Ensemble Filters for Data Assimilation: Flexible, Powerful, and Ready for Prime-Time?

Jeffrey Anderson NCAR Data Assimilation Initiative

A rapidly growing interest in the use of ensemble filters for atmospheric data assimilation has been driven by the relative ease with which filters can be developed for given models and observation sets and the amazing power of these methods when tested in perfect model frameworks. However, developing operational quality ensemble filters has proved to be difficult due to problems with dealing with systematic model error.

The NCAR Data Assimilation Initiative is developing a facility called the Data Assimilation Research Testbed (DART) to foster research and development of ensemble filter assimilation. Several key results from DART will be highlighted to provide a feeling for the ease of development and power of ensemble filters. Results from an 'operational' assimilation system developed with NCAR's CAM2 global climate model will be compared with operational global assimilation systems to assess whether problems with systematic model error have been overcome. A brief analysis of the cost of implementing ensemble filters will be provided.