Development of integrated methodologies for optimal characterization of reactive contaminant sources and monitoring network design in polluted aquifer sites

Groundwater is one of the most important natural resources in many parts of the world; however it is widely polluted due to human activities. Reliable groundwater management and remediation strategies are generally developed following the identification of groundwater pollution sources, where the measured data from monitoring locations are utilized to estimate the unknown pollutant source location, magnitude and duration of activity. However, accurately identifying characteristics of unknown contaminant sources is a challenging task due to uncertainties in terms of predicting source flux injection, hydro-geological and geo-chemical parameters, and the concentration observation field measurement. Although sufficient concentration measurement data are essential to accurately identify source characteristics, available data are often sparse and limited in quantity. Therefore, this inverse problem of characterizing unknown groundwater pollution sources is often considered ill-posed, complex and non-unique. Different methods have been utilized to identify pollution source characteristics; however, the linked simulation-optimization approach is one effective method to obtain acceptable results under uncertainties in complex real life scenarios. With this approach, the numerical flow and contaminant transport simulation models are externally linked to an optimization algorithm, with the objective of minimizing the difference between measured concentration and estimated pollutant concentration at observation locations. Concentration measurement data are very important to accurately estimate pollution source properties; therefore, optimal design of the monitoring network is essential to gather adequate measured data at desired times and locations. A simulation model should be utilized to accurately describe the aquifer processes properties in terms of hydro-geochemical parameters and boundary conditions. However, the simulation of the transport processes becomes complex when the pollutants are chemically reactive. An additional difficulty with linked simulation-optimization models is that an optimal solution generally requires huge computation time, due to iterative repeated solution of the numerical flow and transport simulation models. To address this, Genetic Programming based surrogate models may be used to approximate the numerical simulation model in the linked simulation-optimization model for source characterization. Therefore, the aim of the present study is to demonstrate the feasibility and efficiency of a developed methodology to optimally identify or characterize the unknown distributed pollution sources with chemically reactive species in complex contaminated aquifers. Because the accuracy and reliability of the source characterization process depends on the quality and extent of the spatial and temporal concentration measurement data, a relevant issue is the design and implementation of a suitable and efficient monitoring network under conditions of various uncertainties. This is especially true, where the initial measurement data available are sparse and obtained from arbitrary monitoring locations. Therefore a new two objective Pareto optimal monitoring network design methodology is developed. This design methodology utilizes Fractal Singularity Mapping Technique to determine plume boundaries, information used to select potential monitoring locations for contaminant concentration monitoring. This approach substantially improves the source characterization efficiency as demonstrated for illustrative study areas. In order to improve the efficiency and accuracy of the source characterization methodology in real life sites where contamination is evident, but the monitoring data are very sparse and arbitrary, the monitoring network design model is integrated with the source characterization process by sequentially utilizing the source characterization model to estimate the sources. This information then, is utilized to design and implement a cost effective monitoring network. This sequential and iterative methodology is shown to improve the source characterization efficiency and accuracy, even when dealing with a hydrogeochemically complex aquifer system with multiple reactive species. The performance of the linked source characterization model is also evaluated by limited application to real life sites, which included a complex, abandoned mine site in Queensland, Australia. The sequential source characterization and monitoring network design methodology is applied to a contaminated aquifer in an urban area in Australia. Several techniques are utilized in the proposed methodology to increase the efficiency of the source characterization including trained and tested Genetic Programming based surrogate models, Adaptive Simulated Annealing optimization algorithm, Fractal Singularity Mapping Technique, and Statistical Kriging interpolation. The study includes the following steps: 1. The flow and reactive contaminant transport simulation model is utilized to simulate the aquifer processes; 2. Trained Genetic Programming (GP) based meta-models are developed using the simulated response of the aquifer to randomly generated source fluxes. The selected GP models replace the numerical simulation model in the linked simulation-optimization model for source characterization. 3. Two objectives Pareto-optimal design of a monitoring network for sequential characterization of pollutant sources uses a linked simulation-optimization model incorporating Adaptive Simulated Annealing as the optimization algorithm. 4. Integrated source identification and monitoring network design is carried out to obtain sufficient accuracy in characterization of source properties. 5. The performance of the developed methodologies is evaluated by limited application of the developed methodologies to real life sites