Monday, 12 January 2009
An Examination of Simulated Reflectivity Forecasts from a 10-Member Storm Scale Ensemble Prediction System
Hall 5 (Phoenix Convention Center)
Simulated reflectivity forecasts from the Storm Scale Ensemble Forecast (SSEF) system, including two 4-km control runs, and one additional 2-km high-resolution deterministic forecast, were examined in this study. The SSEF system contains ten WRF-ARW members, eight with perturbed initial conditions and radar data assimilation, and two unperturbed controls with and without radar assimilation. The 2-km high-resolution forecast is otherwise identical to the 4-km SSEF control that includes radar assimilation. All of these model forecasts were compared to the national mosaic base reflectivity (BREF) product to evaluate the following sensitivities: 1) model cloud microphysics schemes, 2) assimilation of radar into the ensemble members, 3) evolution of convection and 4) size of horizontal grid spacing. From here, two techniques were used to compare the simulated and actual reflectivities from ten randomly selected days in May 2008. First, the percent coverage of 20, 30, 40, and 50 dBz and greater was found by counting the number of gridpoints with dBz values at or above the specified threshold and then dividing by the total number of gridpoints. Second, the biases at the four thresholds for the eight SSEF members, two 4 km control runs, and one 2 km run were found by taking the forecasted reflectivity divided by the actual reflectivity. The members with Thompson microphysics had a substantial low bias at all forecast hours. Also, none of the ten members reproduced well the nocturnal convection evident in this 10-day sample.