Documents published in Scipedia

• Machine learning and deep learning for numerical simulation of contaminant transport in water systems: a narrative review

  G. Tinoco Guerrero, F. Domínguez Mota, J. Guzmán Torres, G. Pedraza Jiménez, E. Chagolla Inzunza, J. Gonzalez Figueroa

  SMCCA (2026). 8

  
  

  Abstract

  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.

  Abstract
  Numerical simulation of contaminant transport in water systems enables environmental risk assessment and the design of remediation strategies. These processes combine advection, [...]

  • 56
  •  read