Physical Process Perturbation For Ensemble Forecasts On MCC System

Dynamical mechanisms differ between systems in different scales. According to different systems to design suitable physical process disturbances is necessary. Based on the WRF model, the physical process perturbation schemes are examined by selecting a typical meso-scale process, the MCC, to search for proper physical process perturbation schemes.The main conclusions are as follows:(1) According to the forecast of MCC, diabatic physical processes have less impact on the large-scale circulation variables than the variables characterizing the meso-scale features. The influence of convection parameterization schemes in the integral time on thermal field is greater than the dynamic field, after a period of time the result is opposite. Due to convection trigger mechanism, different convection parameterization schemes is different in the time, place and intensity of precipitation, thus the releasing of condensation latent heat lead to different temperature disturbance, then the atmosphere is different from the dynamic and thermal structure, so the forecast results are different.(2) The ensemble forecast considering the numerical model uncertainties can capture some uncertain characteristics of the MCC system. For the24h accumulated rainfall less than25mm, the BS scores of ensemble forecast are greater than1. Whereas, the BS scores are less than1. Ensemble mean of rainfall, spread of rainfall and the precipitation probability forecast provide valuable information for the rainfall area and precipitation intensity of MCC system. In Grell cumulus parameterization scheme, changing the parameters in "trigger function" will change the area, intensity of precipitation and the vertical profile distribution of variable factors.(3) The method of different physical parameterization, the perturbation of sensitive parameters in cumulus parameterization scheme and stochastically perturbed parameterization tendencies scheme, the precipitation forecast effect of ensemble forecast through these three methods is obviously improved relative to the control forecast. Viewed by the RMSE and spread, the RMSE of stochastic physical perturbation is less the other two methods, while its spread is bigger than other two methods, the stochastic physical perturbation scheme has encouraging effects. Future the stochastically perturbed parameterization tendencies scheme should be given more consideration to construct ensemble forecast system for MCC.