We develop an ensemble data assimilation system using the 4-dimensional Local Ensemble Transform Kalman Filter (LEKTF) for a global hydrostatic Numerical Weather Prediction (NWP) model formulated on the cubed-sphere. Forecast-analysis cycles run stably and thus provide newly updated initial states for the model to produce ensemble forecasts every 6 hours. Performance of LETKF implemented to the global NWP model is verified using the ECMWF reanalysis data and conventional observations. Global mean values of bias and root mean square difference are significantly reduced by the data assimilation. Besides, statistics of forecast and analysis converge well as the forecast-analysis cycles are repeated. These results suggest that the combined system of LETKF and the global NWP formulated on the cubed-sphere shows a promising performance for operational uses.
Ensemble data assimilation; Local ensemble transform Kalman filter (LEKTF); Numerical weather prediction (NWP); Atmospheric global model (AGM)