"The problem with weather is that our knowledge of its initial conditions is highly imperfect"
The Weatherman Is Not a Moron is an excerpt published in yesterday's New York Times from a forthcoming book by Nate Silver entitled "The Signal and the Noise: Why So Many Predictions Fail – But Some Don't."
I nearly bought a copy of the book in advance last week, but I was worried it was going to be a boring treatise of digital signal processing, but that is clearly not the case. The Weather Is Not a Moron is a pretty good read and provides insights into how uncertainty impacts weather forecasting. Hopefully the book delves into ensemble weather forecasting and efforts by meteorologists in the past two decades to quantify uncertainty, as that has become a critically important component of our profession.
The author covers numerical weather prediction in the excerpt, and how forecasters deal with uncertainty, but never states that many recent advances in weather forecasting are due to the development of ensemble modeling systems that attempt to quantify uncertainty. Rather than running a single model with the highest resolution (meaning detail) possible, it is common today to run an ensemble of many models with somewhat lower resolution (meaning less detail) in order to see the range of possibilities in the future. It is somewhat counter intuitive that a suite of less detailed model forecasts can be more valuable than a very detailed model forecast, but this is the case in weather prediction because the atmosphere is so chaotic and sensitive to the initial conditions. This ensemble modeling of uncertainty has helped decrease the number of meteorological predictions that "fail" in recent years. Hopefully this is covered in the book. If not, I look forward to reading it anyway.