Found this article from the Technology Review blog very interesting for those in computer science, applied mathematics, and statistics. (HT: Gene Rohrbaugh, CS prof. here at Messiah College) The predictions of the track of Hurricane Irene were remarkably good. Days ahead of time, forecasters had the track nailed down closely. They knew where it would make landfall, and the track up the coast that it would take. However, they thought the storm would be much stronger than it turned out to be. Here is a taste of the article:
While path prediction has steadily improved over the decades, forecasting the intensity of storms still proves tricky. Irene’s expected monster intensity—much to the nation’s relief—was far less as she weakened a day or so after reaching land. “What made Irene especially difficult for the forecasting models was that she had three landfalls and followed the coastline,” says Heymsfield. “We need a lot more research to understand how to better model this land interaction.”
Others point to the unusual way Irene’s “eye wall”—the inner core of storms surrounding the hurricane’s eye—behaved.
Note the solution to the problem. They need more research. In other words, they need more data, and better statistical modeling. I noticed that sites like weather.com and others included error bars on their track prediction. They had a “most likely” line and then bars on either side to indicate the margin for error. Sounds like statistical modeling to me! Check out the rest of the post for details about the type of data needed to better model storms like Irene.