集合Kalman滤波同化方法在福建沿岸潮汐数值模拟和预报中的应用研究

Application of Ensemble Kalman Filter Data Assimilation Method in Tide Simulation and Prediction in the Taiwan Strait

为了提高潮汐数值模拟和预报的准确度,利用《2009潮汐表》,建立了含集合Kalman滤波同化模块的潮汐数值预报模型,对集合Kalman滤波同化模块中的参数进行确定,并对同化效果进行了检验.为了确定集合Kalman滤波同化模块中的参数(主要是集合数和截断半径),考虑了42种参数组合,在这些组合中,以模拟准确度和计算相同时段的潮汐值所需时间为标准,存在最优的参数组合:其中,东山、厦门、娘宫、崇武4个站最优的截断半径为35 km,三沙站最优截断半径为30 km,5个站的最优集合数均为20.实验结果表明:将集合Kalman滤波同化方法用于潮汐数值预报模拟中是可行的,同化验潮站的潮汐表上的潮位数据对验潮站本身及其邻近网格点的潮汐模拟改进最大;当5个站不同化调和数据进去的时候,东山、厦门、娘宫、崇武4个站的模拟准确度,在未来12 h仍然有所提高.

In order to improve the accuracy of numerical simulation, a numerical model of tide with an Ensemble Kalman Filter（En- KF） data assimilation technique is established. In this article we take into account 5 stations,which are Dongshan, Xiamen, Nianggong,Chongwu,Sansha. The data assimilated in the model is from the book of 2009 Tide Table. In order to decide EnKF the best combination of two parameters, which are the cut radius and ensemble number, we consider 42 kinds of parameter combinations at each station. Based on the accuracy of numerical simulation and the time spent in calculating the same period of tide, we can define the best parameter combination at each station. Numerical experiments show that the best value of cut radius is 35 km for Dongshan, Xiamen, Nianggong and Chongwu stations, and the best value of cut radius is 30 km for Sansha station. The unique best value of ensemble number for 5 stations is 20. It can be seen from numerical experiments that data assimilation with EnKF method can improve the accuracy of tide simulations and tide predictions in the following 12 hours in the Taiwan Strait.