InterPore2026

Physics informed neural network for modeling seawater intrusion in coastal aquifers

Not scheduled

1h 30m

Poster Presentation (MS15) Machine Learning in Porous Media Poster

Speaker

Marwan Fahs (ENGEES-LHYGES)

Description

Physics-informed neural networks (PINNs) is a new approach designed to reduce the dependence of neural network models on data. This technique shows strong potential for groundwater applications, where data are often scarce. PINNs can be used for forward modeling, surrogate modeling, uncertainty quantification, and inverse modeling. For this reason, the groundwater-related applications of PINNs are significantly growing. However, PINNs remain a recent technique, and their transition into operational tools for groundwater management requires substantial effort, particularly to address associated challenges and to adapt the approach to diverse groundwater problems. To the best of our knowledge, PINNs have not yet been applied to seawater intrusion (SWI) in coastal aquifers. This represents a challenging application due to the presence of coupled, nonlinear, multi-physical processes. This study aims to fill this gap by applying PINNs to the well-known SWI benchmark, the Henry problem. Using this benchmark provides insights into the applicability of PINNs for SWI and demonstrates how PINNs can enhance the reliability of neural network models for simulating SWI under limited data conditions.