Tuesday, December 12, 2006

Personalized Weather Forecasts

An IBM supercomputer forecasts weather down to a one-kilometer resolution.

By Duncan Graham-Rowe

IBM has launched a new one-kilometer-resolution weather-forecasting service. This image shows that while some parts of Chicago were hit by severe thunderstorms (blue regions), other parts got off scot-free. Credit: IBM Research

IBM has launched a new weather service called Deep Thunder that can predict the rain, the wind, and temperature conditions down to a one-kilometer resolution. In time, IBM researchers say they should even be able to nail the resolution down to individual streets.

The idea is to provide weather-sensitive businesses in metropolitan areas with information that's more accurate than what government agencies are capable of providing, says Lloyd Treinish, a researcher at IBM's TJ Watson Research Center, in Yorktown Heights, NY.

At a local submetropolitan level, the weather really can vary quite significantly, says Treinish. Yet typical forecasts will often slap a single simplistic symbol, such as the sun, a cloud, or a snowflake, on an area representing a small city.

A huge number of businesses really depend on accurate weather forecasts, says Stephen Lord, director of the National Oceanic and Atmospheric Administration's Environmental Modeling Center, in Camp Springs, MD. Transportation, energy distribution, shipping, and even sporting events are all at the mercy of the weather, he says. Indeed, as much as $1 trillion worth of the U.S. economy is weather sensitive, says Treinish. By providing more-detailed forecasts, IBM hopes to help businesses streamline their operations and save money.

For example, local municipal services such as snowplowing could be deployed more efficiently with more-detailed information about precisely where snow will fall. Similarly, by being more prepared, utility companies could better manage energy demand and better cope with outages caused by severe weather. Even airports and postal services would benefit: they could plan and schedule operations around weather conditions.

Government agencies, such as the National Weather Service (NWS), are currently unable to provide the same level of detail. This is partly because they don't have the technical resources, but it's also because they are mandated to offer a uniform level of service across the nation, preventing them from providing higher resolutions for some areas and not others. Even at a metropolitan level, where local meteorological services try to improve on NWS forecasts by factoring in local measurements and conditions, the resolution is rarely much better than eight kilometers, says Treinish.

When combined, all these factors represent a gap in the market that companies like IBM could fill by tailoring their services to individual businesses. "We want to think about the information in relation to solving particular business problems," says Treinish.