Optimal Groundwater-Use Strategy for Saltwater Intrusion Management in a Pacific Island Country

Abstract

Escalating salinity levels in the Bonriki aquifer due to unplanned groundwater extraction is a major concern for the people of Kiribati. A multiobjective management model capable of providing sustainable optimal groundwater pumping strategies and simultaneously confining salinity concentrations in the aquifer within specified limits is needed for the Bonriki aquifer system. This study applies a regional-scale linked simulation-optimization (S/O) methodology with a Pareto front clustering technique to prescribe optimal groundwater withdrawal patterns from the Bonriki aquifer. A numerical simulation model is calibrated and validated using available field data. For computational feasibility, support vector regression (SVR) surrogate models are trained and tested utilizing input-output data sets generated using the numerical flow and transport simulation model. The developed surrogate models were externally coupled with the multiobjective genetic algorithm optimization (MOGA) model as a substitute for the numerical model. The study area consisted of freshwater pumping wells for extracting groundwater for beneficial use. Pumping from barrier wells installed along the coastlines were also considered as a management option to hydraulically control saltwater intrusion. The executed multiobjective linked S/O model generated 700 Pareto-optimal solutions. Analyzing a large set of Pareto-optimal solutions is a challenging task for the decision maker. Hence, the k-means clustering technique is utilized to reduce the larger original Pareto-optimal solution set for solving the large-scale saltwater intrusion management problem in Bonriki.