Solution of Large-Scale Multi-objective Optimization Models for Saltwater Intrusion Control in Coastal Aquifers Utilizing ANFIS Based Linked Meta-Models for Computational Feasibility and Efficiency

Abstract

Saltwater intrusion in coastal aquifers poses significant challenges in the management of vulnerable coastal groundwater resources around the world. To develop a strategy for regional scale sustainable management of coastal aquifers, solution of large-scale multi-objective decision models is essential. The flow and solute transport equations are also density dependent, where the flow parameters are dependent on salt concentration; hence, the flow and solute transport equations need to be solved as coupled equations. In a linked optimization simulation model, the numerical simulation model as a predictor of the physical processes need to be solved enormous number of times to be able to identify an optimum solution as per the specified objectives and constraints. This problem becomes even more complicated when multiple objectives are included and the Pareto optimal solutions need to be determined. Therefore, to ensure the computational efficiency and feasibility of determining a regional scale strategy for control and sustainable use of a coastal aquifer, meta-models that are trained, tested and validated using randomized solutions of the numerical simulation models can be utilized. These meta-models once trained and tested serves the purpose of an approximate emulator of the complex numerical models rendering the solution of a complex and large scale linked optimization model computationally efficient and feasible. The optimal groundwater extraction patterns can be obtained through linked simulation-optimization (S/O) technique in which the simulation part is usually replaced by computationally efficient meta-models. This study proposes a computationally efficient meta-model to emulate density reliant integrated flow and solute transport scenarios of coastal aquifers. A meta-model, Adaptive Neuro Fuzzy Inference System (ANFIS) is trained and developed for an illustrative coastal aquifer study area. Prediction accuracy of the developed ANFIS based meta-model is evaluated for suitability. The meta-model is then integrated with a multiple objective coastal aquifer management model to demonstrate the potential application of this methodology. The optimization algorithm utilized for solution is the Controlled Elitist Multi-objective Genetic Algorithm. Performance evaluation results show acceptable accuracy in the obtained optimized management strategies. Therefore, use of trained and tested meta-models linked to an optimization model results in significant computational efficacy. It also ensures computational practicability of solving such large-scale integrated S/O approach for regional scale coastal groundwater management.