Monday, 12 January 2004: 9:45 AM
Data assimilation with deterministic ensemble filters
Room 6A
A number of ensemble filter methods for data assimilation
have been developed during the past decade. The theoretical
simplicity and ease of development of these algorithms
makes it possible to consider developing generic filtering
algorithms that can be applied efficiently to a wide array
of models and data streams. A filtering data assimilation
facility has been developed as part of NCAR's new data
assimilation initiative. This facility, the Data Assimilation
Research Testbed, has developed ensemble filter assimilation
capabilities for a hierarchy of models including low order
models, a regional prediction model (WRF), several climate
AGCMs (the GFDL AM-model and the NCAR CAM model), and an
operational global prediction model (NCEP's GFS model). A
survey of the capabilities of ensemble filter assimilations
will be presented, beginning with their ability to extract
vast amounts of information from highly indirect observations
in synthetic observation experiments. Results from
experiments with the operational NCEP global model and data
streams will point out the relative capabilities of current
ensemble filters and operational 3D-variational schemes. A
closing discussion will point out remaining difficulties and
evaluate the potential for using ensemble filters in
operational atmospheric prediction systems.
Supplementary URL: