正则化方法在风海流Ekman模型资料同化中的应用

VARIATIONAL DATA ASSIMILATION OF WIND-DRIVEN CURRENT MODEL

针对修正的风海流Ekman模型,采用变分同化技术结合反问题的正则化技巧,对模式的初始条件、边界条件和模式参数进行最优调整,使模式解与观测场能较好地符合。在对模式参数的估计过程中,引进正则化思想,在目标泛函中增加稳定泛函,从而克服在调整模式参数时造成的不适定。并通过数值试验进一步肯定正则化理论在资料同化中的作用。

The regularization method of inverse problem is used in combination with the variational data assimilation method to adjust the model initial/boundary conditions and parameters so that the analyzed model can agree with the observation. The computational instability will appear as a result of ill-posed inverse problem during variatioanal data assimilation and optimal model parameter estimation.The regularization ideas are introduced by adding a stable function to the cost function. The ill-posed variational assimilation during optimal adjustment of model parameters in some cases are corrected with satisfactory results. Ideal experiments are also conducted to verify the effect of the regularization method on variational data assimilation.