Soden (Annu. Rev. Energy Environ., 25, 441– 475. (2000)), Pierrehumbert et al. ("On the Relative Humidity of the Earth's Atmosphere" in The General Circulation, T. Schneider and A. Sobel, eds. Princeton U. Press 2005,) Pierrehumbert (Subtropical water vapor as a mediator of rapid global climate change. . in Clark PU, Webb RS and Keigwin LD eds. Mechanisms of global change at millennial time scales. American Geophysical Union:Washington, D.C. Geophysical Monograph Series 112, 394 pp1999), and the RealClimate article on the subject). but they are not. I shall try to explain.

In equilibrium, the Earth must lose as much energy out the top of its atmosphere as it gains by absorption of Solar energy. This is the principle of energy balance that controls the climate of all Earthlike planets. Currently our planet is out of equilibrium because the rapid rise of carbon dioxide is more than the slow response time of the oceans can keep up with; even if CO2 increase were halted today, the planet would continue to warm for a while as it comes into equilibrium. Planets only have one way of losing energy, which is by infrared radiation to space, often called "Outgoing Longwave Radiation," or OLR. The next piece of the story is that convection is always lifting air from the ground to high altitudes in the troposphere, causing the air to cool by expansion as it rises. This is the basic reason that temperature goes down with height in the troposphere. Convection and other dynamical heat transport mechanisms link together all the air in the troposphere, so that, to a first approximation, the whole troposphere can be considered to warm and cool as a unit. It doesn't matter much where you put in or take out heat from the troposphere.. It is mainly the net energy budget of the troposphere that counts. Now, if the atmosphere contains a greenhouse gas, the atmosphere will be partly opaque to infrared trying to escape from the surface. Infrared from the surface will be absorbed before it gets very far. As a result, the infrared that escapes to space comes more from the higher, colder parts of the atmosphere. Since infrared radiation increases like the fourth power of temperature, the radiation from these layers is much feebler than the radiation that would escape from the ground. On the other hand, the radiation into the ground comes predominantly from the warm layers nearest the ground.

This situation is illustrated in Figure 1, showing actual values of fluxes which I computed for a sounding over Paris during the August heat wave of 2003 (with an idealized water vapor profile having 80% relative humidity near the ground and 50% aloft). The red arrows in this figure originate at the mean altitude from which radiation escapes upward or downward. Because the radiation to space and the radiation to the ground come from different places, increasing the greenhouse gas concentration of the atmosphere would affect the two radiations in different ways.

If we increase the concentration of a greenhouse gas (say, CO2), then that makes more of the atmosphere opaque to infrared, and so the infrared escapes from yet higher and thinner and colder parts of the atmosphere. This would reduce the OLR, if the temperature of the atmosphere were held fixed at its original value. The planet would then be receiving more Solar energy than it gets rid of. Solar energy is primarily absorbed at the surface and communicated to the troposphere by surface heat fluxes. This energy input stays the same, while the reduction in OLR has reduced the rate at which the atmosphere is losing energy. As a result, the troposphere must warm until the top of atmosphere energy budget is brought back into balance. Remember that the whole troposphere warms more or less as a unit. That means that the air near the ground must warm along with the rest. In this way, we see that the warming of the entire troposphere can mostly be inferred just by thinking about the top of atmosphere budget, without bringing the surface budget into the picture in any detail. So far, all we need to know about the surface budget is that all the energy absorbed at the surface eventually makes its way into the atmosphere.

We are not done yet. We still have to say how this change in the tropospheric temperature translates into a change in the temperature of the solid underlying surface on which we live. This is where the surface energy budget comes in. The complication here is that, while the top-of-atmosphere balance has only one loss term (the infrared), the surface has many ways to exchange energy with the overlying atmosphere:

Sensible heat flux (warming or cooling air in immediated contact with the surface and then mixing it aloft by turbulent motions)
Latent heat flux (cooling the surface by evaporation)
Infrared heat flux (cooling by emission of infrared by the surface, and warming by absorption of downelling infrared from the atmosphere)

with latent heat flux tends to be the dominant term, because evaporation is such an effective way of transferring heat. In fact, in warm, wet places like the Tropical Pacific Ocean, the evaporative heat transfer is so effective that all the surface budget tells us is that the surface temperature must stay quite close to the overlying air temperature. In a case like this, we don't even need a detailed surface heat budget to say what the surface temperature change is -- it is just dragged along with the tropospheric temperature increase. Changes in the surface budget instead affect the amount of evaporation needed to close the budget, and hence affect the precipitation rather than the temperature. The buffering of the surface budget by evaporation limits the leverage of the surface budget on surface temperature over much of the rest of the globe, though not to the same extent as in the tropical oceans.

The preceding reasoning does not mean that changes in the surface budget cannot affect the surface temperature. The right way to view the system is that (approximately) the top of atmosphere budget determines the warming of the low level air temperature, while the surface budget determines the difference between the air temperature and the surface temperature. There are many cases where this could further modulate the primary climate change, adding to or decreasing the primary top-of-atmosphere driven warming. This is particularly the case when a formerly wet land surface dries out. For example, the hot Sahara sands are around 10 degrees C warmer than the overlying air in the daytime, because in the absence of moisture the relatively inefficient sensible and radiative heat transfers need to have a pretty large temperature difference to work with in order to get rid of the necessary amount of heat. This is also why a dry sidewalk (pavement, to UK readers) gets very hot on a hot summer day. If the Sahara were made moister (as it was some thousands of years ago) the surface would cool regardless of what CO2 is doing. Conversely, if the moister parts of North America dry out in response to CO2 increase, the reduction in soil moisture will compound the surface temperature increase. Getting back to the implications of Philipona's results, since Europe is not in a completely evaporation-dominated regime, the downwelling infrared increase could possibly allow the surface temperature to warm more rapidly than the air temperature, compounding the general global warming driven by CO2. Whether or not this happens depends in large measure on how evaporative and sensible heat fluxes adjust. This aspect of the problem was not treated by the paper. Philipona et al find that the observed downward radiation increases by roughly 2.7 Watts per square meter over and above what would be expected from the air temperature increase alone. This would lead to a surface warming of about six tenths of a degree C if it were balanced entirely by an increase in surface infrared cooling. Sensible heat flux would bring the warming down by about a factor of two. Evaporative heat flux would bring the warming down yet more, but at the expense of increasing the evaporation and aggravating the drying of soils. These climate changes are not inconsequential, especially in view of the fact that they have taken place over a relatively short period and come on top of the "normal" global warming driven by the top-of-atmosphere balance.

To see why the anthropogenic greenhouse effect does not, however, rely on the direct perturbation of the surface energy budget by greenhouse gas changes, let's consider an idealized limiting case. Suppose that the lowest dozen meters or so of the atmosphere is so full of water vapor or cloud water that it acts like a perfect black body. It is as opaque as it can be to infrared. Now suppose that we double the atmosphere's CO2 content. This doesn't increase the infrared emission to the ground, because the low level air already has so much greenhouse-substance in it that it is radiating like a perfect blackbody, whose emission is determined by its temperature alone. It is radiating as much as it possibly can, for its given temperature. In radiative transfer-speak, its emission is "saturated." Furthermore, since the low layer is opaque to infrared, the CO2-caused change in downward emission aloft does not reach the ground. Does that mean there can be no further global warming in this case? No! What happens is that the increase in CO2 throws the top-of-atmosphere budget out of kilter, forcing the whole troposphere to warm up to bring the planet back into balance. Convection links the whole troposphere, which means the low level air warms up. The warming of the low level air, in turn, increases the flux of energy into the ground by all three of the mechanisms enumerated previously. In particular, the downward infrared flux increases because the air itself has become warmer -- not because it has become more optically thick in the infrared. The increase in downward flux then communicates the warming to the surface. As Phillipona et al. show, the real midlatitude European boundary layer is not perfectly opaque to infrared, so increases in water vapor content or CO2 can directly increase the infrared heating of the surface. This is very interesting, but it is in no way essential to the anthropogenic greenhouse effect.

The water vapor involved in the effect of water vapor on infrared downwelling to the surface is almost a completely separate issue – a different water vapor, as it were – from the water vapor we speak of when talking about the role of "water vapor feedback" in the context of global warming.. Water vapor feedback of the latter sort is a consequence of the effect of water vapor on the top of atmosphere radiation budget. Water vapor near the surface has very little effect on this. Making the surface layer of the atmosphere a more effective infrared absorber/emitter has little influence on the infrared upwelling into the rest of the atmosphere because the temperature of the ground differs little from the temperature of the overlying air; one is just replacing one radiating surface with another radiating surface of practically the same temperature. In contrast, the relatively small quantities of water vapor aloft have a much greater effect on the top-of-atmosphere budget, because they increase the infrared opaqueness of layers of the atmosphere that are much colder than the surface; they block the infrared flowing upward from the warmer parts of the atmosphere, and replace it with "new" infrared emission from the cold layer.

That brings us to the second of the two recent water vapor papers, which is perhaps the more important of the two, though the subject matter it treats is less novel. . This one ( B. J. Soden, D. L. Jackson, V. Ramaswamy, M. D. Schwarzkopf, X. Huang, Science 310, 841 (2005); October 2005 (10.1126/science.1115602)., subscription required for full text) has been more or less ignored by the media. Soden et al. deal with the aspect of water vapor feedback that affects the top of atmosphere radiation budget. The analysis consists in using various satellite observations to compare the behavior of mid to upper tropospheric water vapor between a general circulation model and reality. The analysis is carried out for the period 1982-2004, corresponding to the period of satellite data availability. The basic technique is the "model to satellite" method, in which the model temperature and humidity are used to directly simulate the brightness of radiation that would be observed by satellites looking at the atmosphere in various wavelength bands. By choosing satellite observations that are sensitive to the higher-altitude water vapor distribution, one can zero in on how well the model is doing in these all-important regions. Because of the relatively short period of the comparison, this exercise should not be regarded as an attempt to detect a trend in atmospheric water vapor and compare it with models. Rather, it is a check on whether the model does the same thing to upper layer water vapor as the real world, under varying year-to-year conditions (which do contain a trend over this period, as well as other things, e.g. El Nino).

By examining infrared satellite data, Soden et al. find that upper-level moisture increases in warmer conditions, in much the same way as predicted by the model. Further, by artificially suppressing moisture changes in the computation of the synthetic satellite data, they decisively reject the hypothesis that the atmospheric upper layer water content stays fixed as temperature changes. Synthetic satellite data computed on the basis of this hypothesis look nothing at all like the real thing. The authors take their analysis even further. Because the radiation measured by the satellites depends both on moisture and temperature,there is the possibility that faults in the climate model's upper level temperature predictions might be leading to spurious agreement with the infrared satellite data. To rule this out, they make use of microwave satellite data that is sensitive to the mid to upper tropospheric temperature, in order to formulate a diagnostic that is primarily sensitive to upper level moisture changes rather than temperature changes. Again, they find that the data demand that the upper troposphere get moister in warmer conditions. They conclude: "Reproduction of the observed radiance record requires a global moistening of the upper troposphere in response to atmospheric warming that is roughly equivalent in magnitude to that predicted under the assumption of constant relative humidity." This is probably the most direct evidence to date that there is nothing terribly wrong about the way general circulation models handle water vapor feedback. This is quite remarkable, given the potential role of small scale cloud processes in moistening the atmosphere. To be sure, the analysis only deals with clear sky regions, but the moisture in these regions originates in the cloudy convective regions, and so it provides a fair test. In any event, within the cloudy regions themselves, the clouds rather than water vapor have the dominant effect on the radiation budget.

There would appear to be less and less room for skeptics to dismiss climate model predictions on the grounds that we aren't sure they do water vapor feedback right. The picture is about to become even clearer, as researchers begin analyzing microwave upper level water vapor data, which will allow the analysis to be taken deeper into the convective, cloudy regions. To be sure, there is still a gap in understanding what the models are actually doing, in that it is far from clear why such complex processes boil down to a simple behavior: that the water vapor over a deep region of the troposphere changes in such a way as to keep relative humidity approximately constant. I have some ideas on this myself, but the general picture is still very much a work in progress. Meanwhile, it becomes increasingly clear that whyever the models do what they do to upper level water vapor, there can't be anything too terribly wrong with what they are doing.

Europe can warm more than the average, and it can still be water vapor that is the aggravating factor, precisely because the water vapor feedback being discussed in the GRL paper is a different water vapor feedback than the one that provides the general amplification of the CO2 effect. The latter water vapor feedback works through its effect on the top of atmosphere budget. The one discussed by Philipona et al works through the surface budget. Quite different things. As for your comment about it being unfair that Europe gets the worst of NH climate change, it's a sad fact that physics is unfair. Global warming has global effects, unevenly distributed, and the worst polluters do not pay the worst price. That's precisely why global warming is a harder problem to tackle politically than more local things, like the old London smog. --raypierre]

There isn't anything in Philipona et al that says that the surface water vapor feedback they note should be specific to Europe. One could think of reasons it could be, having to do with wind or soil moisture effects, but they do not make a case for it. The case that surface water vapor feedback is in some way the root cause of European warming is weak to say the least. What I found interesting in the paper was the analysis showing how much water vapor increases contribute to the increase in downward infrared flux. This is an effect that can amplify temperature trends caused by any mechanism. It's not a new effect, but it is nice to see it come out so clearly in the observations. It would have been nicer if Philipona et al had done a better job of distinguishing surface water vapor feedback from the top-of-atmosphere feedback. --raypierre]

Rolf, it's nice to meet you. I'm glad you're reading this. Welcome to RealClimate. You're among friends here -- we're all just trying to understand how climate works as best we can. Pull up a chair... (raypierre here, and in the following)]

Changing greenhouse effect:
I do not agree with your explanation of measuring "changes" of the greenhouse effect looking from space or from the surface. I agree with you that if we measure the total flux we get different results because as you mention the bulk of the signal measured from the satellite is from a colder region than the signal measured from the ground, hence you show correctly that the flux measured from above is smaller than that from the surface. This is because part of the radiation is fully absorbed by greenhouse gases, or in other words some spectral regions are saturated, the atmosphere is opaque for these wavelengths.

But then your statement that "increasing" greenhouse gases would affect the two radiations in different ways, I do not agree. Let's assume that we are adding ozone in the stratosphere. If an instrument looks from below it sees the additional ozone because ozone absorbs and emits at around 10 microns in the atmospheric window. In the window the atmosphere is not opaque and the instrument sees through the entire atmosphere into space, and hence, it receives the signal from the additional ozone molecules. If an instrument looks from a satellite it sees all the way down to the surface and receives the same signal from the added ozone as the instrument from the surface. The only difference is that it gets a decrease while the one from the surface an increase. The value is the same. Also it is not important where the additional ozone is in the stratosphere or in the troposphere, it can be observed from space or the surface at any location.

Now if we add water vapor to the atmosphere it increases the greenhouse effect in the spectral regions that are not saturated not opaque, which means in the atmospheric window. It contributes to the water vapor continum in the window and this is similar to the ozone, producing a flux change that can be seen from above as from the surface. And this is the same thing for other greenhouse gases too. The warming depends on how much we additionally close the window with more greenhouse gases and this must be observable from both sides.

[Response: But what I have said just stems from radiation physics of the most basic sort. Since (as you agree) radiation emitting to space comes from a different altitude and passes through a different medium from radiation emitted to the surface, it stands to reason that it will be affected by addition of greenhouse gases in different ways. This can be verified an any spectrally resolved radiation model. To use the Paris heat-wave sounding as an example, if I keep everything else the same and increase CO2 by 12ppm, then (using the NCAR radiation model) the OLR goes down by .17 W/m**2, while the surface radiation goes up by .037 W/m**2. One goes down, the other goes up, and the amount one goes down is different from the amount the other goes up. I'd call that "greenhouse gas increases affecting the two radiations in different ways." CO2 is well- mixed, but for a gas like water vapor which has strong vertical gradients, there is the additional effect that changing upper level water vapor affects the emissivity at a level closer to that from which radiation is emitted to space. For example, using my Paris sounding again, if I increase the relative humidity above the boundary layer from 50% to 60%, keeping CO2 fixed, then OLR goes down by 3.8 W/m**2 while the downward IR into the surface goes up by 1.96 W/m**2. Your example would only work if the entire atmosphere were transparent in the unsaturated bands. ]

What is warming what?
You say: "Remember that the whole troposphere warms more or less as a unit. That means that the air near the ground must warm also with the rest. In this way, we see that the warming of the entire troposphere can mostly be inferred just by thinking about the top of atmosphere budget, without bringing the surface budget into the picture in any detail."
From papers in the IPCC report we learn that the lower troposphere is warming and the upper troposphere is cooling.

[Response: No, not at all. It is the stratosphere that is cooling, not the troposphere. The troposphere warms with a vertical profile that is given approximately by the constraint that things stay on the moist adiabat. See Figure 12.8 of the IPCC Third Assessment Report. The stratosphere can be somewhat decoupled from the rest of the atmosphere precisely because convection doesn't reach into the stratosphere. However, the stratosphere accounts for a relatively small portion of the mass of the atmosphere, hence a relatively small portion of the greenhouse effect. The effects of stratospheric cooling are not completely insignificant with regard to the energy budget, but they are a decidedly secondary effect. ]

Also, the surface absorbs about 50% of the solar radiation. Further, you mentioned yourself that a large part of the energy available at the surface is brought into the atmosphere by sensible and latent heat fluxes. So how can you claim that all depends on the top of the atmosphere radiation budget without bringing the surface radiation budget into the picture?

[Response: Because the surface energy budget must close; that means all the energy absorbed at the surface makes its way into the atmosphere. For many purposes one doesn't need to know exactly which term is doing the transfer. Note that I didn't say that the surface budget was irrelevant to everything. The surface budget is important in determining the temperature difference between the free troposphere and the solid surface, and in some cases changes in this temperature jump can add to or take away from the general tropospheric warming. To determine how much this happens, though, it is necessary to look in detail at how the evaporation and sensible heat transfers respond. Where there's a sufficient moisture supply, it's hard for the surface to air temperature jump to get large enough to make much difference. ]

Let us look at the measurements shown in our paper. In figure 4 we show that over the Iberian peninsula temperature decreased while water vapor decreased also. In central Europe however we see a strong increase of temperature over the same time period and a strong water vapor increase. We show the total integrated water vapor increase in the atmosphere but we can show that this is very well related to the increasing moisture at the 2 meter level over ground. In figure 3b we show that in central Europe the annual mean shortwave net radiation (that what gets absorbed) decreased by 1.1 Watt m-2 over the period. Hence the increased temperature is not due to solar radiation. But if we add to the shortwave net radiation the longwave downward radiation then we see in figure 3c that we have now good correlation between total incoming radiation and the temperature. Figure 3 shows always the changes for the different months (trend over the monthly means from 1995 to 2002). Hence, even though this is trivial we prove experimentally that the surface temperature is driven by the radiation fluxes at the surface.

[Response: Not exactly. To determine the change in surface temperature, you need to complete the argument by saying how evaporative and sensible heat transfers respond. Note that I didn't say myself that the increased downward IR flux COULDN'T lead to the temperature increase. It's just that the argument is incomplete. Also, there's nothing in the description you've given here that's incompatible with my description of your paper as dealing with a different water vapor feedback than the top-of-atmosphere feedback discussed in, e.g. Held and Soden.]

Now since shortwave decreased and longwave increased it looks like the warming is due to increased longwave radiation. The longwave radiation from the cloud-free atmosphere can only increase due to increasing surface temperature or due to increasing greenhouse gases, who additionally close the window.

Increased longwave radiation due to increased surface temperature:
The calculations which you present do not match our calculations shown in the paper because you did not take into account the apparent sky emittance. You only calculated the longwave radiation emitted from the surface with a respective temperature change. But the atmosphere absorbs and emits only part of the upward radiation. The apparent sky emittance is about 0.7 . If you use this you will find for the 2.7 Wm-2 the temperature difference of 0.8°C as we show in figure 3a).

[Response: I don't understand which of my arguments you are referring to here. In Figure 1, I use the NCAR radiation model, and don't make any assumption about the apparent sky emittance. It is computed from the model. In Figure 2, I do assume a unit emittance, but that was just to make the point that greenhouse warming does not rely on greehouse gas increases being able to directly increase the downward IR. In cases when the lower atmosphere is transparent enough to allow the greenhouse gas increase to affect downward IR, it can give a bit of additional warming, but the amount depends on evaporation and sensible heat fluxes as well. ]

What else than anthropogenic greenhouse gases and water vapor is increasing the temperature?
After we had made the correction on longwave downward radiation for the surface temperature increase we are left with an increase on the annual means of 1.18 Wm-2 as shown in figure 3e. For individual months this can be up to 5 Wm-2. We show that this shows good correlation with the moisture change at the surface. A radiative transfer model and a sensitivity study helped us to subtract the part which is due to the absolute humidity increase. After that we are left with a longwave radiation increase of 0.35 Wm-2 and this is close to what is expected from anthropogenic greenhouse gases.

[Response: I agree that the most likely root cause of the European warming is anthropogenic greenhouse gas increases, but I don't see anything in your paper that actually supports the attribution. Process of elimination is not a very convincing argument. The only way I see to do the attribution is to have a model that reproduces the European warming, then re-run it without CO2 increases to see how things change. Has anybody actually done this?]

From measurements we know that aerosols rather decreased over Europe over the last ten years and this should have increased the shortwave radiation, but our measurements show a decrease. Hence, increasing water vapor in the atmosphere apparently absorbed more solar radiation and overcompensated the aerosol effect.

What else than greenhouse gases is left as a forcing if solar-, aerosol- and cloud- are not the culpit?

[Response: Well, something like a circulation changed forced by the NAO pattern (which may in turn be affected by greenhouse gases) might cause an increase in European air temperatures, which in turn would allow low level moisture to increase if there is enough moisture supply, which would then constitute an amplification of a signal driven remotely. Again, without looking at the response of evaporation and sensible heat transfer, one can't determine whether this amplification shows up as additional surface warming, or additional evaporation.]

Radiative transfer models:
We have been comparing radiative transfer models to surface longwave radiations in many occasions. As an experimentalist I have learned to appreciate these models, who very well agree with longwave radiation instruments that are traced to absolute standards. These models allow us to demonstrate what I explained above.

[Response: Is the way you do the radiative calculation explained in more detail in some other publication? As I mentioned in my reply to Ferdinand above, there isn't enough detail in your GRL paper to allow me to reproduce the result.]

We are not asking Raypierre nor you nor anybody else to be kind with us. But everybody should be fair and first thoroughly study our work before criticizing it. Whether our arguments are poorly thought out, future will show. Well, for the time being I wait and see whether you can take a climate model and fix it as you suspect, such that it provides better arguments to explain why temperature decreases in the southwest and increases in central and northeastern Europe, than what we have shown with measurements in our paper.

In the first paper the investigation is based only on measurements in the central Alps. This did not allow us to show where the additional water vapor came from. We thought that it might be related to a positive NAO index.
In the second paper we show the temperature in central Europe and compare it to the temperature increase in the northern hemisphere. We show what solar radiation is decreasing since 1980 and that only with the very sunny summer 2003 the slope gets slightly positive. We show the radiation budget for the period 1995-2002 and 1995-2003 in order to show the impact of the hot summer 2003.
In the last paper we extend our analysis over all Europe. Figure 1 and figure 4 shows with data from CRU and ERA-40 how the temperature increases respectively decreases in Europe over the last two decades. Figure 2 shows how temperature and water vapor evolves from 1995-2002 for the individual month. With this picture we show that the water vapor and temperature increase, which show a strong gradient from southwest to northeast can not be due to the NAO but is most likely due to water vapor feedback.

Let me quote what is in our paper. "Stand-alone MODTRAN radiative transfer model calculations show a +0.26 Wm-2 annual mean longwave downward forcing for the 12 ppm CO2 and other greenhouse gas increases in Europe from 1995 to 2002, apart from water vapor." We did not say that we have 0.26 Wm-2 just from the 12 ppm C02 increase! There are other greenhouse gases that did increase also from 1995-2002 and these are included, but again without the water vapor. The water vapor increase alone, which we measured makes 0.83 Wm-2 longwave radiation increase as shown in the paper. This is backed up by a sensitivity study between longwave radiation measurements, radiosonde profiles and MODTRAN.
I do not know what you did with your MODTRAN in order to get only 0.063 for 12 ppm. Since I do not have MODTRAN ad hand right now I did another simple calculation.
From 1750 to now CO2 increased from 280ppm to 375ppm. IPCC calculates for that increase a forcing of 1.4 Wm-2. Proportional to that numbers we get for 12ppm increase a forcing of 0.18 Wm-2. For all the greenhouse gases IPCC gave a forcing of 2.4. Again with the numbers above we get 0.3 Wm-2 forcing from 1995-2002. Hence our 0.26 is not far off.

[Response: You can do Modtran calculations using Dave Archer's web version of the model. Just follow the link in Ferdinand's comment. When I did the calculation using the standard Modtran midlatitude summer profile for clear sky conditions, I find that increasing CO2 by 12ppm only increases the downward IR flux by about .09 W/m**2. (In the model, set the detector to be at the ground, looking upward). I also tried using the NCAR radiation model, with various European soundings. For the warm Paris sounding, the optical thickness of the boundary layer cuts the change in surface IR to well below .09 W/m**2. If I use a colder sounding, I still only get a surface effect of .06 W/m**2. Low clouds would reduce the CO2 effect even more. Maybe methane and other GHG changes could bump the number up somewhat, but I would need to understand the details of what you did in order to understand how you get the number all the way up to .26. By the way, the IPCC radiative forcing numbers you are quoting are for top-of-atmosphere radiative changes, not surface radiative forcing. These are not the same thing, and I am wondering if the Modtran results you mention also used top-of-atmosphere results by mistake. ]

If you can show a direct correlation between the NAO index and temperature in central Europe for the years 1995 to 2002, I would be very interested. We did not find this correlation. In the early 1990 the NAO index was high and the temperature in Europe increased but since the late 1990 the index dropped but the temperature increased even more.

With figure 2 in our paper we show with the monthly changes that there are two different phenomenas which influence temperature in Europe. One phenomena influences temperature and water vapor similarly all over Europe. We have strong evidence that this is due to changes of large scale weather pattern (circulations). But then on top of that we observe a gradient of temperature and humidity changes from southwest towards northeast. There is no way to explain both phenomenas with large scale circulations. Hence we assume that the gradient is because water vapor increases due to water vapor feedback in regions where water is available for additional evapotranspiration.

[Response: Again, to complete this argument, you need to turn the radiative forcing into a surface temperature change using all of the surface fluxes, not just the radiative ones. Further, if the root cause of the European warming is low level water vapor feedback, then why don't GCM's reproduce it? They have all the physics necessary to moisten the boundary layer. It's possible that there is some problem with the parameterizatio of evapotranspiration, but it would have to be a pretty big problem to miss the effect by so much. As I said in my response to Ferdinand, the paper you cite in support of the claim that GCM's underestimate the feedback in fact only says that the particular GCM in question underestimates the European warming, for one reason or another.

Certainly, your results will be valuable to modellers seeking to check whether GCM's are doing low level water vapor correctly. I do hope we can agree on two things though:

(1) The water vapor feedback you are talking about is a completely different thing from the top-of-atmosphere water vapor feedback discussed by Held and Soden, and most other users of that term,

(2) The observed increase in downward infrared due to low level water vapor feedback means it is possible that this effect plays a role in the enhanced European warming, but the attribution of the warming to this cause would require additional analyses that have not yet been done. (--raypierre)]

[Response: Mainly he is not doing a proper intergral across the whole spectrum. When you do you get numbers as seen in the table here. - gavin]

[Response: The calculation Essenhigh describes is done in any standard radiative transfer model, and done very accurately based on well validated spectroscopic data. As the Chem and Eng. News respondent said in his comment on Essenhigh's calculation, and as is pretty much confirmed in Gavin's water vapor post, the 95% figure for water vapor vs. CO2 is bogus. The real figure is more like 2/3 water to 1/3 CO2, and even that gives a misleading impression of the extent to which CO2 controls climate, given that the water vapor acts as a feedback and is more or less controlled by CO2. As a complement to Gavin's post, you can also take a look at my water vapor article from the Caltech general circulation volume. It has a graph showing the relative importance of CO2 and water vapor in determining the Earth's radiation budget. You can find the paper here . From the brief description of the work, I can't tell exactly how Essenhigh did his calculation, but in addition to the error Gavin noted, I believe that Essenhigh did not solve the radiative transfer equation properly, if at all. In essence, he seems to be looking at the amount of radiation from the surface which is absorbed by the whole column of the atmosphere. That is a different thing from computing the effect of water vapor and CO2 on the amount of radiation that escapes to space, since that radiation has contributions from a wide range of different altitudes. --raypierre]

Of course, water vapor is not well mixed. I think the difference must be in the vertical distribution of water vapor. As its concentration depends on temperature, it may be less concentrated higher up where greenhouse gases have the most effect. Does this make sense? Are there any figures for relative concentration of water vapor by altitude?

[Response: I'm pleased you read my paper. Either I was unclear, or you misinterpreted, what I said about the radiative effects of water vapor vs. CO2. Regarding the sensitivity, what I said (or at least meant) was that if you take the present water vapor concentration in "typical" midlatitude sounding, and doubled it or halved it at each level, you change the OLR by about 6W/m**2 as opposed to 4W/m**2.You can't use that figure to extrapolate back to zero water vapor and thus get the total effect of the present amount of water vapor in the atmosphere. That result is given, based on direct radiative calculations, in my Figure 1. I re-did that calculation myself, but the numbers are not particularly novel. They're essentially the same as what Gavin reported in his RC piece, based on a somewhat different calculation with a somewhat different radiation code. Let's not mince words, though -- there is no question that water vapor is extremely important to the radiation budget. As I showed in my paper, if you take out half the water in the atmosphere, you precipitate an ice age. If you take it all out, you through Earth into a globally frozen "snowball earth" state. --raypierre]

[Response: Why don't you think for yourself instead of just parrotting whatever sound-bite you've picked up from the Cato Institute or its equivalent. The answers to your questions have been amply dealt with here, and in any number of elementary and popularized books on climate. I just can hardly begin to count how many ways your are wrong. Nobody is ignoring clouds or cloud cover, not me, not climate models, not Philipona's paper. Clouds don't invariably cause cooling during the day because clouds can have a strong greenhouse effect -- and you've got no cause to ignore the night-time half of the equation. Increased water vapor does not lead to increased clouds, since increased temperature can dissipate clouds. Clouds are not simply related either to water vapor content or to temperature. Hence the idea of a simple cloud thermostat like you are describing is just snake oil. Ferdinand E. has a bit of a skeptical streak, I think, but I enjoy responding to his posts because he makes an effort to educate himself about what is already known, and as a result comes up with interesting arguments. Remarks like yours are just meaningless noise. --raypierre]

When I examine your Figure 1, I notice that the average outgoing longwave radiation seems to exceed the 342 W/m² incoming radiation. Also striking is the huge difference in outgoing radiation between the north and south polar regions. Is this because of the high albedo of the Antarctic ice cap? The other obvious observation is that water vapor has more relative effect in the tropics, where it is moister.

[Response:You're forgetting the cosine weighting of area vs latitude, or maybe looking at the wrong curve. For the OLR computed with observed water vapor, the net emission is 270 W/m**2. Allowing for an albedo of .3, the absorbed solar radiation is 240 W/m**2, not 342. The computed emission is about 30 W/m**2 greater than this because the OLR I show in the figure is the clear-sky value. Clouds bring down the OLR by the required amount. Note that when we say the albedo is .3, that also includes cloud effects. It turns out that clouds have a net cooling effect on the present climate, though that doesn't mean that changes in clouds will have a net cooling effect when you double CO2. As for North Pole vs. South Pole, that's mainly because Antarctica gets colder than the NP, which is because Antarctica is a continent and has a whacking great tall ice sheet on it. It's not primarily an albedo effect. All that, by the way, would apply to the annual mean. However, note that the figure in my paper is for January, when the OLR is lower at the NP because it's winter there, hence colder. Nice comments all 'round. I'm pleased you have gotten so much out of my paper.--raypierre]

In your response to my comment, you are asking whether we can agree on the two following things:

(1) The water vapor feedback you are talking about is a completely different thing from the top-of-atmosphere water vapor feedback discussed by Held and Soden, and most other users of that term.

To point (2) I agree that this requires additional analysis:
Let us look again at your second figure of your initial overview on measuring an increasing greenhouse effect from space or from the ground. You show that in an unperturbed situation (left graph) the outgoing longwave radiation (OLR) is emitted from the troposphere at a temperature of 255K with an irradiance of 240Wm-2. This irradiance is equal to the solar radiation absorbed by the planet. You also show that from low altitude at temperature of 280K longwave downward radiation (LDR) is emitted to the ground with an irradiance of 349Wm-2.

Then you show the out of equilibrium situation with increased CO2 (middle graph). With more CO2 OLR is now emitted from a higher colder region of 254K and hence irradiates 4Wm-2 less or 236Wm-2. From the lower troposphere you still show LDR being emitted from the same level and by the same amount. However, in your response to my answer #23 you agree that with increased CO2 LDR emittance increases even though the increase may not be as large as the decrease of OLR. I hope we agree, that with increasing greenhouse gases LDR is from a lower level with higher temperature and is further increased because the atmospheric window is additionally closed. Your graph should therefore show LDR from a lower level and 1 or 2Wm-2 increased and therefore show about 351Wm-2.

Then you show the equilibrium restored situation (right graph). Here the temperature of the troposphere increased and with it OLR by 4 Wm-2 and is now again equal to the unperturbed situation. Since the temperature increased also at the lower troposphere, LDR increased by about 3Wm-2 and now shows a 5Wm-2 higher value or 354Wm-2 than during the initial unperturbed situation.

If we assume now that the temperature of the entire troposphere increased more or less equally (as shown in IPCC Fig 12.8) then we must conclude that at the top of the atmosphere no change of OLR between 1995 to 2002 could have been observed. The idea of putting 2 times CO2 in the atmosphere is an interesting Gedankenexperiment. However, in reality CO2 increases by about 2 ppm per year and the observations show that the atmosphere with its rather low heat capacity responds to that quite rapidly. With the 90 ppm CO2 increase since 1750 atmospheric temperature already considerably increased. Of course the oceans and the earth itself have a large heat capacity and will therefore respond slowly and with a delay. However, it is the atmosphere with increased greenhouse gases which makes the additional insulation and this is what effects the changing radiative fluxes that we are talking about. The observed fact that temperatures increases slower over the oceans than over land demonstrates that the large heat capacity of the ocean tries to hold back the warming of the air over the ocean and produces a delay at the surface but nevertheless the atmosphere responds quit rapidly to increasing greenhouse gases.

With all this we must conclude that as long as the Earthâ??s albedo (shortwave) does not change, the OLR hardly shows any changes at the top of the atmosphere even if the greenhouse gases increase (the shift of the emission to higher levels is always compensated by increasing temperature in the troposphere). From outside planet Earth is therefore more or less in equilibrium because the atmosphere has responded to the changes. I say more or less because the slow uptake of heat in the ocean and the Earth mantle produces a long delay and therefore a very small decrease of OLR.

If we observe from the ground LDR increases if greenhouse gases increase since LDR is now emitted from lower warmer levels. Over land atmospheric temperature increases faster since the heat uptake over land is lower than over the oceans and with increasing temperature LDR increases even more. This is what we have measured and reported in our paper. Of course in our case not only CO2 increased but also other greenhouse gases and the water vapor, which made the LDR rises large enough that they can be observed even in the short time period.

To point (1): different effect of water vapor feedback at higher versus lower troposphere

Water vapor in the atmosphere increases with increasing temperature (if there is water available on the ground) and is therefore a feedback and with regard to the atmospheric greenhouse an additional insulator. If the temperature increases similarly throughout the entire troposphere then water vapor will increase at any altitude and will increase temperature at the surface. In the saturated spectral part increasing water vapor will shift OLR emission upward in the upper troposphere and LDR emission downward in the lower troposphere. For spectral parts in the atmospheric window additional longwave radiation will be emitted from the water vapor continuum, and this is observable from space and from the ground.

In our experiment we observed that the relative humidity more or less stayed constant. We also show that specific humidity as well as the integrated water vapor increased with temperature such that it more or less follows the Clausius-Clapeyron relation.

Under such circumstances it is not clear why increasing water vapor should be more important in the upper versus the lower troposphere. Water vapor and other greenhouse gases have an insulating effect that must be measurable on both sides. However, as previously shown, if measured from space it is masked by effects of the increasing temperature whereas if measured from the ground it is not masked. I have studied again the paper of Held and Soden (2000) to understand their arguments about different water vapors but I am not convinced. I would however refer to their final remarks where they state that their test of models are limited to observations of natural climate variability. Our paper shows increases of radiative fluxes measured at the surface while greenhouse gases did increase and water vapor strongly increased, most probably not due to natural climate variability.

[Response: Dear Rolf-- Since the comment form is closed and there may not be many readers with us any more, I'll just try to make a few final points by way of clarification and rounding out the discussion, without usurping the "last word." Naturally, I look forward to continuing this discussion in other venues.

With regard to the discussion of my Figure 2, please note that this was put in just to make the point that surface temperature can increase even if the lower atmosphere is so optically thick that increasing CO2 has no direct effect on the downward IR at the surface. I'm not saying that the midlatitudes conform to this idealization, nor am I saying that the sequence of large imbalance followed by large adjustment is a realistic description of what is seen in the atmosphere, which has a more gradual CO2 increase. It was just a convenient way to explain how the greenhouse effect works. One can also explain it via a change of radiating level for a system that stays near equilibrium (that's how I do it in my climate book) but in my experience most people find that explanation a bit harder to grasp. My reason for emphasizing a distinction between the role of the top-of-atmosphere vs. surface effects is that somebody who does a study like yours elsewhere in the world (e.g. in a cloudy part of the Tropics) could well find that the direct effect of greenhouse gases on downward radiation (i.e. the effect apart from temperature change) was insignificant. If they didn't keep the top-of-atmosphere effects in mind, they and their readers would conclude erroneously that the greenhouse effect wasn't operating, and something else was causing any observed temperature change. I hope there is nothing in my article that would seem to devalue your work, which I think is interesting, even though my interpretation of what is significant about it is different from what the press seems to have picked up on it.

On the matter of the radiative effects of upper level vs. lower level water vapor, I only want to emphasize that the effects discussed by Held and Soden and by many others do not derive from the parts of models that are debatable, but from radiation physics such as Kirchoff's Law and from band-models of radiation -- all of which are well confirmed by laboratory experiments, observations, and comparison with detailed line-by-line radiative calculations. The essence of the argument is simply that, in a part of the spectrum where water vapor is a good absorber (hence, by Kirchoff, a good emitter) when one puts some water vapor at a high, cold level one blocks the infrared coming from below, and replaces it with infrared emitted at a lower temperature by the high-cold layer. Putting more water vapor near the ground, where the air temperature is nearly the same as the ground temperature, does not do this because you are replacing one radiating surface with another with nearly the same temperature. In fact, if the air is warmer than the underlying surface, as often happens, putting more water vapor in the boundary layer will actually increase the OLR, until adjustment occurs to bring the system into balance. The other thing to keep in mind is that the radiative effects of water vapor are approximately logarithmic in concentration, like most greenhouse gases. That is why a doubling or halving of the relatively small amounts of water in the mid to upper troposphere is still radiatively significant.

What your paper made me think of, though, is that the midlatitudes are midway between the optically thin and optically thick limit. That means that the radiating temperature to space and the radiating temperature to the ground, are different, but not all that different. The heat wave example I chose in Fig. 1 somewhat exaggerates the difference as compared to cooler, drier profiles. That all means that the increase in infrared opacity due to low level water vapor causes an additional surface forcing that needs to be thought about. The degree of warming this surface IR forcing causes depends on the response of the other terms in the surface energy budget.

I want to conclude by thanking you for taking the time to contribute your comments to RealClimate. They add greatly to the educational value of our site. --raypierre]