集合资料同化方法的理论框架及其在海洋资料同化的研究展望

在海洋动力系统的数值模拟中,海洋资料同化是一种能够有效融合多源海洋观测资料和数值模式的方法。它不仅可以显著地提高数值模拟的效果,构造海洋再分析资料场,还能有效减少海洋和气候预报时模式初始条件的不确定性。因此,海洋资料同化对于海洋研究和业务化应用具有非常重要的意义。资料同化方法的研究一直是大气、海洋科学的热门课题之一。其中,集合卡尔曼滤波器(EnKF)是一种有效的资料同化方法,自提出以来经过了20多年的发展和改进,已经在海洋资料同化中得到了广泛的研究和应用。近年来,随着动力模式的不断发展和计算能力的提高,粒子滤波器由于不受模型线性和误差高斯分布假设的约束,也逐渐成为了当前资料同化方法研究的热点。本文分析和总结了目前关于集合卡尔曼滤波器和粒子滤波器的一些最新理论研究结果,在贝叶斯滤波理论的框架下讨论了这两类算法的关联和区别,以及各自在资料同化实践中的优势和不足。在此基础上,我们探讨了粒子滤波器应用于海洋模式资料同化的主要困难和目前可行的一些解决方法,展望了集合资料同化方法研究的新趋势,为集合资料同化方法的进一步发展和应用提供理论基础。

The theoretical framework of the ensemble-based data assimilation method and its prospect in oceanic data assimilation

In the numerical simulation of the ocean dynamic system,data assimilation is able to use the limited observation data and numerical model to best estimate the ocean state,and effectively reduce the uncertainty from the initial conditions. Therefore,data assimilation plays an important role in the study of modern physical oceanography. The ensemble Kalman filter (EnKF) is an effective data assimilation method,which has attracted broad attention in oceanic data assimilation since it is proposed about twenty years ago. In recent years,the particle filter (PF) has become a hot research field,for it is not restricted by the linear and Gaussian assumption of the model. This paper analyzes and summarizes the current theories about the EnKF and PF,in the framework of Bayesian filtering theory. The EnKF and PF algorithms are proposed and compared. On this basis,we further discuss the major obstacle for applying the particle filter in oceanic data assimilaiton at present. Some feasible solutions are also introduced. This paper is expected to provide theoretical basis for further development and application of the ensemble-based data assimilation method in oceanic data assimilation.