The Application And Discussion Of Ensemble Kalman Filter Method In Groundwater Flow And Solute Transport Data Assimilation

Hydrogeology parameters are one of the key factors in groundwater flow and solute transport models. Since the spatial variability of aquifer medium, hydrogeology parameters cannot be described accurately. Ensemble Kalman Filter(EnKF) is an efficient data assimilation method that can assimilate various kinds of observation to estimate hydrogeology parameters. In recent years, it has gradually began to be applied to the field of hydrogeology.Based on the previous researches, this paper investigates the covariance localization scheme of EnKF first and a two-dimensional synthetic example is constructed for comparing between localized EnKF and standard EnKF. The results show that localized EnKF can improve the assimilation results and solve sampling noise problem better compared with standard EnKF, especially when the ensemble is small.Then, the effects of time/spatial density on EnKF and covariance localization scheme is investigated. The results indicate that with the spatial density increasing, localized EnKF exhibits a promotion of calculation accuracy, while the results of the standard EnKF does not show such trend. Except for the poor results got when the time density is too dense, the general trend shows that the increase of time density leads to better calculation results but the amplitude varies with different numbers of observation wells:the larger the observation well number is, the less remarkable the result will be. Localized EnKF of lower spatial density can exhibit even better than that of higher spatial density with an optimized time density. This study has important guidance meaning for the design of observation wells arrangement.At last, a MODFLOW-MT3DMS-EnKF coupling model is given and concentration observation data is used for the assimilation to estimate hydrogeology parameters. The research indicates that hydrogeology parameters can be estimated only by concentration observation data assimilation, but the results are not as well as CASEs only using head observation data. Concentration observation data can bring much more available information when its observation design is reasonable.