Seasonal climate modeling over the Indian Ocean by employing a 4D-VAR coupled data assimilation approach

We carry out the first attempt to apply an adjoint method to a coupled general circulation model (CGCM) toward enhancing a skill in seasonal climate modeling. Focusing on 10-day mean errors of a CGCM output, we optimize the oceanic initial conditions together with the bulk adjustment factors by employing a four-dimensional variational data assimilation approach. We perform 9-month-long assimilation experiments independently every 6 months between January 1990 and March 2000. When using the optimized values for the initial conditions and the adjustment factors, a set of 9-month-long, 10-member ensemble simulation always displays realistic seasonal cycle and its interannual modulations over the tropical Indian Ocean (e.g., growing, mature, and decaying phases of the Indian Ocean Dipole Mode events). The optimized values of the bulk adjustment factors primarily reduce the model biases in climatological fields, while the optimization of the oceanic initial conditions largely contributes to a realistic representation of the interannual modulations of seasonal cycle. In the overlapped seasons (i.e., January-March and July-September), the ensemble mean states derived from two experiments show only slight differences in seasonal climate variations over most of the Indian Ocean. These results validate that our assimilation approach is generally effective for advancing a seasonal climate modeling and for obtaining a realistic analysis that is compatible between atmosphere and ocean.