HR: 0830h
AN: A41B-09    [Abstracts]
TI: A Multiseason Comparison of the Forecast Skills among Three Numerical Models over Southcentral United States
AU: * LU, D
EM: dlu@twister.jsums.edu
AF: Jackson State University, 1400 Lynch St., Jackson, MS 39217 United States
AU: Reddy, S
AF: Jackson State University, 1400 Lynch St., Jackson, MS 39217 United States
AB: During the summer 2003 and winter 2003-2004, three mesoscale numerical models, the fifth-generation Pennsylvania State University-NCAR Mesoscale Model (MM5), Navy's Coupled Ocean/Atmospheric Mesoscale Prediction System (COAMPS) and the Weather Research and Forecasting model (WRF), were operationally run at a horizontal resolution of 27 km twice daily in Jackson State University (JSU). Three models were run by the initial and lateral boundary conditions from AVN data. The purpose of this paper is to evaluate the performances of three models during these two seasons. It was found that the temporal variation of distribution and strength of mean error (ME) biases at 12, 24 and 36h was rather weak for surface temperature, sea level pressure and surface wind speed. During two seasons, the MM5 underpredicted the seasonal precipitation while the COAMPS and WRF overpredicted. This is consistent with the statistical score analyses of rainfall. The Bias scores revealed that the MM5 yielded an underprediction of precipitation, especially for heavier rainfall events. Due to the under estimate of rainfall areas and strength, the MM5 presented the lower TS, POD and KSS scores at lighter rainfall events compared to the COAMPS and WRF. At moderate to heavier thresholds, three models produced rather low KSS and POD scores that are consistent with the high FAR values. The WRF skills in predicting precipitation heavily depend on the performance of cumulus parameterization scheme. Instead of Kain-Fritsch scheme, using other two schemes, Grell-Devenyi and Bette-Miller-Janjic, in the WRF for warm season 2003 demonstrated that the precipitation overprediction had been efficiently suppressed. Overall, the performances of three models revealed that the best skill is at 12h and the worst at 36h.
DE: 3337 Numerical modeling and data assimilation
SC: Atmospheric Sciences [A]
MN: 2005 Joint Assembly