Development Of Adjoint Assimilation System And Its Experimental Researches

Developing data assimilation system of numeric predict model by using the adjoint code technique is one of new four-dimensional data assimilation methods. In this paper, the results of simulating two rainfall cases show that LASG-REM model has good power of predicting rainfall. The adjoint code technique is employed for projecting the adjoint programs of Betts-Miller cumulus convection scheme of LASG-REM model and their correctness is verified, indicating that the adjoint program and its nonlinear program are converse in theory. The example of a numerical prediction model shows the basic method of developing adjoint assimilation system of numeric predict model. Experimental evidences on ideal and real data show that different weighting and scaling have different influence on the adjoint assimilation system,and the system has good ability of assimilation and the adjoint code method is a very effective tool.