It takes a lot of computer power to predict where an ocean storm will hit, how high the storm surge will be, and the damage that will ensue. The process starts when meteorologists collect thousands of data points from buoys, satellites, ships, and other sources. The data describe surface water temperature, winds, relative humidity, and all of the other things that will affect the storm. Meteorologists load the data into models on a supercomputer, hit a button, and wait. The more models they run, the better their forecasts will be. But turning data into forecasts takes time, and sometimes the fastest computers in the world still are not fast enough.

Three and two-day forecasts for Hurricane Katrina turned out to be remarkably accurate, according to panelists at the Katrina Summit. Yet the limits of computing speed make faster-moving storms harder to forecast, and therefore more dangerous. This is one reason why there are so many government grants aimed at increasing the speed of computers. Faster computers will allow forecasters to issue warnings that are earlier and more specific, saving lives and dollars.

“There are several choke points in current forecasting methods,” says Dan Reed, director of the Renaissance Computing Institute at the University of North Carolina-Chapel Hill. “When you release an evacuation order and it turns out to be inaccurate, you run the risk of lessening compliance with future orders because people will think you’re crying wolf. You’re also making the evacuation bigger and more expensive than it needs to be. And when an evacuation order is not followed, that can be disastrous.”

“When an evacuation order is not followed, that can be disastrous.” School buses in a flooded parking lot in New Orleans, September 1 2005. Photo: Phil Coale, Associated Press (AP).