The Interaction Between Large-scale Circulation And Estimation Of Boundary Conditions For Numerical Model

This paper consists of two parts. Firstly, the large-scale circulation system plays an important role in global climate changing and the mutual effect is often accompanied by climate change. So, the research of the interactions between the large-scale horizontal circulation, the meridianal circulation and the zonal circulation is significant. In addition, the model error is unavoidable under the existing numerical forecast model. The model error is one of the main causes of the inaccurate prediction. There are many ways to revise the model error. But most of them are based on the assumption of that the boundary conditions are satisfied rigorously. But the fact is that it really difficult to obtain the accurate data of the boundary conditions. So if we could solve the estimation problem of lower boundary conditions effectively by using the historical observations. It must be helpful to improve the prediction effect.In the first part of this study, our method is to combine the three-dimensional circula-tion decomposition model with the primitive equations of motion in spherical coordinates and build the large-scale horizontal circulation, the meridional circulation and the zonal circulation dynamic equations. Then on the basis of that the physical properties are in-variant, these equations are simplified under the assumption of non-dissipative and heat insulation. Finally, we obtain a simplified model of the interactions of large-scale circula-tion and expect to reveal the correlations between the large-scale horizontal circulation, the meridianal circulation and the zonal circulation, provide theoretical basis for climate prediction.In another piece of this work, according to the continuous evolution of atmospheric motion, the lower boundary conditions of numerical model must be embedded in the latest observation data. If we regard the lower conditions as unknown terms, then the latest ob-servation data could be regarded as the particular solution of the numerical model which satisfied the accurate boundary conditions. So, the problem of confirming the numerical model’s boundary conditions turns into finding the unknown term of the equation under the condition of knowing the particular solution. It is essentially an inverse partial dif-ferential equation(PDE). In this paper, we use the idea of inverse problem, introduce the regularization theory, transform the numerical model’s boundary problem into optimal control problem. Then we could solve the problem by optimization methods. At last, we have done some numerical experiments, the result shows that by using the latest observation data, transforming the numerical model’s boundary problem into inverse PDE and solving it by optimization methods is an effective way to improve the prediction effect.