A Comparative Experimentation Of Cumulus Parameterization Schemes On Heavy Rainfall

Heavy rainfall can cause mudslides, landslide and flush flood, and always brings heavy loss in lives and wealth, Hence it is important to predict the occurrence of heavy rainfalls as much accurate as possible.In recent years, a mixed-physics ensemble approach has been investigated as a method for a better prediction of heavy rainfall by using Mesoscale Numerical Weather Prediction model. For both design and interpretation of mixed-physics ensemble, knowledge of general impact of various physical schemes and their interactions on heavy rainfall forecasts would be useful. In this thesis, a newly emerging Weather Research and Forecasting (WRF) model is tested in Hunan Province to extend such benefits. To pursue this goal, a matrix of 21 WRF model configurations, created using different physical scheme combinations, was run with a grid spacing of 20 km and a nested grid spacing of 60/20 km for 3 heavy rainfall cases respectively. For each case, 3 different periods of simulation time, 7 different microphysical schemes, and 3 different cumulus parameterization schemes were used. As a measure of forecast accuracy, an Equitable Threat Score (ETS), a Percent of Doom (POD) and a False Alarm Rate (FAR) were calculated.Skill score measures averaged over all 3 cases using 21 configurations indicated that no one configuration was always obviously the best during all periods of simulation time. A configuration with nested grids would be better than one with no nesting and the ETS would be decreased while the simulation time was extended. Specifically, the greatest variability in forecasts was found to come from the changes in the choice of convective scheme in different periods of simulation time, although notable impacts also occurred from the changes in microphysics schemes.Projects with WSM3 microphysics and 3 Cumulus parameterization schemes were distinguished in this study. The results indicated that the locations of heavy rainfall roughly match the observation areas, showing better results with 2-way simulations with a nested grid of 60 /20 km. Compared with BMJ scheme for location departure and over intensity of heavy rainfall, KF scheme gave out results much closer to the observation data, while the simulation results of GD scheme showed a weaker intensity and smaller areas of heavy rainfall. Some falsehood heavy rainfall centers also occurs in southern areas of simulation results with KF and BMJ schemes.