Sixth Conference on Artificial Intelligence Applications to Environmental Science 88 th Annual Meeting of the American Meteorological Society New Orleans, LA January 20-24, 2008 Use of the LEAD Portal for On-Demand Severe Weather Prediction Keith A. Brewster *1 ; and Daniel B. Weber 1 , Kevin W. Thomas 1 , Kelvin K. Droegemeier 1,2 , Yunheng Wang 1 , Ming Xue 1,2 , Suresh Marru 3 , Dennis Gannon 3 , Jay Alameda 4 , Brian F. Jewett 5 , Jack S. Kain 6 , Steven J. Weiss 5 , and Marcus Christie 3 1 Center for Analysis and Prediction of Storms – Univ. of Oklahoma, Norman, OK 2 School of Meteorology – University of Oklahoma, Norman, OK 3 Indiana University Department of Computer Science, Bloomington, IN 4 National Center for Supercomputing Applications, University of Illinois, Champaign-Urbana, IL 5 Dept. of Atmospheric Sciences, University of Illinois, Champaign-Urbana, IL 6 NOAA – National Severe Storms Laboratory, Norman, OK 7 NOAA – Storm Prediction Center, Norman, OK ABSTRACT The Linked Environments for Atmospheric Discovery (LEAD) Portal was used in real time for on-demand forecasting of severe weather during the 2007 Hazardous Weather Testbed (HWT) Spring Experiment. The LEAD portal is web-based and uses service oriented architecture to allow users to access and analyze meteorological data and to prepare, submit, and monitor nu- merical forecasts and then archive, analyze, and verify forecast data. Two 9-hour Weather Re- search and Forecasting model (WRF) forecasts, initialized at 15 UTC over relocatable regional domains, were submitted once a day during a portion of the two month HWT experiment. The forecast domains were centered on areas of elevated risk for severe weather occurrence as de- termined by a LEAD Project meteorologist using information supplied from Mesoscale Discus- sions issued by the NOAA/Storm Prediction Center and/or daily HWT weather briefings. The initial conditions for the two forecasts were either interpolated from the LEAD ARPS Data Analysis System (ADAS) 10-km horizontal grid spacing CONUS 1500 UTC analysis or from the 3-hour North American Model (NAM) 1200 UTC forecast. The lateral boundary conditions were extracted from the 1200 UTC NAM forecasts using the ARPS EXT2ARPS software package. Due to resource limitations, the WRF forecast data were not available in time for comparison with other model guidance prior to afternoon convective activity, but the forecast data such as temperature, dew point temperature, winds and precipitation were available for comparison with observations and other model forecasts in near-real time. This test validated the use of the LEAD Portal for on-demand forecasting of severe weather. Preliminary assessments of fore- cast skill and recommendations for future on-demand testing are presented. * Corresponding Author’s Address: Keith Brewster, CAPS/University of Oklahoma, 120 David L. Boren Blvd, Suite 2500, Norman OK, 73072