基于集合卡尔曼滤波的渤海水位低频波动数值模拟

NUMERICAL SIMULATION OF LOW-FREQUENCY WATER-LEVEL FLUCTUATION IN BOHAI SEA BASED ON ENSEMBLE KALMAN FILTER

风松弛导致亚潮尺度的水位低频波动是渤黄海秋冬季常见的海洋动力过程,这一过程的数值模拟要求模型空间范围要覆盖整个渤黄海区域甚至是整个东部陆架海区域。对于以渤海海峡为开边界的渤海模型,则由于风区过小无法再现这一过程。针对这一问题,本文在渤海模型上使用集合卡尔曼滤波(ensemble Kalman filter,ENKF)进行同化,成功再现了这一现象,并进一步分析了ENKF同化的作用及其影响因素。主要结论如下:渤海的水位低频波动与输入渤海的开边界水位有关,仅使用潮汐数据和风场数据无法准确地模拟出渤海水位的变化,然而通过ENKF同化则可以再现渤海内的水位低频波动,且同化效果优于两类不同开边界输入的结果。ENKF同化的准确性与集合数目以及观测数据采样位置有关。集合数目越多同化效果越强,但超过40个集合后,同化效果改善不明显。使用的观测数据应尽量均匀分布且要远离开边界。

The low-frequency fluctuation of water level at a subtidal scale caused by wind relaxation is a common marine dynamic process in the Bohai Sea and Yellow Sea in autumn and winter.It requires the numerical modelling to cover the entire Bohai Sea and Yellow Sea region and even the East China Sea to reproduce this phenomenon.As in the Bohai Sea model that taking the Bohai Strait as the open boundary,this modelling cannot be reproduced due to limited wind fetch.To solve this problem,we employed the ensemble kalman filter(ENKF)assimilation method,with which this phenomenon was successfully reproduced.Furthermore,the role of ENKF assimilation and its influencing factors was analyzed.It was found that the low-frequency fluctuation of water level in the Bohai Sea is driven by the water-levels at the open boundary.Using tidal data and wind data only could not reproduce the subtidal water level variation in the Bohai Sea.However,the low-frequency fluctuation of water level can be reproduced by the ENKF assimilation;and the assimilation result is better than the results driven sole by the open boundary forcings.The accuracy of ENKF assimilation is related to the number of ensembles and the location of sampling stations.Although more ensembles would lead to better assimilation results,over 40 ensembles could not improve the results significantly.The sampling stations selected should be evenly distributed in the Bohai Sea and kept far enough away from the boundary.