集合最优平滑同化方法的研究与应用

Research and application of Ensemble Optimal Smoothing

海洋状态场的历史变化过程对其分布状态有重要影响。在观测资料稀疏的情况下,合理利用历史观测资料能够为海洋数据同化提供大量有效信息。然而在目前的顺序资料同化过程中,往往只同化当前时刻的观测数据,没有考虑到历史观测资料对当前状态的约束。四维变分虽然可以体现变量在时间维度的演变过程,但引入伴随方程会增加计算代价。本文基于集合最优平滑同化算法(Ensemble Optimal Smoothing,EnOS)探讨了一种在数据同化中加入历史观测资料的简易可行方案,其能够根据历史观测数据估计当前状态,并进行单点同化实验和区域同化实验来验证该方案的有效性。实验结果表明,将历史观测资料引入到同化过程中可以把控时间演变趋势,减小分析数据与真实值之间的偏差,更有效地消除数值模式误差,提高同化质量。

The historical change process of the marine state field has an important influence on its current distribution state.In the case of sparse observational data,rational use of historical observational data can provide a large amount of effective information for ocean data assimilation.However,in the current sequential data assimilation process,the observation data of the current moment is often assimilated,and the constraints of the current state of the current state are not considered.Although in the four-dimensional variable division,although the evolution of the time dimension can be reflected,the calculation cost is introduced with the accompanying equation.Based on the Ensemble Optimal Smoothing(EnOS),the paper discusses a simple and feasible solution for adding historic observation data to data assimilation,and performs single-point assimilation experiments and regional assimilation experiments to verify the effectiveness of the method.Its main advantage is the ability to estimate the current state from observations in historical time.The experimental results show that adding historical observation data to the assimilation process can control the time evolution trend to reduce the deviation between the analysis data and the real value,eliminate the model error more effectively,and improve the assimilation quality.