中文摘要：

为了在水流计算中定量化利用遥感水位数据，基于偏微分方程最优化控制理论建立了变分模型来融合二维平面非恒定流的数学模型和数据。根据遥感数据空间信息密集的特点，提出遥感水位数据同化的新算法。采用人工合成数据考察了面域遥感数据对于定常参数和时变两类参数的反演效果。试验结果表明，遥感数据提供的空间分布式信息有利于空间分布式参数的反演识别，而且通过引入考虑水面空间变化信息的附加项，可以改善观测信息的同化，更好地辨识时变参数（流量过程）。以Moselle河的RADARSAT卫星遥感水位数据检验了模型的实用性。

英文摘要：

Taking the initial condition, the flow boundary condition and the roughness as the control parameters, an adjoint model of 2D horizontal unsteady flow model is established based on the optimal control theory of partial differential equations. A variational data assimilation method is developed by the mathematical model and the field data of flow. According to the spatial density characteristic of remote sensing data, a new item of cost function is introduced in order to improve the assimilation of remote sensing water levels with the emphasis laid on the assimilation of remote sensing water level. The assimilation numerical experiment of synthetical single water level image shows that the spatially distributed information provided by the remote sensing data is available to the identification of spatially distributed parameters, and the additional item is useful for the assimilation of measurement and identification of time-dependent parameters. In the simulation of an actual river, the method is applied to the assimilation of distributed water level data from satellite remote sensing images.