From Complexity to Comprehension: Interactive Real-Time Data Visualization for Geological Models and Uncertainty Analysis

The search for a repository site and its design rely on digital models and simulations. An integral part of the site selection process is the assessment of the integrity of the containment-providing rock zone. Computational models assessing the integrity of geological barriers in repository systems often yield complex results typically accessible only to experts in these disciplines. An extra layer of complexity is added to these results by considering input uncertainties. As the site-selection process progresses, it is becoming increasingly important to make this kind of results accessible to a broader audience to support and facilitate decision-making and improve the acceptance of pre-selected and rejected sites.

Deterministic finite element simulation results are resolved in time and space for each primary variable and each additional post-processed quantity, such as pore water pressure, stresses, temperatures, and integrity criteria. If uncertainty – or, in other words, variability – in the input parameters is considered, the dimensionality of the state space grows with each parameter included in the stochastic or parametric model. A surrogate model is constructed by adaptive sparse grid sampling, efficiently mapping the complete state space. No data is reduced in this step such that a functional dependency between uncertain inputs and complete finite element results is established, enabling finite element result interpolation for any location in the state space. With modern software technologies developed at the Federal Institute for Geosciences and Natural Resources (BGR), the time for querying the surrogate model has been reduced to such an extent that complete finite element results are reproduced in fractions of a second, laying the foundation for a real-time visualization dashboard by which the effect of the change of any uncertain input parameter can be investigated in the complete physical and time domain.

This contribution uses the geological model ANSICHT-II [3] as an example. It represents a generic repository in a clay rock formation of greater thickness without a fixed local reference. Individual simulation runs were performed by OpenGeoSys 6 [4], and the stochastic model was generated by OpenGeoSysUncertaintyQuantification.jl [1]. The interactive dashboard was generated with the help of Makie.jl [2], a flexible, high-performance, cross-platform plotting ecosystem for the Julia programming language. With this dashboard, the user can freely select the most important input parameters (thermal, hydraulic, and mechanical properties of the host rock) within realistic ranges, and the corresponding results are displayed in real time. In addition, the area where integrity violation is to be expected is marked.

[1] Bittens, M. (2024). OpenGeoSysUncertaintyQuantification.jl: a Julia library implementing an uncertainty quantification toolbox for OpenGeoSys. Journal of Open Source Software, 9(98), 6725.

[2] Danisch, S., & Krumbiegel, J. (2021). Makie.jl: Flexible high-performance data visualization for Julia. Journal of Open Source Software, 6(65), 3349.

[3] Maßmann, J., et. al. (2022): ANSICHT-II – Methode und Berechnungen zur Integritätsanalyse der geologischen Barriere für ein generisches Endlagersystem im Tongestein. Bundesanstalt für Geowissenschaften und Rohstoffe (BGR), Ergebnisbericht; Hannover. DOI:10.25928/n8ac-y452.

[4] Kolditz, O., et al. (2012). OpenGeoSys: an open-source initiative for numerical simulation of thermo-hydro-mechanical/chemical (THM/C) processes in porous media. Environmental Earth Sciences 67 (2012): 589-599.