[t vago]

And another money quote:

Back in the 1980s, climate models were very crude simulations of the greenhouse effect. The main test of a climate model is to start sometime in the past and "predict" the present, with all the temperature swings and ice ages and so forth in between. When scientists tried this out on their early models, they got silly results, such as severe ice ages occurring in the 20th century. To avoid this kind of "drift," scientists applied a sort of fudge factor to ensure a sensible outcome. This doesn't do much good when it comes to predicting the future, which may be why 1988 predictions of rapid warming by 2000 never panned out. The average temperature hasn't climbed at all.

The change adds to the models' credibility, but does it mean they are reliable in predicting the future? It doesn't, Lindzen argues. For one thing, added complexity does not ensure that the models reflect what nature is doing. Take the case of aerosols--dust and other particles in the atmosphere. Scientists realized only a few years ago that aerosols reflect light and may exert a cooling influence; their effects are poorly understood. Putting them in climate models is essentially the same thing as adding a fudge factor. "There are no records of aerosol production before the 1960s," Lindzen says. "So you have complete freedom to adjust the amount of aerosols to make the models replicate the temperature record."