基于局域化EnKF同时同化地下水水流和溶质运移数据的渗透系数场识别研究

局域化改进集合卡尔曼滤波（EnKF）可以克服EnKF方法在使用小集合时，对参数识别精度较低的缺陷，其能同化地下水位观测数据有效识别渗透系数场。实际工作中，溶质运移数据也较容易获得。崔凯鹏（2013）尝试增加溶质运移数据以改进只同化水流数据对渗透系数的估计结果，但是精度提高有限。本文在其基础上修改模型，进一步增加溶质注入井，探究同时同化水头和溶质运移数据，对渗透系数场识别效果，之后对比了局域化EnKF与非局域化EnKF参数识别结果，并分析了溶质影响范围与参数识别的关系。结果表明：同时同化溶质运移和水头资料，比同化单一种类观测数据识别的渗透系数精度更高；相同实现数目下，局域化EnKF比EnKF对渗透系数场的估计结果与真实场更为接近；仅考虑溶质影响范围内的渗透系数，同化水头数据在最后时刻参数识别结果好于同化溶质运移数据参数识别结果，但差别不大。

On Localized Ensemble Kalman Filter to Calibrate Heterogeneous Hydraulic Conductivity Field by Assimilating Hydraulic Head and Solute Concentration Observations

Localized Ensemble Kalman Filter (EnKF) is used to overcome the defect of EnKF that the accuracy of parameter estimationis low when a small ensemble is used. There are numerous study on using EnKF to estimate heterogeneous hydraulic conductivity byassimilating hydraulic head observations. In practice, solute concentration data are easy to collect. Cui (2013) used localized EnKF toestimate heterogeneous hydraulic conductivity field in a 2-D confined aquifer by assimilating the hydraulic head and solute concentrationobservations simultaneously, but the result was not satisfying for some unknown reason. In this study we use two solute injection wells(instead of one solute injection well) so that more effective solute concentration data can be collected during the assimilation process.In addition, we compare the results with the fields estimated based only on head and solute transport data, respectively. It is found that:(1) Localized EnKF yields more accurate hydraulic conductivity estimates than EnKF; (2) assimilating hydraulic head observationsand solute concentration observations together results in better estimates than assimilating any single type of observations. In addition,hydraulic conductivity estimates in the reach of plumes can be well estimated by using solute concentration or head observations.