What bothers me about the climate models is that there is almost no opportunity for that sort of feedback loop. We only have a few decades of hard data, and they are spotty. Worse still: those are the data that are being used to develop the models, so they can't honestly be used to test the models. All you really get is an empirical fit to the existing data, whether you set out to get that or not. (If you've worked with ensembles of neural networks, you know what I mean: a trained network that performs beautifully on its training set can fail infuriatingly on a new data set.)
Moreover, the Earth's climate is complicated. Hadronic shower development, by contrast, is a well-understood process governed by simple equations, but it is still challenging to model properly. There are always approximations to be made, series to be truncated, and numerical instabilities to be overcome.
If I am skeptical of the models, it's a skepticism I've earned through years of experience. I worry that most of the scientists who accept the models do so from a position of ignorance, basing their trust on some of the spectacular, public successes of computer models in other contexts. Those who made the models, while not ignorant, of course, believe the models because they have convinced themselves they're right. (It hurts to work that hard and be wrong. This I also know through experience.) Time will tell either way.
Thanks for the feedback. What I obviously forgot is that a computer model is just as prone to error or misinterpretation as any other attempt at physical modeling.
The problem is not the scientists or their studies but the socialist who push public policy based on these climate models.
You have large foundations giving huge sums of money to enviro groups who in turn file law suits to change public policy. These law suit serve to cut off the domestic competition to the large foundations overseas logging, mining, and agricultural interests. They also serve to socialize our natural resources.
Well said. There is also a community bias to modeling, which few modelers are aware of.
Whenever there is a modeling community for a problem, there will emerge an expected answer (sometimes from nonmodeling sources). When models produce results different from the expected answer, the models will be looked at carefully. New elements will be brought in. Gee, we're just using a multiplier here, we'd better go into more depth.
But when models produce the expected answer, they won't be looked at so closely. This introduces a heavy bias to the process, independent of any more deliberate bias, e.g., political or financial.
It's as if you take a coin from your pocket and want to prove it is weighted towards heads. If it's heads on the first flip, you stop. Otherwise, keep flipping until you get more heads than tails.
2. Damn lies
3. Statistics
Mark Twain