First off, let it be noted the mistakes of the IPCC are not in the technical studies. These are largely flawless. The colossal mistakes of the IPCC are in what was not done, what was left out, what was ignored. These are the mistakes of the administrative function of the IPCC and not the scientific contributors. The first category of the crucial missing parts concerns the role of water in the Earth's climate system. The climate projection models presume that as CO₂ and the other non-water greenhouse substances increase there will be an increase in temperature which will result in more water vapor in the atmosphere and hence a positive feedback effect. But, since the amount of water vapor in the atmosphere is 25 to 100 times as much as the amount of CO₂ a small downward fluctuation in the amount of water vapor can cancel out the entire effect of the CO₂. It is a very small tail that climate modelers are presuming will wag the dog.

Fundamentally the major problem with the IPCC is that it did not function as a truth seeker; instead it functioned as a true believer focused upon anthropongenic CO2. In doing so it left out the autonomous anthropogenic water vapor and anthropogenic cloudiness. All of the water vapor put into the air through irrigation and the increased transpiration from farm crops compared to native vegetation and the mining of the aquifers such as the Ogalala.

The IPCC should have put a major effort into compiling times series of statistics on water vapor content of the atmosphere around the globe. Because the absorption of thermal radiation by the atmosphere is a nonlinear phenomenon (see Saturation) the information has to be compiled by at least latitude rather than relying upon global averages. However, it is not just water vapor that needs to be considered. Clouds and their content, denseness and altitude also enter into the picture. It is indicative that the climate modelers prematurely resorted to computations and later they had to incorporate the effects of aerosols and clouds. But for all the billions of dollars spent on global warming research there is still no figure available for the average water vapor content of the atmosphere and no graph showing the total effective content of greenhouse gases in the atmosphere over some period, say 1970 to 2000. And, to repeat what was noted above, because of the nonlinearity of the greenhouse effect what is needed is the global profile of the time average of water vapor content by latitude and longitude.

The climate modeling should not have begun until the statistics on the water regime of Eatht's climate system were available. Fundamentally the problem is that the IPCC started off with the notion that that its assignment was to prove that anthropogenic CO₂ was the problem rather than to seek the truth. Anthropogenic CO₂ is part of the problem but so is anthropogenic water vapor and anthropogenic cloudiness. These latter two phenomena are so highly correlated with anthropogenic CO₂ that it is difficult if not impossible to separate their effects statistically.

However, even with the flawed models the analysis could have been vastly improved by backcasting. That is to say, beforer the modelers started generating projections a century into the futures they should have generated backcasts of the climate over the past century and half. A comparison of the backcast with the recorded data would have revealed how much confidence could be placed in the raw projections of the models. However the proper thing to do is regress the recorded values on the backcast values. The estimates based upon the regression equation have the systematic biases in the projections eliminated. One can call this process the calibration of the model. one could have a greater degree of confidence in these calibrated projections of the future climate than the raw projections. Additionally the regression would give the statistical information needed to compute confidence intervals for the future projections, somethng that the IPCC projections now sorely lack. As it is now the IPCC gives the range of projections of about 15 different climate models and asserts that has some relevance for the variability of the Earth's future climate rather than being merely a reflection of how uncertain the IPCC is of what constitutes a good climate projection model.

Another really interesting topic concerning backcasting of climate models is how the methodology for estimating the variability of weather holds up under such tests. The IPCC gets the variability of future weather from a tabulation of the projections from the estimate of the initial state of the world. Small errors in the initial state can lead to big differences in the projections. I would love to see that methodology tested by backcasting.

Backcasting on a regional level would also be highly interesting. There is a good chance that the correlation of the regional backcasts with the actual regional data is so low as to dictate the abandonment of such regional future projections. The IPCC now confesses to such a wide range of uncertainty about such regional projections that their use can only be misleading.

Correction of just the above two categories of flaws would greatly enhance the acceptibility of the IPCC projections, if in fact they deserve acceptance. Here is the time series for global temperatures and backcast values for a model from the Canadian Centre for

The model does not do very well and overestimates the temperature change from 1905 to 1993 by about 100 percent. The regression of recorded temperatures on the model temperatures confirms that the model has a bias of overestimating temperature change by about 100 percent. When this bias is corrected for the picture looks much better.

The above case indicates that the climate modelers should not avoid backcasting but should embrace it to calitrate their models. Of couse the calibrated Canadian model will not produce the extreme and sensationalistic projections it has become notorious for.

The use of regression analysis does not have to be limited to the correction of bias in the individual models. It can be use to generate an average for the 15 models that gives greater weights to the models whose backcasts are more highly correlated with the past data. As it is the IPCC uses a simple average that does not differentiate between the better models and the poorer models. This would be a worthwhile exercise.

(To be continued.)