Ensemble Prediction Systems and Application at China Meteorological Administration

China Meteorological Administration (CMA) has been developing Global/Regional Ensemble Prediction Systems (EPSs) based on the CMA operational Numerical Weather Prediction (NWP) model, Global/Regional Assimilation PrEdiction System (GRAPES), since 2008. To convey forecast uncertainty, four key techniques have been developed and used in the CMA's Global/Regional EPSs. For initial condition uncertainty, the Singular Vector (SV) initial perturbations are used in the global EPS and the multiscale blended initial perturbations based on Ensemble transform Kalman Filter(ETKF) are developed and applied to the regional EPS. For model uncertainty, Stochastic Physical Tendency Perturbation (SPPT) and Stochastic Kinetic Energy Backscatter (SKEB) techniques were introduced into the global and regional EPSs with different scale definition in a space-time correlation random function. For EPS Post-processing and application, we focus on the improvement of Extreme Weather Forecasting. Three methods have been developed: (1) The first is subtracting a bias tendency term during model integration to reduce model mean bias errors. (2) The second is incoporating a new algorithm for Radar Reflectivity of Sub-grid Convective Precipitation in the regional EPS. (3) The third is developing three types of Extreme Weather Products: (a) Anomaly Probabilistic Forecasts (APF) for 500hPa geopotential height and 850hPa temperature and wind speed with decaying average-based adjustment approach, (b) Extreme Forecast Index (EFI) for 2m temperature, 10m wind speed and precipitation of 10-day forecast, and (c) Probabilistic Forecasts for 2m Temperature above or below a threshold with station topography calibration.

Based on the CMA high performance computer, we have implemented the above key techniques and established the CMA operational Global/Regional EPSs ranging from prediction of regional weather with 3-day lead time to large-scale phenomena with 15-day lead time scales. The performance of the CMA Global/Regional EPSs and some comparisons with other NWP centers in terms of verification metrics (latitude-weighted RMSE, ACC and Spread, etc.) will be presented. The applications to the national, provincial and local weather offices of China will be introduced and the challenges and future plans of the CMA Global/Regional EPSs will also be discussed.

Keywords: GRAPES model, ensemble prediction, perturbation method, verification, application