1United Nations Development Programme, Abu-Dhabi, UAE

A series of experiments based on the recently implemented WRF Numerical Weather Prediction (NWP) system, and especially its component dealing with data assimilation had been carried out over Middle-East and Arabian Peninsula areas. The ultimate goal of this work is to put into operations a complete NWP configuration running in both assimilation and prediction modes, taking advantage of the best numerical and scientific techniques offered by WRF system, and achieving better forecast performance when compared to the native WRF forecasts based on the Global Forecasting System (GFS) initial states. The conducted experiments investigate the impact of using different background errors, tuning observation errors, multiple outer loops, First Guess at Appropriate Time (FGAT) and ingesting some high resolution non conventional data like Radar reflectivities. It was found that, when the positive impacts of these different ingredients are associated together, forecasts initiated from WRF 3D-Var initial states, in a warm cycling mode, outperform those based on GFS Grid point Statistical Interpolation (GSI) final analyses