Locating monitoring wells in groundwater systems using embedded optimization and simulation models.

In this paper, a new methodology is proposed for optimally locating monitoring wells in groundwater systems in order to identify an unknown pollution source using monitoring data. The methodology is comprised of two different single and multi-objective optimization models, a Monte Carlo analysis, MODFLOW, MT3D groundwater quantity and quality simulation models and a Probabilistic Support Vector Machine (PSVM). The single-objective optimization model, which uses the results of the Monte Carlo analysis and maximizes the reliability of contamination detection, provides the initial location of monitoring wells. The objective functions of the multi-objective optimization model are minimizing the monitoring cost, i.e. the number of monitoring wells, maximizing the reliability of contamination detection and maximizing the probability of detecting an unknown pollution source. The PSVMs are calibrated and verified using the results of the single-objective optimization model and the Monte Carlo analysis. Then, the PSVMs are linked with the multi-objective optimization model, which maximizes both the reliability of contamination detection and probability of detecting an unknown pollution source. To evaluate the efficiency and applicability of the proposed methodology, it is applied to Tehran Refinery in Iran.