Modeling Ground Water Quality Sampling Decisions

Abstract

Questions such as what, where, when, and how often to sample play a central role in the development of monitoring strategies. Limited resources will not permit sampling for many contaminants at the same frequency at all well sites. Therefore, a resource allocation strategy is necessary to arrive at answers for the preceding types of questions. Such a strategy for a ground water quality monitoring program is formulated as an integer programming model (an optimization model). The model will be of use in the process of deciding what constituents to sample and where to sample them so as to maximize a given objective, subject to a set of budget, sampling, and regulatory constraints. The maximization objective in the model is defined as a weighted function of population exposure to a scaled measure of observed chemical concentrations. The sampling constraints are based on the observed variability of contaminants in the aquifer, needed precision in estimates, a chosen level of significance, the available budget for implementing the program, and selected regulatory constraints. The model is tested with field data obtained for 10 selected constituents from more than 650 wells in the Cambrian‐Ordovician aquifer in Iowa. Results from two alternative formulations of the model are compared, analyzed, and discussed. Further avenues for research are briefly outlined.

Long-term Monitoring Plan for the Central Nevada Test Area

“…They listed two general types of approaches to network design, namely the hydrogeologic and the statistical approaches, where the latter type can be further divided into simulation, variance-based, and probability-(or risk-) based techniques. Andricevic (1996) pointed out that the monitoring network design approaches can in general be based on geostatistical methods (e.g., Carrera et al, 1984;McLaughlin and Graham, 1986;Rouhani, 1985;Rouhani and Hall, 1988), optimization methods (e.g., Olea, 1984;Loaiciga, 1989;Andricevic, 1990;Hudak and Loaiciga, 1992;Hsueh and Rajagopal, 1988;Hsu and Yeh, 1989), methods based on extensive simulation (e.g., Meyer and Brill, 1988;Massmann and Freeze, 1987b), the transfer function method (e.g., Andricevic and Foufoula-Georgiou, 1991), or the Bayesian decision theory methods (e.g., Grosser and Goodman, 1985;James and Freeze, 1993;James and Gorelick, 1994).…”

“…Andricevic (1990) indicates that the two commonly used approaches to analyze and design monitoring networks are optimization and simulation. Andricevic and Foufoula-Georgiou (1991) categorize the monitoring network design approaches into mixed-integer programming approaches (Hsu and Yeh, 1989), kriging and co-kriging application (Carrera et al, 1984;McLaughlin and Graham, 1986), variance-reduction analysis (Rouhani, 1985), nearest neighbor approach (Olea, 1984), and methods based on optimization (e.g., Hsueh and Rajagopal, 1988;Loaiciga, 1989;Andricevic, 1990) and simulation (e.g., Meyer and Brill, 1988;Massmann and Freeze, 1987a, b).…”

“…Therefore, there is always an objective function, such as minimizing the estimation variance of groundwater quality indicators such as contaminant concentration (Knopman and Voss, 1988a;Loaiciga, 1989). The objective function is normally subject to different constraints such as resource constraints, the governing equations of the physical processes (e.g., hydrodynamic dispersion), the statistical constraints (e.g., accuracy of groundwater quality parameter estimates (Hsueh and Rajagopal, 1988)), and areal coverage of the monitoring network. The mathematical programming problem, represented by the optimization of the objective function, subject to these constraints, is then solved by appropriate algorithms.…”

“…Ultimate objectives explicitly consider the value of groundwater quality information in achieving monitoring goals such as environmental protection (Hsueh and Rajagopal, 1988), the reduction of remediation costs, and minimizing exposure risks (Massmann and Freeze, 1987a) or health hazards. Surrogate objectives are substitutes for ultimate objectives, and are used in many cases to bypass the difficulties posed by the formulation of network design problems in terms of ultimate objectives.…”

Cost‐effective long‐term groundwater monitoring design using a genetic algorithm and global mass interpolation

“…Meyer and Brill [1988] built upon this work by combining Monte Carlo simulation with an optimization model, which was formulated as a maximal covering location problem. Analytical optimization formulations have since been used in several studies to identify monitoring network designs at landfill sites that maximize the likelihood of intercepting plumes from leaking landfill leachate [Hsueh and Rajagopal, 1988;Loaiciga, 1989;Hudak and Loaiciga, 1993]. Hudak and Loaiciga [1992] present a heuristic approach using facility location theory to augment preexisting monitoring networks.…”