Parameter Estimation and Well Installation Optimization for Contaminant Plume Delineation

Abstract

Groundwater contamination by dense non-aqueous phase liquids (DNAPLs) continues to be a significant environmental problem. DNAPL site characterization forms the basis of DNAPL site remediation success and requires attention. Mathematical modelling was utilized to delineate a trichloroethene (TCE) plume resulting from the dissolution of DNAPL in a subsurface environment. A numerical model (DNAPL3D-RX) was used to generate high-resolution field-scale three-dimensional groundwater concentration datasets (referred to as “true” plumes) and semi-analytical models (SCOToolkit3D and SWIF routine) were used to predict the maximum “true” plume extent. The study of SCOTookit3D found that the applied model can predict the maximum “true” plume extent using different monitoring framework settings and having different uncertainties in model input parameters, yet this tool had the limitation of over predicting concentrations by one to three orders of magnitude. The failure of SCOToolkit3D to accurately characterize the “true” plumes motivated the development of the stochastic well installation framework (SWIF) routine, which not only mitigated the earlier discovered failure of SCOToolkit3D but also provided a formal decision support tool for groundwater practitioners to reduce cost and uncertainty in developing a conceptual site model (CSM) by supporting monitoring well or membrane interface probe (MIP) installation decisions through a probabilistic approach.

The study of the SWIF routine found that installing monitoring wells or MIPs within low and high probability target zones (a target zone is a specific probability contour range on the probability map generated by the routine) will likely result in a large quantity of monitoring unit installations and a corresponding high site characterization cost, compared to installing monitoring wells or MIPs in a medium level probability target zone. The results also demonstrated the successful estimation of key attributes of a CSM in both homogenous and heterogeneous porous media using monitoring well and MIP installations. The results of this research provide the necessary basis to understand the model input parameter uncertainty and the site investigation strategy by optimizing the well installations required for the development of a meaningful CSM.