Lessons Learned with the NASA/GMAO OSSE

The Global Modeling and Assimilation Office at the National Aeronautics and Space Administration (NASA/GMAO) has developed an Observing System Simulation Experiment framework. This framework includes a 24 month run of the 7 km Global Earth Observing Model (GEOS-5) for the Nature Run, synthetic observations for radiance, conventional, GPS-RO, and atmospheric motion vector data types, simulated observation errors with correlated and uncorrelated components, and the 50 km GEOS-5 forecast model with Gridpoint Statistical Interpolation (GSI) data assimilation system (DAS). The GMAO OSSE has been calibrated and validated to match the statistics of data assimilation as closely as possible to real data statistics.

The GMAO OSSE has been used to investigate different aspects of DAS and model behavior, using the Nature Run as verification to allow direct calculation of analysis and forecast errors as well as statistics of the DAS process. Findings from recent research into model and initial condition error growth, the role of correlated and uncorrelated observation errors, and adjoint derived observation impacts will be presented. Lessons learned from refining the OSSE calibration process will also be discussed.