Stochastic analysis for optimal management strategies applied to the remediation of contaminated groundwater/aquifer systems

The design and operation of effective remediation systems that comply with technical, economic, regulatory, and social constraints is typically an extremely expensive and challenging process given the variety of hydrogeological and contamination settings of hazardous waste sites. The task requires the implementation of innovative site characterization programs, the use of state-of-the-art data assimilation and visualization techniques, and in-situ and ex-situ technologies. One of these technologies is mathematical optimization, which can be applied to both the design of new Pump-and-Treat schemes for the remediation of contaminated aquifers, and the improvement of the efficiency of currently active cleanup systems. Because of high cost normally involved in constructing and operating these systems, the use of mathematical optimization has the potential to yield substantial saving and to improve remediation systems designed for the cleanup of large and complex plume distributions.; Typical groundwater remediation design problems consider the placement of a number of injection and extraction wells, and determination of flow rate schedules in order to identify the best management alternatives while considering management objectives and constraints. The modelling approach to this problem considers the coupling of optimization algorithms with flow and transport numerical models to determine optimal remedial designs. The approach is limited by the uncertainties characterizing groundwater flow and transport models. Lack of data about hydrogeological settings, subsurface heterogeneities, contaminant sources and plume distributions, reaction pathways and rates, ultimately can lead to remediation systems that are either over-designed or have a certain probability of failure. Failure of the system may be defined as the violation of the established performance criteria, that is, the violation of constraints in the optimization formulation. Under conditions of uncertainty the optimal remediation design inevitably acquires a stochastic nature, whose ultimate goal is that of determining remedial strategies that tradeoff increasing levels of reliability against increasing cleanup costs.; There are four major thrusts of this work. First, the parameter uncertainty problem is tackled in terms of the tradeoff between cost-optimality and reliability. The management constraints are reformulated into additional objective functions represented by either the probability of failure of a given remediation policy, or by the expected penalty incurred in case of constraint violations. Second, one of the major limitations of stochastic optimization approaches lies in the heavy computational cost associated with linking Monte Carlo flow and transport simulation with optimization algorithms. An innovative methodology is thus developed to drastically reduce the computer effort by calibrating surrogate forms of the management objectives. Third, since cost-optimality and reliability of a cleanup system may be improved with the collection of data, which reduces parameter uncertainty and consequently the risk, the optimal tradeoff between increasing site-investigation costs and decreasing management is identified. Fourth, an adaptive management framework is devised to dynamically improve the efficiency of the remediation system based upon data collected during the actual cleanup process.