关键词: 流域河网, 水动力模型, 集合卡尔曼滤波, 数据同化

Abstract: Accurate simulation of large-scale river networks is still a difficult issue at present, and uncertainty in real environments is responsible for the inefficiency in improving the accuracy of traditional simulation methods. Data assimilation, which fuses data and observations, is an important method to improve the accuracy of practical predictions. This paper develops an assimilation model integrating the Taihu basin river network model and ensemble Kalman filtering algorithm to explore how field observations impact the assimilation effectiveness. The results show the water conservancy zones in this basin are relatively independent, thereby requiring a certain amount of observations in each zone to achieve effective assimilation results. And for each zone, two or three observation points with a sampling interval of 1-2 days will be enough to reduce the simulation errors by about 40% in comparison to the traditional method. The predicted stages of Taihu basin agree with observations with the maximum prediction error of flood stage no more than 0.02 m at stations of Yixing, Danyang, and Changzhou.

Key words: basin river network, hydrodynamic model, ensemble Kalman filter, data assimilation