G. Tinoco Guerrero, F. Domínguez Mota, J. Guzmán Torres, G. Pedraza Jiménez, E. Chagolla Inzunza, J. Gonzalez Figueroa
Numerical simulation of contaminant transport in water systems enables environmental risk assessment and the design of remediation strategies. These processes combine advection, hydrodynamic dispersion, sorption, and chemical reactions, governed by mass balances and coupled flow equations. Although traditional numerical simulators are standard practice in hydrogeology, their computational cost limits applications that require repeated evaluations, such as parameter calibration, uncertainty quantification, and source identification. This review examines the integration of machine learning (ML) and deep learning (DL) in this domain, focusing on two main methodological families: data-driven surrogate models or emulators, and physics-informed neural networks (PINNs). We discuss their mathematical formulations, underlying assumptions, acceleration capabilities, and extrapolation limits. Evidence shows that statistical accuracy does not guarantee physical validity; mass conservation, numerical stability across advection-dominated regimes, and training overhead determine practical applicability. Ultimately, ML and DL complement classical numerical methods rather than replace them.
Keywords:
You do not have permission to edit this page, for the following reason:
You are not allowed to execute the action you have requested.
You can view and copy the source of this page.
Return to Tinoco Guerrero et al 2026a.
Published on 20/10/26Submitted on 14/07/26
Licence: CC BY-NC-SA license
Views 70Recommendations 0