Climate models predict one thing, observations show another:
This research, published on-line Wednesday in the Royal Meteorological Society’s International Journal of Climatology*, raises new concerns about the reliability of models used to forecast global warming.
“The usual discussion is whether the climate model forecasts of Earth’s climate 100 years or so into the future are realistic,” said the lead author, Dr. David H. Douglass from the University of Rochester. “Here we have something more fundamental: Can the models accurately explain the climate from the recent past? “It seems that the answer is no.”
I looked at that paper and found it most interesting.
The models agreed pretty well on the surface temperature records and diverged significantly away from the surface.
This would tend to indicate that the models are not based on first principles but are tuned to the surface record. If that is in fact the case the odds of any of them predicting the future are about nil.
Why? Such divergence indicates that the modelers do not understand what they are modeling. It is also possible that the data used to develop the models has significant uncorrected error (some suspect the urban heat island effect). If you tune the model based on erroneous data, obviously the models will not reflect reality.