The bad news about hurricane forecasting is that long-term projections of frequency, intensity, and landfall still have a long way to go. The good news is that near-term forecasting has improved enormously in recent years, saving countless lives and many millions of dollars. Hurricane Katrina, despite the tremendous problems with the evacuation of New Orleans, provided a vivid example of today's more skillful hurricane predicting. "Katrina's 48-hour forecast was right on geographically, and only a couple of hours off in time," says Bob Gall, a scientist at the National Center for Atmospheric Research (NCAR). "It was an excellent forecast."

"A most powerful hurricane with unprecedented strength," was the headline on the U.S. National Weather Service's alert issued a day before the hurricane hit the Gulf Coast, almost precisely where it was expected to. "Most of the area [hit] will be uninhabitable for weeks, perhaps longer" [see photograph, "Everybody Out"].

It was because of the weather service's accurate forecasts that the decision was made to evacuate New Orleans and adjacent areas. Without the forecasts, the human toll from the storm would have been far greater, though that does not take the edge off all the serious errors made in the evacuation. "I am a bit bitter," says Gall, "because we had this forecast, and they didn't have a clue how to use it."

Gall is one of the 30 or so scientists at NCAR in Boulder, Colo., who are developing advanced atmospheric models that weather forecasters use for storm prediction. NCAR scientists don't make the predictions themselves but move forward the state of the art that makes predictions possible.

And that state of the art has come a long way. Hurricane forecasts today are a full day ahead of those of 10 years ago—which means that a forecast as accurate as that given 48 hours before Katrina could only have been made 24 hours ahead of Andrew in August 1992 [see map, "On Target"].

Those extra 24 hours came from many advances. Weather satellites are more sophisticated, with more and better sensors. The understanding of the science of weather is better, with researchers able to consider small variations—say a few puffy clouds—in the big picture. The models—the complex numerical representations of atmospheric forces—take more variables into account. And the supercomputers on which these models run have gotten faster and faster.

This last advance is key, because the more data forecasters add into a model, the longer it takes to run. For example, if forecasters use models that represent weather parameters such as wind, temperature, and pressure every 10 kilometers instead of every 30 km, the number of calculations increases by a factor of 30—as does the time it takes a computer to make them.

Forecasts that take longer to process than the actual weather takes to evolve are of no use to anyone, so the scientists who develop these systems must continually balance accuracy—the more information, the more accurate the forecast—against the limitations of processing speed. Forecasts like the one for Katrina take several hours to run on the world's fastest supercomputers.

Gall expects advances on all these fronts to continue, moving accurate forecasts out to three and potentially even four days. "Maybe four days is what we really need," he says. "If we had had four days before Katrina, they might have cut the number of people left in town by a factor of two. They could have brought in pumps and positioned them. Maybe they could have raised the levees. There are all sorts of ways lead times can help."

Besides giving people more warning, more accurate forecasts can save money. Evacuating more coastline than is necessary wastes a lot of money. The figure most often quoted for average evacuation cost is US $1 million per mile, which disregards differences in population density and probably is a gross underestimate, according to Roger Pielke Jr., a professor of environmental studies at the University of Colorado, Boulder. The actual cost, he says, could be far more. Even using the $1 million figure, evacuating, say, 515 km instead of 645 km, which would take only a 20 percent increase in forecasting accuracy, would save $80 million per storm, Pielke calculates.

But evacuation decisions depend on much more than just forecasts, Pielke notes. Because of factors like development, the amounts of coastline evacuated have actually increased in recent years, even as the accuracy of predictions has improved.