Stochastic modeling of saltwater intrusion in highly heterogeneous coastal aquifers.

Abstract

This research presents a novel method for a reliable quantification of uncertainties with the prediction of saltwater intrusion (SWI) in poorly field-characterized heterogeneous coastal aquifers. The method uses crude prior information about hydrogeological input parameters, and simultaneously incorporates parameter uncertainty and imprecision in prior information to infer the required statistics for a Monte Carlo (MC) simulation into the uncertainty analysis. Compared with other methods, it is a computationally effective method for uncertainty analysis in highly heterogeneous coastal aquifers. In this method, the Sequential Gaussian Simulation (SGS) is used to create random parameter fields while fuzzy set theory accounts for the imprecision pertaining to prior information. Prediction uncertainties are evaluated by generating a large number of calibration-constrained models using the Null Space Monte Carlo (NSMC) method. This research presents another novel method that is used to design additional hydrogeological field-investigations towards reducing the prediction uncertainties in a SWI system. The method provides flexibility concerning model dimensionality, allows for any desired task-oriented formulation, targets any measurement type, accounts for various sources of uncertainty while also ensuring that it is cost-effective. It expands on the existing linear data-worth analysis method through the incorporation of Bayesian model averaging (BMA) and genetic algorithms (GA) when conducting a three-dimensional location search for sampling new data in non-linear systems. Both methods can use any model independent tools for parameter estimation and any variable-density simulation codes. The efficiencies of these methods in quantifying and reducing prediction uncertainties were demonstrated using the SEAWAT model and data from an actual heterogeneous coastal aquifer with limited hydrogeological field-data. Located along the Eastern Mediterranean (Beirut), the pilot aquifer consists of karstified limestone of Cre