Jed Rothwell wrote [to Rick Monteverde]:
> ... as you and I agree it [global warming] is happening. The cause is the > only question. Yes, you and Rick agree, and only argue over the cause. However, part of the reason I posted my comments about Alaska and Canada, and almost posted a snide comment about those who think a few cool months in 2008 "prove" global warming isn't happening, is because not all participants in this thread agree with the two of you. In a message early in the thread it was stated -- *not* by Rick: >> Could a significant global cooling effect be taking place.? I notice >> there is a deafening silence from Pope Algore and his Church of >> Global Warming on this subject. It would be very inconvenient for >> the selling of carbon indulgences, oops... that's offsets. Nothing >> is made of the fact that 2007 saw the largest one year drop in >> average global temperature in recorded history. Didn't hear about >> that did you? Again, that was not Rick talking. However, Rick did make a point here which you, Jed, may have overlooked. You said: > And I say [ ... ] if people can predict the weather > tomorrow in six months or a year in advance they can darn well predict > it 10 or 20 or even 50 years in advance, although obviously not for any > particular spot on earth. If you understand how the atmosphere will work > in the next 24 hours you can understand to some extent how it will work > cumulatively for the next 20 years. Evidence suggests that the climate on Earth is a chaotic process, and chaotic processes may behave in such a way that they are simply *not* *predictable* over the long term, save within very broad bounds. A common example seems to be El Nino. Its behavior can be predicted for a few months, but trying to predict whether there will be an El Nino event in progress as few as 24 months from now is hopeless -- it's chaotic, and flips from one mode to another as a result of tiny perturbations. So, while it makes intuitive sense to say "if people can predict the weather tomorrow ... they can predict it 10 years in advance", it doesn't actually follow from the science. The process could be such that error accumulation renders predictions worthless when trying to look more than a small number of months out into the future. This sort of effect is not an artifact of current computers; it's apparently a fundamental feature of the process being modeled. It's like an NP-complete problem -- you can solve it for small datasets, but the nature of the problem makes it intractable when the data set grows large. In the case of an NP-complete problem the complexity (and time to solve the problem) grows geometrically with the dataset size. In the case of a chaotic process, the precision required in the calculations (and measurements) grows rapidly with the length of time over which you want your prediction to be good. In both cases, no matter what sort of hardware you're running on you'll run out of horsepower in short order. It's no coincidence that one of the most powerful machines in the world is named "thunder". Weather prediction consumes incredible numbers of computrons.
