Stochastic investigation of flow and transport parameters via assimilation of concentration and water table data

Modelling groundwater flow and transport contaminant require heterogeneous parameters such as hydraulic conductivity, dispersivity and porosity which commonly are only sparsely available, if at all. State variables such as hydraulic head and concentration are generally more extensively sampled and can be assimilated to improve the characterization of the parameters. In the last decades, many works have focused on the characterization of hydraulic conductivity heterogeneity by assimilating piezometric heads using variations of the ensemble Kalman filter (EnKF). More recently, several authors have worked on the assimilation of concentrations to identify contamination source, aquifer geometry and K. The characterization of the variability of dispersivity and porosity, however, has not been addressed. In this work, the objective is to investigate the capability of the restart EnKF (r- EnKF) to identify hydraulic conductivities, dispersivities and porosities by simultaneously assimilating concentration and water table data in a two-dimensional numerical experiment of an aquifer vertical cross section. Reference data of hydraulic conductivity, dispersivity and porosity are generated. Groundwater flow and transport equations are solved using these reference data. The concentration and the water table obtained from the numerical model are used as the reference aquifer response. Values of the state variables of the reference aquifer are sampled at a limited number of points to serve as assimilating data for the inverse problem. Prior variograms functions of the hydraulic conductivity, dispersivity and porosity are assumed and three hundred equally-likely realizations, conditioned to data from reference aquifer, are generated. Stochastic inverse modelling is conducted using the r-EnKF for the identification of hydraulic conductivity, dispersivity and porosity. The results are analyzed by using the average absolute bias (AAB) that represents a measure of accuracy between the reference values and the realization values. Our results show the importance of the concentration and water table data for improving the characterization of hydraulic conductivity, dispersivity and porosity