5.3 Recent development and research on GSI based hybrid ensemble-variational data assimilation system for GFS

Tuesday, 8 January 2013: 11:30 AM
Room 9B (Austin Convention Center)
Xuguang Wang, University of Oklahoma, Norman, OK; and T. Lei, J. Whitaker, and D. Kleist

A dual resolution GSI-based 3D ensemble-variational hybrid data assimilation system (3DENSVAR) is implemented operationally for NCEP GFS since May 2012. The 4D extension (4DENSVAR) was further developed and tests at single resolution have shown that the 4DENSVAR significantly improved upon the 3DENSVAR. Further development and research on the 3DENSVAR and 4DENSVAR system are ongoing. The impact of the resolution of the ensemble covariance and the impact of the static covariance on the hybrid are studied by comparing and diagnosing the performances of the hybrid at both the single and dual resolution. It is found that the hybrid with the reduced resolution covariance performed worse; the use of the static covariance shows much more benefit at the dual resolution than at the single resolution; 4DENSVAR can still show benefit relative to 3DENSVAR at dual resolution mode. Further development and tests of the 4DENSVAR system including the time localization and use of the digital filter weak constraint are also ongoing and results will be reported in the conference.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner