Multicriteria management of groundwater quality under uncertainty

The primary purpose of this research project is to incorporate parameter uncertainty into the development of multicriteria planning and management tools for groundwater quality problems. We have focused upon three criteria--cost, water quality, and sensitivity . 'We have also focused upon a particular type of management strategy--the use of injection and extraction wells to control and remove a contaminant plume. Two different stochastic management tools have been developed in this project--the Marginal Sensitivity Technique (MST), and the Parameter Configuration Technique (PCT) . The former technique designs a "best" pumping scheme with respect to cost and parameter sensitivity and determines the tradeoff between these two criteria. The latter technique seeks to identify unfavorable (but physically plausible) spatial distributions of groundwater parameters. The MST uses an efficient method to compute the sensitivity of the hydraulic gradient along the plume boundary with respect to changes in transmissivity values throughout the flow domain. The maximum sensitivity is constrained to be less than or equal to a user-supplied parameter . A parametric linear programming algorithm is used to determine the tradeoff between cost and sensitivity. The PCT finds a "bad" set of spatially varying transmissivity values by solving a constrained optimization problem. The constraints, which guarantee that the transmissivity field is phsically reasonable, are based upon geostatistical concepts. The MST has been applied to a simple hypothetical problem involving uniform flow through a two-dimensional, homogeneous aquifer. The MST shows that it is possible to increase pumping (i.e., cost) in such a way so as to manipulate the groundwater system into states of low sensitivity to parameter changes. A hypothetical example problem was constructed to illustrate the use of the PCT. Two pumpout schemes were designed, under the assumption of uniform transmissivity; one scheme was based on one extraction well, the other on one extraction and one injection well. The least-cost design (pumping scheme) of each was then subjected to a PCT-generated transmissivity field. For the dataset used, the least-cost one-well scheme captured 85% of the contaminant and the two-well design captured 86%.U.S. Department of the InteriorU.S. Geological Surve