Ground penetrating radar (GPR) is a widely used geophysical technique for detecting and characterizing subsurface structures such as fractures. While it is effective for determining fracture geometry, accurately characterizing heterogeneous aperture fields remains a challenge. We investigate the latter through a combination of numerical simulations and controlled laboratory experiments. We created a fracture aperture field by modelling its rough-wall surfaces with spatially correlated random fields. We used this digital model in numerical simulations and to fabricate a physical replica via 3-D printing and gypsum casting. We then simulated and measured GPR reflection data on a 2-D measurement grid above the fracture, and after amplitude normalization and averaging, computed both relative and absolute reflection coefficients. This data set allowed us to assess how aperture heterogeneity is expressed in the GPR response and to extract spatial statistics. In particular, our results show that the correlation length of the aperture field can be reliably inferred from reflection data through variogram analysis, even after normalization. Our results are significant for subsurface fracture characterization, where alternative methods to remotely estimate apertures do not exist.

Aperture is a key property of fractures, as it controls fluid flow and mechanical behaviour in fractured rock, yet it remains difficult to observe directly. Ground Penetrating Radar (GPR) offers non-invasive means of imaging fractures and is commonly used to determine their geometry. The amplitude of the reflected GPR signal, which contains valuable information about fracture aperture and filling material, is often underutilized in quantitative interpretation. In this study, we present a workflow for estimating heterogeneous fracture aperture fields by inverting GPR reflection data in the frequency domain. The approach relies on the thin-bed approximation and uses the effective-dipole forward model. We first validate the workflow using laboratory data with a known aperture field, demonstrating accurate recovery of aperture distributions in high-sensitivity regions. We then apply the method to field data from the Bedretto Laboratory, where it successfully reconstructs a 1-D relative aperture profile along a target fault. From this, we extract the correlation length, providing additional insight into fracture structure. These results demonstrate that the approach can deliver robust, spatially resolved estimates of fracture aperture and its variability, even under realistic field conditions.

Fault zone geological and geometrical complexities are prime parameters playing a fundamental role in controlling the characteristics of both natural and induced seismicity. In the Bedretto tunnel (Switzerland), the Fault Activation and Earthquake Rupture (FEAR) project aims at triggering a Mw = 1 seismic event through fluid injection and stimulation of a natural fault zone situated in a large-scale (> 106 m3) fractured granite reservoir. The limited exposure of the fault zone in the tunnel, however, restricts the possibility to constrain in detail the geometrical and geological characteristics of the experimental target. Therefore, in order to constrain the geological and geometrical characteristics of the target fault zone, we have integrated structural analyses, borehole and core logging, and borehole ground penetrating radar (GPR). Preliminary field investigations in the tunnel allowed to identify the complex fault structure characteristics, fault rock properties, and slip tendency in the current stress field of the selected fault zone. These results were compared to the structural observations obtained from field surveys and remote sensing, constraining the slip history, and lateral extent of the set of natural fault zones occurring on the surface above the Bedretto tunnel. Indeed, the lateral extent of the selected fault has been confirmed through the logging (optical/acoustic televiewer, fracture intensity, fracture typology) of exploration boreholes and the analyses of the related cores. The comparison between the geological characteristics of fault zones in the cores and the characteristics of the selected fault zone exposed in the tunnel allowed to confirm the occurrence of the same typology of fault zone further away from the exposure in the tunnel. In addition, GPR logging of the exploratory boreholes provided fundamental insights on the lateral continuity of the identified fault zones on the tunnel wall, as well as those identified in the borehole/core logging. All geological and geometrical information have been integrated into a preliminary 3D geometrical model (in Leapfrog Geo), representing the overall geometry of the selected fault zone. This preliminary geometrical model has been validated against synthetic GPR profiles, computed through GPR forward modelling along the exploration boreholes. The integrated results define the selected fault zone as a 3-7 m wide zone of higher density (up to 5/m), of variably oriented secondary fractures, and bounded by two main slip surfaces. The slip surfaces are irregularly decorated by phyllosilicate-rich gouge patches, filling the roughness of the fault surface. The lateral extension of each discrete fracture does not exceed 30 m in length, but the overall lateral continuity of the fault zone exceeds several hundreds of meters. The presented integrated characterization approach allowed us to constrain a geologically-sound, first-order 3D geometrical model of a complex natural fault zone, validated against geophysical forward modelling. These preliminary results have fundamental implications for the expected experimental planning and outcomes, modelling and injection strategies, project logistics, as well as the design and deployment of the monitoring network around the stimulated fault zone.

Fracture curvature has been observed from the millimetre to the kilometre scales. Nevertheless, characterizing curvature remains challenging due to data sparsity and geometric ambiguities. As a result, most numerical models often assume planar fractures to ease computations. To address this limitation, we present a novel approach for inferring fracture geometry from travel-time data of electromagnetic or seismic waves. Our model utilizes co-kriging interpolation of control points in a three-dimensional surface mesh to simulate fracture curvature effectively, resulting in an unstructured triangular grid. We then refine the fracture surface into a structured grid with equidistant elements so that both small-scale heterogeneities and large-scale curvature can be modelled. To constrain the fracture geometry, we perform a deterministic travel-time inversion to optimally place these control points. We validate our methodology with synthetic data and address its limitations. Finally, we infer the geometry of a large (more than 200 m) fracture observed in single-hole ground-penetrating radar field data. The fracture surface closely agrees with borehole televiewer observations and is also constrained far from the boreholes. Our modelling approach can be trivially adapted to multi-offset ground-penetrating radar or active seismic data.

Virtual reality concepts have been widely adapted to teach geoscientific content, most notably in virtual field trips-with increased developments due to recent travel restrictions and challenges of field access. On the spectrum between real and fully virtual environments are also combinations of digital and real content in mixed-reality environments. In this category, augmented-reality (AR) sandboxes have been used as a valuable tool for science outreach and teaching due to their intuitive and haptic interaction-enhancing operation. Most of the common AR-sandboxes are limited to the visualization of topography with contour lines and colors, as well as water simulations on the digital terrain surface. We show here how we can get beyond this limitation, through an open-source implementation of an AR-sandbox system with a versatile interface written in the free and cross-platform programming language Python. This implementation allows for creative and novel applications in geosciences education and outreach in general. With a link to a 3-D geo-modelling system, we show how we can display geologic subsurface information such as the outcropping lithology, creating an interactive geological map for structural geology classes. The relations of subsurface structures, topography, and outcrop can be explored in a playful and comprehensible way. Additional examples include the visualizations of geophysical fields and the propagation of seismic waves, as well as simulations of Earth surface processes. We further extended the functionality with ArUco-marker detection to enable more precise and flexible interaction with the projected content. In combination, with these developments, we aim to make AR-sandbox systems, with the additional dimension of haptic interactions, accessible to a wider range of geoscientific applications for education and outreach.

Geological modelling is an essential aspect of a wide variety of geophysical and geological investigations related to geo-energy exploration and monitoring. A commonly-applied procedure is to use 3D geological models (often referred to as static models) to characterise the spatial distribution of material properties, which are then used in subsequent process simulations. The physical processes are described with partial differential equations that can be solved using different numerical methods by creating a discretisation in the space of the geometric object (i.e., a mesh). However, mesh generation can be a time-consuming step that generally only allows an approximation of the true geometric model. Several methods have been proposed to resolve these issues. We investigate here the use of the isogeometric analysis (IGA) technique, which exploits the finite element method (FEM) to numerically solve differential equations without the need of creating a mesh. Instead, it uses computer-aided design (CAD) tools, specifically Non-Uniform Rational B-splines (NURBS), to accurately represent any form of conic sections geometry. This presentation shows the link between NURBS representing geological interfaces and subsequent geothermal process simulations. The link is implemented in a user-friendly Python package (https://github.com/danielsk78/pygeoiga) with a simple but clear interface. It differs from other implementations by dealing with multipatch structures and focusing on geological modelling with multiple subdomains. A series of numerical examples are presented to show the use of the technique for solving the two-dimensional heat conduction problem. Results are contrasted to the results of a traditional FEM approach. The comparison shows that IGA requires fewer degrees of freedom (DoF) for convergence of the solution. Further, IGA provides a way to ease the workflow from the geological modelling to the results of process simulations, enabling tighter integration between modelling and simulation. Lastly, we describe shortly how the IGA concept can be implemented on top of existing standard FEM libraries.