Fractal Singularity–Based Multiobjective Monitoring Networks for Reactive Species Contaminant Source Characterization

Abstract

The first step in effective design of contaminated aquifer site remediation is the accurate characterization of contaminant sources, which requires a large amount of concentration measurement data. However, in real-world scenarios contamination monitoring wells are generally arbitrary in location, monitoring data are sparse in time and space, and there are various uncertainties in predicting the transport process. It is a very challenging problem to optimally design effective monitoring networks intended for accurate unknown contaminant source characterization, with multiple potential source locations. In this study, the local singularity mapping technique is utilized to obtain potential monitoring well locations, which are used as input to the optimal network design model. This set of potential monitoring locations is utilized for selecting the subset of the optimal monitoring locations. This method of selecting the set of potential locations can improve the efficiency of the designed monitoring network for source characterization. The proposed methodology utilizes a multiobjective optimization algorithm for solving a two-objective optimal monitoring network design model. The optimization model is linked to a numerical simulation model simulating flow and transport processes in the aquifer. While constraining the maximum number of permissible monitoring locations, the designed optimal monitoring network improves the accuracy of unknown contaminant source characterization. The designed monitoring network can decrease the degree of nonuniqueness in the measured set of possible aquifer responses to geochemical stresses. The potential application of the developed methodology is demonstrated by evaluating the performance for an illustrative contaminated mine site aquifer. These performance evaluation results show the improved efficiency in source characterization when concentration measurements from the designed monitoring network are utilized.