Climate variables are described by non-linear stochastic relations, which in simple language means that they cannot be used to project or forecast anything. However, like econometricians, many climate scientists are in awe of the processing power of computers, but utterly unaware that garbage-in = garbage-out. This is a continuing problem with the social sciences - using techniques of the physical sciences without understanding the limitations of those techniques.
Post Modernists would, however, disagree with this statement. As they would.
Louis at Bizarre Science
Well, that explains why the Bureau of Meteorlogy has such a hard time predicting Melbourne's weather; basically they're a bunch of jumped up sociologists. Maybe it's about time they employed some real scientists.
Commenters are cordially invited to speculate on possible post-modernist approaches to meteorology or to compose short excerpts from academic papers by deconstructionist climate physicists etc.
Update: there's also a chuckle or two to be had from the comments thread, where Louis questions the statistical validity of atmospheric temperature measurements:
It is not too difficult to show that measurements of temperatures, as published, are statistically unsound. The science of statistics was initially conceived to deal with measureable attributes of discrete objects, be they humans, cannon balls, marbles, tennis balls, etc. When restricted to these categories, statistical analysis is sound.
However when extended to mining and other phsyical objects, statistics hit a conceptual wall. When one measures the temperature of the air at a particular place, say in Australia, what is one measuring?
Is one measuring the attribute of some physical object? If so, what is that object.
In the absence of an objective basis, measurements of "un-objects", while measuring something, cannot be amenable to statistical analysis, since the measurements do not have, in the statistical sense, uniform "support".
What was that about post-modernists again?