My favorite chapter in Nate Silver’s “The Signal and the Noise” thus far is definitely the chapter on weather forecasting. In this chapter, Silver discusses how over the last 25 years, we have become much better at predicting the weather. One major example of to show this idea is that a quarter century ago, forecasters could determine where a hurricane could make landfall within a 350-mile radius. Today, they can predict where it will hit land within a 100-mile radius – an area much easier to evacuate. He also goes on to mention that models that predict trends in weather patterns are much more accurate than they used to be as a result of the enormous amount of computing power behind the models today. One of the points that Silver concludes is that the steady, constant improvement of weather forecasting is one of the successes stories of statistical analysis in the modern era.
This was an interesting statement since everybody’s favorite past time is complaining that the meteorologist’s on the news are never right. Silver discusses this in the chapter, describing how local forecasters are the least accurate with a “wet” bias. What was so interesting to me, is that your opinion of the weather forecasters can be so different whether you are approaching it from a statistical perspective or not. To a statistician who is trying to predict something like elections; or the stock market; or earthquakes, the gains in weather prediction are an impressive achievement, but most people, myself included before reading this book, love to mock them for being wrong as often as they are. I was impressed that the most cutting edge, powerful computers are running the models.
Silver continuously talks about how predicting events is very difficult, especially when human actions are involved. Something like weather, governed by forces we have a decent understanding of and can not meddle in (climate change notwithstanding), seems like a pretty straight forward task, and yet still requires the fastest computers out there. This certainly puts into perspective how complex the predictions of normal events can be.
The next time a friend or colleague complains that the weather forecasters got it wrong again, I will definitely think twice. It would take continued improvement in weather forecasting to change peoples prejudices about weather forecasting, and a shift in mentality will be slow, much in the same way that people refused to allow the weather channel to change “rain” on the radar maps from green to blue back in 2001.