Groundwater Management Optimization and Saltwater Intrusion Mitigation under Uncertainty

Abstract

Groundwater is valuable to supply fresh water to the public, industries, agriculture, etc. However, excessive pumping has caused groundwater storage degradation, water quality deterioration and saltwater intrusion problems. Reliable groundwater flow and solute transport modeling is needed for sustainable groundwater management and aquifer remediation design. However, challenges exist because of highly complex subsurface environments, computationally intensive groundwater models as well as inevitable uncertainties. The first research goal is to explore conjunctive use of feasible hydraulic control approaches for groundwater management and aquifer remediation. Water budget analysis is conducted to understand how groundwater withdrawals affect water levels. A mixed integer multi-objective optimization model is constructed to derive optimal freshwater pumping strategies and investigate how to promote the optimality through regulating pumping locations. A solute transport model for the Baton Rouge multi-aquifer system is developed to assess saltwater encroachment under current condition. Potential saltwater scavenging approach is proposed to mitigate the salinization issue in the Baton Rouge area. The second research goal aims to develop robust surrogate-assisted simulation-optimization modeling methods for saltwater intrusion mitigation. Machine learning based surrogate models (response surface regression model, artificial neural network and support vector machine) were developed to replace a complex high-fidelity solute transport model for predicting saltwater intrusion. Two different methods including Bayesian model averaging and Bayesian set pair analysis are used to construct ensemble surrogates and quantify model prediction uncertainties. Besides. different optimization models that incorporate multiple ensemble surrogates are formulated to obtain optimal saltwater scavenging strategies. Chance-constrained programming is used to account for model selection uncertainty in probabilistic nonlinear concentration constraints. The results show that conjunctive use of hydraulic control approaches would be effective to mitigate saltwater intrusion but needs decades. Machine learning based ensemble surrogates can build accurate models with high computing efficiency, and hence save great efforts in groundwater remediation design. Including model selection uncertainty through multimodel inference and model averaging provides more reliable remediation strategies compared with the single-surrogate assisted approach.