As the costs of Hurricane Katrina continue to spiral higher -- to date, $125 billion in damages and 1,200 deaths - there's keen interest in perfecting the science of hurricane forecasting. The insurance industry in particular has a big stake in learning where and how the next big one is likely to hit. The problem is that traditional methods of statistical analysis, relying on previous landfalling storms, only go so far in generating useful risk assessments. “We have a bad time predicting in real time when and where hurricanes will develop,” says Emanuel. "It's not even easy to state over a long period of time what the probability is."

Emanuel is refining risk assessment by adding physics to the mix. His laptop-run program takes into account not only a century's worth of actual storms, but also the temperature at the ocean’s surface, from which a hurricane derives much of its energy, as well as air currents, to generate tens of thousands of potential hurricane tracks. For instance, only 29 hurricanes have landed within 100 kilometers of Miami in the past century - relatively little data to help predict potential future damage. Emanuel can conjure up thousands of possible storms evolving in the Atlantic and pounding that city with winds of a given intensity. The real trick, says Emanuel, will involve factoring in climate change. He's found a correlation between sea surface temperature and wind speed that poses serious consequences for a world that's rapidly heating up, with "a greatly increased hurricane destructive potential," says Emanuel. However, from a "U.S.-centric point of view, on a 50-year timescale, this probably doesn’t mean much at all." The likelihood of another superstorm like Katrina or Rita hitting our coast will be a matter of bad luck. But gazing beyond a 50-year horizon, "then you have to worry about global trends," he says.