Climate Change & Models

Here is an excellent reason why I don't have a lot of faith in "climte warming" based on GCM (General Circulation Models). This is from a posting by Ben Herman on the Watts Up With That? blog:

We are looking for effects to cause temperature trends on the order of 0.1 deg C per decade. I will limit the following discussion to the transfer of solar energy only. Years ago when we were developing our radiative transfer models, we required solar energy conservation from a model to within about 0.1%, with no solar absorption by the atmosphere.

To accomplish this degree of conservation required an angular grid of 108 angles, and an optical depth increment of 0.01 for the numerical integrations of the relevant equations. Since it was necessary to also include polarization to avoid errors of up 10%, this introduced a 4×4 matrix into the calculations, increasing the number of equations by a factor of 16.

For a nominal optical depth of 0.1, this made 10 levels, so there was a system of about 16,000 simultaneous, highly coupled equations to solve. We assessed the resulting degree of energy conservation by summing the outgoing irradiance at the top and bottom of the atmosphere with the incoming solar at the top. This was with model that I would guess, and I stress the word guess, was much more accurate than what they use in the climate models. Now, if we assume that there was an absorption of solar energy in addition to scattering, we could not assess the accuracy because the irradiances at the top and bottom no longer equal the incoming solar due to internal losses from absorption.

Now,this was with a flat atmosphere model. With a more realistic spherical atmosphere, for various reasons due to the complexity of the model, it was very difficult to check for conservation of energy but undoubtedly, the conservation was not as good as with the simpler one dimensional flat atmosphere. We could check for programming errors in the spherical model by letting the radius of the sphere get very large (approximating a flat atmosphere) and comparing results to the flat atmosphere model. The agreement was very good but not exact (I can’t recall the deviations). The small differences between conservation and model results were due to numerical errors.The transfer equations must be solved numerically and this leads to the errors.

The above was for one wavelength. Now, to integrate over wavelength, other approximations are made which introduce additional errors. Other uncertainties due to the addition of aerosols with unknown composition, number density, shape effect, vertical profile, optical depth, complex index of refraction, etc have been well discussed and I will not go into that here. We have also assumed no clouds. The above applies just to the solution of the transfer equation, assuming all of the quantities mentioned above are known. While all of these errors may be quite small (some are undoubtedly not small) are they small when trying to predict resulting temperature trends of 0.1 deg C/per decade as I stated above? I don’t know but they certainly cannot be dismissed without careful consideration. The equations are non-linear in optical depth so the trends themselves will also have errors if the irradiances are incorrect.

The above applies primarily to transfer in the solar spectrum. In the IR, where scattering (except by clouds and precipitation) can be ignored, the problem is less complex, but the integration over wavelength is still a major issue. Again, I will not go into that here, but I believe I have raised some questions which could be of importance in obtaining the accuracies required for climate change issues.

That gives you a taste of the immense complexities of trying to "compute" climate. The above is a part of the climate that isn't even addressed in GCMs but is highly relevant to getting the right answer. That should give you a pretty good reason why you should be skeptical of claims of "accuracy" in predicting global warming.