6.1
Application of Local Ensemble Kalman Filter: perfect model experiments with NASA fvGCM model

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 1 February 2006: 8:30 AM
Application of Local Ensemble Kalman Filter: perfect model experiments with NASA fvGCM model
A405 (Georgia World Congress Center)
Junjie Liu, University of Maryland, College Park, MD; and E. Klein, H. Li, I. Szunyogh, B. Hunt, E. Kalnay, E. J. Kostelich, and R. Todling

Presentation PDF (361.2 kB)

The Local Ensemble Transform Kalman Filter (LETKF) applies the Ensemble Transform Kalman Filter technique (Bishop et al. 2001) locally to update the analysis ensemble members. By performing data assimilation on each local patch (following the LEKF approach of Ott et al, 2004), LETKF can utilize the low-dimensional subspace to reduce the required ensemble size. Due to this characteristic, it is very easy to apply LETKF parallely.

Due to the accuracy and fast computational speed, LETKF is applied to NASA fvGCM to assimilate simulated model variables and rawinsonde observations. The results from simulated observations of every model grid point show that the analysis RMS error is below observation error after ten days. Even with 11% observations, the analysis RMS error drops below observation error after about 20 days spin up time. We will present the comparison of the results from this scheme and those obtained from the results with the operational NCEP model (Szunyogh et al., 2004 ) and the operational fvGCM PSAS scheme. Ultimately, LETKF will be applied to assimilate real observations at the observation time (Hunt et al, 2004), estimate and correct model errors present in the fvGCM model, and assimilate AIRS observations .