Application of Sequential Data-Assimilation Techniques in Groundwater Contaminant Transport Modeling

Abstract

Groundwater contaminant transport modeling is basically performed to predict contaminant concentration and to understand the biochemical and physical processes that happen in the subsurface of porous media. Modelers have been faced with the challenge of accurately modeling the behavior and fate of contaminants in groundwater with models and techniques that incorporate the appropriate noise statistics and estimates the hydrogeologic parameters effectively. Unaccounted noise and uncertainties in the modeling greatly affect the accuracy of these predictions. In this paper, two Monte Carlo-based techniques, particle filter (PF) and Ensemble Kalman filter (EnKF), were applied to a three-dimensional (3D) groundwater contaminant transport model to accurately estimate the first-order decay rate and contaminant concentration at each time step. The PF and EnKF are embedded with Sampling Importance Resampling (SIR) and Singular Value Decomposition (SVD) concepts to avoid degeneracy and matrix singularity, respectively. The simulation is performed with a specified domain space and with ensembles and particles size of 50 for parameter estimation and concentration prediction. A set of sparse observation points selected at specific locations were used to update the predictions from the filter at each time step. An analytical solution is generated as true solution to test the accuracy of the predicted values. Algorithms to generate the simulation results were run. The first-order decay rate estimated using PF and EnKF all converged at