Reply #32: Here is the difference

[Nederland]

You can create a computer model that models a nuclear reactor under various conditions. You can then actually go to a real live nuclear reactor, duplicate the conditions, and observe the result. You can then compare the model results to the actual results to verify that the model is correct. You can fix all the variables in the reactor except one, or all except two, or any combination that you want because you have control over the system. You can run these tests over and over and over using countless variations until you have a great deal of confidence that your computer model is correct. You can compare your model to reality.

You cannot do that with a climate model.

You cannot fix solar radiance, methane, clouds and aerosols and play with the level of CO2 and observe the results to verify that the way you are modeling CO2 is correct. You can't run a virtually unlimited set of tests over and over to verify that your model is correct. The only thing you have is the last thirty years of actual climate data. The only thing you can do is plug the actual numbers of those last thirty years into your model and see if your model accurately predicts the observed results. That's it. Want to verify that your model correctly predicts what happens when CO2 hits 450ppm, aerosols decrease, and cloud cover stays constant? Too bad, you can't. Want to verify that your model when CO2 stays where it is but methane concentrations double? Sorry, you're shit out of luck. The only variations you can test are the ones that have already happened--and those aren't the ones you care about.

That is why you can't claim predictions that come out of a computer model are "science". They cannot be verified. They are not falsifiable. At least not until they actually happen.