Application of techniques in inverse problems to variational data assimilation in meteorology and oceanography

There is an international focus on the developments of data assimilation systems for meteorology and physical oceanography models and there has been considered interests in the "Inverse Problems" of determining poorly known initial boundary conditions and model parameters by incorporating measured data into the numerical model, taking into account both the information about dynamics about the model and the information about the true state which is constrained by a set of measurements. In this paper the data assimilation. problem in meteorology and physical oceanography is re-examined using the adjoint methods in combination with regularization ideas in inverse problem, then two sets of numerical experiments are performed to examine whether the proposed approach is capable to reconstruct the accurate initial boundary conditions and model parameters. One set of experiments are using global observations and the other with local observations, the numerical experiments show that variational data assimilation with regularization techniques contribute a lot to the stability and accuracy of the numerical calculation.