中文摘要：

由于地下水流动的发生环境通常非常复杂,通过稀少的钻孔资料难以对含水层特征进行准确的描述,而数据同化技术可以利用地下水位等动态数据反演出含水层的特征.为减小大尺度问题的抽样误差,提出了以确定性抽样技术为基础的卡尔曼滤波方法,讨论了确定性卡尔曼滤波方法在强烈非均匀介质和大尺度问题中的应用效果.研究结果表明：确定性卡尔曼滤波方法生成了唯一的样本集合,避免了传统集合卡尔曼滤波方法预测结果的不确定性;该方法能够缓解小样本条件下的系统方差快速衰减现象,并在强烈非均质介质中表现出良好的计算性能;结合局部化技术,确定性集合卡尔曼滤波方法能够很好地解决大尺度地下水系统的参数反演问题.

英文摘要：

The circumstance of groundwater flow is very complex;and the sparse borehole data can not provide a detailed description to aquifer properties.Data assimilation technique is commonly used to infer the properties according to dynamic observation data such as piezometric head.In this study,the ensemble Kalman filter method based on deterministic sampling technique is proposed to reduce the sampling error.The ability of deterministic ensemble Kalman filter is discussed for strongly heterogeneous porous media and large-scale groundwater problem.The results show that deterministic ensemble Kalman filter method generates unique sample ensemble;so it avoids the estimation uncertainty that traditional enesemble Kalman filter suffers.The proposed method alleviates the rapid decrease of system variance when the sample size is not sufficiently large.It is able to obtain satisfactory hydraulic conductivity with few realizations even for strongly heterogeneous porous media.By combining with localization technique,this method can efficiently deal with the inverse modeling in large-scale application.