Only_Truth,
“With the computer models, failing to account for even just one variable adequately (or even at all) results will skew the results years from now far more than it will in the short term.”
Obviously, the coin analogy is much simplified. With respect to your point, I think it a makes some sense, but the climate models are a boundary problem and not an initial state problem. So what matters is that the physical processes are modeled appropriately. To model every physical process is, as many note, impossible. However, the idea that some short term errors will propagate exponentially is not supported by looking at the models runs themselves, which are quite stable on timescales of centuries and much longer. Moreover, as I noted above, models have been accurate in terms of predicting the Mt. Pinatubo response and other things, like response to ENSO and paleo and 20th century hiundcasts.
So, I agree that your concern has merit in theory, but the evidence suggests it is not a problem.
