Numerical rainfall simulation with different spatial and temporal evenness by using WRF multi-physics ensembles
Abstract. The Weather Research and Forecasting (WRF) model is used in this study to simulate six storm events in two semi-humid and semi-arid catchments of Northern China. The six storm events are classified into four types based on the rainfall evenness in the spatial and temporal dimensions. Two microphysics, two planetary boundary layers (PBL) and three cumulus parameterizations are combined to develop 12 physical ensembles for rainfall generation. The WRF model performs the best for Type 1 event with relatively even distributions of rainfall in both space and time. The average relative error (ARE) for the cumulative rainfall amount is 16.98 %. For the spatial rainfall simulation, the lowest root mean square error (RMSE) is found with event II (0.3989) which has the most even spatial distribution, and for the temporal simulation the lowest RMSE is found with event I (1.0171) which has the most even temporal distribution. It is found to be the most difficult to reproduce the very convective storm with uneven spatiotemporal distributions (Type 4 event) and the average relative error (ARE) for the cumulative rainfall amounts is up to 68.07 %. The RMSE results of Event III with the most uneven spatial and temporal distribution are 0.9363 for the spatial simulation and 2.7769 for the temporal simulation, which are much higher than the other storms. The general performance of the current WRF physical parameterisations is discussed. The Betts-Miller-Janjic (BMJ) is found to be unsuitable for rainfall simulation in the study sites. For Type 1, 2, and 4 storms, ensemble 4 performs the best. For Type 3 storms, ensemble 5 and 7 are the better choice. More guidance is provided for choosing among the physical parameterisations for accurate rainfall simulations of different storm types in the study area.