Trained meta-models and evolutionary algorithm based multi-objective management of coastal aquifers under parameter uncertainty

Abstract

Abstract Meta-model based coupled simulation-optimization methodology is an effective tool in developing sustainable saltwater intrusion management strategies for coastal aquifers. Such management strategies largely depend on the accuracy, reliability, and computational feasibility of meta-models and the numerical simulation model. However, groundwater models are associated with a certain amount of uncertainties, e.g. parameter uncertainty and uncertainty in prediction. This study addresses uncertainties related to input parameters of the groundwater flow and transport system by using a set of randomized input parameters. Three meta-models are compared to characterize responses of water quality in coastal aquifers due to groundwater extraction patterns under parameter uncertainty. The ensemble of the best meta-model is then coupled with a multi-objective optimization algorithm to develop a saltwater intrusion management model. Uncertainties in hydraulic conductivity, compressibility, bulk density, and aquifer recharge are incorporated in the proposed approach. These uncertainties in the physical system are captured by the meta-models whereas the prediction uncertainties of meta-models are further addressed by the ensemble approach. An illustrative multi-layered coastal aquifer system is used to demonstrate the feasibility of the proposed approach. Evaluation results indicate the capability of the proposed approach to develop accurate and reliable management strategies for groundwater extraction to control saltwater intrusion.