| This dissertation focuses upon the development and testing of flow-and-transport models for laboratory and field experiments involving the in-well vapor-stripping system. The in-well vapor stripping system combines air-lift pumping and groundwater recirculation to remediate groundwater laden with volatile organic compounds (VOCs) without lifting the groundwater above the ground surface. The primary motivation for using numerical modelling is to understand the field and engineered systems well enough to obtain reasonable agreement, if possible, between simulated and observed measurements of head and concentration at specific monitoring points.; A detailed, numerical analysis was carried out for a laboratory experiment of the in-well vapor-stripping system conducted in a 2-meter long tank containing homogeneous sand. A three-dimensional flow model with equilibrium transport was able to match most experimentally-derived concentrations histories of three VOCs (chloroform, toluene, and trichloroethylene) at 12 monitoring locations, with disparities arising primarily from differences between the actual and estimated initial concentrations at the downgradient end of the tank.; A detailed, numerical analysis was also carried out for a field demonstration conducted for 191 days at Edwards Air Force Base, California. A single vapor-stripping well was used to remediate a small portion of a large trichloroethylene plume. A layered flow model and a one-dimensional, nonequilibrium streamline transport model provided reasonably good matches of simulated to observed heads and concentrations. Refining the estimates of hydraulic conductivity through inverse modelling allowed the simulated measurements of head and concentration to more accurately match the observations. The improved estimates of conductivity show both elongated zones of high conductivity and localized zones of low conductivity, both of which are geologically plausible. The inverse modelling exercise confirms that the heterogeneity in the conductivity of the porous media is a primary factor affecting both head and concentration observations.; The inverse modelling of the field demonstration was aided by an adaptive data methodology for geostatistically-based inverse problems. Simulating field data requires that model parameters be estimated from sparse and spatially distributed data. The dissertation contains a short chapter on how the adaptive methodology utilizes a subset of quasi-redundant data to reduce computational expense. |
