SQuaRe

Stress predictions – quantification and reduction of uncertainties with geomechanical-numerical subsurface models

The aim of the project is an improved estimation of uncertainties in the material parameters, boundary- and initial stress conditions as well as the influence of the mesh resolution in the context of geomechanical subsurface models. GFZ Potsdam and RWTH Aachen University are co-operation partners in the SQuaRe project.

Facts about the project

Project manager: Dr. Karsten Reiter +++ Duration: 1.1.2023 until 31.12.2025 +++ Project funding: Federal Ministry for Economic Affairs and Climate Action (BMUV) Projektträger Karlsruhe (PTKA)

Robust predictions of the stress state in the earth's crust are of central importance for the characterisation of potential sites for the final disposal of highly radioactive waste. In order to assess the confidence interval of geomechanical-numerical models, it is also necessary to specify the uncertainties in the predicted stress magnitudes and orientations. This requires a systematic quantification of the uncertainties of the parameters included in the modelling. The aim is a comprehensive quantitative consideration of the uncertainties of stress prediction, which can be used both for site characterisation and for site comparison.

A systematic quantification of the uncertainties of the parameters used in the modelling, such as underground geometry, material properties, boundary conditions and the stress data used for calibration, is to be carried out. In particular, the uncertainties of the material parameters used, the boundary- and initial stress conditions, as well as the influence of the mesh resolution will be analysed. For this purpose, generic models are created, by means of which the respective influence of the parameters is analysed separately.

Deposition scenarios for the options of repository products resulting from the photoelectrochemical approaches.
Deposition scenarios for the options of repository products resulting from the photoelectrochemical approaches.

Geomechanical-numerical model geometries are converted into a mesh during discretisation. This consists of many elements, which are the smallest homogeneous units of such a model. In simplified terms, it can be assumed that a better resolution, i.e., a better representation, can be achieved with more elements. However, a balance must be struck between resolution and the necessary computing time. In addition to the purely quantitative number, other factors such as element type, aspect ratio of elements, the spatial distribution of the discretisation, etc. have an influence on the results. Using generic models, these factors are examined separately in order to subsequently draw up recommendations for action to enable an optimal discretisation of a model.

  • Reiter, K., Ziegler, M., Heidbach, O., Desroches, J., Fjær, E., and Giger, S.: Calibration of geomechanical models – Examples from the site selection process for a nuclear waste repository in northern Switzerland, The Geological Society – Tectonic Stress: from the lithosphere to the wellbore, London, United Kingdom, 21–22 May 2024, 2024.
  • Ziegler, M. O., Seithel, R., Niederhuber, T., Heidbach, O., Kohl, T., Müller, B., Rajabi, M., Reiter, K., and Röckel, L.: The effect of stiffness contrasts at faults on stress orientation, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-1109, 2024.
  • Reiter, K., Degen, D., Ziegler, M., Heidbach, O., Wellmann, F., Henk, A., Projekt SQuaRe, Projektstatusgespräche 2023 zu BMUV-geförderten FuE-Projekten zur Entsorgung radioaktiver Abfälle, 2023

Status Quo & Outlook

The project began in spring 2023 with intensive collaboration between the working groups. After defining the geometry and the material properties to be used for a generic model to test the reduced basis method, initial tests were successfully carried out by the project partners. As classic tests of uncertainty analysis are impossible with complex models due to excessive computing times, a method was developed to estimate the uncertainty of model results on the basis of the distribution function of material properties using just a few model runs. A model geometry was created on the basis of site selection models in Switzerland, which is to be used for the complex reference model for the final test of the methods developed in the project in the last third of the project period. In addition, a 2-D model geometry was derived from it in order to compare different methods of discretising a computational mesh. Furthermore, this model will be used to investigate the influence of variable material properties, the mesh resolution, element type and element order.