Sustainable Management of Coastal Aquifers subject to Seawater Intrusion using Reduced-Order Groundwater Flow Models

The simulation of seawater intrusion (SWI) in coastal aquifers under complex hydrogeological conditions typically requires using "variable-density" models, which simulate the groundwater flow and the transport of salt dissolved in water. When combined with optimization algorithms, variable-density models constitute powerful tools to support the management of groundwater resources in coastal systems vulnerable to SWI, sea-level rise and unstainable groundwater abstraction. However, the application of simulation-optimization (SO) to SWI problems has so far been limited by the prohibitive computational effort required by full-scale variable-density models that simulate the aquifer response to proposed groundwater abstraction strategies. A viable solution is thus to develop “surrogate” models that emulate full-scale model responses at a fraction of their computational cost. In this study, a surrogate model of SEAWAT, a popular variable-density groundwater flow model, will be presented. This surrogate is based on the proper orthogonal decomposition (POD) method, which is a projection-based approach where the coefficient matrices and the right-hand side vectors derived through finite-difference discretization of the coupled flow and transport equations, are mapped onto a space of size significantly smaller than the model grid. Preliminary results show that the POD-based surrogate model is remarkably faster than the full-scale model, and provides results of comparable, and thus acceptable, accuracy. These features make the surrogate ideally suited for substituting the full-scale variable-density model within the SO framework adopted to support the management of coastal aquifers.