Stochastic groundwater quality management: Role of spatial variability and conditioning

Abstract

In this paper a methodology for the stochastic management of groundwater quality problems is presented. A stochastic algorithm to solve the coupled flow and mass transport inverse problem is combined with a stochastic management approach to design a reliable pump‐and‐treat scheme. Three main differences exist between this research and previously published research in the field of groundwater management. The first is the method of parameterization, where we have explicitly considered spatial variability of flow and mass transport parameters in our method. Second, we have used adjoint and/or sensitivity equations to calculate first and second derivatives of state variables with respect to hydraulic parameters. Finally, the third point is the nonlinear multiobjective formulation we adopt which enables us to explore the whole trade‐off curve between reliability and cost of remediation. A number of key results are listed here. The stochastic inverse algorithm used here, the representers method, makes it possible to estimate heterogeneity as reflected by the measurements of parameters and state variables. Meanwhile, using the adjoint operator to accurately estimate sensitivity of state variables with respect to the parameters has facilitated the use of the head and concentration measurements in the inverse algorithm. Moreover, stochastic management, which is conditioned on head and concentration measurements simultaneously are shown to provide more realistic and reliable remediation schemes. Finally, with the aid of the multiobjective formulation of the management problem and analysis of its results, it is possible to identify a significant remediation scheme. Such a scheme has the advantage of producing maximum possible reliability with the least possible increase in the total cost of remediation.