EGU General Assembly 2020
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Sensitivity analyses and uncertainty quantification in THM models: a benchmark study

Jörg Buchwald1,2, Aqeel Chaudhry2, Keita Yoshioka1, Olaf Kolditz1,3, and Thomas Nagel2
Jörg Buchwald et al.
  • 1Department of Environmental Informatics, Helmholtz Centre for Environmental Research UFZ, Leipzig, Germany
  • 2Geotechnical Institute, Technische Universität Bergakademie Freiberg, Freiberg, Germany
  • 3Environmenal Systems Analysis, Technische Universität Dresden, Dresden, Germany

Coupled thermo-hydro-mechanical (THM) models are used for the assessment of nuclear waste disposal, reservoir engineering, and other branches of geo-environmental engineering. Model-based decision-making and design optimization in these domains require sensitivity analyses (SA) and uncertainty quantification (UQ) methods that are suitable for coupled THM problems on an engineering scale. Due to different coupling levels, non-linearities, and large spatial and temporal extents, these analyses can often be challenging both conceptually and computationally.

For an initial evaluation in a setting relevant to nuclear waste disposal we start by employing an analytical solution for thermal consolidation around a point heat source which encompasses the most relevant primary couplings and allows us to cover the entire parameter space robustly and efficiently. For uncertainty quantification, we applied an experimental design (DoE-) based history-matching approach. This approach uses DoE methods to construct a proxy model, which is used later for efficient Monte Carlo sampling and subsequent filtering of the uncertainty space of the history-match error. As a result, we obtain a family of curves that is compatible with the prior parameter set and experimental data to match, which then enables further uncertainty quantification. In our work, we demonstrate the applicability of the workflow and discuss its particular suitability to this problem class, including its (in-)sensitivity to prior parameter distribution assumptions.

For SA, we contrast the conclusions drawn via two different approaches: local one variable at a time (OVAT) and global sensitivity analysis (GSA) based on Sobol indices for different spatio-temporal settings to observe near and far-field effects as well as early- and late-stage system response. The conducted studies can serve as a benchmark for UQ and SA software designed around numerical THM simulators.

How to cite: Buchwald, J., Chaudhry, A., Yoshioka, K., Kolditz, O., and Nagel, T.: Sensitivity analyses and uncertainty quantification in THM models: a benchmark study, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-22535,, 2020.


Display file