Physics-Informed Machine Learning for Evaluating Contaminant Transport and Long-Term Barrier Performance in Geosynthetic Clay Liners

Authors

  • Sander Uit de Willigen Department of Computer Science and Engineering, University of Nevada, Reno, Reno, NV, USA. Author
  • Ewan O'Dwyer Department of Computer Science, University of New Hampshire, Durham, NH, USA. Author
  • Osmo Korpela Department of Electrical Engineering and Computer Science, University of Missouri, Columbia, MO, USA. Author

Keywords:

geosynthetic clay liner; contaminant transport; physics-informed machine learning; Bayesian uncertainty; hydraulic conductivity; barrier reliability

Abstract

Geosynthetic clay liners (GCLs) are low-permeability barrier components used in waste containment and environmental containment systems, but their long-term performance is difficult to evaluate because contaminant transport is coupled to stress-dependent hydraulic conductivity, geochemical alteration, sorption, and diffusive flux. Conventional advection-dispersion models require constitutive parameters that are sparse, scale-dependent, and often calibrated under conditions that do not represent long service life. Purely data-driven models can fit available observations but may violate mass balance, extrapolate unrealistically beyond tested permeants, and provide weak support for design decisions. This paper develops a physics-informed machine-learning framework for evaluating contaminant transport and long-term barrier performance in GCLs. The framework couples a neural surrogate for dissolved concentration with constrained constitutive closures for hydraulic conductivity and sorption, embeds the advection-dispersion-reaction residual in the training objective, and enforces nonnegativity and monotonicity constraints justified by sodium-bentonite behavior. Bayesian inference propagates uncertainty in transport and closure parameters, while a reliability formulation translates posterior concentration and hydraulic-conductivity predictions into time-dependent barrier-performance probabilities. Analytical assessment addresses identifiability, extrapolation, data requirements, baseline comparisons, and domain-specific error modes. The central insight is that physically constrained learning can preserve contaminant mass balance and known aging trends while retaining flexibility to assimilate column and field data. Interpretability and global sensitivity analysis identify which geochemical and stress variables dominate predicted service life, providing a defensible basis for scenario analysis, monitoring prioritization, and design decisions.

References

1. Rowe, R. K. (2005). Long-term performance of contaminant barrier systems. Géotechnique, 55(9), 631–678.

2. Shackelford, C. D., Benson, C. H., Katsumi, T., Edil, T. B., & Lin, L. (2000). Evaluating the hydraulic conductivity of GCLs permeated with non-standard liquids. Geotextiles and Geomembranes, 18(2–4), 133–161.

3. Bouazza, A. (2002). Geosynthetic clay liners. Geotextiles and Geomembranes, 20(1), 3–17.

4. Benson, C. H., Thorstad, P. A., Jo, H.-Y., & Shackelford, C. D. (2007). Hydraulic performance of geosynthetic clay liners in a landfill final cover. Journal of Geotechnical and Geoenvironmental Engineering, 133(7), 814–827.

5. Rowe, R. K., Quigley, R. M., Brachman, R. W. I., & Booker, J. R. (2004). Barrier systems for waste disposal facilities. Taylor & Francis.

6. Li, D., Jiang, Z., Tian, K., & Ji, R. (2025). Prediction of hydraulic conductivity of sodium bentonite GCLs by machine learning approaches. Environmental Geotechnics, 12(2), 154-173.

7. Raissi, M., Perdikaris, P., & Karniadakis, G. E. (2019). Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics, 378, 686–707.

8. Karniadakis, G. E., Kevrekidis, I. G., Lu, L., Perdikaris, P., Wang, S., & Yang, L. (2021). Physics-informed machine learning. Nature Reviews Physics, 3(6), 422–440.

9. Kennedy, M. C., & O'Hagan, A. (2001). Bayesian calibration of computer models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 63(3), 425–464.

10. Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32.

11. Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems 30 (pp. 4765–4774). Curran Associates.

12. Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M., & Tarantola, S. (2008). Global sensitivity analysis: The primer. Wiley.

13. Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A., & Rubin, D. B. (2013). Bayesian data analysis (3rd ed.). CRC Press.

14. Tartakovsky, A. M., Marrero, C. O., Perdikaris, P., Tartakovsky, G. D., & Barajas-Solano, D. (2020). Physics-informed deep neural networks for learning parameters and constitutive relationships in subsurface flow problems. Water Resources Research, 56(5), e2019WR026731.

15. Willard, J., Jia, X., Xu, S., Steinbach, M., & Kumar, V. (2022). Integrating scientific knowledge with machine learning for engineering and environmental systems. ACM Computing Surveys, 55(4), 1–37.

16. Lu, L., Meng, X., Mao, Z., & Karniadakis, G. E. (2021). DeepXDE: A deep learning library for solving differential equations. SIAM Review, 63(1), 208–228.

17. Jo, H. Y., Katsumi, T., Benson, C. H., & Edil, T. B. (2001). Hydraulic conductivity and swelling of nonprehydrated GCLs permeated with single-species salt solutions. Journal of Geotechnical and Geoenvironmental Engineering, 127(7), 557–567.

18. Katsumi, T., Ishimori, H., Ogawa, M., Maruyama, S., & Fukagawa, R. (2007). Hydraulic conductivity of nonprehydrated geosynthetic clay liners permeated with inorganic solutions and waste leachates. Soils and Foundations, 47(5), 917–929.

Downloads

Published

2026-11-06