Andrew Mason said:
My original question was: how does an increase in CO2 increase convection? Or, looking at it another way, how does adding CO2 to the atmosphere increase convection beyond what adding the same amount of nitrogen or oxygen to the air?
I have no idea how CO2 can influence convection all by itself. But what is true is that an increased CO2 content will increase the "opacity" of the atmosphere to IR, and hence as such redistribute absorbed and emitted power fluxes. This can have an influence on the temperature distribution in the atmosphere, which by itself is responsible for the drive behind convection.
But that is not the main point. IMO, the main point is that convection is a negative feedback mechanism (independent of whether there is CO2 or not) for radiative forcing of the surface, in that if the radiative forcing increases, then the surface temperature will also increase, and this will drive more convection, which will allow for a larger heat flux to transit through the atmosphere than when no such convection were present.
You are of course right that the total outward flux must equal the inward solar flux that is not reflected directly as visible light (albedo). So the total outward IR flux is fixed (at fixed albedo), and this, independent of what is the composition, and what are the heat transport mechanisms in the atmosphere. At least in a simple 1D model.
So we know (for fixed albedo) what is going to be the outward radiant IR flux. However, what we are interested in, is the total temperature gradient in the atmosphere needed to obtain that IR radiant flux. This net flux is everywhere going to be the same, at every point in the atmosphere. It will of course be composed of different partial fluxes: a radiant upward flux, a radiant downward flux, heat convection, and heat conduction. But the total balance, at every point in the atmosphere, must equal the same, fixed, outward IR flux. As such, the atmosphere (and even the vacuum) act as a kind of "resistor" in which a radiant flux is driven by temperature differences. For the "vacuum", that "resistor" is simply given by the black body formula: for a certain temperature (difference: with the CMB, but that's neglegible), we have a certain outward radiant flux. The atmosphere adds "resistance" to this: we need a bigger temperature difference to obtain the same radiant flux. That extra resistance is the greenhouse effect. It is due to the partial absorption and re-emission of IR radiation by layers of the atmosphere, which cause also a downward IR flux, and hence we need a higher upward flux to compensate, and arrive at the same net outward flux.
As such, we can think of heat to "make its way" through the atmosphere, and needing an extra "delta-T" each time it crosses a layer of atmosphere. The more the atmosphere absorbs, the more delta-T there is, and that's the basis of the extra greenhouse effect due to greenhouse gasses.
Also, these "delta-T"s will influence the temperature differences in the atmosphere itself.
So radiatively speaking, heat gets emitted from the Earth surface, is radiated a bit upward, then a bit downward, then a bit upward again, etc... and makes its way all the way up to the highest layers, where it is eventually emitted in space. The sum of all these final contributions must make up the fixed outward radiant flux. The more opaque the atmosphere, the "more difficult" this outward way is, and the higher the overall delta-T that establishes it. Hence, heating of the surface. Each atmospheric layer is a thermal resistor that increases a bit more the overall thermal resistance of this atmosphere.
But here's my point about convection: if there is a "second way" by which heat can be transported upward through the atmosphere, then it "shunts" part of those resistors. Heat can then, by this second way, reach the higher layers more easily than just by radiation from layer to layer, with each time a partial down radiation. And as such, it will lower the delta-T as compared when there were no such convection.
And now here's my point about feedback. If we look at the purely optical effect of increasing CO2 content of the atmosphere, as calculated by MODTRAN, without altering in any other way 1) the rest of the composition of the atmosphere and 2) any convection or whatever, then we find that for a doubling of the CO2 content, we need to increase the surface temperature by about 0.8K in order to restore the same outward IR flux as before (which, we agree upon, is fixed by the solar influx, and albedo).
Now, if you take it that the atmospheric composition also changes concerning water vapor, and you keep fixed relative humidity (instead of fixed total water vapor), which means that you suppose that at the surface, the wet surface will keep a similar equilibrium as the ratio between partial vapor pressure and temperature in the equilibrium case, but without more cloud formation or convection or anything, then you have, IMO, the maximal possible positive feedback from water vapor. MODTRAN then calculates that you need about 1.5 K surface temperature increase for a CO2 doubling, and the increase in water vapor (due to 1.5 K temperature increase) to have again the same outward IR flux. One would expect the right answer to be somewhere between the two. More water vapor probably means more cloud formation and so on (which increase albedo), but that effect is difficult to quantify. Water vapor is also lighter than air, so this might increase convection (what happens in a cooling tower). This might also decrease the effect of water vapor. Water will not evaporate more than given by the partial pressure equilibrium, so this case is the maximal water vapor feedback. So, without taking into account clouds or convection, according to MODTRAN, a CO2 doubling should result in a surface temperature increase between 0.8K and 1.5K.
Now, convection is a negative feedback which should reduce this needed temperature increase of the surface. I have no idea by how much, but I haven't seen any treatment of this.
But the IPCC talk about an average value of 3 K for CO2 doubling, in an interval of 1.5K to 6K. So there needs to be an extra positive feedback, which is not water vapor, and which is capable to bring this 0.8K to 1.5 K interval or smaller to the 1.5 K to 6 K interval. However, those mechanisms are not really explained. This is where I still have my question marks.