论文部分内容阅读
海洋环流的定常解一般是通过直接积分模式一段足够长的时间而得到．近年来在海洋气候资料同化研究中，在代价函数（ｃｏｓｔｆｕｎｃｔｉｏｎ）中引入了时间导数项，利用伴随法求解，使时间导数项和与观测的方差之和最小，以保证解是准定常的并与观测吻合。本文将代价函数中的观测项和时间导数项分别考虑，来测试算法效能。如果在代价函数中仅有时间导数项，则问题是求海洋环流的定常解。本文的目的是通过与直接积分方法相比较，来考察伴随法求定常解的效率。利用ＩＡＰ浅水方程模式及其伴随模式进行的数值试验也表明伴随法求解定常解的效率不优于直接积分的方法，需要进行大数量的迭代才能使解收敛到准定常解。如果在代价函数中仅有观测项，并且资料是完全的，伴随法可以很快地求出与资料拟合的解。这说明了如只求定常解，伴随法并不合算，其优点在于资料同化。
The general solution to ocean circulation is usually obtained by direct integration of a long enough period of time. In recent years, in the study of ocean climatic data assimilation, the time derivative term is introduced into the cost function, which is solved by the adjoint method to minimize the sum of the time derivative and the observed variance so as to ensure that the solution is quasi-steady and Observed consistent. In this paper, the cost function in the observation items and time derivative items were considered separately, to test the efficiency of the algorithm. If there is only a time derivative term in the cost function, the problem is to find a steady solution to the ocean circulation. The purpose of this paper is to compare the efficiency of the adjoint method to solve the constant solution by comparing with the direct integral method. Numerical experiments with IAP shallow water equations and their accompanying modes also show that the efficiency of the adjoint method to solve stationary solutions is not better than that of direct integrals. A large number of iterations are needed to make the solutions converge to quasi-stationary solutions. If there are only observations in the cost function, and the data is complete, the adjoint method can quickly find the data fit solution. This shows that as long as the solution is given, the concomitant method is not cost-effective, its advantage lies in data assimilation.