Improvement in the Mesoscale Ensemble Forecast System in East China and A Precipitation Experiment in the 2020 Meiyu Season

By considering the various sources of uncertainty in regional model forecasts, the initial condition uncertainty (IC), lateral boundary condition uncertainty (BC), and model uncertainty (PHY) are introduced to construct a new generation of East China regional mesoscale ensemble forecast systems (SWARMS-ENV2s). Experiments were performed during the 2020 Meiyu season. Selecting typical cases to adjust the parameters of the stochastically perturbed parameterization tendency (SPPT) shows that the parameter selection has certain universality, and as the influence of the random process is strengthened, the low-level wind and humidity fields in the system have obvious feedback, and the ensemble spread can be improved. The influences of the above three parameters on the forecast were as follows: variance in grid point space, spatial length scale (or spatial decorrelation), and temporal decorrelation time. Comparing SWARMS-ENV2 with SWARMS-ENV1 shows that the root mean square error (RMSE) of SWARMS-ENV2 is reduced, and the ensemble spread is obviously increased, the precipitation forecast capability is improved in all forecast periods for different magnitudes of precipitation in terms of the TS score and the probability forecast score, the uncertainty of physical processes has an obvious influence on heavy precipitation events, and the forecast reliability of the system is improved.