For example, recent studies have found that clouds below a certain altitude act to trap heat while clouds above that altitude reflect it back into space. None of the models come close to factoring in these cloud effects. From all that I've read, none of the existing models can start with the known climate conditions at some point in the past, say 100 or 1000 years ago, and can then work to predict what the climate is today. If that's true, then the models are at best unvalidated and at worst useless. If the model can't show what actually happened over the past 100 or 1000 years, who in his right mind would believe it would somehow magically be able to predict the next 100 years? In short, if a model can't "predict" what actually happened, it's foolish to believe it can predict what will happen.

The ability to validate the model against the known climatic past is key to accepting any predicitions from the model about the future. To give an idea just how difficult it is to model atmospheric behavior, NOAA has been working for years to develop computer models of hurricane behavior. Several models are in use but no one of them seems to be up to the job. NOAA has a comprehensive effort underway to evaluate the models predictions against actual behavior, either singly or in combination, in order to find which models are the most accurate. All of this is done to predict where a hurricane might go in a few days - a much simplier task than modelling the climate as a whole. I've done work for NOAA and found them to be a group of dedicated and intelligent professionals. This stuff is far more difficult than most people imagine. In addition to things like clouds, there are factors like solar variability, volcanoes, ocean factors, etc. that just aren't very well understood. Again, if we don't fully understand the factors that caused the many climate changes in the past, how can we develop accurate models to measure the effects of the change that's currently underway?