Data Assimilation of the Global Ocean using the Local Ensemble Transform Kalman Filter

Steve Penny
UMD

Abstract:

Ocean data assimilation combines forecasts from computational models with observations such as XBT and Argo float profiles, and satellite measurements of surface data. Specifically for the ocean, observations are limited and special techniques are required to manage the sparseness of the data. Two data assimilation schemes have been successfully implemented with a global ocean model: (1) the Local Ensemble Transform Kalman Filter (LETKF) of Hunt et al, and (2) an Optimal Interpolation (OI) method via the Simple Ocean Data Assimilation (SODA) of Carton et al, which has been shown to be mathematically equivalent to 3D-Var. In a 7-year historical reanalysis (1997-2003) using only observed vertical profiles, LETKF is shown to outperform the OI method - particularly in the data-rich equatorial regions. The LETKF system has since been upgraded to include the capability to assimilate observations of sea surface temperature, sea surface height, sea surface salinity, and drifter velocities. LETKF is currently being integrated with GODAS to facilitate the development of a hybrid EnKF/3D-Var data assimilation system.