For five decades, weather forecasting in the United States has relied on models that run on the latest computer technology at the National Weather Service’s National Center for Environmental Prediction in Camp Springs, MD. The models use more than 100 million daily measurements of temperature, moisture, air pressure, wind speed, and wind direction gathered from different locations around the world. Based on this data, forecasts are calculated on global, national, and regional scales every six hours for areas as small as 12 kilometers by 12 kilometers.Real weather, of course, can vary quite a bit over distances as short as a few kilometers, says Craig Burfeind, a meteorologist who cofounded Digital Cyclone with Douglas. “Winter storms can have a precise line, and a few miles to either side of that line can mean the difference between rain and six inches of snow.” Burfeind notes that one day this past winter, temperatures in the southern suburbs of Minneapolis hit the low 20s C (70s F) while the northern suburbs remained around freezing-resulting in temperature differences of 6 C or more over a six-kilometer range. And Minneapolis isn’t even near a large geological feature, such as a mountain or a valley, that can affect wind speed and direction, humidity, and temperature, and create measurable differences across a small area. Even a sizeable body of water can create sharp temperature contrasts that contribute to lake-effect snow, heavy coastal fog, or unexpected thunderstorms.
The Weather Service would be happy to produce forecasts for everyone’s locale if it were practical. But increasing the resolution of the forecast grid to, say, six kilometers by six kilometers actually requires eight times as much calculation. “Our next step in that realm is to 10 kilometers, which won’t be operational until the end of 2004,” says Lauren Morone, operations officer at the National Center for Environmental Prediction.
But there is another way. In the 1990s, researchers at Pennsylvania State University began incorporating the raw data collected from the National Weather Service into their own PC-based models. “Running a model used to be a large centralized operation, like the Manhattan Project,” says Penn State climatologist Paul Knight. In contrast, he says, the new generation of PC models complete fewer calculations for a smaller area of the globe and are therefore able to produce high-resolution, localized weather forecasts that can be churned out relatively quickly.
Digital Cyclone, for one, is capitalizing on Penn State’s success. The company provides forecasts for a number of metropolitan U.S. areas, using a single PC to turn out a weather prediction for a particular city. The forecasts are twice as frequent as those coming from the National Weather Service and cover a smaller area; that is, they run every three hours and have a resolution of six kilometers within about a 120-kilometer radius of the city.
Customers of Digital Cyclone can access the information from a Web site. By keying in their locations, they can get weather maps centered on their towns, complete with radar images and projected storm tracks. But the real value of Digital Cyclone’s service, says Douglas, is that people can acquire the information from their Internet-enabled mobile phones. Later this year, those same phones will emit audible alerts sent out by the company and tailored to people’s needs. And at a time when more and more mobile phones are equipped with Global Positioning Satellite software and signal receivers that provide information about their geographical location, Digital Cyclone is developing software-expected to become available in the next few years-that would use GPS data to offer high-resolution weather-forecasting maps automatically centered on the phone’s location.