Development and validation of a system identification methodology for the characterization of contaminated sites

Abstract

The quality of groundwater has become a major concern in the United States and many other industrialized countries since the discovery of numerous sites with contamination from hazardous wastes and leaked fuels. Investigation and monitoring have begun at many of these sites, but execution of remedial plans has often been delayed due to regulatory and financial constraints as well as limited understanding of the processes that control the distribution of contaminants in the subsurface. One of the primary difficulties encountered in the site remediation process is the inability to determine the site heterogeneity in an adequate manner. In this research, a mathematical technique known as System Identification Methodology (SID) is used in conjunction with the flow and contaminant transport equations to address the above problem.

In this mathematical procedure, a finite volume formulation with an upwind velocity scheme was used to discretize the flow and transport equations. In order to determine site heterogeneity, independent contaminant transport parameters were assigned to each rectangular element in the flow domain. The finite volume formulations of groundwater flow and contaminant transport equations were then reorganized into the standard state-space form that is commonly used in system identification procedures. The determination of the distribution of unknown parameters at the site was then accomplished by minimizing the error between both measured contaminant concentration and hydraulic head and calculated contaminant concentration and hydraulic head by the above finite volume models for groundwater flow and contaminant transport. The Levenberg-Marquardt algorithm was used as the optimization scheme.

Once the mathematical model was developed, test cases were run to verify the mathematical accuracy of the model. Sensitivity analyses were performed to determine the relative significance of the heterogeneities in dispersion, retardation, and decay terms on contaminant flow. The model was validated by applying it to actual observations from four selected case studies. As the first step in the validation, the SID methodology was applied to contaminant concentration data obtained from large-scale sand tank tests. Subsequently field validation was accomplished by applying the SID methodology to site in the Southern High Plains of Texas, Jordan aquifer in Iowa and landfill site at Borden.