This paper describes an information system that uses National Weather Service computer-generated forecasts to automatically produce 1- and 2-day fire severity forecasts for 49 States. There are two computer models - the limited fine mesh model and the nested grid model. Both models are run twice daily and each produces a 48-hour forecast with data at 6-hour intervals. Our system inputs weather elements directly, interpolates between forecast times, extrapolates from available data, or transforms available data into elements needed by the severity index. The forecast data are calibrated to eliminate bias relative to observational data. The system has been 99 percent reliable during 8 months of operational testing. During 73 days of national evaluation (28,178 individual forecasts), 91 percent of 1-day forecasts were within 1 point of the observation (87 percent for 2-day forecasts),. These results presumably reflect persistence to some degree. Automated forecasting of fire severity appears to be an important and useful addition to fire management technology. © by the Society of American Foresters. Abstract reproduced by permission.