Groundwater Forecasting using a Numerical Flow Model Coupled with Machine Learning Model for Synthetic Time Series

Abstract

In this work, the processes that characterize the phenomena of flow in porous media are studied. The necessary different aspects for the resolution of the groundwater equation by the finite difference method using MODFLOW have been discussed to establish a model of groundwater. The calibrated model has been the subject of exploitation, which has allowed the quantification of groundwater balance and the determination of inputs/outputs areas of the groundwater. The work tends to simulate the hydrological time-series through the artificial neural networks (ANN) to study the climate scenarios and their likely impact on the groundwater. Within the same context, the work has developed a new concept that operates the ANN model output as an input of the flow model, to define the flow model boundary conditions again.