White-sand forests contribute significantly to dissolved organic matter (DOM) production in the central Amazon, forming blackwater rivers that dominate organic matter export from the Amazon basin to the ocean. Despite their importance in controlling DOM export, white-sand forests are understudied, and it remains unclear whether systematic changes in the formation of blackwater DOM occur and how seasonal variations and extremes like El Niño-associated droughts impact them. We collected soil porewater from two central Amazon white-sand forests for 2 years, spanning a wet La Niña year followed by an El Niño drought year. The molecular composition of DOM was analyzed using high-resolution mass spectrometry, and correlation network analysis was employed to identify ecologically meaningful DOM subsets. Using additional chemical characterization, database annotations, correlation with 14C-age of DOM and climatic variables, and ecological null modeling, we propose five distinct DOM sources: plant litter and throughfall, soil organic matter (SOM) decomposition, root exudation, and two drought response subsets of likely microbial and plant origin. During drought conditions, aboveground plant-derived compounds decreased, while SOM products, root exudates, and drought response compounds increased. These drought responses were qualitatively similar in both years but notably amplified in the drier El Niño year. Drought amplified deterministic control over DOM composition, indicating that DOM reflected directed biological responses and that future droughts are likely to generate similar shifts. Overall, drought substantially altered belowground carbon cycling by shifting DOM sources and inducing stress responses, effects expected to recur and potentially intensify under future climate scenarios.

The novel Python-based spICP-MS data processing algorithm ‘Sparta’ is presented and benchmarked against existing software for characterising nanoparticles (NPs) in terms of element-specific mass, size, and particle number concentration (PNC).

Thermodynamic benchmark calculations have been performed to better understand the behavior of 75Se(VI), 99Tc(VII), 233U(VI), 237Np(V), 241Am(III), Th(IV) and 242Pu(IV)) in the evolving geochemical conditions of the Long-term In-situ Test (LIT) at the Grimsel Test Site (GTS) and corresponding mock-up experiment. It also aims to identify the status of the geochemical speciation models and databases for these elements. The experiments are simulating the near-field conditions in some radioactive waste repository concept including a bentonite engineered barrier emplaced in crystalline rock and the findings are contributing to the long-term safety assessment of these facilities.In general, the calculations of all six modelling groups who participated in the benchmark agree well and differences in results could always be traced back either to differences in input data or to modelling assumptions. Such assumptions concern for example the characteristics of solid phases, especially the selection of a crystalline, aged or fresh mineral phase, which directly impacts the predicted radionuclide (RN) solubility as illustrated in the results for thorium and plutonium. In a similar way, the inclusion/exclusion of aqueous species leads to an increase/decrease of the solubility like for the polyselenides in case of selenium. In particular, the selection of recently approved species such as ternary silica actinide complexes or ternary uranyl carbonate complexes are impacting, for example, the speciation and solubility of U, Pu, Am or Np in the systems investigated here. It is crucial that all such data selection and modelling decisions need to be transparent and well justified or at least documented.The results from the thermodynamic benchmark contribute to a better understanding of the mobility of the respective RNs in the LIT and accompanying mock-up test. The observed mobility of significant fractions of 99Tc, 237Np and 233U in the mock-up experiment, but not in the LIT, together with the results from the thermodynamic benchmark calculations, suggest that conditions in the mock-up experiment are less reducing, since they are not buffered by surrounding minerals in the shear zone as is the case in the LIT. Rather low mobility 241Am(III) and 243Pu(III/IV) redox states are predicted in the calculations for all waters in accordance with the fact that no release of these nuclides is observed in either LIT or in the mock-up experiment. Ongoing analysis of the over-cored material of the LIT, namely the characterization of RN distribution profiles in the bentonite, will contribute to a further verification of the benchmark results.

Characterizing subsurface reservoirs, more specifically naturally fractured subsurface reservoir rocks, is essential for the study of subsurface reservoir properties. Image segmentation is an important aspect of digital rock physics (DRP) workflows. Traditional image segmentation techniques are less accurate than deep learning-based segmentation algorithms. In this paper, we investigate the segmentation accuracy of a convolutional neural network U-net and compare it with traditional methods of Watershed and multi-Otsu thresholding for multiphase segmentation of grayscale images from a naturally fractured coal sample. The segmentation target involved multiphase classification of the matrix, fully-filled fractures with minerals, and open fractures. The results reveal that U-net outperformed the others with an Intersection over Union metric of 94.9%, a Dice metric of 97%, and a Recall metric of 97.5%. The results support the importance of multiphase, deep learning-based segmentation techniques to support DRP studies of naturally fractured rocks.

Microplastic (MP) pollution is a growing concern for soil health, water quality, and biogeochemical processes. Evaluating the long-term environmental impacts of MPs requires understanding the processes controlling their mobility in soils. However, conventional approaches mainly rely on column outlet measurements or destructive sampling, leaving pore-scale MP behaviour poorly resolved. Here, we developed a systematically optimised, non-destructive workflow that combines small-scale column experiments, high-resolution X-ray micro-computed tomography (µCT), and digital rock physics. This approach enables three-dimensional pore-scale visualisation and quantification of retained small-sized MPs (down to 2 µm) within saturated soil-relevant porous media under controlled hydraulic conditions. Our results show that MP retention behaviour does not decrease monotonically with increasing flow velocity. Instead, high-flow conditions cause more localised MP accumulation, resulting in measurable decreases in soil permeability. At low flow velocities, MP retention was primarily associated with diffusion-enhanced delivery toward grain surfaces. With increasing flow velocity, advective transport became dominant, resulting in lower but more evenly distributed MP retention. Under high-flow conditions, however, hydrodynamic multi-particle bridging blocked pore throats, resulting in a permeability reduction of up to 4.6%. By linking pore-scale retention mechanisms with changes in hydraulic properties, this study provides new mechanistic insight into MP transport in saturated porous media. The proposed workflow provides a basis for future studies aimed at improving the prediction of MP transport in subsurface porous media.

Stylolites are complex geological features that usually appear as irregular surfaces. Fromed by pressure-driven dissolution, they are commonly filled with insoluble materials compared to surrounding host rocks. In underground reservoir flow (e.g., hydrogen, carbon dioxide, or natural gas storage), stylolites can influence reservoir fluid flow. They can act either as flow barriers or as flow pathways in reservoirs. Therefore, identifying stylolites in reservoir rocks is an essential task for the correct prediction of underground fluid flow. In this study, we develop a hybrid deep-learning workflow to automatically detect and classify stylolites in slabbed whole-core images. First, the “You Only Look Once” object-detection architecture is used to locate core-rock boundaries in raw whole core images containing multiple cores. These core images are then subdivided into 1,827 smaller image patches. We employ two convolutional neural network architectures of ReNet-50 and ResNeXt-50 for the final classification task. As the dataset, 150 m cores from three wells of a carbonate reservoir are used. The core image patches are first classified manually into five different image classes of stylolites, fractures, vertical plugs, horizontal plugs, and intact rock. As the dataset is imbalanced with respect to the number of images in each class, data-augmentation techniques such as flipping, cropping, rotation, and adjustments to brightness and contrast are implemented. This expands the dataset images to almost 16,000 images in total. Both networks are pre-trained on the ImageNet dataset and fine-tuned on the augmented dataset. To control overfitting, additional regularization techniques like dropout and adaptive learning-rate scheduling are used. Results show that ResNeXt-50 achieves the best classification performance of 92% on previously unseen whole-core images.

Crystalline rocks are considered host rocks for high-level radioactive waste (HLW) disposal because of their mechanical, thermal, and chemical properties. However, fractures and associated mineral precipitates, e.g. calcite, introduce significant heterogeneity that complicates predictions of radionuclide (RN) transport in fractured crystalline rock systems. This study combines positron emission tomography (PET) imaging and transport modeling to investigate the effects of calcite-filled fractures on predominantly diffusive transport in crystalline rocks. Diffusion experiments using a conservative iodine tracer (124I) were performed on rocks from the BUKOV underground research facility (Czech Republic), and CT-based calcite-filled fracture geometries were incorporated into RTM simulations. The results show two types of transport modifications due to calcite precipitation in fractures: enhanced transport in porous calcite generations and suppressed transport in low-porosity calcite fracture fillings. The spatial distribution of fracture-filling calcite, its porosity, and the porosity of the surrounding matrix materials can lead to heterogeneous or anisotropic diffusion behavior. This suggests that important transport characteristics in heterogeneous systems may be overlooked by conventional modeling using homogenized diffusion coefficients. The apparent diffusivity of the fractured and variously sealed BUKOV rock ranges from 10-13 to 10-10 m2/s. We conclude that exploiting the variability and anisotropy of diffusive flux behavior is beneficial for improving the accuracy of RN migration predictions in fractured and mineralized crystalline rock systems. This work demonstrates the value of integrating advanced imaging and modeling techniques to better characterize solute transport in fractured crystalline host rocks.

Identifying rock properties at the pore scale plays a crucial role in understanding larger-scale properties. For this purpose, the digital rock physics technique is used to model rock images at the pore scale. Achieving high-resolution (HR) images with a large field of view (FoV) is essential for pore-scale modeling of heterogeneous rock samples, which presents significant challenges due to their complex structures. However, because of the trade-off between resolution and FoV, it is not possible to acquire large HR images. Multi-scale image reconstruction methods enable modeling images at different resolutions and FoVs. Despite various approaches being introduced, a common limitation is the high computational cost. In this study, a novel approach based on Octree structures is introduced to minimize computational cost while maintaining accuracy. A Berea sandstone (BS) and an Edward Brown Carbonate (EBC) sample were scanned at both HR and low resolution (LR) using X-ray microtomography. Our method involves splitting the unresolved porosity in rock images into smaller sections of unresolved templates using the watershed algorithm and considering the optimized parameters. We then applied a cross-correlation based simulation technique to find the best match of each unresolved template. The novelty of our approach lies in the use of an Octree structure to perform calculations on LR images, significantly reducing computation time and memory consumption due to the fewer number of pixels in Octree LR structures. The accuracy of the images thus reconstructed using our approach was compared with those from previous methods by evaluating geometric properties and single- and two-phase flow properties. The results were promising, demonstrating that our approach achieved a permeability close to the real value, while the previous method had an error of approximately 4% for both BS and EBC rocks. More importantly, our approach was approximately three times faster and reduced memory usage by 20 to 130 times. The findings of this study facilitate dual- or multi-scale modeling and evaluate heterogeneous rock images at a significantly lower computational cost. In particular, for heterogeneous rocks, where multi-scale image reconstruction demands substantial memory and runtime, the use of the Octree technique enables accurate reconstruction with lower computational cost.

Digital Rock Physics can significantly enhance our understanding of rock behavior. However, modeling heterogeneous rocks remains challenging because of the trade-off between resolution and field of view. To address this, researchers have developed multi-scale pore network models (PNMs), which integrate PNMs from different scales to create unified multi-scale PNM. Various methodologies exist for merging PNMs from different resolutions, but they often suffer from inaccuracy, high runtime and significant memory consumption, particularly when microporosity is integrated into larger scales. This study introduces a novel fusion and an innovative upscaling approach for efficient multi-scale PNM reconstruction of rocks containing microporosity. Our methods separate resolved and unresolved porosities using different voxel sizes from CT scans at multiple resolutions. Resolved regions have larger voxel sizes, while unresolved areas retain smaller voxel sizes. We extract macroPNM from the resolved regions and generate stochastic micro-PNM for the unresolved areas. An artificial neural network (ANN), trained on micro-PNM, links micro- and macro-PNMs. The multi-scale PNMs generated using the ANN method had an average permeability of 252 +/- 3 mD, closely matching the laboratory-measured permeability of the rock (257 mD). In contrast, the average permeability of multi-scale PNMs reconstructed using the statistical method was significantly higher, at 308 +/- 38 mD. Consequently, the ANN-based reconstruction method, owing to the proper connection between scales, improved the accuracy of permeability prediction by approximately 90% compared to the statistical reconstruction method. In the next step, each microPNM is upscaled to a base pore based on its effective hydraulic conductance. These base pores are then connected to the macro-PNM using a novel approach. We utilized synchrotron CT images of an Indiana limestone rock at two resolutions as our training dataset. The single- and multi-phase flow analysis of the fused PNM demonstrated excellent agreement with laboratory-measured rock properties. Our upscaling method also reduced runtime by up to 40% (from 312 to 190 CPU-seconds) and memory consumption by approximately 68% (from 25 GB to 8 GB), all without compromising predictive accuracy.

Whole core photography is an essential step in core analysis, offering complementary insights that improve accuracy of lithological assessments. Cores are photographed to document their features accurately, preserve a visual record for future analysis, and facilitate better identification of lithological units. Visible (or white) light photography provides detailed visual information about rock physical characteristics, while ultraviolet- (UV-)light images are commonly used due to the fluorescence properties of certain minerals, enhancing mineral detection and identification. Both approaches improve the accuracy of lithological assessments and inform subsurface management decisions. With the rise of artificial intelligence and the demand for precise and automatic rapid predictions, machine learning techniques, particularly convolutional neural network (CNN) architectures, have gained prominence in lithology identification research. This study implemented multi-input CNNs to automatically predict lithology from whole core images. To improve image classification accuracy, we combined UV- and white-light images as input to the CNN, allowing the network's filters to learn richer features automatically. We used 176-m core data from two formations in the Middle East. Data augmentation techniques were used to create 6000 images. The dataset was randomly divided into three parts for training, validation, and testing the networks. We selected the ResNeXt-50 architecture for its superior efficiency in classifying three lithologies: sandstone, loose sand, and limestone. This model was compared to the EfficientNet architecture. The network parameters were initialized using transfer learning. We optimized the network hyperparameters including learning rate, batch size, and optimizer, achieving 99% accuracy in predicting unseen data. This study establishes an accurate and rapid procedure for automatic lithology classification, outputting a lithology column.

Bentonite plays a critical role in engineered barrier systems designed for radioactive waste storage in geological repositories especially in crystalline formations. Ensuring its long-term stability under realistic hydrogeochemical conditions is vital for evaluating the safety of these repositories. This study investigated the influence of controlled water flow in a shear zone on the erosion of bentonite through a 4.5-year Long-Term In-Situ Test (LIT) at the Grimsel Test Site, Switzerland. Compacted Ca-Mg-type FEBEX bentonite rings (with 90 % montmorillonite content) were positioned in-situ in an emplacement borehole intersecting a water-conducting shear zone providing direct contact with low-mineralized glacial meltwater. X-ray computed tomography scanning, along with digital rock physics methods, were used to quantify bentonite mass loss and the contact shear zone aperture distribution on over-cored LIT samples. A Random Forest classifier, a machine learning technique, was used for segmentation, which enabled more precise quantification of bentonite mass loss and improved fault characterization. This approach used multiphase segmentation, allowing accurate distinction between different material phases in the cored interval, which is essential for resolving complex interactions in heterogeneous systems. The selection of the correct region of interest was crucial for minimizing segmentation errors and improving mass loss quantification by reducing interferences from non-relevant structures. The aperture distribution between the three boreholes over-cored within the shear zone was evaluated with a mean thickness of 2.90 +/- 1.09 mm (2 sigma). Furthermore, the bentonite mass loss was computed from the scanned images and compared with mobilised montmorillonite colloid masses, continuously sampled in the water from observation boreholes (0.11-0.12 m and 6 m distance) measured by inductively coupled plasma mass spectrometry (ICP-MS) and laser-induced breakdown detection (LIBD) techniques. The data evaluation of both techniques used in this study provided erosion rates <2 kg/m(2)/y, which are at least two orders of magnitude below the mass loss assessment rates of 500 to 1500 kg/m(2)/y defined by safety case considerations of the Swedish Nuclear Fuel and Waste Management Company (Svensk K & auml;rnbr & auml;nslehantering Aktiebolag, SKB) and the Finnish company POSIVA handling the final disposal of the spent nuclear fuel generated by its owners, the nuclear plant operators Teollisuuden Voima and Fortum. The creation of a digital twin model for the bentonite-water-shear zone system provided new insights into the erosion processes showing inhomogeneous erosion in contact with real fracture geometries.

Variability in fracture geometry and its complex surface characteristics are major contributors to solute transport and retention effects in rocks such as granite. Understanding the effects of cross-scale fracture geometry on solute transport modeling is critical for reliable quantitative predictions in applications such as geothermal energy use and nuclear repository safety. Here, we systematically investigated the sensitivity of fracture surface topography variability to the flow field and solute transport behavior. Specifically, we investigated the role of multiscale fracture surface roughness in solute transport modeling. As a starting point, our study utilized a 3D fracture geometry derived from CT scans and employed COMSOL Multiphysics software for solute transport modeling. By introducing increasingly lower spatially resolved geometries, while maintaining a constant high resolution mesh for simulation calculations, we investigated the consequences for transport on surfaces composed of superimposed building blocks of different sizes and shapes. The results indicate that fracture geometry simplifications with reduced spatial frequency information of well-defined, specific domains do not have a clear trend to alter the BTCs tailing. Instead, this type of model simplification can cause both increased and decreased tracer residence times, leading to misleading interpretations. We explain this by a complex superposition of surface building blocks of different sizes, such as single crystal surface pits, grain boundaries between crystals, and fracture curvature. For model sensitivity analyses, we suggest the use of concentration difference and acceleration maps to identify local transport heterogeneities introduced by geometric simplifications. In addition, we conclude that power spectral density (PSD) analysis provides a means of defining a range of surface spatial frequencies that helps to avoid oversimplification in geometric models of reactive transport.

During subsurface storage in geological underground reservoirs, operators implement various well configurations, also known as injection patterns, to inject target invading fluids into a reservoir. Each pattern consists of several injection and production wells arranged uniquely. Therefore, reservoir connectivity plays a key role in determining the optimal well location for each pattern. In this study, we apply concepts of percolation theory to investigate the impact of well locations on overall reservoir connectivity within injection patterns. We generate 104 reservoir realizations using Monte Carlo simulations. We then examine reservoir connectivity for two well-configuration scenarios and compare the results with the literature values of the conventional line-to-line (L2L) percolation connectivity model. In the first scenario, we investigate reservoir percolation properties and connectivity by fixing one well at the corner of the reservoir and placing the second well at the opposite side representing the boundary connectivity model. In the second scenario, we examine reservoir connectivity for wells located within the reservoir boundaries, known as the off-boundary connectivity model. To verify the algorithm, we compute the infinite percolation thresholds and compare them with values reported in the literature. Our results indicate that the most challenging connectivity occurs when wells are located at the corners of the reservoir, as in the boundary connectivity model. Comparing the mean connectivity curves and the connectivity exponent values of the boundary, off-boundary, and L2L connectivity models reveals that the off-boundary connectivity model has characteristics between the other two models. The results demonstrate that connectivity in underground storage systems is highly dependent on well placement, permeability heterogeneity, and reservoir boundaries. By comparing various connectivity models, we establish a relationship between injection well locations and reservoir fluid migration pathways, providing insights into optimizing storage and retrieval efficiency.

Post-mining landscapes worldwide often remain impacted by high heavy metal or radionuclide mobility and acid mine drainage. The “Gessenwiese” test site in the former uranium mining area of Ronneburg, Germany, was established to develop land reclamation strategies for such post-mining soils. The approach combined calcareous substrate amendments and microbial inoculation with mycorrhiza and Streptomyces, alongside lignocellulose production through short rotation forestry using birch (Betula pendula Roth), alder (Alnus incana (L.) Moench), and willow (Salix triandra × viminalis “Inger”), without competing with land-use for food production. Physico-chemical and hydrological parameters, as well as trace element concentrations in the mobile < 0.45 µm fraction of porewater and shallow groundwater were measured. Trees were harvested after two consecutive growth phases of 2–4 years. Trace element content in the aboveground biomass was assessed for suitability as solid biofuel. Soil amendments, particularly calcareous substrates, effectively reduced mobile metal concentrations in porewater, though some residual contamination persisted in groundwater. Metal content in birch and alder biomass generally met solid biofuel thresholds and was further reduced by soil treatments; however, Cd and Zn in willow exceeded permissible levels. Alders produced the highest biomass across soil treatments, while birch growth picked up only after the first harvest and willow generally performed worst on the acid mine drainage impacted soil. This study proves the feasibility of combining phytostabilization with lignocellulose production on moderately contaminated post-mining sites. Observation periods exceeding six years are essential for predicting the long-term success of such strategies.

AbstractA research programme has been conducted jointly by GRS together with BRIUG and BGR to characterize GMZ bentonite as buffer material in comparison with the well-known MX80 bentonite.