INTERCONNECTS

A torrent of satellite-based data has pushed weather forecasting to new heights of accuracy, according to Richard Anthes, president of the University Corporation for Atmospheric Research (UCAR). What's lacking, says Anthes, is the human and computing power to wring the most value from these and other observing systems. On Monday, January 12, at the annual meeting of the American Meteorological Society in Seattle, Anthes will deliver an update of his vision of the state of global weather services by the year 2025. Anthes delivered his original outlook at the 1999 AMS meeting. The fine-scale detail made possible by the Weather Research and Forecasting (WRF) model will depict weather features such as showers and thunderstorms with far greater fidelity. WRF is being created through cooperation among six government agencies, including NCAR. The steady progress in forecast skill over the last several decades has yet to wane, says Anthes. For instance, the National Weather Service's three-day heavy rain outlook, introduced in 2000, is proving as accurate as the two-day forecasts of 1990. "There's no evidence the rate of improvement is slowing down," Anthes says. "We're moving toward a useful two-week weather forecast, and by 2025 we should be very close to that goal." A key driver of progress in the last five years, says Anthes, has been the incorporation, or assimilation, by computer forecasting models of data from a wealth of satellites, aircraft, and other tools that keep a continuous watch on Earth's atmosphere. "Satellites are finally making a huge impact in forecasting," says Anthes. This progress comes with a price, though, he adds. Global support for research and operation of weather satellites will total roughly $25 billion from 2001 to 2007, Anthes notes. In contrast, the leading players in weather modeling-including centers in North America, Europe, Asia and Australia-together invest less than $1 billion per year for computing and personnel at their national forecasting centers. "We need to balance this," says Anthes, "by making sure we have the computing power and the people to fully exploit our huge investments in space hardware." The bedrock of recent achievements in forecasting, according to Anthes, are the steadily improving computer models that depict the atmosphere, oceans, land, and ice, as well as advanced techniques for the models to assimilate satellite and other data. Pioneered in the 1950s, three-dimensional models of wind, temperature, and moisture map out the broad-scale contours of weather days in advance. Increased computing power has enabled scientists to build models that are more detailed and far-sighted than ever. For instance, a recent study by the European Centre for Medium-Range Weather Forecasts shows that the center's flagship model can track large-scale weather features in the Northern Hemisphere with 80 percent accuracy almost six days in advance. The comparable lead time in 1982 was less than four days. Although satellites have given human forecasters a bird's-eye view of weather features for over 40 years, only in the last decade have atmospheric, mathematical, and computer scientists figured out reliable ways to convert the vast amount of satellite data-several billion bytes a day-into forms that models can use. Global models have long been challenged by the Southern Hemisphere, because few weather balloons are launched there and the distances between them are vast. Thanks largely to the inclusion of more satellite data, global forecast models can now depict large-scale features in the Southern Hemisphere with nearly the same accuracy as in the Northern Hemisphere. Other types of remote sensing are now based on the Global Positioning System, the military's high-altitude satellite network. More than 200 ground-based GPS moisture sensors have blossomed across the United States in the last few years. These small devices measure the tiny, atmosphere-induced delays in GPS signals. Scientists can then infer the amount of water vapor responsible for the signal lag. As several academic, federal, and military groups have built regional clusters of these moisture sensors, a national network is taking shape. "It's amazing how fast it's developed," says Anthes. Meanwhile, GPS receivers on low-altitude satellites are now in the works at UCAR and elsewhere; they will complement other satellites, canvassing the globe and yielding temperature and moisture profiles. Tomorrow's computer models will use these and other data to build increasingly fine-scale depictions. Anthes foresees a smorgasbord of local and regional models, some of them with telescoping grids to zero in on areas as small as a few city blocks. These models will track wildfires, airborne chemicals, and disease vectors, as well as severe thunderstorms, hurricanes, and heavy rains. In addition, forecasts based on probabilities, such as those now issued for tornado likelihood by the National Oceanic and Atmospheric Administration, will help society anticipate some weather risks days ahead of time. "Models won't replace humans," says Anthes. Well-trained scientists are especially needed to improve on the skill of computer-generated forecasts-particularly during extreme weather-and to devise new ways of using satellite-based data in future models. "We need increased investment in human resources and computer power in order to take advantage of the present and even greater future potential of satellites. We also need to improve our rate of transferring research into operations," says Anthes. He recently chaired a National Academy of Sciences panel on the transition of NASA research to operations at NOAA (the National Oceanic and Atmospheric Administration), which includes the National Weather Service. UCAR's president since 1988, Anthes is a highly regarded atmospheric scientist, author, educator, and administrator. He previously directed the National Center for Atmospheric Research (operated by UCAR) and taught at Pennsylvania State University. Anthes helped create the PSU/NCAR mesoscale model (now known as MM5), a leading tool for research and prediction of regional weather.