Research activities of the local ensemble transform Kalman filter (LETKF) at JMA
A local ensemble transform Kalman filter (LETKF), originally developed at the University of Maryland, has been developed with the Earth Simulator global model known as the AFES (AGCM for the Earth Simulator), as well as the Japan Meteorological Agency (JMA) global and mesoscale models. This talk covers most recent results with the JMA global model, especially focusing on the comparison with the operational system using a four-dimensional variational (4D-Var) method. "Ensemble Kalman filter or 4D-Var?" is one of the most excited and controversial questions in data assimilation research community, and this is the first attempt to compare LETKF with 4D-Var in a quasi-operational setup. In addition, the probabilistic forecast skills by LETKF are compared with the current (bred-vector) and next (singular-vector) operational ensemble prediction systems.
The latter part of this talk covers research activities using the Earth Simulator supercomputer. Using the AFES-LETKF system, experimental reanalysis has been performed; the analysis ensemble dataset is now accessible through the Internet for researchers for free of charge. Moreover, we found some dynamical meanings of the analysis errors. Such error products have not been obtained before since variational schemes, used in reanalysis projects, cannot produce them, whereas ensemble data assimilation methods are capable. Furthermore, an observing impact study has been performed in collaboration with observing scientists. This study attempts to pioneer collaborative research using data assimilation, linking two important methodologies of the atmospheric and oceanic science: observing and modeling studies.