For most people who study global warming only casually it is well known that the greenhouse effect acts to increase the surface temperature of the planet (currently) by about 33 K (or 60 F) above the so-called “effective temperature”; this is the temperature value that a planet would need to have in order for the infrared energy it emits to space to balance the energy it absorbs from the sun (assuming the sun is the only important source of energy, which is true enough for Earth and neighboring planets like Venus and Mars). This is simple enough, yet there are still many popular misconceptions out there concerning the relative roles of individual greenhouse gases and the total mean climatology of the greenhouse effect, and some of these confusions have admittedly not been explicitly corrected in the literature very well.
A matter of curiosity from this point is to decide how much of the total greenhouse effect is partitioned between various radiatively active substances in our atmosphere. That is, how much of the natural greenhouse effect is fractionally supported by water vapor, by CO2, etc
There are a number of sources of confusion out there on this issue. For instance, this source claims that water vapor makes up 95% of the total greenhouse effect (in fact, it does so confidently that it actually says 95.000%, a good lesson in abusing significant figures and precision for lower-level science students, see Robert Grumbine’s post a couple years ago). Other secondary sources give numbers like 97 or 98%. Lindzen (1991, Quart. J. Roy. Met. Soc) gives an estimate in this range, although it is not clear where he gets his value from. Coby Beck here in a rebuttal to this claim asserts that “CO2 contributes anywhere from 9% to 30% to the overall greenhouse effect,” presumably giving the impression that there is a 21% disagreement amongst sources and experts out there. Still further, many people incorrectly extrapolate the effect of CO2 on the total greenhouse effect to deduce the forcing you’d expect with a doubling of CO2, or use similar arguments in relation to how we expect feedbacks to behave as the current climate warms.
Table 3 of the famous Kiehl and Trenberth (1997) [PDF] energy budget paper attempts to partition the various gases/clouds by percentage; RealClimate breaks down the contributions in this link while pointing out how water vapor is a feedback and not a forcing (see here for a summary). Values presented here are similar to Ramanathan and Coakley (1978).
Nonetheless, getting a clear account of all of this has remained elusive, especially as the most relevant source seems to be a 2005 blog posting. Some of the folks at NASA GISS including Gavin Schmidt and the radiative transfer guru Andy Lacis (along with Ruedy and Miller) have attempted to correct this situation, with a 2010 paper in JGR that is in press here [PDF]. I will attempt to summarize here.
Sorting this problem out is actually not very straightforward although it can be done with a radiation model. As an analogy, imagine having a large pile of laundry on the floor, with dirty shirts, towels, and pants. Suppose we’re interested in asking what fraction of the floor is covered by each individual item.
The total extent of the whole laundry pile has a rather well-defined value. However, asking about the individual contribution for each item is a tougher question, in part owing to the complex overlap between the various clothes. You can pull out all of the shirts (for example) from the pile and spread them on the floor individually and get an estimate for the area that they now cover. This would, however, dramatically overestimate the fractional contribution that the shirts originally had in the pile. Alternatively, you can pull out an item, and examine the extent of the new pile and you might come up with an underestimate for the importance that clothing article previously had.
Similar to the dirty laundry pile, greenhouse gases exhibit complex spectral overlap (primarily by water vapor and clouds, and in second by water vapor and CO2). The maximum effect would be if the greenhouse gas were acting individually, while the minimum effect would be when only that agent is removed, and these numbers can be quite different (as in the Coby Beck example). What’s more, if you removed two agents together (say water vapor + CO2) the effect would be different than if you remove CO2, put it back, remove water vapor and put it back, and then record the sum of those two effects (ignoring feedbacks such as water vapor dependence on temperature). In particular, the sum of the effect of each absorber acting separately is greater than if they act together.
The greenhouse effect is defined by the difference in upwelling radiation flux at surface and the flux at the top of the atmosphere. With no greenhouse effect, this difference is zero. In the present-day climate, this difference is about 155 W m-2; this atmospheric absorption and emission is what drives the ~33 K enhancement of surface temperatures above the no-greenhouse (and constant albedo) case. The authors define the change in this long wave flux reduction as their metric for the greenhouse effect, and add various gases individually to a greenhouse-free atmosphere or remove them individually from the modern (well, 1980) greenhouse atmosphere.
So what do they find?
First off, water vapor accounts for 39% of the net LW absorption if removed (so taking out the vapor would make the longwave absorption go down to about 60% of its present value), and 62% if acting alone; in that order, clouds make up 15 and 36% and CO2, 14 and 25%. All of the other greenhouse gases are very minor. In terms of the percent contributions after allowing for overlap effects when discussing the individual agents in the atmosphere, here are the numbers that should be cited:
Water Vapor: 50%
Other (ozone, methane, etc): 7%
For a cloud-free atmosphere, the numbers are 67%, 24% and 9% for H2O vapor, CO2, and others. These numbers are within a few percentage points of previously published estimates (or multiply by 155 to get the contribution in W m-2 flux reduction).
As one would expect, there is also variation over the globe. Water vapor for instance accounts for ~55% of the greenhouse effect in the tropics and ~40% at the poles where it is much drier.
The radiative forcing for a doubling of CO2 in this paper is about 4 W m-2, slightly above the detailed line-by-line calculations used in Myhre et al (1998) and cited in the IPCC 2001 and 2007 reports. The “forcing” for a doubling of water vapor would be about 12 W m-2, although clearly we don’t speak of water vapor as a forcing since its concentration is rapidly regulated by temperature. Indeed, the extra long wave absorption in the atmosphere is not 4 W m-2 when you double CO2, but more around 20 W m-2 illustrating the importance of positive feedbacks.
Some words of caution now. You cannot linearize about the greenhouse effect and project CO2’s percent contribution to the total greenhouse effect onto what you’d expect for a doubling of CO2.
To illustrate this, consider what happens if you remove all the CO2 from the current atmosphere. With no feedbacks operating, the planet would cool by ~7 K, as opposed to warm about 1 K if you double CO2.
Once you include feedbacks, removing CO2 from the current atmosphere in the GISS model at least cools the planet by ~35 K after water vapor and albedo kicks in, and triggers a snowball Earth where the whole planet is ice covered. This is consistent with other model studies (Voigt and Marotzke, 2009); in this case a snowball Earth is initiated by either reducing the solar insolation by 6-9% or by reducing the CO2 to 0.1% of its pre-industrial value, although for the CO2 this is a just a single run and they do not consider what other CO2 levels might also trigger a snowball. CO2 levels below, say, 100 ppm do not appear to be realistic in Earth’s history.
It should be noted following this that the canonical “33 K” temperature enhancement by the greenhouse effect is artificial, since it assumes the planetary albedo does not change when you add or remove the greenhouse effect. In reality, removing the greenhouse effect would greatly enhance the surface albedo from expanded ice cover allowing the planet to cool well below the “255 K” effective temperature.
Finally, all of this further reinforces the importance of feedbacks on climate, and that the very popular claim of “water vapor being the most important greenhouse gas” is a bit misguided, even if it is the largest source of infrared absorption in the current atmosphere.
Removing all of the water vapor from the atmosphere (and not replenishing it) would trigger a snowball Earth as well, but the non-condensable greenhouse gases (those which don’t precipitate from the atmosphere under current Earthlike temperature and pressures) such as CO2 would still be able to support a surface temperature of about 10 K higher than it otherwise would be. If you remove the CO2 and other GHG’s however, then you’d also lose a substantial part of the water vapor and cloud longwave effects, resulting in a near collapse of the terrestrial greenhouse effect. A significant water vapor greenhouse effect would not be sustainable without the “skeleton” provided by the non-condensable greenhouse gases, although it is obviously a significant amplification factor, both for the total greenhouse effect and its change in the future. It’s thus like the “skin” on a human or animal which needs the skeleton to hold it up, but provides the extra form and protection that we need to survive. This forcing-feedback distinction also makes CO2 the fundamental driver of global climate change (at least insofar as alterations to the optical characteristics of the atmosphere are concerned). See for example, Richard Alley’s AGU talk which focuses on CO2 as the largest control knob of climate change over geologic timescales. The water vapor is just dragged along with the temperature change, but then substantially amplifies any forcing to help provide the full magnitude of the temperature fluctuations; this is also a reason cold climates tend to be much drier than warmer ones.
All of this is moreso academically interesting than anything. Obviously we don’t live in a world where we are plucking out CO2 all together and then adding water vapor, or having a world where you can have clouds without water vapor, etc…but it should help to put into context the primary (water vapor, clouds, CO2) and secondary factors to the greenhouse effect, and put into perspective the important distinction between a forcing and a feedback.
Finally, claims that water vapor is 95% of the greenhouse effect in our atmosphere is just wrong, and the number cited in the Schmidt et al (2010, still in press) paper is so far the most explicit and detailed partitioning between the various gases/clouds (note that clouds on net cool the planet through albedo, although the study focuses on long wave greenhouse effects). The numbers might make the idea that CO2 is important to climate change more intuitive. For example, if the mean value is 20% of the greenhouse effect instead of just a percent or so, then you might think changing its concentration may be more meaningful…but keep in mind that sensitivity arguments must be evaluated on the basis of the change, and not the total greenhouse effect.