Application of dedicated monitoring: network design for unknown pollutant-source identification based on dynamic time warping

Implementation of monitoring strategy for increasing the efficiency of groundwater pollutant source characterization is often necessary, especially when inadequate and arbitrary concentration measurement data are initially available. The research reported in this paper focuses on estimating three main parameters that are essential for efficient and accurate characterization of groundwater pollution sources, as follows: (1) location of source, (2) its starting time of release, and (3) duration of its activity. Most of the methodologies developed so far for unknown pollutant source identification have not adequately addressed the complexities involved with estimation of starting time of release and the duration of activity. Estimation of the time gap between the first observation of contamination in the aquifer at a location and the starting time of release is important for source identification. The main complexity arises due to the fact that the spatial location and the duration of activity of a pollutant source are interrelated. Therefore, explicitly specifying one and solving for the other simplifies the source characterization problem. In the research reported in this paper, both the source location and starting time of release are treated as explicit unknowns. The developed methodology uses dynamic time warping (DTW) distance as a cost function in the linked simulation–optimization model to design a monitoring network to efficiently estimate source characteristics including the starting time of release of unknown groundwater pollutant source. Performance of the developed methodology is evaluated with data obtained from a real aquifer. The evaluation results demonstrate that pollutant source characterisation based on pollutnat concentration measurements obtained from a designed monitoring network consisting of a fraction of total observation wells available compares very well with that based on all concentration information recorded at all the 74 monitoring wells over a period of 4 years. These evaluation results demonstrate the potential use of the developed methodology for efficient identification of unknown contaminant source in an aquifer