Comparison of rainfall pattern and statistical verification of CPS for extreme rainfall simulation on 13 April 2013

Abstract

The Numerical Weather Prediction (NWP) plays a vital role in short-range rainfall forecasting. The atmospheric physical parameters have been continuously developed for the NWP. These parameters are implemented into NWP models based on equations derived from theoretical assumptions based on observations and physics of the event. The aim of study is to investigate the use of three cumulus parameterization schemes (CPS), namely, Kain-Fritsch (KF), Bett-Miller-Janjic (BMJ) and Grell-Devenyi (GD), in the version 3.5.2 of Weather Research and Forecasting (WRF) model for the off-season heavy rainfall caused by cold surge during the pre-monsoon (April) of 2013 in the east coast of southern Thailand. The initial conditions preceding the heavy rainfall events, 24, 48, and 72 forecasts were simulated using the WRF with three different schemes. The performance of the schemes was measured by the comparison with observed rainfall data from Thai Meteorological Department and Global Satellite Mapping of Precipitation (GSMaP). The result shows that all of schemes can detect the rainfall amount with the forecast from the GD being higher than that of the KF and BMJ scheme. The statistical indicators utilized to compare the performance of the three schemes include critical success index (CSI), equitable threat score (ETS). Two statistical indicators support the finding that the BMJ scheme has a slightly better performance than KF and BMJ, except Probability of detection (POD) that KF has the best skill when used in off-season rainfall forecasts over the east coast of southern Thailand.