Modeling Toward Accurate Storm Forecasting


Hurricanes and tropical storms are back in the public mind, especially in North and Central America, where they did more than US $100 billion in damages last year. Some see the inescapable side effects of global warming, noting that 2005 was the worst year for Atlantic hurricanes in recorded history. Others see a financial opportunity: online gambling establishments have opened betting pools on this year’s storms—you can wager on how many hurricanes will hit the United States, how many will hit Florida, and even on what the storms’ Saffir-Simpson intensity categories will be.

While it sometimes feels like storm forecasting is one bad bet after another, hurricane forecasting, and weather forecasting in general, have actually improved enormously in the half century since John von Neumann and Jules Charney used 24 hours of ENIAC computing time to produce the first 24-hour numerical weather forecast. And according to our authors Robert Gall and David Parsons from the U.S. National Center for Atmospheric Research, in Boulder, Colo. [” It’s Hurricane Season: Do You Know Where Your Storm Is?”], it’s going to get significantly better very soon.

Why? Supercomputing power, sophisticated sensors, weather-specific satellites, advances in modeling techniques, and better understanding of how weather works will converge and ­dramatically improve our ability to know precisely what’s coming our way, and when.

Collecting weather data is a complicated business. Meteorologists gather readings from instruments on balloons, planes, ships, satellites, and terrestrial weather stations. The amount of data is staggering and getting bigger all the time; it will increase by a factor of 10 000 in the next decade, thanks to the deployment of new sensing devices and the launch of several new weather satellites. Six alone were launched last April in a joint venture between the United States and Taiwan. The European Union and several Asian nations also have ambitious plans on the launchpad. Forty-four countries are now developing plans to share their combined collected weather data as part of the Global Earth Observation System of Systems, or GEOSS.

More data means better weather models, but these mushrooming measurements also call for lots of computing power if they are going to be analyzed in anything resembling real time. Supercomputers now make possible a technique called ensemble forecasting. A weather prediction system runs the same weather model over and over again but tweaks the initial state each time—relative humidity might be increased by a fraction of a percent and so on. A range of forecasts emerges, and as the iterations continue, one pops up more frequently than the rest. As a result of international efforts to coordinate modeling, large supercomputing grids will be able to run larger and larger data-encompassing ensembles even before 10- or 20â''teraflop computers come online.

What it will all mean is that when the TV meteorologist says take your umbrella, you’d better do it. And more important, when she says Hurricane Zelda is going to hit Tampa on Wednesday at 3:00 p.m. EST, it undoubtedly will.

Ensemble modeling makes obvious another difficulty in predicting the weather. Weather is dynamic, while current weather prediction technologies aren’t—they can’t adapt to changing weather conditions, which makes things like predicting a storm’s changing intensity virtually impossible. The ultimate challenge will be to create forecasting systems that can react to their own changing measurements and observations in real time.

In North and Central America, the 2006 Hurricane season is now upon us—August to October is peak season. The U.S. National Oceanic & Atmospheric Administration expects a busy one, predicting 13 to 16 named storms, with 8 to 10 becoming hurricanes. How much damage and human misery these storms bring with them will depend on how strong they are and whether or not they make landfall in heavily populated areas. How ­valuable it would be to understand how they will behave days or even weeks in advance, so people would really know when to prepare for the worst and when to continue business as usual.

For the time being, however, getting the weather right will still require a large dose of good fortune. But continued improvements in our weather predicting capabilities will eventually take the guessing out of storm prediction and bring us to the moment when wagering on the weather will be a safe bet indeed.

The editorial content of IEEE Spectrum magazine does not reflect official positions of the IEEE or its organizational units. Please address comments to Forum

Related Stories