Sensitivity of WRF model parameterizations in simulating extreme rainfall over complex terrain – case of July 2020 over the Poyang Lake basin, China

In recent years, the Weather Research and Forecasting (WRF) model has been used to obtain reliable rainfall with higher spatial and temporal resolutions. The selection of physical parameterization schemes largely influences the WRF simulation. In this study, the performance of the WRF model for an extreme rainfall simulation during July 2020 in Poyang Lake basin was evaluated using relative error (RE), critical success index (CSI), percentage root mean square error (P-RMSE), and CSI/P-RMSE. The WRF model was configured with 27, 9, and 3 km three-level nesting for 36 combinations of physical parameterization schemes using three microphysics (MP) and three cumulus (CU) parameterization schemes, two planetary boundary layer (PBL) and two land-surface model (LSM) parameterization schemes. The results show that: (1) WRF tends to underestimate rainfall amount in this rainfall event. Increasing the horizontal resolution has a limited improvement in rainfall simulation (1.3%), while choosing a suitable parameterization schemes combination can significantly improve the simulation accuracy by 34.3%. (2) Among the 36 experiments, c5 containing parameterization schemes of Morrison, BMJ, YSU and Slab shows overall best performance in simulating accumulated areal rainfall as well as spatial and temporal distributions. (3) CU schemes have the largest impact on the rainfall simulation, followed by MP, PBL and LSM.