Geo-modelling and geophysical inversion are powerful tools for addressing these challenges, enabling the integration of geological, geophysical, and petrophysical data into consistent three-dimensional subsurface representations. Recent advances in modelling algorithms, inversion techniques, and computational approaches, including machine learning and AI, allow for improved accuracy, uncertainty quantification, and interpretability across multiple spatial and temporal scales, even in data-scarce environments.
This session highlights recent developments in 3D geological modelling, reservoir and fluid-flow simulation, and model-based inversion, with a focus on applications supporting the energy transition and sustainable subsurface use. Both methodological contributions and case studies are welcome.
Topics of interest include, but are not limited to:
• 3D geological and structural modelling approaches
• Geo-modelling for geothermal energy, CCS, and hydrogen storage
• Fluid-flow and reservoir modelling (single- and multiphase)
• Model-based geophysical inversion and data integration
• Multi-scale modelling, upscaling, and representative elementary volumes (REVs)
• Uncertainty quantification and computational efficiency
• Machine learning and AI in geological modelling and inversion
Orals: Wed, 6 May, 08:30–10:15 | Room -2.43
Faults may act either as preferential migration pathways or as effective sealing barriers during hydrocarbon accumulation. Therefore, robust evaluation of fault sealing capacity is critical for assessing trap integrity and reducing exploration risk in fault-controlled reservoirs. Conventional fault-sealing assessments are commonly restricted to two-dimensional cross-sections or single-point analyses, which are insufficient to represent the pronounced spatial heterogeneity and structural complexity of fault systems.To address these limitations, this study proposes a three-dimensional quantitative evaluation method for fault sealing capacity based on displacement pressure difference. Taking the K area gas field in the step-fault zone of the Pinghu Slope, Xihu Depression, as a case study, we integrate three-dimensional structural modeling with geological property modeling. Fault planes are discretized, and key controlling attributes—including in-situ stress, shale content, burial depth, and fault activity timing—are assigned to individual fault elements, allowing calculation of both fault rock displacement pressure and the displacement pressure of juxtaposed reservoirs in three-dimensional space. This framework enables simultaneous evaluation of vertical and lateral sealing capacities.The results indicate that faults within the P5 and P6 members of the Pinghu Formation exhibit relatively strong sealing capacity and constitute the principal sealing intervals in the study area, consistent with existing exploration outcomes. In addition, faults located adjacent to the depocenter generally display enhanced sealing capacity, highlighting favorable zones for future exploration. Compared with traditional fault-sealing evaluation methods, the proposed approach significantly improves spatial resolution and visualization of sealing capacity, enhances evaluation efficiency, and reduces the subjectivity inherent in point-based analyses. This method is therefore well suited for fine-scale characterization of fault-controlled hydrocarbon accumulation in structurally complex basins.
Keywords：Fault sealing capacity; Displacement pressure difference; Three-dimensional fault model; Xihu Depression
How to cite: Yan, J. and Xue, H.: Three-Dimensional Evaluation of Fault Sealing Capacity Based on Fault Attributes and Displacement Pressure Difference: Application to the K Area of the Pinghu Slope, Xihu Depression, East China Sea Shelf Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2441, https://doi.org/10.5194/egusphere-egu26-2441, 2026.
Three-dimensional geological modelling is increasingly used to analyze and investigate the geological evolution of complex areas, offering advantages over classical 2D maps and cross-sections, to test and validate geometries and structural (topological) relationships against sparse field data.
Within this context, 3D modelling in polymetamorphic belts poses different challenges, first the absence of formally defined stratigraphic surfaces that are transposed and cancelled by multi-stage tectono-metamorphic events. Alternative tectonostratigraphic or tectonometamorphic units are used when mapping in these environments, but the non-formal definition of these units and of their boundaries can lead to topological ambiguities and even inconsistencies in geological legends, that in turn lead to geo-ontological deficiencies in the modelling process – i.e. deficiencies in the explicit and formal shared conceptualization of the geological meaning and role assigned to units and boundaries.
To address these issues, it is essential to explicitly integrate topological and ontological reasoning into the modelling workflow, building a consistent geological legend, in order to generate valid 3D models both in implicit and explicit modelling approaches.
Here, we present the Structural Topology model (STm), a workflow grounded in classical structural geology’s thinking and field mapping knowledge, which systematically analyses scale-dependent topological relationships between surfaces and volumes to reconstruct a geologically valid and internally consistent 3D legend based on the concept of a generalized structural polarity. This is a vector that, depending on the geological environment and modelling purpose, can be defined as the younging direction (when relative or absolute ages are available), but also structural position with respect to some convenient reference, and cross-cutting relationships allowing to constrain a sequence of geological events. In this framework, units are classified as tectonometamorphic (TMU), tectonostratigraphic (TSU), stratigraphic (SU), intrusive (IU), or shear zone (SZ) according to their origin and evolution. Their boundaries may be conformal to the main foliation (in the broadest sense, including bedding) or discordant, e.g. at some tectonic contacts, shear zones and unconformities.
The formal definition of units with internal polarity, conformal vs. discordant boundaries with polarity, and cross-cutting relationships, allow connecting geological ontology with a topological model that can be implemented in a 3D model. In addition, properly defining polarity for each model entity allows using implicit surface methods (that operate by interpolating a scalar field whose gradient is the polarity) at each stage of the modelling workflow.
Here we present an implementation of the STm within the PZero open-source software (https://github.com/gecos-lab/PZero), including a lightweight graphical interface that enables the construction of STm-based geological legends from a Polarigram, where units are plotted against polarity. Examples from complex polymetamorphic areas in the Alps demonstrate that geological topology can be robustly defined even when geological ontology remains ambiguous and scale-dependent, providing a consistent foundation for 3D geological modelling.
How to cite: Monti, R., Bistacchi, A., Hussain, W., Favaro, S., Herwegh, M., Drvoderić, S., Furlan, M., Musso Piantelli, F., Dal Piaz, G., and Monopoli, B.: The Structural Topology model: integrating topology and ontology in 3D geological modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18878, https://doi.org/10.5194/egusphere-egu26-18878, 2026.
Geological models are needed for subsurface engineering purposes, and it is crucial to identify their uncertainties. However, uncertainties in their input are easily overlooked. Through a data review of Paleozoic geology in the Eurogion Meuse-Rhine, sources of uncertainty were identified in geomodelling input. Causes of uncertainty were classified into four groups: (a) stratigraphic interpretation, (b) fault interpretation, (c) transferring data, and (d) uncertainty in legacy materials. The causes of uncertainty are interlinked, causing the uncertainty chain in geomodelling to be more complex than generally considered.
The Paleozoic geology in the study region is structurally complex and geomodelling is hampered by limited outcrops and scattered input data. This study compiles geomodelling input and examines data inconsistencies. We collected legacy literature and maps, conducted fieldwork, and compiled a dataset of 738 boreholes. New borehole data are included, and two legacy boreholes (Kastanjelaan-2 and RWTH-1) were re-evaluated. Differences are observed between various stratigraphic profiles for these two boreholes among different sources. Here, we propose updated stratigraphic interpretations for them. Comparing a newly drilled borehole with an existing geological cross-section reveals a >1 km depth mismatch between stratigraphic units. Comparing the stratigraphy of the borehole dataset with different geologic maps shows various degrees of agreement. The identified inconsistencies demonstrate the necessity of validating input data before embarking on any geomodelling exercise.
How to cite: Maars, J., Hupkes, J., Houben, A. J. P., Vis, G.-J., Martinius, A. W., Geel, C. R., de Ceukelaire, M., and Abels, H.: Causes of uncertainties in geomodelling inputs: data review of Paleozoic geology of the Euregion Meuse-Rhine, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9578, https://doi.org/10.5194/egusphere-egu26-9578, 2026.
We will never cease to be interested in fluid flow in the subsurface. Only the fluid changes.
For the energy transition the emphasis of our modelling efforts is changing, however. Aspects we could accept as ‘reasonable representation’ in oil and gas production projects (especially gas) are less acceptable for storage projects.
This talk will pick out two key elements which differ for ‘transition work’ from a modelling perspective:
- The need for multi-scale modelling (the REV, sometimes requires for production, always required for storage), and
- The need for better reservoir-scale structural representation – we’re good at sedimentary heterogeneity but much less so, in practice, for structural heterogeneity.
The expertise with scenario-based workflows, familiar from decades of production projects, applies directly to storage projects. The principal difference is the lack of calibration data for aquifer-scale storage projects, as these are operating at scales more familiar from regional exploration groups, yet requiring a representation of physics more comparable to km-scale EOR production projects. With such consequent uncertainty, this means the need for scenarios increases and the requirement for a base case decreases, to the point that ‘base case’ modelling becomes effectively meaningless for storage projects.
For geothermal projects the requirements change again, which even more emphasis on structural modelling. The challenge here is marginal economics of geothermal projects and hence a different approach to project management (and resulting need for modelling support). In the extreme case, the argument can be made for ‘no modelling, just fund a pilot project and learn from experience’. A more nuanced approach would be to take learnings from existing modelling work for other projects (production or storage) and apply selectively and sensibly to the geothermal arena.
The talk will illustrate the above with reference to some example model workflows.
How to cite: Bentley, M.: Modelling Workflows for the Energy Transition - New Tricks, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21861, https://doi.org/10.5194/egusphere-egu26-21861, 2026.
Accurate subsurface characterization is essential for energy transition technologies including geothermal systems, carbon capture and storage (CCS), and hydrogen storage. Reservoir models face a critical challenge: core measurements of petrophysical properties at centimetre scale are used to populate simulation cells which are typically 10-100 m, creating a very large scale gap (1010). Inappropriate upscaling methods lead to systematic errors in flow predictions and fail to preserve the impact of heterogeneity at different levels, particularly in heterolithic depositional environments such as tidal systems where mud drapes create extreme vertical flow barriers. Here, we present a novel approach for multi-scale study that uses virtual outcrop analogues including standard virtual outcrops and high-resolution mini-models collected using smartphone-based lidar. The Sego Sandstone Formation from Sego Canyon (the Book Cliffs, Utah, USA) serves as the case study, representing a tide- and wave-influenced shoreline succession. These deposits serve as analogues to the Garn Formation in the mid-Norwegian Continental Shelf and similar tidal reservoirs.
The workflow comprises three steps. The first step includes small-scale models (1-5 m extent; 1 cm grid resolution) where virtual mini-model data captures centimetre-scale sedimentary heterogeneity. These models were then upscaled to 1 m horizontal and 0.1 m vertical resolution, followed by statistical regression analysis. The second step involves meso-scale models (25×25×15 m model size; 1×1×0.1 m cell size) where these regression relationships are applied, enabling systematic testing of upscaling methods under controlled conditions. These meso-scale models were then upscaled to 25 m horizontal and 1 m vertical resolution, representing typical cell dimensions for reservoir models. Statistical and regression analysis were repeated to derive reservoir-scale upscaling parameters. The final step comprises macro-scale reservoir models with outcrop-scale dimensions and 25×25×1 m cell size, applying validated upscaling parameters from the previous step. Outcomes were compared with original small-scale data to quantify the impact of multi-resolution heterogeneity and identify which geological features have the most influence on upscaled values.
Results demonstrate that depositional architecture fundamentally controls upscaling behaviour, with heterogeneity significantly affecting permeability predictions at all levels. Clean sand facies (tidal bars, shoreface) show predictable behaviour with minimal scale effects on horizontal permeability and moderate vertical anisotropy controlled primarily by cross-bedding dip. Heterolithic facies (inter-bar, offshore transition zones) display moderate horizontal permeability variation but extreme vertical permeability reduction due to continuous mud drapes creating severe vertical flow barriers. Overall, permeability shows complex behaviour at different scales, which cannot be captured by placing data from centimetre-scale core plug measurements directly into simulation cells - a critical limitation for subsurface studies. This methodology is transferable across all depositional environments and directly applicable to energy transition projects requiring accurate multi-resolution flow predictions.
How to cite: Kapustina, I., Howell, J. A., and Kelly, S.: A multi-scale analysis bridging the gap from centimetres to reservoir simulation cell size for heterogeneous tidal reservoirs, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21157, https://doi.org/10.5194/egusphere-egu26-21157, 2026.
Among the geometrical parameters that can be calculated from an outcrop analogue, fracture areal intensity (P21), defined as the ratio between the total sum of fracture trace length and the sampling area, represents the stopping criterion of 2D marked point process stochastic Discrete Fracture Network models (DFN) or 3D simulations if properly upscaled to its volumetric equivalent (P32). Given the heterogeneous nature of fracture networks, P21 calculation is inherently scale and position dependent, meaning that its value changes depending on the scan area size and its position within the outcrop boundary. For this reason, P21 calculation cannot be separated from the concept of Representative Elementary Volume (REV) or Area (REA), in case of 2D outcrop studies. Depending on the field of application, REV or REA size definition changes, adapting to the different sampling strategies and parameters specific to that research field. We propose a novel approach to define the REA for the P21 parameter as a range bounded by a lower and an upper limit. The upper limit, often overlooked but nonetheless theorized, identifies the largest representative domain, which is crucial for optimizing computational efficiency. To determine the REA range, we evaluate the shape, mean, and variance of P21 value statistical distributions across scan areas of increasing radius. Each statistical parameter is assessed by combining formal statistical tests and diagnostic plots. Within a multi-parametric framework, the method enables a detailed analysis of the statistical behavior of the dataset, facilitating more objective and informed decisions when defining the REA range.
How to cite: Casiraghi, S., Bertacchi, D., Benedetti, G., Mittempergher, S., and Bistacchi, A.: Statistical Assessment of Fracture Areal Intensity (P21) Representative Elementary Area in Digital Outcrop Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12711, https://doi.org/10.5194/egusphere-egu26-12711, 2026.
Geological carbon sequestration (GCS) is a key technology for mitigating CO₂ emissions from hard-to-abate industrial sources, it has been tested in various subsurface formations including basalts, coal seams, shales, carbonate rocks, sandstones and salt formations. With carbonates and sandstones being the most widely utilized reservoirs for long-term storage. Although these formations may exhibit favorable porosity and permeability, they are typically heterogenous because of various depositional processes and diagenesis modifications. Such heterogeneity has significant impact on CO₂ injectivity, migration, and storage efficiency.
Bioturbation, the reworking and modification of sediments by organisms represents an additional and often underexplored source of heterogeneity in both carbonate and sandstone reservoirs. Burrow networks can locally enhance or impede fluid flow, thereby influencing CO2 injectivity, migration behavior, and storage performance. This study investigates the role of bioturbation, represented by Thalassinoides networks, in controlling CO2 storage behavior in tight sedimentary strata, with the Upper Jurassic Hanifa Formation of Saudi Arabia serving as a representative case study.
High-resolution X-ray computed tomography scans of Thalassinoides-bearing carbonate rock samples were used to capture the three-dimensional geometry and connectivity of the burrow networks. These data served as training images for multipoint statistics modeling, allowing the construction of a realistic fine-scale rock model that preserve burrow morphology and spatial continuity. To facilitate dynamic flow simulations, the model was upscaled to a coarser grid while maintaining the nature of the burrow network. In this study, three different burrow permeability values (1, 10, and 100 mD) were tested while maintaining the matrix permeability constant (0.1 mD). CO₂ injection simulations were performed using a numerical reservoir simulator, testing the three different scenarios: 1. high burrow permeability (100 mD), 2. medium burrow permeability (10 mD), and 3. low burrow permeability (1 mD).
The results demonstrate that the permeability contrast between the Thalassinoides burrow network and its surrounding matrix has a major control on CO₂ plume diffusion. A high permeability contrasts promote rapid injectivity while leading to a strong channelized flow confined to the burrow networks with poor CO₂ penetration into the matrix. A medium permeability contrast allows for a balanced CO₂ flow and efficient CO2 diffusion into the matrix. A low permeability contrast results in a more homogeneous CO₂ diffusion and an improved storage efficiency due to high penetration into the rock matrix.
These findings highlight the necessity of incorporating bioturbation-induced heterogeneity into GCS assessments. Explicitly accounting for ichnological assemblages can improve simulation accuracy, optimize injection strategies, and support more robust site selection for GCS projects. Similar refinements can be applied in Saudi Arabia and in analogous sedimentary settings worldwide.
How to cite: Colmenares, J., Eltom, H., and Ayranci, K.: CO2 Injection Simulation in Thalassinoides-bearing rocks: Implications for geological carbon sequestration, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9598, https://doi.org/10.5194/egusphere-egu26-9598, 2026.
CO2 storage could benefit hard-to-abate industries that face significant challenges in reducing their greenhouse gas emissions. To mitigate global warming without affecting industrial production, there is an urgent need to develop large-scale CO2 storage projects, which, among other factors, depend on reliable forecasts of subsurface CO2 behavior.
Knowledge of the fluid flow in depleted or producing reservoirs provides an important background, but cannot be directly applied to forecast the shape and position of the injected CO2 mass. In deep saline aquifers, one of the main promising storage sites with relatively high storage capacity, injecting CO2 (a lower-viscose fluid) into brine (a higher-viscose fluid) is sensitive to reservoir heterogeneities. Due to this viscosity contrast, CO2 retention is affected by permeability variations of less than one order of magnitude. Consequently, any permeability contrast within the reservoir can favor structural-stratigraphic and residual trapping mechanisms. Moreover, injection affects not only flow behavior near the wellbore but also geomechanical responses over larger areas, requiring a multiscale approach to represent the deep saline aquifer's heterogeneity. Given these particularities and considering that CO2 project forecasts rely primarily on reservoir modeling, questions commonly arise about which technique to use for constructing 3D geological models.
We present a comparative analysis of two stochastic methods for modeling reservoir properties: (1) Sequential Indicator Simulation/Sequential Gaussian Simulation (SIS/SGS) and (2) Multiple-Point Statistics (MPS). Both methods were used to build geological models of the Jureia-Ponta Aguda Formation, a deep saline aquifer in the offshore Santos Basin, Brazil. The formation is a 2,000 m-thick reservoir composed of fluvio-deltaic to shallow marine sediments occurring at depths below 800 m. Based on a dataset of 40 wells and 2D/3D seismic data, an area of 4,000 km2 was modeled at a 1:100,000 mapping scale, with representative geologic elements having a minimum dimension of 1 km. The comparison focuses on the dynamic response of each model under CO2 injection. Key inputs for decision-making in a storage project, including the well injectivity and the area affected by pressure variations and CO2 saturation, are quantified to assess the impact of the reservoir modeling technique on CO2 subsurface behavior.
The SIS/SGS model exhibits a more continuous distribution of reservoir properties, whereas the MPS model better captures the geometry of geological elements, resulting in a more discretized spatial distribution of facies, porosity, and permeability. In a direct comparison, the two models produce different fluid-flow behavior, and the MPS technique appears to be the best choice at first glance because it more accurately represents the inputs. However, the sparse subsurface dataset carries a high level of uncertainty, and different geologic scenarios have a greater impact on the CO2 plume geometry, the pressure front size, and well injectivity than the modeling methodology itself.
As the choice of modeling algorithm becomes less critical in the uncertainty process, addressing CO2 subsurface behavior should focus on the range of possible geologic scenarios. The structural-stratigraphic-sedimentologic framework, along with capillary pressure and drainage/inhibition permeability curves, is the key factor in reservoir models that support decision-making for a storage project.
How to cite: Lira, F., Erdtmann, M., Hajibeygi, H., Martinius, A., and Geiger, S.: The impact of reservoir modelling techniques on CO2 storage, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13931, https://doi.org/10.5194/egusphere-egu26-13931, 2026.
Subsurface reservoirs are used for various applications, driving the energy transition towards zero-carbon energy. Making optimal use of subsurface reservoirs is a great challenge for society these days. Geological CO2 Sequestration (GCS) can play a significant role in reducing anthropogenic CO2 emissions while allowing society to slowly phase out traditional energy sources. An accurate representation of GCS requires computationally expensive modelling of complex physical phenomena at various scales. These models involve many uncertain reservoir parameters and imprecise input information, demanding the generation of representative ensembles of models, thus making the computational cost even higher. In this talk, I will share our experience in the simulation of GCS applications using high-fidelity physics-based computational models. I will present parametrization technology, which allows us to develop a unified modelling framework with multiphase thermal-compositional formulation capable of covering a wide variety of GCS challenges. Several reservoir engineering applications relevant to the CO2 sequestration portfolio will be shown as well.
How to cite: Voskov, D.: Modeling of reservoir applications relevant to the CO2 sequestration portfolio, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12605, https://doi.org/10.5194/egusphere-egu26-12605, 2026.
Posters on site: Wed, 6 May, 16:15–18:00 | Hall X4
The growing urgency of global atmospheric decarbonization and reaching net zero carbon emission underscores the importance of shifting our dependency from fossil fuels to relatively cleaner renewable resources. Though India’s average annual CO2 emission growth from fossil fuels decreased from 6.4 % (2005-2014) to 3.6% (2015-2024) in recent times, India still stands as the third largest emitter of carbon at 3.2 billion tonnes per year (2024), reflecting the importance of expanding its renewable energy portfolio. Among the renewable energy resources, geothermal energy, which uses the natural heat stored inside the Earth, holds considerable potential for India, particularly along the Himalayan belt. The Ladakh Himalaya, which forms the northwestern sector of the India-Eurasia collision zone, is the focus of the current study. Ladakh is considered a highly promising geothermal province due to its active tectonics, crustal deformation, and widespread occurrences of numerous hot springs, like Puga, Chumathang, Panamik, Changlung, and Demchok where temperature reaches up to 90 oC. However, despite its strong geothermal potential, the region has limited subsurface thermal characterization, which poses challenges for effective resource assessment and sustainable exploitation.
This study presents an integrated approach that combines field and laboratory based geophysical datasets and geodynamic context to improve understanding of subsurface thermal structure for sustainable geothermal assessment. A comprehensive thermophysical dataset including thermal conductivity, porosity, density, specific heat capacity and radiogenic heat production has been generated in the present study from representative rock types. These include granitoid, sandstone, limestone, ophiolite, phyllite, schist and gneiss belonging to different geological formations such as Higher Himalayan Crystalline, Zanskar Formation, Lamayuru Formation, Indus formation, Ladakh Batholith and Khardung-Shyok Formation. Temperature measurements are carried out in boreholes for determining geothermal gradient, which is essential for calculating surface heat flow. These parameters serve as crucial inputs and boundary conditions, along with the available geological and geophysical information for constructing numerical thermal model and the quantification of crustal heat generation.
The thermal modelling simulates temperature distribution with depth and enhances the understanding of the lithospheric thermal structure of Ladakh, helping to delineate prospective zones for geothermal energy exploration. This work demonstrates the value of multi parameter geomodelling to transform sparse field observations into a robust geothermal assessment. The outcomes significantly contribute to future clean energy strategies, support a promising pathway towards global energy transition to achieve decarbonization goals and provide a framework for regional energy security of a remote mountainous region like Ladakh.
Keywords: Geothermal energy; Heat flow; Thermal conductivity; Radiogenic heat production; Thermal modelling; Ladakh Himalaya.
How to cite: Dutta, A., Ray, L., Chopra, N., Sidagam, E. R., Prajapati, S. K., and Podugu, N.: Thermophysical Characterization of Major Rock Types and Subsurface Thermal Modeling in the Ladakh Himalaya, India: Implications for Geothermal Energy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-939, https://doi.org/10.5194/egusphere-egu26-939, 2026.
A characterization regarding the mechanical response of subsurface reservoirs is of increasing interest for energy-related applications, including geothermal energy production and storage of georesources and waste. Modelling the dynamic response of geological formations to fluid injection often relies on fully coupled thermo-hydro-mechanical (THM) models, which provide a high-fidelity representation of the governing physical processes. These models support operational and design decisions under significant geological and parametric uncertainties. However, their high computational cost severely limits their applicability in large-scale statistical analysis and thus limiting the potential to account for these uncertainties.
Still, understanding how uncertainties in reservoir and operational parameters influence application-relevant outcomes is essential for stimulation design and risk mitigation. Global sensitivity analysis offers a quantitative framework to identify the controls on selected quantities of interest (QoIs). The choice of a QoI is inherently problem-dependent and reflects the specific operational objective or risk-related question being addressed, making it a central element in the interpretation of model results.
To overcome the computational demands of full-order THM simulations, we employ non-intrusive reduced-order modeling techniques to efficiently and accurately approximate the transient reservoir response. Projection-based model reduction methods target accurate, physics-based response characterization, resulting in interpretable, physics-consistent, and scalable surrogate models. We train surrogate models using solutions of the coupled THM equations. These surrogates are then used to perform global sensitivity analyses for different choices of QoIs. Finally, we demonstrate the proposed workflow through an application to the Groß Schönebeck geothermal field, featuring a multistage injection scenario, that provides a basis for future analyses targeting induced seismicity.
How to cite: Hayek Valencia, J. N., Cacace, M., and Degen, D.: Global sensitivity analysis of multistage injection in geothermal reservoirs using surrogate models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5316, https://doi.org/10.5194/egusphere-egu26-5316, 2026.
Subsurface temperature is a critical parameter when assessing the geothermal energy potential of a region. Regardless of how favorable an aquifer may be in terms of porosity, permeability, or depth, geothermal exploitation is not economically viable if temperatures are insufficient. As part of the ThermoGIS project, we produce nationwide estimates of geothermal energy potential for the entire Netherlands, which requires a high-resolution statistically robust model of subsurface temperature.
This research adopts a combined physics-based and data-driven approach to estimate the three-dimensional temperature field beneath the Netherlands down to a depth of 10 km. The model is discretized on a regular mesh with a horizontal resolution of 1 × 1 km in longitude and latitude, and a variable vertical resolution that averages between 10 and 30 m in the upper 5 km, and 200 m down to 10 km. This represents a five-fold increase in resolution compared to the most recent published temperature model of the Netherlands (Bekesi et al., 2020).
We first construct a three-dimensional lithological model of the Netherlands comprising 101 distinct litho-stratigraphic layers. Based on expert stratigraphic knowledge, lithological compositions are assigned to each layer. These layers are then populated with thermal conductivity and radiogenic heat production values derived from standard reference data (Hantschel and Kauerauf, 2009), yielding an initial prior model.
Using these prior conductivity and radiogenic heat fields, we solve the three-dimensional steady-state heat diffusion equation using centered finite differences. The model parameters are subsequently updated using Ensemble Smoother Multiple Data Assimilation (Emerick and Reynolds, 2013) to match a high-quality dataset of mostly corrected bottom-hole temperature measurements, and DST and geothermal production temperatures.
A key innovation distinguishing this model from previous temperature models of the Netherlands (Bonte et al., 2012; Bekesi et al., 2020) is the use of efficient numerical solvers combined with a more accurate and detailed lithological model, enabling an order-of-magnitude increase in spatial resolution. Our model was written entirely in python and the code will be made open source upon publication.
How to cite: Brett, H., Veldkamp, H., van Wees, J.-D., Limberger, J., and Thieulot, C.: An updated 3D Temperature model of the Netherlands, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10312, https://doi.org/10.5194/egusphere-egu26-10312, 2026.
Subsurface reservoir models typically use grid cells of tens to hundreds of meters in the horizontal directions, and several to tens of meters in the vertical direction. The effects of small-scale heterogeneities (below grid cell size) on fluid flow are often ignored or assumed to have an effective ’upscaled’ average in reservoir simulations. The used grid-cell size might not correspond to the scale at which these fluid-rock interactions can be accurately averaged. The smaller scale geological heterogeneities (from pore scale up to meter scale and tens of meters, or outcrop scale) and fluid behavior (e.g., capillary vs gravitationally dominant flow regimes) play an important role in CO2 sequestration and hydrogen storage. Ideally, the grid cell size of a reservoir model is determined by the representative elementary volume (REV) which accurately captures the net effect of smaller scale structures on a certain fluid for a certain (representative) volume. We perform a rigorous analysis on the influence of sub-meter scale heterogeneities on REV scales for single-phase flow. Using generated 3D models of 2x2x2m size we vary the bed thickness, dip angle, azimuth angle, and permeability distribution (values and fining or coarsening upwards), and determining their impact on the REV. To do this a sensitivity analysis is performed on the parameters and the three calculated dimensions of the REV.
How to cite: Steinvoort, J., Daniilidis, A., Abels, H., and Geiger, S.: Model-Parameter Sensitivity Analysis on Representative Elementary Volumes for Small-Scale Geological Heterogeneities., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4835, https://doi.org/10.5194/egusphere-egu26-4835, 2026.
Tidal reservoirs exhibit complex sedimentary architectures that remain a major challenge to capture in subsurface geological models. In particular, predicting the influence of sedimentary heterogeneities on fluid-flow behavior across multiple spatial scales still remains a major challenge. Simplified or generic modeling approaches often fail to represent the multiscale elements that are characteristic to tidal deposits, resulting in uncertainty in flow predictions and reservoir performance. This limitation is especially critical for geo-energy applications, where reliable forecasts are required to design efficient injection, storage, or drainage strategies.
This study adopts a scale-aware approach based on the concept of the Representative Elementary Volume (REV), defined as the minimum volume over which a heterogeneous property, such as permeability, can be considered effectively homogeneous. Identification of the REV at relevant modelling scales enables consistent upscaling of petrophysical properties and reduces uncertainty in geological models and flow simulations. As such, REV-based approaches are essential for building robust geo-models that capture key geological heterogeneities and support reliable performance forecasting for subsurface energy applications.
The methodology is demonstrated using reservoir rocks from the Viking Graben (Norway), comprising the Middle Jurassic Brent Group, where the reservoir interval represents a highly heterogeneous tidal depositional system. Detailed sedimentological core logging of selected intervals with pronounced heterogeneity is used to identify the principal lithofacies within the reservoir. For each lithofacies, sketch-based geological models are constructed to capture characteristic heterogeneity, such as key architectural elements, facies proportions, and spatial relationships. This sketch-based approach enables transparent and concept-driven representation of the geological complexity for each lithofacies. For each lithofacies model, the REV is systematically calculated for single-phase flow, resulting in a lithofacies-specific REV. Together, these models form a REV library in which each lithofacies is associated with a representative scale that captures its characteristic heterogeneity. This library provides a transferable framework that can be applied to similar reservoirs worldwide, supporting improved upscaling and flow modeling in data-limited subsurface energy applications.
How to cite: Weert, A., Geiger, S., and Martinius, A. W.: A sketch-based REV library for tidal lithofacies, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11468, https://doi.org/10.5194/egusphere-egu26-11468, 2026.
Geothermal reservoir modeling often focuses on fractured or single-layer systems, even though multilayer porous aquifers offer an additional opportunity for geothermal energy extraction. This study examines how specific modeling assumptions influence the flow field and thermal evolution in stratified geothermal systems. It focuses on a representative multilayer aquifer of the Bückeberg Formation in the North German Basin. The targeted interval contains stacked sandstone units separated by claystone between depths of 1200 and 1400 m with reservoir temperatures around 70 °C and injection rates of several tens of litres per second.
A three dimensional numerical model is developed in the open source software OpenGeoSys to evaluate groundwater flow and heat transport in this layered system. As the fluid viscosity is temperature-dependent, the resulting flow field evolves over time. To investigate the associated water distribution around the well, injection and extraction are represented using three numerical approaches. First, a pipe based implementation is used to explicitly model the flow through the wells so that the distribution of injected and produced water between the sandstone layers is not prescribed but governed by the geological and hydraulic properties of the multilayer aquifer. Further, two imposed injection concepts are applied for comparison: a line based source term and a source term defined on a cylindrical borehole surface. Moreover, the approximation of constant viscosity is assessed by comparison with simulations using both constant and temperature-dependent viscosity for the three well implementations.
Preliminary results show that temperature-dependent viscosity noticeably alters the flow field and affects the evolution of production temperature. These tendencies confirm the relevance of viscosity formulation when analysing thermal behaviour in multilayer geothermal systems. The ongoing comparison of injection and extraction well approaches extends these findings by including the influence of inflow and outflow patterns on the flow field near the wells, where flow paths in multilayer aquifers are inherently more complex. This highlights the importance of choosing appropriate inflow and outflow conditions for modeling the thermo-hydraulic response of stratified reservoirs.
How to cite: Niehage, K., Graf, T., and Neuweiler, I.: Effects of deep geothermal injection and extraction well modeling approaches on flow and heat transport in multilayer aquifers, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10942, https://doi.org/10.5194/egusphere-egu26-10942, 2026.
How to cite: Banerjee, I., Crinière, A., Sánchez-León, E., and Zosseder, K.: Comparison of modelling practices for groundwater flow and heat transport in heterogeneous deep geothermal systems, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11230, https://doi.org/10.5194/egusphere-egu26-11230, 2026.
The demand for energy transition is growing rapidly as climate change accelerates, and concerns about energy security have also increased due to heightened geopolitical tensions. Geothermal energy is a reliable and sustainable solution that produces both heat and electricity with low greenhouse gas emissions and reduces dependence on fossil fuels. The Upper Rhine Graben (URG) is widely recognised as one of Europe’s most promising regions for geothermal development. However, the risk of induced seismicity associated with fluid injection and production processes poses a significant challenge to public acceptance and project viability. Therefore, understanding the crustal stress state is crucial to ensuring a safe and efficient operation of a geothermal plant.
This study employed a three-dimensional (3D) geomechanical-numerical modelling approach to predict the local in situ stress distribution and fracture networks in a faulted reservoir located near Karlsruhe, Baden-Württemberg, Germany. The structural model showing the subsurface geometry was built using a horizon and fault interpretation derived from 3D seismic data provided by an industry partner. This structural model was discretised and parameterised utilizing Visage (Petrel Reservoir Geomechanics software from SLB). Rock mechanical properties, including modulus of elasticity, Poisson’s ratio, bulk density, Biot coefficient, tensile strength, unconfined compressive strength, cohesion, friction angle, and hydraulic properties, were assigned to each formation. These properties were derived from samples taken in outcrops with Muschelkalk, Buntsandstein, Rotliegend, and crystalline basement rocks for laboratory testing.
The boundary conditions of the Finite Element model were calibrated using the minimum horizontal stress magnitudes measured in a nearby well, and the orientation of the maximum horizontal stress obtained from the World Stress Map database. After validation, modelling results provide a prognosis of the complete 3D stress tensor in the entire model domain. Among others, this can be used for well path planning and optimal well placement. To evaluate the probability of reactivation of the faults under the modelled stress conditions, a slip tendency analysis was performed. In particular, faults within the Muschelkalk formation exhibited a higher slip tendency compared to other target units, indicating zones of elevated seismic risk. These findings provide critical insights for geothermal reservoir development and contribute to risk mitigation strategies aimed at minimising induced seismicity.
How to cite: Wittayasettakul, S., Mebrahtu, T., and Henk, A.: Geomechanical modelling of tectonic stresses in deep geothermal reservoirs of the Upper Rhine Graben, Germany, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22110, https://doi.org/10.5194/egusphere-egu26-22110, 2026.
The creation of reliable structural geological models is often a crucial component of geoscientific workflows. Challenges emerge not only from the availability of data and model construction but also regarding the knowledge and accessibility of software, coding abilities (particularly for open-source tools), and geological expertise. These obstacles hinder the exploration, evaluation, and comparison of diverse modeling methods, often leading to highly customized workflows for specific scenarios that are labor-intensive to create and hard to reuse in other settings.
To mitigate these issues, we present a workbench for digital geosystems that employs a component-and-connector software architecture alongside both textual and graphical domain-specific languages (DSLs) to establish a modular framework. Within this framework, we define fixed interface formats for each workflow step, allowing components responsible for specific tasks to be interchangeable. Structural modeling serves as the initial step in these workflows, which also encompass 3D mesh generation, simulation, and visualization, thereby representing a typical geoscientific workflow.
Within this design, multiple components can be integrated for each workflow step, facilitating straightforward method comparison. Additionally, the DSLs enhance usability for users who may not have extensive coding experience.
We will showcase the software architecture and DSL system through a series of simple models with an emphasis on structural geological modeling and comparisons among multiple implicit modeling methods. A cloud-based version of the graphical DSL will be provided to test the workbench with a curated set of input datasets.
How to cite: von Harten, J., Lüpges, A., Baes, M., Niederau, J., Wellmann, F., Rumpe, B., and Cacace, M.: Integrating and comparing structural modeling methods within a digital workbench, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10020, https://doi.org/10.5194/egusphere-egu26-10020, 2026.
Reliable characterization of subsurface geology is a key prerequisite for reducing uncertainty in geoscientific studies and for lowering costs and risks in geothermal drilling. In this study, we apply GeoBUS (Geological modeling by Bayesian Updating of Scalar fields), a probabilistic structural geological modeling workflow, to a synthetic benchmark that represents the characteristic succession of geological units in the canton of Thurgau, Switzerland. As an initial test case, we construct a representative one-dimensional geological model based on available legacy data.
In the first step, a geological prior model is created by introducing epistemic structural uncertainty, perturbing the depths of geological interface points within predefined bounds. For each realization, implicit geological modeling is performed using radial basis function interpolation, resulting in an ensemble of scalar fields from which geological interfaces are represented as isolines of common scalar values.
In a second step, we calculate synthetic surface-wave dispersion curves based on the geological models using representative literature values and considering uncertainties and model variations. The dispersion curves are then inverted for subsurface velocity profiles to estimate biases and resolution limits of inversion schemes compared to the ground truth. We will test an ensemble of plausible subsurface models that is consistent with the dispersion data rather than as a single deterministic solution.
In the third step, literature-based seismic velocities are assigned to the geological units in the prior ensemble of geological model to enable comparison with the seismic data inversion results. An ensemble-based Bayesian update step is then applied to the scalar field ensemble, resulting in a posterior ensemble that is consistent with the assimilated seismic information. By evaluating each scalar field to derive geological interfaces, we obtain a posterior ensemble of geological models that consistently integrates information from both geological modeling and geophysical inversion and allows structural uncertainty to be quantified.
Using a synthetic example, we assess the performance of the GeoBUS workflow with respect to (1) the structural uncertainty in the geological model and (2) the value of information contained in the seismic data, including the influence of measurement sensitivity and prior constraints that may lead to updates in model regions weakly constrained by the assimilated seismic data. Validating the approach in this controlled one-dimensional setting provides an essential benchmark before extending the study to higher-dimensional and more complex geological settings.
This work was funded by the European Union’s Horizon Europe Framework Programme for Research and Innovation under the GeoHEAT project (Grant Agreement No. 101147571)
How to cite: Satiroglu, E., Bobe, C., Finger, C., Muñoz-Burbano, F., and Wellmann, F.: Probabilistic Evaluation of Structural Uncertainty in a Synthetic Geological Benchmark Using GeoBUS, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10778, https://doi.org/10.5194/egusphere-egu26-10778, 2026.
Uncertainty quantification is a key component of geological modeling for mining, exploration, and civil engineering. While uncertainty estimation workflows for implicit structural modeling and inversion are well established for fixed parameter spaces, they require the number of model parameters to be defined a priori. This assumption is often unjustified and subjected to bias, as the number of geological layers or phases is commonly unknown, leading to models that are either overly complex or overly simplistic. Trans-dimensional Markov chain Monte Carlo methods provide a powerful framework for model selection by favoring parsimonious representations in settings with high uncertainty. In particular, Reversible Jump Markov chain Monte Carlo (RJ-MCMC) has recently gained attention for solving inverse problems with variable dimensionality. In this study, we investigate the applicability of RJ-MCMC to parameters governing geological interpolation functions. By automatically inferring the optimal number of parameters, the method reduces reliance on subjective user choices. We generate synthetic geophysical data from simple structural models to establish ground truth and perform geophysical inversion (gravity) by updating ensembles of prior structural models. This probabilistic framework enables likelihood-based model evaluation and supports further inference as new data become available. Generated candidates can be grouped to identify model archetypes that fit the data, while parsimony is maintained.
How to cite: Chudalla, N., Nathan, D., and Wellmann, F.: Adaptive Geological Model Parameterization Using Reversible Jump MCMC, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18061, https://doi.org/10.5194/egusphere-egu26-18061, 2026.
Structure-based inversion offers a geologically informed alternative to conventional voxel-based approaches by explicitly representing subsurface interfaces. This enables inversion results to be interpreted in terms of meaningful geological structures rather than smoothed property distributions. Extending this concept to a joint inversion framework further allows multiple geophysical data sets to update a single shared geological model, exploiting complementary sensitivities to better constrain subsurface structure and reduce interpretational ambiguity.
Here, we investigate the effectiveness of structure-based joint inversion for imaging tectonic features in a faulted near-surface environment by jointly inverting electrical resistivity tomography (ERT) and travel-time tomography data. The inversion framework is built around an implicit geomodel in which fault-related interface points are included directly in the model vector (Balza Morales et al., 2025). Both geophysical methods contribute to a single objective function, enabling tectonic interface geometry and associated physical property distributions to evolve consistently during inversion.
The workflow is evaluated using (i) synthetic experiments in a crosshole setting and (ii) field data acquired in the Southern Erft block, a structurally complex tectonic setting in the Lower Rhein Embayment (Menzel et al., 2024). This work provides two contrasting environments: one in which ERT and SRT exhibit strongly complementary sensitivities, leading to improved interface recovery and increased stability of fault geometry updates in joint inversion; and a second in which limited coverage from one method restricts the degree of complementarity, so that while joint inversion can still be performed, it offers only minor improvements over the single-method structure-based inversion. The field case is used to assess how structure-based inversion improves fault interpretation relative to voxel-based inversion through explicit parameterization of geological interfaces derived from an initial conceptual geomodel, and to test whether joint inversion produces more robust and consistent updates to the shared geomodel under realistic acquisition conditions and noise.
By systematically contrasting voxel-based, single-method structure-based, and joint structure-based inversions, the analysis examines how increasing levels of geological coupling influence the stability, interpretability, and geological plausibility of inferred fault architecture while maintaining consistency with an optimized (joint) data misfit. While demonstrated here using ERT and seismic travel-times the proposed evaluation strategy and inversion framework are transferable to other geophysical methods and subsurface applications where structural complexity limits conventional interpretation.
References:
Menzel, N. and Klitzsch, N. and Altenbockum, M. and Müller, L. and Wagner, F.M. (2024): Prospection of faults on the Southern Erftscholle (Germany) with individually and jointly inverted refraction seismics and electrical resistivity tomography. Journal of Applied Geophysics. https://doi.org/10.1016/j.jappgeo.2024.105549
Balza-Morales, A., Förderer, A., Wellmann, F., Maurer, H., & Wagner, F. M. (2025). Structure-based geophysical inversion using implicit geological models. Geophysical Journal International, https://doi.org/10.1093/gji/ggaf445
How to cite: Balza Morales, A., Menzel, N., Maurer, H., and Wagner, F. M.: Joint structure-based inversion of electrical resistivity and seismic travel-time data for fault characterization., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21192, https://doi.org/10.5194/egusphere-egu26-21192, 2026.
Fault-controlled karstic systems hosted in carbonates strongly influence groundwater flow and morphological evolution, including the development of associated basins. Characterizing the geometry of the karst system and related sedimentary basins may contribute to modelling the groundwater system, determining hazards related to collapse, and understanding the relationship with associated geological structures. However, subsurface imaging and geophysical characterization may be challenging due to the depth of the hosting rocks, the presence of saturated layers, and the dimensions of the area affected by the systems. In the case of large and morphologically complex areas, 3D deep full-waveform geo-electrical surveys using wireless devices help to overcome the limitations of conventional small-scale electrical surveys. In particular, the ability to generate integrated Electrical Resistivity Tomography (ERT) and Induced Polarization (IP) models has proven effective in imaging karst features, allowing the detection of cavities and structural complexity.
The Piani di Monte Lago basin, in the Umbria–Marche sector of the Central Apennines, is an intramountain karst–tectonic basin characterized by the development of seasonal lakes and presents a poorly understood karst system with fast water discharge. Its evolution has been shaped by karst processes, with poljes and active ponor drainage developing under the combined influence of tectonic deformation, lithological contrasts, and Pleistocene geomorphological changes.
This work aims to characterize fracture zones and karst features through a combined application of deep ERT and Induced Polarization (IP). The approach integrates acquisition, processing, modelling, and geological interpretation to achieve a more accurate subsurface image in this structurally complex setting.
The survey was conducted using the FullWaver® wireless system (IRIS Instruments), deploying 15 dual-channel receivers and a 5-kW transmitter across an 800 × 800 m area. Flexible quadrupole configurations and dipoles up to 1000 m enabled investigation depths of about 200 m, while GPS synchronization ensured precise time-domain measurements of resistivity, IP, and self-potential, with a timing accuracy of 250 µs.
The generated models, constrained by geological and topographic data, reveal sharp resistivity contrasts between the carbonate substratum and overlying lacustrine deposits. IP variations highlight fracture zones and possible buried structures that control hydraulic connectivity. These results provide new insights into the structural influence on karst development and clarify subsurface drainage patterns. By integrating geophysical inversion with geological constraints, this approach reduces uncertainty and refines 3D models in structurally complex carbonate settings. The findings have direct implications for groundwater resource management, environmental assessment, and land-use planning in tectonically active regions.
How to cite: Zambrano, M., Lucci, N., Ramos, S., Baena, J., Arellano, H., Eriza, J., Arias, A., Mateus, Y., Jablonska, D., and Torres, D.: Imaging fault-controlled karst systems through integrated Resistivity and IP inversion in the Umbria-Marche Apennines, Piani di Montelago (Sefro, MC), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21479, https://doi.org/10.5194/egusphere-egu26-21479, 2026.
Predictions of the three-dimensional in-situ stress state are crucial for the site selection process for deep geological repositories for nuclear waste and their long-term safety. However, the geological configurations relevant to a potential siting region, such as fault offsets, salt intrusions, and intersecting sedimentary units, create increasing structural complexities. These complexities, especially in the form of intersecting lithological boundaries and interfaces, present significant challenges for the discretization using the Finite Element Method (FEM). In the FEM, lithological boundaries and interfaces are commonly modeled using spline-based representations, which are then geometrically approximated by finite elements. This process introduces an additional layer of geometrical approximations that can lead to discretization errors, mesh distortions, and the need for repeated geometry regeneration when testing different model scenarios.
This study investigates whether the Isogeometric Analysis (IGA) as a discretization method can enhance and facilitate geometric fidelity and contribute to an automated modeling workflow. Unlike FEM, IGA employs the same spline basis functions (e.g., NURBS) for both the geometrical representation and numerical approximation. This direct application of splines eliminates the need for a geometry-to-mesh approximation step, allowing for an exact representation of both lithological boundaries and structural features, such as faults. The workflow for IGA differs from traditional FEM primarily in the preprocessing, solver implementation, and postprocessing stages: geometry is handled directly through control points, spline basis functions replace conventional shape functions, and the numerical solution is stored at these control points before being mapped back to the physical domain. While IGA does not necessitate a separate meshing step, refining the spline representation may still be required.
To evaluate this approach, we begin with a three-layer benchmark model previously used in sensitivity analyses and introduce a fault that offsets the lithological layers. IGA is utilized to compute stress, strain, and displacement fields, with its performance compared to that of the FEM, focusing on the impact of the geometrical approximation. The results aim to illustrate how exact geometrical representation and spline refinement influence stress predictions, particularly in areas where faults or salt contacts create sharp geometrical variations.
This work represents a significant advancement toward a more automated and reliable geomechanical modeling workflow. By reducing the need for manual geometrical regeneration and directly integrating spline-based representations into the analysis, IGA can streamline model scenario exploration and support more consistent gemechanical modeling for repository-scale studies.
How to cite: Achour, R., Degen, D., Heidbach, O., Reiter, K., Cacace, M., and Wellman, F.: Automatization of Geomechanical Modeling for Complex Geological Structures Using Isogeometric Analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10705, https://doi.org/10.5194/egusphere-egu26-10705, 2026.
Implicit neural representations (INR) have emerged as a flexible tool for implicit modelling of subsurface structures. Works such as GeoINR (Hillier et al., 2023), and curlew (Kamath et al., 2025) have laid the foundation for building increasingly complex geological models with neural fields. Linking these modelling approaches to geophysical forward models would enable better constraints on the 3D structural geological models (SGM) widely used to predict subsurface geometry for mining, engineering and energy applications.
Specifically, within curlew, geological structures are defined as distinct neural fields. Each field can “learn” arbitrary geometries that fit the available constraints, including geological and petrophysical data. The different fields are then chained together with offsetting and overprinting relationships to derive geological complexity. In this contribution, we combine the spatio-temporal model building capabilities of curlew with a highly optimized FFT-quadrature based gravity forward model (Wang et al., 2023) to generate gravity data from implicit fields. The entire framework is built within PyTorch, which allows us to update SGMs of subsurface geometry populated with property distributions through inversions of gravity datasets. Our preliminary results show that the ability to incorporate several different kinds of losses, as well as constrain both the geometry and property, dramatically improve the inversion results compared to standard inversion techniques.
References:
Hillier, M., Wellmann, F., de Kemp, E. A., Brodaric, B., Schetselaar, E., and Bédard, K.: GeoINR 1.0: an implicit neural network approach to three-dimensional geological modelling, Geosci. Model Dev., 16, 6987–7012, https://doi.org/10.5194/gmd-16-6987-2023, 2023.
Kamath, A., Thiele, S., Moulard, M., Grose, L., Tolosana-Delgado, R., Hillier, M., & Gloaguen, R. (2025). Curlew 1.0: Spatio-temporal implicit geological modelling with neural fields in python. doi:10.31223/x5kx81
Wang, X., Liu, J., Li, J. et al. Fast 3D gravity and magnetic modelling using midpoint quadrature and 2D FFT. Sci Rep 13, 9304 (2023). https://doi.org/10.1038/s41598-023-36525-2.
How to cite: Kamath, A., Thiele, S., Grose, L., and Gloaguen, R.: (Auto) Differentiating geophysics: gravity modelling with spatio-temporal neural fields, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11917, https://doi.org/10.5194/egusphere-egu26-11917, 2026.
Three-dimensional structural geological models are widely used to describe subsurface geometry, but their quality strongly depends on how geological complexity and data uncertainty are handled during model construction. In this study, we present a stepwise workflow for building an uncertainty-aware 3D structural model of the Devonian subsurface in the Eastern Eifel region (Germany) using universal co-kriging implemented in the open-source modelling software GemPy.
The modelling approach follows a gradual and controlled strategy. The model construction starts with a simplified stratigraphic framework based on surface geological data. Major Devonian units are added sequentially, followed by the introduction of a main fault structure. Although the modelling steps are applied sequentially, the model is always fully constructed from the input data and can therefore be completely reproduced on this basis. This enables transparent model building and supports future integration into forward uncertainty quantification and sensitivity analysis workflows. This stepwise procedure allows continuous validation of the model and helps to isolate the effect of individual modelling choices, such as unit simplification, fault geometry, and orientation constraints.
Fault modelling is based on a limited number of geometrically constrained fault points and orientations. This setup reproduces a meaningful displacement across the fault while keeping layer surfaces smooth and geologically plausible on both sides. The focus is not on producing a final deterministic model, but on creating a reproducible baseline model that can be extended towards uncertainty quantification.
The resulting 3D structural framework provides a robust basis for future integration of additional geological or geophysical data and for uncertainty analysis using stochastic or ensemble-based approaches. This work demonstrates how stepwise 3D structural modelling with GemPy supports geological consistency while preparing the model for uncertainty-aware subsurface analysis in structurally complex regions.
How to cite: Rahimi, A., Von Harten, J., Chudalla, N., and Wellmann, F.: Stepwise and uncertainty-aware 3D structural modelling of the Devonian subsurface in the Eastern Eifel region (Germany), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6945, https://doi.org/10.5194/egusphere-egu26-6945, 2026.
3D geological modelling is an essential tool for visualization, interpretation and retrodeformation of orogenic systems. Cross-sections restoration allows improved correlations between present-day structures and their paleogeographic evolution. Despite such advantages, accurately representing polydeformed orogenic belts remains challenging due to their structural complexity, large spatial extent, and multiphase tectonic histories.
To address these challenges, the swisstopo-funded swissAlps 3D (SA3D) project (2024–2030) aims to develop a consistent, large-scale 3D geological model of the major lithostratigraphic and structural boundaries of the Swiss Alpine region. As part of SA3D, the Helvetics 3D (HL3D) project focuses on the 3D reconstruction of the Swiss Helvetic domain. This work focuses specifically on the 3D geometry of the External Crystalline Massifs (ECM), from the Aar Massif to the Aiguilles Rouges–Mont Blanc massifs and the overlaying Helvetic nappe stack.
The Helvetic nappes overlying the ECM – composed of allochthonous Mesozoic marine limestones, marls, shales, and sandstones – experienced multiple deformation phases from ~39 Ma to the present (Burkhard, 1988). These events produced complex structural geometries, including recumbent and isoclinal folds and major thrust systems, making the Helvetic domain a key natural laboratory for verifying and reconstructing 3D geological structures.
The elaborated 3D model is based on 2D geological datasets, interpreted and re-validated cross-sections, borehole data, other published 3D geological models, and geophysical datasets. Except for the Aar Massif (Musso Piantelli et al. 2026), most of the Helvetic realm is characterized by limited subsurface constraints, with sparse borehole information, seismic profiles, and mainly geological cross-sections. To address this limitation, we developed a workflow combining explicit and implicit 3D modelling techniques, preserving the accuracy of detailed geological observations while increasing modelling efficiency.
Here, we present the preliminary results from the HL3D project, illustrating the 3D modelling of the ECMs in the westernmost Swiss Helvetic domain and their progressive retrodeformation from the present-day configuration back to the Miocene Grindelwald deformation phase (Handegg, Oberaar, and Pfaffenchopf phases; Herwegh et al., 2023). The 3D geometry of this area indicates the ECMs as elongated domes, with the long axes of the Aiguilles Rouges/Mont-Blanc and Aar/Gastern massifs respectively, plunging to the ENE and WSW. Their histories characterized by differential uplift, combined with an a-cylindrical and en-echelon arrangement of the basement units, affected the overlaying Helvetic nappe stack, and continues to control large-scale structures such as the Rawil depression.
The retrodeformation of this Miocene uplift shows that the Rawil depression formed in response to the inversion of a complex paleogeographic geometry of the former European passive margin during late-stage collision between the Adriatic and European plates. In this context, the ECMs and the overlying nappe stack experienced an uplift exceeding 8 km (Herwegh et al., 2023; Mercier et al., 2023).
In summary, this first HL3D model and its retrodeformation (i) provides new insights into the geometry and structural evolution of the western ECM, (ii) demonstrates the necessity and strength of 3D modelling in unravelling Alpine complex tectonic evolution, and (iii) reveals the initial extent and paleogeographic configuration of the central part of the European passive continental margin.
How to cite: Furlan, M., Herwegh, M., Berger, A., Schlunegger, F., Brisson, S., Diehl, T., Monti, R., Garefalakis, P., Drvoderic, S. P., Strasky, S., Kurmann, E., and Musso Piantelli, F.: 3D Modelling and Retrodeformation of the Western Helvetics: Insights from the swissAlps 3D Project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5432, https://doi.org/10.5194/egusphere-egu26-5432, 2026.
Over the past few decades, 3D geometrical modelling of the subsurface geological structure has become an essential tool in structural geology, significantly improving the visualization and interpretation of complex geological architectures. However, due to their structural complexity and multi-phase deformation histories, orogenic belts remain a major challenge for 3D modelling. The swissAlps3D project (2024-2030), led by swisstopo, aims to build a consistent, large-scale underground 3D geological model of the main geological and structural boundaries of the Swiss Alps and neighbouring regions. The project is subdivided into eight domain-specific sub-projects, including the Lepontine 3D project (LP3D). Here, we present the datasets, modelling strategy, and preliminary results of a 3D geological model of the eastern Lepontine dome.
The Lepontine dome consists predominantly of amphibolite-facies metamorphic rocks of late Eocene–Oligocene age and covers approximately 4,500 km² at the surface. At depth, it extends from the Simplon Fault Zone to the west and from the Forcola Line to the east. It is further bounded by the Periadriatic Lineament to the south and the Northern Steep Belt to the north. Structurally, it comprises a tectonic window within the Central Alps, which exposes the deepest Alpine nappes in the core of the bell-shaped architecture. It is made up predominantly of polycyclic basement gneisses, intruded by Permo-Carboniferous granitoids and locally bounded by a thin autochthonous or parautochthonous Mesozoic-Cenozoic sedimentary cover. Moreover, ductile isoclinal sheath folding, complex fold interference structures and a strongly debated tectono-stratigraphy (swisstopo 2024) make the Lepontine dome a challenging area for testing and refining its tectonic evolution and 3D modelling techniques capable of representing such units.
The presented 3D geological model, targets the major lithostratigraphic and structural units of the dome and is constructed by combining explicit and implicit 3D modelling approaches that interpolate a heterogeneous input dataset composed of newly acquired and published geological and structural maps, cross-sections, borehole data, existing 3D geological models, and available geophysical constraints from the literature. Preliminary modelling results depict from bottom to top the Leventina-Lucomagno, Simano, Adula, Piz Terri-Lunschania, Vals, Aul, Grava and Tomül nappes. These Lepontic and Lower Penninic units dip eastward at low angle beneath the overlying Pennine and Austroalpine nappe stack. Emphasis is placed on the 3D-architecture of the Lepontic-Penninic boundary, which is dissected by the Forcola extensional shear zone and is interpreted to form an eastward-developing ramp–flat geometry.
How to cite: De Guglielmi, M., Schenker, F., Schmalholz, S., Samsu, A., Gouffon, Y., and Musso Piantelli, F.: The eastern Lepontine dome nappe stack 3D model (swissAlps3D), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21718, https://doi.org/10.5194/egusphere-egu26-21718, 2026.
The Tauern Window in the Eastern Alps represents an ideal natural laboratory to investigate the three-dimensional architecture due to deformation processes related to continental collision and indentation. It exposes deeply subducted and subsequently exhumed European crust and remnants of the Alpine Tethys ocean beneath nappes derived from the Adriatic upper-plate, which frame the window (Austroalpine units). South of the Tauern Window, the Austroalpine Rieserferner and Drau–Möll blocks acted as indenting wedges during Oligocene-Miocene Adria–Europe convergence, while the Southern Alps form the leading edge of the Adriatic indenter. Lower-plate units exposed in the Tauern Window record a complex tectono-metamorphic evolution from Late Cretaceous to Miocene times, including accretion and subduction (D1-D2), exhumation and isoclinal folding of the Alpine Tethys ophiolites of the Penninic nappes (D3), formation of the crustal-scale Venediger duplex and nappe stacking of the European crust (D4), and overprint of earlier duplex structures by late-stage indentation, doming and lateral escape (D5). The final configuration of the Eastern and Western Tauern domes is then represented by several doubly plunging, upright antiforms deforming the D4 roof thrust of the Venediger duplex and the overlying units. In order to validate these nappe-scale structures, three-dimensional modelling was performed using the open-source software PZero (https://github.com/gecos-lab/PZero).
Geological 3D modelling in such metamorphic belts is hindered by both mathematical and geological complexities, including the interpolation of polydeformed surfaces and the definition of a consistent geological legend. Traditional explicit modelling approaches often generate inconsistencies, while fully time-aware implicit modelling is difficult to apply due to poorly constrained or heterogeneous ages of tectono-metamorphic boundaries. In this project we use the Structural Topology model (STm) approach that integrates conceptual geological interpretation with topological analysis of these units (i.e. volumes) and their boundaries (i.e. surfaces), systematically classifying (i) units as tectono-metamorphic, tectono-stratigraphic, or intrusive, and their boundaries (ii) as being conformal or discordant with the internal foliation of units, and (iii) according to crosscutting relationships that reveal the tectonic evolution. For all these model entities, a polarity is defined that, in addition to constrain stratigraphic and structural relationships, allows constraining the gradient of the scalar field used for implicit interpolation.
We believe that this strategy allows reconstructing topologically and geologically consistent 3D models despite polyphase deformation and reactivation of structures. The resulting 3D geomodel also provides new insights into the architecture and exhumation history of the Eastern Tauern Dome and offers a transferable framework for geological 3D modelling in complex orogenic belts.
How to cite: Favaro, S., Monti, R., Schuster, R., and Bistacchi, A.: 3D geomodel of the Eastern Tauern Dome (Tauern Window - Eastern Alps), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17661, https://doi.org/10.5194/egusphere-egu26-17661, 2026.
This study presents a high-resolution tectono-stratigraphic characterisation of the Palaeocene–Eocene Sylhet–Kopili interval on the Upper Assam Shelf using a 371.5 km² 3D post-stack time-migrated seismic volume integrated with two deep wells. Targeting the 2.5–3.5 s TWT interval, time-slices (3.048–3.200 s) combined with geometric (dip, curvature), discontinuity (similarity), and amplitude-based (reflection magnitude) attributes reveal two distinct structural domains. The north-western sector is dominated by syn-depositional normal faults forming tilted blocks and fault-bend folds, whereas the south-eastern sector is characterized by E–W to ENE–WSW sinistral strike-slip faults, S-shaped bends, relay splays, and NE–SW transfer faults that generate localized transtensional pull-apart structures. Dip and curvature attributes enhance fault-plane continuity and fault-block geometries, while reflection magnitude highlights deformation-controlled variations in reflector strength and stratigraphic contrast within the Kopili and Prang intervals. Similarity and dip-magnitude co-interpretation sharply delineate fault segmentation, linkage zones, and deformation intensity, significantly improving structural resolution compared to conventional seismic interpretation. Synthetic ties to wells W-01 and W-02 (correlation 0.6–0.72) validate key horizons and strengthen structural control. This work provides the first high-resolution 3D multi-attribute imaging of the Sylhet-Kopili interval, resolving deformation styles previously undocumented in the Upper Assam Shelf. The resulting framework delineates multiple trap geometries-tilted blocks, horsts, and fault-bend closures—and demonstrates the value of integrated dip, curvature, similarity, and magnitude attributes for improving tectono-stratigraphic interpretation in foreland basin settings.
How to cite: Kalyani, K. and Kirti Rao Gautam, P.: Tectono-Stratigraphic Characterization of the Paleocene–Eocene Interval in the Upper Assam Shelf Using Integrated 3D Seismic Attribute Analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-972, https://doi.org/10.5194/egusphere-egu26-972, 2026.
Implicit structural geological modelling methods can integrate various geological constraints to rapidly constrain subsurface geometries, and are widely used for resource evaluation, geotechnical and hazard assessment, and reservoir characterisation. However, established approaches based on conventional interpolators (e.g., radial basis functions or co-kriging) often suffer from interpolation artefacts (“bubbles”) and can struggle to incorporate common constraints like stratigraphic relationships (inequalities) and geophysics data.
In this contribution we present an update on progress developing curlew, an open-source python package for structural geological modelling using neural fields (https://github.com/samthiele/curlew/). This flexible modelling framework incorporates various local constraints (value, gradient, orientation and (in)equalities) and tailored global loss functions to ensure data-consistent and geologically realistic predictions. Progressive Random Fourier Feature encodings are adopted as a tool for improving the convergence and reliability of neural fields, and drop-out based approaches to uncertainty assessment are explored. We also present a newly developed method for deriving non-interpolated (analytical) geological prototype models and illustrate how these can be used as useful priors for hyperparameter optimization and the creation of subsequent data-driven (interpolated) models.
Finally, the applicability of these approaches to real-world data is demonstrated through several case-studies, including the Altenberg-Teplice Caldera (Germany) and Stonepark-Pallas Green region (Ireland).
How to cite: Thiele, S., Kamath, A., Grose, L., Tolosana-Delgado, R., Hillier, M., and Gloaguen, R.: 3D geological modelling with curlew and neural fields, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11921, https://doi.org/10.5194/egusphere-egu26-11921, 2026.
The southernmost island of the Republic of Korea, Jeju Island, is composed of multi-layered lava flows formed by repeated volcanic activity and sedimentary layers deposited during quiescent periods between eruptions. Approximately 96% of the island’s total water use is supplied by groundwater, making groundwater the most critical water resource in the region. Therefore, accurate identification of the spatial distribution of aquifers and the major groundwater flow pathways is essential for sustainable groundwater management.
The objective of this study is to establish a 3D geological model for the mid-mountain area of Jeju Island and to quantitatively characterize the spatial distribution of the volcanic aquifer system and the major groundwater flow pathways. This approach enhances the understanding of groundwater recharge and flow mechanisms and provides a scientific basis for future groundwater conservation and management.
The 3D geological modeling results are summarized as follows. The basal Seogwipo Formation occurs below approximately 50 m above mean sea level and has an average thickness of about 50 m. It is overlain by basaltic lava flows with a total thickness of approximately 400 m, emplaced by at least 14 eruptions over the past 500,000 years. Intercalated sedimentary layers mainly consist of mudstone and silty sandstone and are generally less than 5 m thick. These fine-grained, silt–clay–dominated layers act as semi-confining units that retard downward infiltration of rainfall. Scoria and clinker layers occur mainly at the upper and lower boundaries of lava flows and increase in frequency from high-elevation zones to low-lying areas. In contrast to the dense basalt, these porous layers serve as major groundwater flow pathways.
How to cite: Choi, Y., Kim, J.-W., Ham, M.-S., Park, J.-B., Koh, H.-J., and Moon, S.-H.: Three-Dimensional Geological Model of Scoria and Clinker as Major Groundwater Flow Media in Volcanic Aquifers, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1660, https://doi.org/10.5194/egusphere-egu26-1660, 2026.
The Julong'an deposit is a large volcanic rock type uranium deposit located in the western part of the Xiangshan uranium ore field in Fuzhou City, Jiangxi Province. This study comprehensively utilized topographic maps, geological maps of mineral deposits, 26 exploration line profiles, 5 mid section plans, 65 drill holes etc. Using the “show-hide” interactive 3D geological modeling method, a 3D geological model of the Julong'an uranium deposit was constructed based on GOCAD software (Figure 1). Through 3D modeling, the following deep geological features were revealed:
- The second memberof the Ehuling Formation is thin in the north and thick in the south, with an exposed elevation of -610 to + Between profile lines 68 and 74, the bottom interface rises sharply from south to north, with a drop of 893m, and the north-south profile is in an "S" shape.
- The second memberof the Daguding Formation is thick in the north and thin in the south, with an exposed elevation of -700 to + Between lines 68 and 74, the thickness of the rhyolitic dacite increases sharply from an average of 15m to over 764m, and within a range of 240m from north to south, the thickness increases by over 749m. The top interface drops by more than 768m, forming an east-west "lava waterfall distribution".
- The fault structures mainly include the nearly north-south F7 fault, a hidden fault, and the east-west Niutoushan-Julong'an-Chuankeng volcanic collapse structure. The hidden fault is located on the west side of F7, and the two merge to the north with a gradually widening distance to the south. The cutting depth gradually becomes shallower from north to south. The shallow part of the F7 fault is a steeply dipping main fracture zone, while the deep part gradually transitions into a gently dipping crack dense zone. The fracture appears in an "Open fork" shape on the east-west cross-section and in an inverted "Open fork" shape on the horizontal cross-section (Figure 2). When the fissure zone approaches the intersection of different geological interfaces, mineralization significantly increases, forming a dense vein ore bodies. In the diamond shaped block of Julong'an, other nearly north-south and eastward dipping faults have similar characteristics and mineralization patterns. So, deep mineral exploration should focus on the intersection of the fracture zones of north-south faults and the interfaces between volcanic rock formations, angle unconformities, and intrusion interfaces, especially in the composite areas of east-west volcanic collapse structures on the southern part, which are key target areas for searching for large and rich minerals.
Funding: National Natural Science Foundation of China (42472130), ECUT Research Development Fund (K20240018), and the Natural Science Foundation of Jiangxi Province (20242BAB25183).
How to cite: Wu, Z., Guo, F., Li, B., and Li, G.: 3D geological modeling of the Julong'an uranium deposit in Jiangxi Province and implications for deep geological feature analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8900, https://doi.org/10.5194/egusphere-egu26-8900, 2026.
Geological modeling is an essential component of reservoir characterization in geothermal exploration. A geological model aims to understand the spatial relation between geological features such as rock unit boundaries, horizons and discontinuities (unconformities, faults) at various scales. However, geological models can contain significant uncertainties – often due to limited information at depth. It is therefore imperative to use all available information, including legacy data. In the KarboEx2-project, legacy seismic data from former coal exploration in the region of North Rhine Westphalia are digitized and reprocessed with modern seismic processing workflows. In our contribution here, we investigate how uncertainties in the interpretation of this legacy data can be considered in subsequent geological modeling workflows.
In the context of model construction, this type of uncertainty relates to the real position of the input points, commonly associated to data uncertainties (seismic processing, picking and interpretation, etc.). Several position points result from the picking of the horizons on legacy seismic data. A simple way to address this type of uncertainty is to perform sampling from the data treating it as fully correlated (i.e., moving all points simultaneously) or fully uncorrelated (i.e., moving all points independently). However, geological errors are commonly correlated with distance. One possibility to consider spatial correlations is to generate a geological surrogate model with a lower-dimensional representation of modelled interface. In addition to accounting for different uncertainties in space, such a low-dimensional representation allows to perform inference, sensitivity analysis, etc. We explore here a workflow based on the application of a variational Gaussian process (VGP) model and universal co-kriging for implicit geological modeling from inducing points using two open-source Python packages (GeoML, GemPy).
Our results show that it is possible to create surrogate models efficiently for a range of geological settings – with a balance between the dimension (input points) of the surrogate model and the level of complexity of the original interface. In addition, due to a variational approach, uncertainties in the input data can also be represented in the surrogate model. In next steps, the generated surrogate models will then be integrated into geothermal exploration workflows, including the uncertainties in the legacy seismic data.
How to cite: Satizabal, D., Gomes Gonçalves, Í., von Harten, J., Chudalla, N., Nathan, D., and Welmann, F.: 3-D Geological Modeling from Legacy Seismic Data with Consideration of Uncertainties, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1427, https://doi.org/10.5194/egusphere-egu26-1427, 2026.
