基于局域化集合卡尔曼滤波的含水层弥散度场识别研究

An approach based on localizaed ensemble Kalman filter to identify groundwater dispersivity field

溶质运移模型对下水污染物运移预测有重要意义,但是准确获取模型参数具有一定难度.集合卡尔曼滤波（EnKF）方法可以融合多来源观测数据对同化系统进行优化修正,从而得到与真实情况接近的参数.将二维承压含水层理想算例的溶质观测数据应用于局域化集合卡尔曼滤波同化系统,估计含水层的弥散度场,并探讨了模型实现数目、初始猜想场的统计特征、观测点数目及时空分布、观测误差对参数估计结果的影响.结果表明,通过同化浓度观测资料可较好地估计溶质运移模型的弥散度场;对于所用模型,实现数目在100~700时,参数估计结果最好;初始猜想场与实际场越接近、观测数据误差越小,越能快速获得较好的估计结果.

Solute transport simulation is significantly important to predict contaminant transport,however,the parameters used in the model are very difficult to obtain precisely.Ensemble Kalman Filter（EnKF）assimilates observation data from multiple sources to get parameters approach to the real values.In this paper,we assimilate solute transport data observed from a two-dimensional confined aquifer with a covariance localized Ensemble Kalman Filter system to estimate the real dispersivity field.The effects of the number of realizations,statistical characteristic of initial assumed field,the number of observations,configuration of observations,and observation error on the efficiency of this method are investigated.Results indicate that a well estimated dispersivity field can be obtained by assimilating transport data with localized EnKF.The optimal number of realization is from 100 to 700 for the model we study,and excessive or insufficient number of realizations affect the accuracy of the assimilation results.A better estimation results can be obtained if initial assumed field is similar with the real logarithmic dispersivity field and the observation error is small.