Exploratory Case Studies Evaluating Convective Parameterization Schemes for Model Predictions of Heavy Rainfall in Thailand

Abstract

Numerical weather predictions have been made to evaluate selected convective parameterization schemes in forecasts of typical heavy rainfall events in Thailand. The Betts-Miller (BM), Grell (GR), Kain-Fritsch (KF), and new Kain-Fritsch (KF2) schemes were used in the Penn State/NCAR non-hydrostatic mesoscale model (MM5) with grid resolutions of 45 km and 15 km. The examples studied were in four important convective conditions, namely days with heavy rain in the hot season (NE region), the cold season (S region), the rainy season (E region), and a tropical depression (NE region).Total rain patterns from the models were compared qualitatively with rain patterns derived from records at rain gauge stations, and with cloud patterns derived from geostationary meteorological satellite infrared images. Quantitative evaluation was obtained by calculating the bias and the Peirce skill score in small study areas with heavy rain as a function of accumulated rainfall over thresholds in the range zero to 80 mm every 12 h up to 48 h.The four convective parameterization schemes gave widely different results. In the hot season case, none of the schemes predicted the rainfall well. In the other cases, the GR scheme was usually the best. The new KF2 scheme usually gave better results than the original KF scheme, indicating a potential for improvement in this scheme. The BM scheme was less successful than the other schemes.