In "The Weatherman is Not a Moron," Nate Silver gives a fascinating tour of how forecast models have evolved and delves into how uncertainty is communicated to the public.
First up is the idea of modeling and an idea that Hari Selden himself would feel at home with. Snip:
"In 1814, the French mathematician Pierre-Simon Laplace postulated that the movement of every particle in the universe should be predictable as long as meteorologists could know the position of all those particles and how fast they are moving. Unfortunately, the number of molecules in the earth’s atmosphere is perhaps on the order of 100 tredecillion, which is a 1 followed by 44 zeros. To make perfect weather predictions, we would not only have to account for all of those molecules, but we would also need to solve equations for all 100 tredecillion of them at once."
Strikingly, it's the integration of the unknown into models that is making them more accurate. Like stock traders, meteorologists have begun to play the odds, acknowledging that they're an inherent part of the system.
Unfortunately, communicating that uncertainty carries cultural baggage, motivating experts to hold back their full analysis. Private sector forecasters have a "wet" bias for predictions of rain, because the public is happy if it's sunny when rain is predicted, and angry if it rains when sun is forecast. In a more serious situation, Silver reports there was a storm in 1997 in Grand Forks North Dakota where the forecast's margin of error wasn't shared with the public. The result? Millions in damage that could have been averted through mitigation efforts like sandbagging.
"The forecasters later told researchers that they were afraid the public might lose confidence in the forecast if they had conveyed any uncertainty."