What did the IPCC climate computer models of 30, or 20, or 10 years ago predict the temps would be in 2008? The article says the predictions “don’t match well” with what actually happened.
It's OK... you just keep changing the models each year to allow for what actually happened... get the terms right, and you can still predict upcoming disaster.
That's the nice thing about the “garbage-in, what-you-want-out” approach.
The criteria of “correct” is “do I get next years grant?”.
It's not necessary to use the older models, they could take their latest and best model and just load in the data up to thirty years ago and have the computer calculate a "projection" of the "next" thirty years. The projection could then be compared to the actual data from available records. I fully expect that such "calibration" runs have in fact been made as part of the modeling process. Since they have never been made public, I expect they would revel a lack of correlation sufficient to invalidate the model.
There are some problems that are considered to be "computationally unsolvable". That is, the math is not all that difficult but a solution requires such a vast number of simple calculations that the time required starts to approach the expected life of the sun or some equally daunting interval. One such problem is factoring very large prime numbers, that simple fact gives an unbreakable cipher system which is in use everyday to secure everything from email to bank and credit card transactions. The protection being the vast amount of time it would take to decipher the code.
I expect that modeling the earth's climate accurately may be such a problem. If true the only way to get a model to work in a reasonable amount of time would be to simplify the model by plugging various dimensionless coefficients into the equations and tweak them until the results matched the historical record to the desired accuracy. The problem with this approach is that the model is only accurate within the time interval you used to establish the values of the coefficients (sometimes known as "Fudge Factors"). A Case in point:
A person could take the Dow Jones closing averages for every Friday in 2006 and fit a sixth order polynomial equation with excellent accuracy. Such an equation would allow you to come up with reasonable estimates for the DGA at any date within the year 2006 but would be wildly in error either before or after that year. The same would be true for climate modeling.
Regards,
GtG