MULTIPLE OBJECTIVE MANAGEMENT STRATEGIES FOR COASTAL AQUIFERS UTILIZING NEW SURROGATE MODELS

Abstract

Coastal aquifers are hydraulically connected to the sea and therefore susceptible to saltwater
intrusion problems. This study proposes the utilization of a new surrogate model within coupled simulationoptimization (S/O) model for the management of coastal aquifers subjected to density-dependent saltwater
intrusion processes. The simulation of the transient 3-dimensional density-dependent flow and transport model
is based on the solution of an implemented numerical simulation model. Direct coupling of the numerical
simulation model into the multi-objective genetic algorithm (MOGA) is computationally expensive. Hence,
the solution of the numerical simulation model with random input variables are used to train and test the support
vector machine regression (SVMR) surrogate models for approximately simulating the flow and transport
processes. The performances of the new surrogate models are evaluated using various performance evaluation
criteria. The resulting validated SVMR surrogate models are coupled to the MOGA and implemented for an
illustrative coastal aquifer with an aim to develop efficient coastal aquifer management strategies. Based on
the objective functions, execution of S/O model presented a set of optimal groundwater withdrawal rates from
the simulated aquifer. It also ensured salinity levels at the designated monitoring wells are constrained within
specified limits. The efficiency of the new SVMR surrogate models is also demonstrated. Evaluation results
suggested that the projected S/O model is an effective way of developing feasible and reliable coastal aquifer
management strategies. It also enhances the possibility of solving more realistic large-scale problems and
developing regional-scale coastal aquifer management methodologies.