Risk Management of Groundwater Contamination in a Multiobjective Framework

摘要

This paper addresses the issue of uncertainty in groundwater contamination by applying risk analysis concepts to the problem of industrial chemical spills. A hypothetical aquifer system is considered that includes a factory and two water supply wells. Accidental spills of solvent at the factory enter the aquifer, causing well solute concentrations to exceed a mandated limit. Regulation forces the company owning the factory to reduce the frequency and magnitude of the spills. Its managers need to determine the optimal levels of investment in spill control technologies that will achieve three objectives: minimize the cost of contamination prevention, minimize the proportion (ratio) of time in which a maximum contaminant limit (MCL) is exceeded, and minimize the sensitivity of the MCL exceedance ratio to uncertainties in aquifer dispersivity. Simulation with a stochastic time series of spills gives sample values of the MCL exceedance ratio for values of the investment decision variables and dispersivity; the investment decisions determine the statistics of the time series. Use of regression enables calculation of a continuous function relating the contamination time ratio objective to investments and dispersivity. The third objective is an approximation to the standard deviation of the MCL exceedance ratio and is computed through the risk dispersion index method (RDIM). The RDIM incorporates the surrogate worth trade‐off method for optimizing the resulting multiple objectives. The simulations assume that the aquifer is in a steady state and behaves linearly. The concentration impulse response at the wells for a single spill is computed via a mass transport model. The well solute concentration over time, which is determined from the convolution of a series of spills, provides the basis for calculating the exceedance ratio. This ratio is defined as the portion of time that the pollution concentration limit is exceeded in some chosen time span. To obtain credible values of the exceedance ratio, several simulations, each encompassing a 20‐year planning horizon, are run for each scenario (policy opti