Articles | Volume 16, issue 6
https://doi.org/10.5194/se-16-477-2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/se-16-477-2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Research article
| 
23 Jun 2025
Research article |
| 23 Jun 2025
| 23 Jun 2025
About the trustworthiness of physics-based machine learning – considerations for geomechanical applications
Denise Degen
CORRESPONDING AUTHOR
Computational Geoscience, Geothermics and Reservoir Geophysics (CG3), RWTH Aachen University, 52074 Aachen, Germany
GFZ Helmholtz Centre for Geosciences, 14473 Potsdam, Germany
Institute of Applied Geosciences, TU Darmstadt, 64287 Darmstadt, Germany
Moritz Ziegler
Professorship of Geothermal Technologies, TU Munich, Arcisstr. 21, 80333 Munich, Germany
GFZ Helmholtz Centre for Geosciences, 14473 Potsdam, Germany
Oliver Heidbach
GFZ Helmholtz Centre for Geosciences, 14473 Potsdam, Germany
Institute of Applied Geosciences, TU Berlin, 10587 Berlin, Germany
Andreas Henk
Institute of Applied Geosciences, TU Darmstadt, 64287 Darmstadt, Germany
Karsten Reiter
Institute of Applied Geosciences, TU Darmstadt, 64287 Darmstadt, Germany
Florian Wellmann
Computational Geoscience, Geothermics and Reservoir Geophysics (CG3), RWTH Aachen University, 52074 Aachen, Germany
Fraunhofer Research Institution for Energy Infrastructures and Geotechnologies IEG, 44801 Bochum, Germany
Related authors
Denise Degen, Ajay Kumar, Magdalena Scheck-Wenderoth, and Mauro Cacace
EGUsphere, https://doi.org/10.5194/egusphere-2025-1925, https://doi.org/10.5194/egusphere-2025-1925, 2025
Short summary
Short summary
Geodynamical simulations cover a wide spatial and temporal range and are crucial to understand and assess the evolution of the Earth system. To enable computationally efficient modeling approaches that can account for potentially unknown subsurface properties, we present a surrogate modeling technique. This technique combines physics-based and machine-learning techniques to enable reliable predictions of geodynamical applications, as we illustrate for the case study of the Alpine Region.
Denise Degen, Daniel Caviedes Voullième, Susanne Buiter, Harrie-Jan Hendricks Franssen, Harry Vereecken, Ana González-Nicolás, and Florian Wellmann
Geosci. Model Dev., 16, 7375–7409, https://doi.org/10.5194/gmd-16-7375-2023, https://doi.org/10.5194/gmd-16-7375-2023, 2023
Short summary
Short summary
In geosciences, we often use simulations based on physical laws. These simulations can be computationally expensive, which is a problem if simulations must be performed many times (e.g., to add error bounds). We show how a novel machine learning method helps to reduce simulation time. In comparison to other approaches, which typically only look at the output of a simulation, the method considers physical laws in the simulation itself. The method provides reliable results faster than standard.
Denise Degen, Cameron Spooner, Magdalena Scheck-Wenderoth, and Mauro Cacace
Geosci. Model Dev., 14, 7133–7153, https://doi.org/10.5194/gmd-14-7133-2021, https://doi.org/10.5194/gmd-14-7133-2021, 2021
Short summary
Short summary
In times of worldwide energy transitions, an understanding of the subsurface is increasingly important to provide renewable energy sources such as geothermal energy. To validate our understanding of the subsurface we require data. However, the data are usually not distributed equally and introduce a potential misinterpretation of the subsurface. Therefore, in this study we investigate the influence of measurements on temperature distribution in the European Alps.
Denise Degen and Mauro Cacace
Geosci. Model Dev., 14, 1699–1719, https://doi.org/10.5194/gmd-14-1699-2021, https://doi.org/10.5194/gmd-14-1699-2021, 2021
Short summary
Short summary
In this work, we focus on improving the understanding of subsurface processes with respect to interactions with climate dynamics. We present advanced, open-source mathematical methods that enable us to investigate the influence of various model properties on the final outcomes. By relying on our approach, we have been able to showcase their importance in improving our understanding of the subsurface and highlighting the current shortcomings of currently adopted models.
Friedrich Carl, Peter Achtziger-Zupančič, Jian Yang, Marlise Colling Cassel, and Florian Wellmann
EGUsphere, https://doi.org/10.5194/egusphere-2025-3203, https://doi.org/10.5194/egusphere-2025-3203, 2025
This preprint is open for discussion and under review for Solid Earth (SE).
Short summary
Short summary
A method for shape quantification based on geometrical parameters is proposed alongside a set of regular geometries established as geomodeling benchmarks. Dimensions, gradient and curvature data is obtained on cross-sections. Data analyses provide insight into the main geometrical characteristics of the benchmark models and visualizes geometrical dis-/similarities between bodies. The method and benchmarks are usable in geomodeling workflows and structural comparisons based on sparse data.
Steffen Ahlers and Andreas Henk
Earth Syst. Sci. Data Discuss., https://doi.org/10.5194/essd-2025-320, https://doi.org/10.5194/essd-2025-320, 2025
Preprint under review for ESSD
Short summary
Short summary
The paper presents a 3D geological underground model of Germany and some neigbouring countries combining 27 individual models. It has been set up for the prediction of the recent crustal stress state of Germany by geomechanical-numerical modelling. It contains 146 units and is provided as a point data set with a resolution of 1x1 km2. It enables to create 3D finite element models within very short times A comprehensive supplement and 157 figures documents the results and assumptions made.
Denise Degen, Ajay Kumar, Magdalena Scheck-Wenderoth, and Mauro Cacace
EGUsphere, https://doi.org/10.5194/egusphere-2025-1925, https://doi.org/10.5194/egusphere-2025-1925, 2025
Short summary
Short summary
Geodynamical simulations cover a wide spatial and temporal range and are crucial to understand and assess the evolution of the Earth system. To enable computationally efficient modeling approaches that can account for potentially unknown subsurface properties, we present a surrogate modeling technique. This technique combines physics-based and machine-learning techniques to enable reliable predictions of geodynamical applications, as we illustrate for the case study of the Alpine Region.
Sarah Diekmeier, Karsten Reiter, Andreas Henk, and Colin Friebe
Earth Syst. Sci. Data Discuss., https://doi.org/10.5194/essd-2024-489, https://doi.org/10.5194/essd-2024-489, 2025
Manuscript not accepted for further review
Short summary
Short summary
The study explores the potential for storing carbon-rich products in Germany to support climate goals. Using geological data, we identified old mining sites suitable for storing products like graphite and oxalate from negative emissions technologies. Results show significant storage potential, both above and below ground, offering a sustainable solution. By reusing existing mining areas, Germany can advance towards carbon neutrality, reducing costs and environmental impact.
Antonio J. Olaiz, José A. Álvarez Gómez, Gerardo de Vicente, Alfonso Muñoz-Martín, Juan V. Cantavella, Susana Custódio, Dina Vales, and Oliver Heidbach
EGUsphere, https://doi.org/10.5194/egusphere-2024-4126, https://doi.org/10.5194/egusphere-2024-4126, 2025
Short summary
Short summary
Understanding the stress and strain conditions in the Earth crust is crucial for various activities, such as oil and gas exploration and assessing seismic hazards. In this article, we have updated the database of moment tensor focal mechanisms for the Greater Iberia. We conducted kinematic and dynamic analyses on the selected populations, determining the average focal mechanism, strain and stress orientations, and tectonic regime. The orientation for horizontal compression is primarily N154° E.
Moritz O. Ziegler, Robin Seithel, Thomas Niederhuber, Oliver Heidbach, Thomas Kohl, Birgit Müller, Mojtaba Rajabi, Karsten Reiter, and Luisa Röckel
Solid Earth, 15, 1047–1063, https://doi.org/10.5194/se-15-1047-2024, https://doi.org/10.5194/se-15-1047-2024, 2024
Short summary
Short summary
The rotation of the principal stress axes in a fault structure because of a rock stiffness contrast has been investigated for the impact of the ratio of principal stresses, the angle between principal stress axes and fault strike, and the ratio of the rock stiffness contrast. A generic 2D geomechanical model is employed for the systematic investigation of the parameter space.
Karsten Reiter, Oliver Heidbach, and Moritz O. Ziegler
Solid Earth, 15, 305–327, https://doi.org/10.5194/se-15-305-2024, https://doi.org/10.5194/se-15-305-2024, 2024
Short summary
Short summary
It is generally assumed that faults have an influence on the stress state of the Earth’s crust. It is questionable whether this influence is still present far away from a fault. Simple numerical models were used to investigate the extent of the influence of faults on the stress state. Several models with different fault representations were investigated. The stress fluctuations further away from the fault (> 1 km) are very small.
Denise Degen, Daniel Caviedes Voullième, Susanne Buiter, Harrie-Jan Hendricks Franssen, Harry Vereecken, Ana González-Nicolás, and Florian Wellmann
Geosci. Model Dev., 16, 7375–7409, https://doi.org/10.5194/gmd-16-7375-2023, https://doi.org/10.5194/gmd-16-7375-2023, 2023
Short summary
Short summary
In geosciences, we often use simulations based on physical laws. These simulations can be computationally expensive, which is a problem if simulations must be performed many times (e.g., to add error bounds). We show how a novel machine learning method helps to reduce simulation time. In comparison to other approaches, which typically only look at the output of a simulation, the method considers physical laws in the simulation itself. The method provides reliable results faster than standard.
Michael Hillier, Florian Wellmann, Eric A. de Kemp, Boyan Brodaric, Ernst Schetselaar, and Karine Bédard
Geosci. Model Dev., 16, 6987–7012, https://doi.org/10.5194/gmd-16-6987-2023, https://doi.org/10.5194/gmd-16-6987-2023, 2023
Short summary
Short summary
Neural networks can be used effectively to model three-dimensional geological structures from point data, sampling geological interfaces, units, and structural orientations. Existing neural network approaches for this type of modelling are advanced by the efficient incorporation of unconformities, new knowledge inputs, and improved data fitting techniques. These advances permit the modelling of more complex geology in diverse geological settings, different-sized areas, and various data regimes.
Oliver Heidbach, Karsten Reiter, Moritz O. Ziegler, and Birgit Müller
Saf. Nucl. Waste Disposal, 2, 185–185, https://doi.org/10.5194/sand-2-185-2023, https://doi.org/10.5194/sand-2-185-2023, 2023
Short summary
Short summary
When stresses yield a critical value, rock breaks and generate pathways for fluid migration. Thus, the contemporary undisturbed stress state is a key parameter for assessing the stability of deep geological repositories. In this workshop you can ask everything you always wanted to know about stress (but were afraid to ask), and this is divided into three parts. 1) How do we formally describe the stress field? 2) How do we to actually measure stress? 3) How do we go from points to 3D description?
Moritz O. Ziegler, Oliver Heidbach, and Mojtaba Rajabi
Saf. Nucl. Waste Disposal, 2, 79–80, https://doi.org/10.5194/sand-2-79-2023, https://doi.org/10.5194/sand-2-79-2023, 2023
Short summary
Short summary
The subsurface is subject to constant stress. With increasing depth, more rock overlies an area, thereby increasing the stress. There is also constant stress from the sides. Knowledge of this stress is fundamental to build lasting and safe underground structures. Very few data on the stress state are available; thus, computer models are used to predict this parameter. We present a method to improve the quality of the computer models, even if no direct data on the stress state are available.
Karsten Reiter, Oliver Heidbach, Moritz Ziegler, Silvio Giger, Rodney Garrard, and Jean Desroches
Saf. Nucl. Waste Disposal, 2, 71–72, https://doi.org/10.5194/sand-2-71-2023, https://doi.org/10.5194/sand-2-71-2023, 2023
Short summary
Short summary
Numerical methods can be used to estimate the stress state in the Earth’s upper crust. Measured stress data are needed for model calibration. High-quality stress data are available for the calibration of models for possible radioactive waste repositories in Switzerland. A best-fit model predicts the stress state for each point within the model volume. In this study, variable rock properties are used to predict the potential stress variations due to inhomogeneous rock properties.
Luisa Röckel, Steffen Ahlers, Sophia Morawietz, Birgit Müller, Tobias Hergert, Karsten Reiter, Andreas Henk, Moritz Ziegler, Oliver Heidbach, and Frank Schilling
Saf. Nucl. Waste Disposal, 2, 73–73, https://doi.org/10.5194/sand-2-73-2023, https://doi.org/10.5194/sand-2-73-2023, 2023
Short summary
Short summary
Stress data predicted by a geomechanical–numerical model are mapped onto 3D fault geometries. Then the slip tendency of these faults is calculated as a measure of their reactivation potential. Characteristics of the faults and the state of stress are identified that lead to a high fault reactivation potential. An overall high reactivation potential is observed in the Upper Rhine Graben area, whereas the reactivation potential is quite low in the Molasse Basin.
Tobias Hergert, Steffen Ahlers, Luisa Röckel, Sophia Morawietz, Karsten Reiter, Moritz Ziegler, Birgit Müller, Oliver Heidbach, Frank Schilling, and Andreas Henk
Saf. Nucl. Waste Disposal, 2, 65–65, https://doi.org/10.5194/sand-2-65-2023, https://doi.org/10.5194/sand-2-65-2023, 2023
Short summary
Short summary
In numerical geomechanical models, an initial stress state is established before displacement boundary conditions are applied in order to match calibration data. We present generic models to show that the choice of initial stress and boundary conditions affects the final state of stress in areas of the model domain where no stress data for calibration are available. These deviations are largest in the vicinity of lithological interfaces, and they can be reduced if more stress data exist.
Steffen Ahlers, Karsten Reiter, Tobias Hergert, Andreas Henk, Luisa Röckel, Sophia Morawietz, Oliver Heidbach, Moritz Ziegler, and Birgit Müller
Saf. Nucl. Waste Disposal, 2, 59–59, https://doi.org/10.5194/sand-2-59-2023, https://doi.org/10.5194/sand-2-59-2023, 2023
Short summary
Short summary
The recent crustal stress state is a crucial parameter in the search for a high-level nuclear waste repository. We present results of a 3D geomechanical numerical model that improves the state of knowledge by providing a continuum-mechanics-based prediction of the recent crustal stress field in Germany. The model results can be used, for example, for the calculation of fracture potential, for slip tendency analyses or as boundary conditions for smaller local models.
Michal Kruszewski, Alessandro Verdecchia, Oliver Heidbach, Rebecca M. Harrington, and David Healy
EGUsphere, https://doi.org/10.5194/egusphere-2023-1889, https://doi.org/10.5194/egusphere-2023-1889, 2023
Preprint archived
Short summary
Short summary
In this study, we investigate the evolution of fault reactivation potential in the greater Ruhr region (Germany) in respect to a future utilization of deep geothermal resources. We use analytical and numerical approaches to understand the initial stress conditions on faults as well as their evolution in space and time during geothermal fluid production. Using results from our analyses, we can localize areas more favorable for geothermal energy use based on fault reactivation potential.
Mohammad Moulaeifard, Simon Bernard, and Florian Wellmann
Geosci. Model Dev., 16, 3565–3579, https://doi.org/10.5194/gmd-16-3565-2023, https://doi.org/10.5194/gmd-16-3565-2023, 2023
Short summary
Short summary
In this work, we propose a flexible framework to generate and interact with geological models using explicit surface representations. The essence of the work lies in the determination of the flexible control mesh, topologically similar to the main geological structure, watertight and controllable with few control points, to manage the geological structures. We exploited the subdivision surface method in our work, which is commonly used in the animation and gaming industry.
Michal Kruszewski, Gerd Klee, Thomas Niederhuber, and Oliver Heidbach
Earth Syst. Sci. Data, 14, 5367–5385, https://doi.org/10.5194/essd-14-5367-2022, https://doi.org/10.5194/essd-14-5367-2022, 2022
Short summary
Short summary
The authors assemble an in situ stress magnitude and orientation database based on 429 hydrofracturing tests that were carried out in six coal mines and two coal bed methane boreholes between 1986 and 1995 within the greater Ruhr region (Germany). Our study summarises the results of the extensive in situ stress test campaign and assigns quality to each data record using the established quality ranking schemes of the World Stress Map project.
Michał P. Michalak, Lesław Teper, Florian Wellmann, Jerzy Żaba, Krzysztof Gaidzik, Marcin Kostur, Yuriy P. Maystrenko, and Paulina Leonowicz
Solid Earth, 13, 1697–1720, https://doi.org/10.5194/se-13-1697-2022, https://doi.org/10.5194/se-13-1697-2022, 2022
Short summary
Short summary
When characterizing geological/geophysical surfaces, various geometric attributes are calculated, such as dip angle (1D) or dip direction (2D). However, the boundaries between specific values may be subjective and without optimization significance, resulting from using default color palletes. This study proposes minimizing cosine distance among within-cluster observations to detect 3D anomalies. Our results suggest that the method holds promise for identification of megacylinders or megacones.
Luisa Röckel, Steffen Ahlers, Birgit Müller, Karsten Reiter, Oliver Heidbach, Andreas Henk, Tobias Hergert, and Frank Schilling
Solid Earth, 13, 1087–1105, https://doi.org/10.5194/se-13-1087-2022, https://doi.org/10.5194/se-13-1087-2022, 2022
Short summary
Short summary
Reactivation of tectonic faults can lead to earthquakes and jeopardize underground operations. The reactivation potential is linked to fault properties and the tectonic stress field. We create 3D geometries for major faults in Germany and use stress data from a 3D geomechanical–numerical model to calculate their reactivation potential and compare it to seismic events. The reactivation potential in general is highest for NNE–SSW- and NW–SE-striking faults and strongly depends on the fault dip.
Denise Degen, Cameron Spooner, Magdalena Scheck-Wenderoth, and Mauro Cacace
Geosci. Model Dev., 14, 7133–7153, https://doi.org/10.5194/gmd-14-7133-2021, https://doi.org/10.5194/gmd-14-7133-2021, 2021
Short summary
Short summary
In times of worldwide energy transitions, an understanding of the subsurface is increasingly important to provide renewable energy sources such as geothermal energy. To validate our understanding of the subsurface we require data. However, the data are usually not distributed equally and introduce a potential misinterpretation of the subsurface. Therefore, in this study we investigate the influence of measurements on temperature distribution in the European Alps.
Moritz Ziegler and Oliver Heidbach
Saf. Nucl. Waste Disposal, 1, 187–188, https://doi.org/10.5194/sand-1-187-2021, https://doi.org/10.5194/sand-1-187-2021, 2021
Short summary
Short summary
The Earth's crust is subject to constant stress which is manifested by earthquakes at plate boundaries. This stress is not only at plate boundaries but everywhere in the crust. A profound knowledge of the magnitude and orientation of the stress is important to select and build a safe deep geological repository for nuclear waste. We demonstrate how to build computer models of the stress state and show how to deal with the associated uncertainties.
Luisa Röckel, Steffen Ahlers, Sophia Morawietz, Birgit Müller, Karsten Reiter, Oliver Heidbach, Andreas Henk, Tobias Hergert, and Frank Schilling
Saf. Nucl. Waste Disposal, 1, 77–78, https://doi.org/10.5194/sand-1-77-2021, https://doi.org/10.5194/sand-1-77-2021, 2021
Karsten Reiter, Steffen Ahlers, Sophia Morawietz, Luisa Röckel, Tobias Hergert, Andreas Henk, Birgit Müller, and Oliver Heidbach
Saf. Nucl. Waste Disposal, 1, 75–76, https://doi.org/10.5194/sand-1-75-2021, https://doi.org/10.5194/sand-1-75-2021, 2021
Steffen Ahlers, Andreas Henk, Tobias Hergert, Karsten Reiter, Birgit Müller, Luisa Röckel, Oliver Heidbach, Sophia Morawietz, Magdalena Scheck-Wenderoth, and Denis Anikiev
Saf. Nucl. Waste Disposal, 1, 163–164, https://doi.org/10.5194/sand-1-163-2021, https://doi.org/10.5194/sand-1-163-2021, 2021
Sophia Morawietz, Moritz Ziegler, Karsten Reiter, and the SpannEnD Project Team
Saf. Nucl. Waste Disposal, 1, 71–72, https://doi.org/10.5194/sand-1-71-2021, https://doi.org/10.5194/sand-1-71-2021, 2021
Short summary
Short summary
Knowledge of the crustal stress state is important for the assessment of subsurface stability. In particular, stress magnitudes are essential for the calibration of geomechanical models that estimate a continuous description of the 3-D stress field from pointwise and incomplete stress data. We present the first comprehensive and open-access stress magnitude database for Germany, consisting of 568 data records. We introduce a quality ranking scheme for stress magnitude data for the first time.
Steffen Ahlers, Andreas Henk, Tobias Hergert, Karsten Reiter, Birgit Müller, Luisa Röckel, Oliver Heidbach, Sophia Morawietz, Magdalena Scheck-Wenderoth, and Denis Anikiev
Solid Earth, 12, 1777–1799, https://doi.org/10.5194/se-12-1777-2021, https://doi.org/10.5194/se-12-1777-2021, 2021
Short summary
Short summary
Knowledge about the stress state in the upper crust is of great importance for many economic and scientific questions. However, our knowledge in Germany is limited since available datasets only provide pointwise, incomplete and heterogeneous information. We present the first 3D geomechanical model that provides a continuous description of the contemporary crustal stress state for Germany. The model is calibrated by the orientation of the maximum horizontal stress and stress magnitudes.
Alexander Schaaf, Miguel de la Varga, Florian Wellmann, and Clare E. Bond
Geosci. Model Dev., 14, 3899–3913, https://doi.org/10.5194/gmd-14-3899-2021, https://doi.org/10.5194/gmd-14-3899-2021, 2021
Short summary
Short summary
Uncertainty is an inherent property of any model of the subsurface. We show how geological topology information – how different regions of rocks in the subsurface are connected – can be used to train uncertain geological models to reduce uncertainty. More widely, the method demonstrates the use of probabilistic machine learning (Bayesian inference) to train structural geological models on auxiliary geological knowledge that can be encoded in graph structures.
Karsten Reiter
Solid Earth, 12, 1287–1307, https://doi.org/10.5194/se-12-1287-2021, https://doi.org/10.5194/se-12-1287-2021, 2021
Short summary
Short summary
The influence and interaction of elastic material properties (Young's modulus, Poisson's ratio), density and low-friction faults on the resulting far-field stress pattern in the Earth's crust is tested with generic models. A Young's modulus contrast can lead to a significant stress rotation. Discontinuities with low friction in homogeneous models change the stress pattern only slightly, away from the fault. In addition, active discontinuities are able to compensate stress rotation.
Denise Degen and Mauro Cacace
Geosci. Model Dev., 14, 1699–1719, https://doi.org/10.5194/gmd-14-1699-2021, https://doi.org/10.5194/gmd-14-1699-2021, 2021
Short summary
Short summary
In this work, we focus on improving the understanding of subsurface processes with respect to interactions with climate dynamics. We present advanced, open-source mathematical methods that enable us to investigate the influence of various model properties on the final outcomes. By relying on our approach, we have been able to showcase their importance in improving our understanding of the subsurface and highlighting the current shortcomings of currently adopted models.
Stephanie Thiesen, Diego M. Vieira, Mirko Mälicke, Ralf Loritz, J. Florian Wellmann, and Uwe Ehret
Hydrol. Earth Syst. Sci., 24, 4523–4540, https://doi.org/10.5194/hess-24-4523-2020, https://doi.org/10.5194/hess-24-4523-2020, 2020
Short summary
Short summary
A spatial interpolator has been proposed for exploring the information content of the data in the light of geostatistics and information theory. It showed comparable results to traditional interpolators, with the advantage of presenting generalization properties. We discussed three different ways of combining distributions and their implications for the probabilistic results. By its construction, the method provides a suitable and flexible framework for uncertainty analysis and decision-making.
Cited articles
Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z., Citro, C., Corrado, G. S., Davis, A., Dean, J., Devin, M., Ghemawat, S., Goodfellow, I., Harp, A., Irving, G., Isard, M., Jia, Y., Jozefowicz, R., Kaiser, L., Kudlur, M., Levenberg, J., Mané, D., Monga, R., Moore, S., Murray, D., Olah, C., Schuster, M., Shlens, J., Steiner, B., Sutskever, I., Talwar, K., Tucker, P., Vanhoucke, V., Vasudevan, V., Viégas, F., Vinyals, O., Warden, P., Wattenberg, M., Wicke, M., Yu, Y., and Zheng, X.: TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems, https://www.tensorflow.org/ (last access: 17 February 2025), 2015. a
Ahlers, S., Henk, A., Hergert, T., Reiter, K., Müller, B., Röckel, L., Heidbach, O., Morawietz, S., Scheck-Wenderoth, M., and Anikiev, D.: 3D crustal stress state of Germany according to a data-calibrated geomechanical model, Solid Earth, 12, 1777–1799, https://doi.org/10.5194/se-12-1777-2021, 2021. a
Amadei, B. and Stephansson, O.: Rock Stress and its Measurement, Springer Science and Business Media, ISBN 10 0412447002, ISBN 13 978-0412447006, 1997. a
Bär, K., Reinsch, T., and Bott, J.: The PetroPhysical Property Database (P3) – a global compilation of lab-measured rock properties, Earth Syst. Sci. Data, 12, 2485–2515, https://doi.org/10.5194/essd-12-2485-2020, 2020. a
Bergen, K. J., Johnson, P. A., de Hoop, M. V., and Beroza, G. C.: Machine learning for data-driven discovery in solid Earth geoscience, Science, 363, eaau0323, https://doi.org/10.1126/science.aau0323, 2019. a
Blöcher, G., Cacace, M., Jacquey, A. B., Zang, A., Heidbach, O., Hofmann, H., Kluge, C., and Zimmermann, G.: Evaluating micro-seismic events triggered by reservoir operations at the geothermal site of Groß Schönebeck (Germany), Rock Mech. Rock Eng., 51, 3265–3279, 2018. a
Bond, C. E., Johnson, G., and Ellis, J. F.: Structural model creation: the impact of data type and creative space on geological reasoning and interpretation, Geol. Soc. Sp., 21, 83–97, https://doi.org/10.1144/SP421.4, 2015. a
Brady, B. H. and Brown, E. T.: Rock Mechanics: For Underground Mining, Springer Science and Business Media, ISBN 10 1402020643, ISBN 13 978-1402020643, 2006. a
Cacace, M. and Jacquey, A. B.: Flexible parallel implicit modelling of coupled thermal–hydraulic–mechanical processes in fractured rocks, Solid Earth, 8, 921–941, https://doi.org/10.5194/se-8-921-2017, 2017. a, b, c
Cosenza, A., Mannina, G., Vanrolleghem, P. A., and Neumann, M. B.: Global sensitivity analysis in wastewater applications: a comprehensive comparison of different methods, Environ. Modell. Softw., 49, 40–52, 2013. a
Crisci, E., Laloui, L., and Giger, S.: Arbeitsbericht NAB 20-08 TBO Bülach Dossier IX Rock-mechanical and Geomechanical Laboratory Testing, Arbeitsbericht, National Cooperative for the Disposal of Radioactive Waste (NAGRA), Wettingen, https://nagra.ch/en/downloads/arbeitsbericht-nab-20-08-2/ (last access: 17 February 2025), 2021. a
Degen, D. and Cacace, M.: Effects of transient processes for thermal simulations of the Central European Basin, Geosci. Model Dev., 14, 1699–1719, https://doi.org/10.5194/gmd-14-1699-2021, 2021. a
Degen, D., Veroy, K., and Wellmann, F.: cgre-aachen/DwarfElephant: DwarfElephant 1.0 (1.0), Zenodo [code], https://doi.org/10.5281/zenodo.4074777, 2020b. a
Degen, D., Spooner, C., Scheck-Wenderoth, M., and Cacace, M.: How biased are our models? – a case study of the alpine region, Geosci. Model Dev., 14, 7133–7153, https://doi.org/10.5194/gmd-14-7133-2021, 2021a. a, b, c
Degen, D., Veroy, K., Freymark, J., Scheck-Wenderoth, M., Poulet, T., and Wellmann, F.: Global sensitivity analysis to optimize basin-scale conductive model calibration – a case study from the Upper Rhine Graben, Geothermics, 95, 102143, https://doi.org/10.1016/j.geothermics.2021.102143, 2021b. a, b, c, d, e, f
Degen, D., Veroy, K., Scheck-Wenderoth, M., and Wellmann, F.: Crustal-scale thermal models: revisiting the influence of deep boundary conditions, Environ. Earth Sci., 81, 1–16, 2022b. a
Degen, D., Caviedes Voullième, D., Buiter, S., Hendricks Franssen, H.-J., Vereecken, H., González-Nicolás, A., and Wellmann, F.: Perspectives of physics-based machine learning strategies for geoscientific applications governed by partial differential equations, Geosci. Model Dev., 16, 7375–7409, https://doi.org/10.5194/gmd-16-7375-2023, 2023. a, b, c, d, e, f, g, h, i, j, k
Degen, D., Ziegler, M., Heidbach, O., Henk, A., Reiter, K., and Wellmann, F.: Non-Intrusive Reduced Basis Code for Elastic Geomechanical Models, Zenodo [data set] and [code], https://doi.org/10.5281/zenodo.13767010, 2024. a
Desroches, J., Peyret, E., Gisolf, A., Wilcox, A., Di Giovanni, M., de Jong, A. S., Sepehri, S., Garrard, R., and Giger, S.: Stress measurement campaign in scientific deep boreholes: focus on tool and methods, in: SPWLA Annual Logging Symposium, May 2021, virtual symposium, SPWLA-2021-0056 https://doi.org/10.30632/SPWLA-2021-0056, 2021. a, b
Ellsworth, W. L.: Injection-induced earthquakes, Science, 341, 1225942, https://doi.org/10.1126/science.1225942, 2013. a
Faroughi, S. A., Pawar, N., Fernandes, C., Das, S., Kalantari, N. K., and Mahjour, S. K.: Physics-guided, physics-informed, and physics-encoded neural networks in scientific computing, arXiv [preprint], https://doi.org/10.48550/arXiv.2211.07377 2022. a, b
Fischer, K. and Henk, A.: A workflow for building and calibrating 3-D geomechanical models – a case study for a gas reservoir in the North German Basin, Solid Earth, 4, 347–355, https://doi.org/10.5194/se-4-347-2013, 2013. a, b
Gaucher, E., Schoenball, M., Heidbach, O., Zang, A., Fokker, P. A., van Wees, J.-D., and Kohl, T.: Induced seismicity in geothermal reservoirs: a review of forecasting approaches, Renew. Sust. Energ. Rev., 52, 1473–1490, 2015. a
Geuzaine, C. and Remacle, J.-F.: Gmsh: a 3-D finite element mesh generator with built-in pre-and post-processing facilities, Int. J. Numer. Meth. Eng., 79, 1309–1331, 2009. a
Gölke, M. and Coblentz, D. D.: Origins of the European regional stress field, Tectonophysics, 266, 11–24, https://doi.org/10.1016/S0040-1951(96)00180-1, 1996. a
Gonus, J., Bailey, E., Desroches, J., and Garrard, R.: Nagra technical report 22-02, TBO Stadel-3-1: Data Report, Dossier VI: Wireline Logging, Micro-hydraulic Fracturing and Pressure-meter Testing, https://nagra.ch/en/downloads/arbeitsbericht-nab-22-02-2/ (last access: 17 February 2025), 2022a. a, b
Gonus, J., Bailey, E., Desroches, J., and Garrard, R.: Nagra technical report 22-02, TBO Stadel-2-1: Data Report, Dossier VI: Wireline Logging, Micro-hydraulic Fracturing and Pressure-meter Testing, https://nagra.ch/en/downloads/arbeitsbericht-nab-22-02-2/ (last access: 17 February 2025), 2022b. a, b
Guo, M. and Hesthaven, J. S.: Data-driven reduced order modeling for time-dependent problems, Comput. Method. Appl. M., 345, 75–99, 2019. a
Guo, T., Rokoš, O., and Veroy, K.: A reduced order model for geometrically parameterized two-scale simulations of elasto-plastic microstructures under large deformations, Comput. Method. Appl. M., 418, 116467, https://doi.org/10.1016/j.cma.2023.116467, 2024. a, b, c
Henk, A.: Perspectives of geomechanical reservoir models–why stress is important, European Magazine, 4, 1–5, 2008. a
Jacquey, A. B. and Cacace, M.: GOLEM, a MOOSE-based application, Zenodo [code], https://doi.org/10.5281/zenodo.999401, 2017. a
Jordan, M. I. and Mitchell, T. M.: Machine learning: trends, perspectives, and prospects, Science, 349, 255–260, 2015. a
Kotsiantis, S. B., Zaharakis, I., and Pintelas, P.: Supervised machine learning: a review of classification techniques, Emerging Artificial Intelligence Applications in Computer Engineering, 160, 3–24, 2007. a
Kumar, A., Hu, R., and Walsh, S. D.: Development of reduced order hydro-mechanical models of fractured media, Rock Mech. Rock Eng., 55, 1–14, 2021. a
Li, Y. E., O'Malley, D., Beroza, G., Curtis, A., and Johnson, P.: Machine Learning Developments and Applications in Solid-Earth Geosciences: Fad or Future?, J. Geophys. Res.-Sol. Ea., 128, e2022JB026310, https://doi.org/10.1029/2022JB026310, 2023. a
Lindsay, A. D., Gaston, D. R., Permann, C. J., Miller, J. M., Andrš, D., Slaughter, A. E., Kong, F., Hansel, J., Carlsen, R. W., Icenhour, C., Harbour, L., Giudicelli, G. L., Stogner, R. H., German, P., Badger, J., Biswas, S., Chapuis, L., Green, C., Hales, J., Hu, T., Jiang, W., Jung, Y. S., Matthews, C., Miao, Y., Novak, A., Peterson, J. W., Prince, Z. M., Rovinelli, A., Schunert, S., Schwen, D., Spencer, B. W., Veeraraghavan, S., Recuero, A., Yushu, D., Wang, Y., Wilkins, A., and Wong, C.: 2.0 – MOOSE: enabling massively parallel multiphysics simulation, SoftwareX, 20, 101202, https://doi.org/10.1016/j.softx.2022.101202, 2022. a
Loh, W.-L.: On Latin hypercube sampling, Ann. Stat., 24, 2058–2080, 1996. a
Mahesh, B.: Machine learning algorithms-a review, International Journal of Science and Research (IJSR), 9, 381–386, 2020. a
Martin, D.: Quantifying in situ stress magnitudes and orientations for Forsmark – Formsark stage 2.2, Tech. Rep. SKB R-07-26, Swedish Nuclear Fuel and Waste Management Co, Stockholm Sweden, technical report, https://www.osti.gov/etdeweb/servlets/purl/925966 (last access: 17 February 2025), 2007. a
NAGRA: In-situ stress field in the siting regions Jura Ost, Nördlich Lägern and Zürich Nordost, NAGRA, Wettingen, NAGRA Arbeitsbericht NAB 24-19, 91 pp., https://nagra.ch/downloads/arbeitsbericht-nab-24-19 (last access: 18 June 2025), 2024. a
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S.: PyTorch: An Imperative Style, High-Performance Deep Learning Library, in: Advances in Neural Information Processing Systems 32, edited by: Wallach, H., Larochelle, H., Beygelzimer, A., d' Alché-Buc, F., Fox, E., and Garnett, R., Curran Associates, Inc., http://papers.neurips.cc/paper/9015-pytorch-an-imperative-style-high-performance-deep-learning-library.pdf (last access: 17 February 2025), 8024–8035, 2019. a
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E.: Scikit-learn: machine learning in Python, J. Mach. Learn. Res., 12, 2825–2830, 2011. a
Permann, C. J., Gaston, D. R., AndrÅ–, D., Carlsen, R. W., Kong, F., Lindsay, A. D., Miller, J. M., Peterson, J. W., Slaughter, A. E., Stogner, R. H., and Martineau, R. C.: MOOSE: enabling massively parallel multiphysics simulation, SoftwareX, 11, 100430, https://doi.org/10.1016/j.softx.2020.100430, 2020. a
Quarteroni, A., Manzoni, A., and Negri, F.: Reduced Basis Methods for Partial Differential Equations: An Introduction, UNITEXT, Springer International Publishing, https://doi.org/10.1007/978-3-319-15431-2, 2015. a, b, c
Razavi, S. and Gupta, H. V.: What do we mean by sensitivity analysis? The need for comprehensive characterization of “global” sensitivity in Earth and environmental systems models, Water Resour. Res., 51, 3070–3092, 2015. a
Reiter, K.: Stress rotation – impact and interaction of rock stiffness and faults, Solid Earth, 12, 1287–1307, https://doi.org/10.5194/se-12-1287-2021, 2021. a
Reiter, K. and Heidbach, O.: 3-D geomechanical–numerical model of the contemporary crustal stress state in the Alberta Basin (Canada), Solid Earth, 5, 1123–1149, https://doi.org/10.5194/se-5-1123-2014, 2014. a, b, c
Röckel, L., Ahlers, S., Müller, B., Reiter, K., Heidbach, O., Henk, A., Hergert, T., and Schilling, F.: The analysis of slip tendency of major tectonic faults in Germany, Solid Earth, 13, 1087–1105, https://doi.org/10.5194/se-13-1087-2022, 2022. a
Saltelli, A.: Making best use of model evaluations to compute sensitivity indices, Comput. Phys. Commun., 145, 280–297, 2002. a
Saltelli, A., Annoni, P., Azzini, I., Campolongo, F., Ratto, M., and Tarantola, S.: Variance based sensitivity analysis of model output. Design and estimator for the total sensitivity index, Comput. Phys. Commun., 181, 259–270, 2010. a
Santoso, R., Degen, D., Cacace, M., and Wellmann, F.: State-of-the-art physics-based machine learning for hydro-mechanical simulation in geothermal applications, in: European Geothermal Congress, Berlin, Germany, 17–21 October 2022, https://www.researchgate.net/profile/Ryan-Santoso-2/publication/368396013_State-of-the-art_physics-based_machine_learning_for_hydro-mechanical_simulation_in_geothermal_applications/links/63 e638e5c002331f726b5401/State-of-the-art-physics-based-machine-learning-for-hydro-mechanical-simulation-in-geothermal-applications.pdf (last access: 17 February 2025, 1–10, 2022. a
Sarrazin, F., Pianosi, F., and Wagener, T.: Global sensitivity analysis of environmental models: convergence and validation, Environ. Modell. Softw., 79, 135–152, 2016. a
Schmitt, D. R., Currie, C. A., and Zhang, L.: Crustal stress determination from boreholes and rock cores: fundamental principles, Tectonophysics, 580, 1–26, 2012. a
Schulz, E., Speekenbrink, M., and Krause, A.: A tutorial on Gaussian process regression: modelling, exploring, and exploiting functions, J. Math. Psychol., 85, 1–16, 2018. a
Segall, P. and Fitzgerald, S. D.: A note on induced stress changes in hydrocarbon and geothermal reservoirs, Tectonophysics, 289, 117–128, 1998. a
Sin, G., Gernaey, K. V., Neumann, M. B., van Loosdrecht, M. C., and Gujer, W.: Global sensitivity analysis in wastewater treatment plant model applications: prioritizing sources of uncertainty, Water Res., 45, 639–651, 2011. a
Smart, K. J., Ofoegbu, G. I., Morris, A. P., McGinnis, R. N., and Ferrill, D. A.: Geomechanical modeling of hydraulic fracturing: why mechanical stratigraphy, stress state, and pre-existing structure matter, AAPG Bull., 98, 2237–2261, 2014. a
Song, X., Zhang, J., Zhan, C., Xuan, Y., Ye, M., and Xu, C.: Global sensitivity analysis in hydrological modeling: review of concepts, methods, theoretical framework, and applications, J. Hydrol., 523, 739–757, 2015. a
Tang, Y., Reed, P., Wagener, T., and van Werkhoven, K.: Comparing sensitivity analysis methods to advance lumped watershed model identification and evaluation, Hydrol. Earth Syst. Sci., 11, 793–817, https://doi.org/10.5194/hess-11-793-2007, 2007. a
Tingay, M., Heidbach, O., Davies, R., and Swarbrick, R.: Triggering of the Lusi mud eruption: earthquake versus drilling initiation, Geology, 36, 639–642, 2008. a
Vanrolleghem, P. A., Mannina, G., Cosenza, A., and Neumann, M. B.: Global sensitivity analysis for urban water quality modelling: terminology, convergence and comparison of different methods, J. Hydrol., 522, 339–352, 2015. a
Wang, Q., Hesthaven, J. S., and Ray, D.: Non-intrusive reduced order modeling of unsteady flows using artificial neural networks with application to a combustion problem, J. Comput. Phys., 384, 289–307, https://doi.org/10.1016/j.jcp.2019.01.031, 2019. a, b
Willcox, K. E., Ghattas, O., and Heimbach, P.: The imperative of physics-based modeling and inverse theory in computational science, Nature Computational Science, 1, 166–168, 2021. a
Zhao, H.: A reduced order model based on machine learning for numerical analysis: An application to geomechanics, Eng. Appl. Artif. Intel., 100, 104194, https://doi.org/10.1016/j.engappai.2021.104194, 2021. a
Ziegler, M. O. and Heidbach, O.: The 3D stress state from geomechanical–numerical modelling and its uncertainties: a case study in the Bavarian Molasse Basin, Geothermal Energy, 8, 11, https://doi.org/10.1186/s40517-020-00162-z, 2020. a, b, c
Ziegler, M. O. and Heidbach, O.: Increasing accuracy of 3-D geomechanical-numerical models, Geophys. J. Int., 237, 1093–1108, https://doi.org/10.1093/gji/ggae096, 2024. a
Ziegler, M., Reiter, K., Heidbach, O., Zang, A., Kwiatek, G., Stromeyer, D., Dahm, T., Dresen, G., and Hofmann, G.: Mining-induced stress transfer and its relation to a Mw 1.9 seismic event in an ultra-deep South African gold mine, Pure Appl. Geophys., 172, 2557–2570, 2015. a
Ziegler, M. O., Heidbach, O., Reinecker, J., Przybycin, A. M., and Scheck-Wenderoth, M.: A multi-stage 3-D stress field modelling approach exemplified in the Bavarian Molasse Basin, Solid Earth, 7, 1365–1382, https://doi.org/10.5194/se-7-1365-2016, 2016. a
Ziegler, M., Heidbach, O., Morawietz, S., and Wang, Y.: Manual of the Matlab Script FAST Calibration v2.4, WSM Technical Report, https://doi.org/10.48440/WSM.2023.002, 23-02, 2023. a
Zoback, M. D.: Reservoir Geomechanics, Cambridge University Press, https://doi.org/10.1017/cbo9780511586477, 2007. a, b
Short summary
Obtaining reliable estimates of the subsurface state distributions is essential to determine the location of, e.g., potential nuclear waste disposal sites. However, providing these is challenging since it requires solving the problem numerous times, yielding high computational cost. To overcome this, we use a physics-based machine learning method to construct surrogate models. We demonstrate how it produces physics-preserving predictions, which differentiates it from purely data-driven approaches.
Obtaining reliable estimates of the subsurface state distributions is essential to determine the...
