An OSSE-based Evaluation of Hybrid Variational-Ensemble Data Assimilation for the NCEP GFS, Part I: System Description and 3D-Hybrid Results

An observing system simulation experiment (OSSE) has been carried out to evaluate the 1 impact of a hybrid ensemble-variational data assimilation algorithm for use with the National 2 Centers for Environmental Prediction (NCEP) global data assimilation system. An OSSE 3 provides a controlled framework for evaluating analysis and forecast errors since a truth is 4 known. In this case, the nature run was generated and provided by the European Centre for 5 Medium-Range Weather Forecasts as part of the international Joint OSSE project. The 6 assimilation and forecast impact studies are carried out using a model that is different than the 7 nature run model, thereby accounting for model error and avoiding issues with the so-called 8 identical twin experiments. 9 It is found that the quality of analysis is improved substantially when going from 3DVar 10 to a hybrid 3D ensemble-variational (EnVar) based algorithm. This is especially true in terms of 11 the analysis error reduction for wind and moisture, most notably in the tropics. Forecast impact 12 experiments show that the hybrid-initialized forecasts improve upon the 3DVar based forecasts 13 for most metrics, lead times, variables, and levels. An additional experiment that utilizes 14 3DEnVar (100% ensemble) demonstrates that the use of a 25% static error covariance 15 contribution does not alter the quality of hybrid analysis when utilizing the tangent linear normal 16 mode constraint on the total hybrid increment. 17