The uncertainties related to climate modeling should come at a
surprise to no one. Butterflies flapping their delicate wings have
made weathermen the butt of jokes for decades.
Perhaps climate model predictions would be better viewed in the
context of the success, or lack thereof, of other complex systems
Macromolecular simulations have extensive data for parameterization
and ample opportunity for testing and fine tuning a model. Moreover,
the models are fundamentally physics based -- that is based on our
most surely understood first principles. Yet these models have often
been unsuccessful in even the seemingly most-sound simulations.
Financial models provide another case in point. Bankers on Wall Street
have for years taken advantage of plummeting computational costs to
model complicated financial transactions, but even Nobel-Prize winning
economists applying these models have failed, phenomenally. Like
macromolecular simulations, financial models have a lot of data to
work with, but in contrast to physically based models, financial
models do not rest on as firm a ground of first-principles.
Climate modeling too provides a substantial amount of data for
parameterization but in a considerable amplification of the difficulty
of the modeling relative to the prior two examples, it has little, if
any, opportunity for formal testing -- e.g., we have only one Earth.
Perhaps climatologists should, in appreciating the step up in the
difficulty of their modeling from other complex systems that are
racked with failures, moderate the certainty in their predictions
Unpublished letter in response to:
Anil Ananthaswamy's "Casting a critical eye on climate models,"
New Scientist, Issue 2795 17/1/2011