A Simulation-Optimization Model for Seawater Intrusion Management at Pingtung Coastal Area, Taiwan

The coastal regions of Pingtung Plain in southern Taiwan rely on groundwater as their main source of fresh water for aquaculture, agriculture, domestic, and industrial sectors. The availability of fresh groundwater is threatened by unsustainable groundwater extraction and the over-pumpage leads to the serious problem of seawater intrusion. It is desired to find appropriate management strategies to control groundwater salinity and mitigate seawater intrusion. In this study, a simulation-optimization model has been presented to solve the problem of seawater intrusion along the coastal aquifers in Pingtung Plain and the objective is using injection well barriers and minimizing the total injection rate based on the pre-determined locations of injection barriers. The SEAWAT code is used to simulate the process of seawater intrusion and the surrogate model of artificial neural networks (ANNs) is used to approximate the seawater intrusion (SWI) numerical model to increase the computational efficiency during the optimization process. The heuristic optimization scheme of differential evolution (DE) algorithm is selected to identify the global optimal management solution. Two different management scenarios, one is the injection barriers located along the coast and the other is the injection barrier located at the inland, are considered and the optimized results show that the deployment of injection barriers at the inland is more effective to reduce total dissolved solids (TDS) concentrations and mitigate seawater intrusion than that along the coast. The computational time can be reduced by more than 98% when using ANNs to replace the numerical model and the DE algorithm has been confirmed as a robust optimization scheme to solve groundwater management problems. The proposed framework can identify the most reliable management strategies and provide a reference tool for decision making with regard to seawater intrusion remediation.