The underlying laws governing elections or the stock market—the rules of mass human behavior that determine the outcomes—are not known well, if they exist at all. The models need to be built on assumptions that past experience will be a guide to future performance. If weather prediction were still done in this way, it would have been simply impossible to predict, days ahead of time, that Hurricane Sandy would turn left and strike the coast while moving westward. No forecaster had ever seen something like that occur, because no storm had ever done it. For the same reason, no statistical model trained on past behavior would have produced it as a likely outcome. In Sandy’s case, forecasters not only could see this outcome as a possibility over a week ahead of time, but they were quite confident of it by four or five days before the storm hit. Forecasts such as the ones we had as Sandy formed and moved up the coast don’t come from the heavens. They’re the result of a century of remarkable scientific achievement, beginning in Norway in the early 1900s. The intellectual foundation of the whole enterprise of weather prediction was the idea that the laws of physics could be used to understand the weather, a radical idea in the early twentieth century. Carrying this out required multiple conceptual advances, over decades, and improvements in technology (especially digital computers). realclimate.org/index.php?p=17591
Posted on: Wed, 05 Nov 2014 20:56:43 +0000
Trending Topics
Recently Viewed Topics
© 2015