A weakly coupled ocean-atmosphere ensemble data assimilation system

The Data Assimilation Research Testbed ensemble Kalman filter has been combined with the Community Atmosphere Model (CAM) version 4 and the Parallel Ocean Program (POP) to create weakly coupled ensemble analyses for the last decade. First, DART/CAM is used to create an 80-member reanalysis assimilating all observations from the NCEP/NCAR reanalysis. Each member of the CAM assimilation is forced from below by the same ocean analysis. A 48-member reanalysis with DART/POP is created using all temperature and salinity observations available in the World Ocean Database. The atmospheric forcings required for POP are obtained from the first 48 ensemble members from the CAM reanalysis. Each POP member sees a different sample estimate of the atmospheric forcing.

The impact of using an ensemble of atmospheric forcings is assessed by comparing to a case in which a single atmospheric forcing is used. The impact of using the adaptive inflation and sampling error correction algorithms in DART is discussed. Details of the analyses near the northern hemisphere western boundary currents and in the ocean near the equator are examined. A brief look at the impact of using analyses as initial conditions for decadal predictions with a coupled CAM/POP model is presented.