Steel corrosion plays a major role in the geochemical evolution at the canister/bentonite interface of the engineered barrier systems of geological radioactive waste repositories. The interactions between corrosion products and bentonite can significantly affect bentonite properties and performance. These interactions have been investigated by resorting to in situ tests conducted in underground laboratories, such as the FEBEX (Full-scale Engineered Barrier Experiment) test. The FEBEX in situ test, which was conducted at the Grimsel underground research laboratory in Switzerland from 1997 to 2015, demonstrated substantial corrosion of the steel liner in areas without a heater, primarily due to the presence of O2. Here we report a reactive transport model that simulates steel corrosion products and their interactions with bentonite. The model builds on a previously published conceptual geochemical model and addresses its limitations by integrating a more detailed representation of temperature and unsaturated flow conditions, leveraging prior thermo–hydrodynamic–mechanical–chemical (THMC) models. Given the prevailing uncertainties in O2 and redox conditions during the test and the limited data on liner corrosion and gas conditions at the liner–bentonite interface, liner corrosion was modeled by using a prescribed time-dependent function for the corrosion rate. Goethite, hematite, and magnetite were the Fe minerals allowed to precipitate in the model. The corrosion rate and the specific surface area of the hematite and magnetite were calibrated based on the profiles of goethite, hematite, and total Fe (including dissolved, exchanged and sorbed forms) observed at the post mortem analysis of the FEBEX in situ test. The model reproduces the observed goethite and hematite precipitation near the liner but underestimates the measured values at greater distances from the liner. The pattern of total calculated Fe concentrations reproduce the measured values except at a distance between 15 and 50 mm from the liner. Goethite is the predominant corrosion product in the model results, even under reducing conditions, owing to kinetic constraints on magnetite and hematite precipitation and to the enhanced stability of goethite driven by pH increase and thermal evolution.

Thanks to the recent progress in numerical methods and computer technology, the application fields of artificial intelligence (AI) and machine learning methods (ML) are growing at a very fast pace. The field of geochemistry for nuclear waste management has recently started using ML for the acceleration of numerical simulations of reactive transport processes, for the improvement of multiscale and multiphysics couplings efficiency, and for uncertainty quantification and sensitivity analysis. Several case studies indicate that the use of ML based approaches brings an overall acceleration of geochemical and reactive transport simulations between one and four orders of magnitude. This paper presents a benchmarking exercise that aims at providing a set of reference data and models for developing and applying ML techniques for geochemical and reactive transport simulations. Several well-known geochemical speciation codes are used to generate systematically a consistent set of high-quality chemical equilibrium data, to be used as input for the training of several ML methods. Two benchmarks are formulated, each with multiple levels of gradually increasing degree of complexity. The first benchmark focuses on cement chemistry, while the second one considers uranium sorption on a clay mineral. The performance of different ML techniques is then evaluated in terms of their numerical efficiency and accuracy. A speedup of several orders of magnitude is observed. The benefits and the limitations of different ML based techniques are then analysed and highlighted.

Fluid flow and mass transport parameters are greatly impacted by the fracture surface's heterogeneous properties. By quantifying the spatial correlation patterns of fracture apertures, this study intends to establish an upscaling framework utilizing Lagrangian-based transport models to estimate dispersivities of naturally fractured granite cores. To do this, the surface morphology of the fragmented core sample was photographed using a 3D laser profile scanner with varied degrees of accuracy. Following a geostatistical analysis of the fracture aperture data, common covariance models were fitted to determine the integral scale of log fracture aperture. This information was used to estimate dispersivity using the Lagrangian-based transport models, which only required the collection of fractured geostatistical data. The study also assessed how the dispersivities are affected by fracture lengths and the measurement accuracy of aperture data in this model. The findings showed that the fracture aperture spatial bivariate correlation structures follow the exponential covariance model. Additionally, while the variance of log fracture apertures is significantly influenced by both the data resolutions and the core samples themselves, the integral scale of log fracture apertures primarily depends on the length of the fracture. In terms of the fracture length, the upscaled dispersivities range from 2.35% to 7.21%. The development of upscaling techniques for estimating dispersivities in fractured and heterogenous granitic rocks and scaling them up to the field size for the accurate prediction of solute transport mechanisms can be guided by these findings.

The assessment of the long-term performance of the engineered barrier systems of high-level radioactive waste (HLW) repositories requires the use of reactive transport models. Montenegro et al. (2023) presented a nonisothermal reactive transport model of the long-term geochemical evolution of a HLW disposal cell in a granitic host rock corresponding to a generic reference concept. The model accounted for the vitrified waste, the carbon-steel canister, the bentonite buffer and the reference granitic rock. Here we extend their model by considering model variants (V), sensitivity cases (SC) and model abstractions (MA). Variants V1, V2 and V3 consist of considering MX-80 bentonite instead of FEBEX bentonite (V1), a larger groundwater flux through the granite (V2) and the Czech reference crystalline rock as a host rock (V3). Cases SC1 and SC2 consider a decrease of the silica concentration threshold value in the glass dissolution rate (SC1) and an earlier canister failure time (SC2), respectively. Runs MA1 to MA4 consider smectite as an unreactive mineral phase (MA1), the porosity feedback effect on chemical and transport parameters (MA2), a time-varying corrosion rate (MA3), and a coarser finite element grid (MA4), respectively. Model results of V1 show a larger pH, a smaller precipitation of magnetite, siderite and greenalite and a slightly smaller dissolution of ISG and smectite than the base run of Montenegro et al. (2023). Model predictions are very sensitive to the increase in the groundwater flow through the granitic host rock (V2). However, predictions are not sensitive to the chemical composition of the granite porewater (V3). The decrease in the silica saturation threshold from 1 & sdot;10- 3 to 5 & sdot;10- 4 mol/L in SC1 leads to a significant decrease in glass dissolution. Glass dissolution after 50,000 years in SC2 (earlier canister failure) is much larger than that of the base run. Model results are not sensitive to considering smectite as an unreactive mineral phase (MA1). However, model results are very sensitive to the porosity feedback effect (MA2). A 60% volume fraction of Fe(s) remains uncorroded after 50,000 years when a variable corrosion rate is considered in MA3. In this case the precipitation of corrosion products is much smaller than that of the base run. The general patterns of the numerical results in MA4 (coarser grid) are similar to those of the base case.

The Strategic Research Agenda (SRA; https://www.ejp-eurad.eu/publications/eurad-sra) of the European Joint Programme on Radioactive Waste Management (EURAD; https://www.ejp-eurad.eu/) describes the scientific and technical domains and sub-domains and knowledge management needs of common interest between EURAD participant organizations. Theme number 7 is entitled “Performance assessment, safety case development and safety analyses.” A list of research and development priorities and activities of common interest to be addressed within EURAD for theme 7 have been established. Amongst others, the Understanding and modelling of multi-physical Thermo-Hydro-Mechanical-Chemical coupled processes (THMC) occurring in radioactive waste disposal is a major and permanent issue to support optimization of design and safety case abstraction. To tackle this challenge a research work package entitled “DONUT: Development and improvement of numerical methods and tools for modelling coupled processes” has been conducted within the EURAD join programming initiative. The purpose of this work package is to improve/develop methods or numerical tools in order to go a step further in development of (i) relevant, performant and cutting-edge numerical methods that can easily be implemented in existing or new tools, in order to carry out high-performance computing to facilitate the study of highly coupled processes in large systems, (ii) numerical scale transition schemes for coupled processes, (iii) innovative numerical methods to carry out uncertainty and sensitivity analyses. In this paper the work carried out within the DONUT work package is put in perspective regarding the existing concept and literature on the field. It does not pretend to be exhaustive but rather to put emphasis on particular issues tackled during the project.

The organic pollutants disposed at the Sardas landfill in Sabinanigo (Huesca, northeastern Spain) by the INQUINOSA lindane factory have reached the Gallego alluvial aquifer and could affect the Sabinanigo reservoir. The daily oscillations of the reservoir water level produce a tidal effect on the piezometric heads of the aquifer. These oscillations are transmitted in a damped way with a time lag, thus attesting that the silting sediments of the reservoir and the natural silts of the Gallego alluvial are interposed between the reservoir water and the layer of sands and gravels. A 2D finite element groundwater flow and total dissolved hexachlorocyclohexane (HCH) transport model through the Gallego alluvial aquifer is presented here. The flow model was constructed to: (1) Quantify the tidal effect, produced by the daily fluctuations of the reservoir water level on the aquifer; (2) Estimate the hydrodynamic parameters of the layer of sands and gravels; and 3) Estimate the vertical hydraulic conductivity of the silting sediments and silts; and (4) Quantify aquifer/reservoir interactions. The flow model reproduces the dynamics of the tidal effect and attests that groundwater velocity and flow direction changes daily in response to the oscillations of the reservoir level. Model results reproduce the measured well hydrographs and the Darcy velocity derived from tracer tests and confirm the validity of the conceptual model. The transport model of total dissolved HCH simulates the time evolution of the contaminant plume. The computed concentrations of total dissolved HCH and the contaminant mass outflux are very sensitive to changes in the source terms and the distribution coefficient, Kd of HCH. The best fit to the measured HCH plumes in September 2010 and December 2020 is obtained with a Kd ranging from 1 to 3 L/kg. The computed flux of dissolved HCH leaving the Sardas site in 2020 towards the Sabinanigo reservoir ranges from 0.6 kg/year for Kd = 3 L/kg to 3.1 kg/year for Kd = 1 L/kg. The findings of this study will be most useful for planning and designing remedial and containment actions at the Sardas site and other similar lindane-affected sites.

Groundwater flow and transport models are essential tools for assessing and quantifying the migration of organic contaminants at polluted sites. Uncertainties in the hydrodynamic and transport parameters of the aquifer have a significant effect on model predictions. Uncertainties can be quantified with advanced sensitivity methods such as Sobol’s High Dimensional Model Reduction (HDMR) and Variogram Analysis of Response Surfaces (VARS). Here we present the application of VARS and HDMR to assess the global sensitivities of the outputs of a transient groundwater flow model of the Gállego alluvial aquifer which is located downstream of the Sardas landfill in Huesca (Spain). The aquifer is subject to the tidal effects caused by the daily oscillations of the water level in the Sabiñánigo reservoir. Global sensitivities are analyzed for hydraulic heads, aquifer/reservoir fluxes, groundwater Darcy velocity, and hydraulic head calibration metrics. Input parameters include aquifer hydraulic conductivities and specific storage, aquitard vertical hydraulic conductivities, and boundary inflows and conductances. VARS, HDMR, and graphical methods agree to identify the most influential parameters, which for most of the outputs are the hydraulic conductivities of the zones closest to the landfill, the vertical hydraulic conductivity of the most permeable zones of the aquitard, and the boundary inflow coming from the landfill. The sensitivity of heads and aquifer/reservoir fluxes with respect to specific storage change with time. The aquifer/reservoir flux when the reservoir level is high shows interactions between specific storage and aquitard conductivity. VARS and HDMR parameter rankings are similar for the most influential parameters. However, there are discrepancies for the less relevant parameters. The efficiency of VARS was demonstrated by achieving stable results with a relatively small number of simulations.

Compacted bentonite is part of the multi-barrier system of radioactive waste repositories. The assessment of the long-term performance of the barrier requires using reactive transport models. Here we present a multiphase flow and reactive transport benchmark for radioactive waste disposal. The numerical model deals with a 1D column of unsaturated bentonite through which water, dry air and CO_2(g) may flow and with the following reactions; aqueous complexation, calcite and gypsum dissolution/precipitation, cation exchange and gas dissolution. INVERSE-FADES-CORE V2, DuMu^X , TOUGHREACT and iCP were benchmarked with 6 test cases of increasing complexity, starting with conservative tracer transport under variably unsaturated conditions and ending with water flow, gas diffusion, minerals and cation exchange. The solutions of all codes exhibit similar trends. Small discrepancies are found in conservative tracer transport due to differences in hydrodynamic dispersion. Computed CO_2(g) pressures agree when a sufficiently refined grid is used. Small discrepancies in CO_2(g) and pH are found near the no-flow boundary at early times which vanish later. Discrepancies are due differences in the formulations used for gas flow at nearly water-saturated conditions. Computed CO_2(g) pressures show a fluctuation between 10^-4 and 10^-3 years which slows down the in-diffusion of CO_2(g) . This fluctuation is associated with chemical reactions involving CO_2 . There are discrepancies in solute concentrations due to differences in the Debye–Hückel (DH) formulation. They are overcome when all codes use the same DH formulation. The results of this benchmark will contribute to increase the confidence on multiphase reactive transport models for radioactive waste disposal.

This opinion paper describes the major coupled T(Thermal)-H(Hydro)-M(Mechanical)-C(Chemical) processes in geological repository systems and the frontier of related model development. Particular focus is made on the analysis of existing approaches and open research questions with respect to the further development of coupled codes and models for realistic multi-scale simulations of repository systems. These include the use of machine learning and artificial intelligence in acceleration of computer codes; sensitivity analysis, inverse modelling and optimisation; software engineering and collaborative platforms for model development.

Data science (digitalisation and artificial intelligence) became more than an important facilitator for many domains in fundamental and applied sciences as well as industry and is disrupting the way of research already to a large extent. Originally, data sciences were viewed to be well-suited, especially, for data-intensive applications such as image processing, pattern recognition, etc. In the recent past, particularly, data-driven and physics-inspired machine learning methods have been developed to an extent that they accelerate numerical simulations and became directly usable for applications related to the nuclear waste management cycle. In addition to process-based approaches for creating surrogate models, other disciplines such as virtual reality methods and high-performance computing are leveraging the potential of data sciences more and more. The present challenge is utilising the best models, input data and monitoring information to integrate multi-chemical-physical, coupled processes, multi-scale and probabilistic simulations in Digital Twins (DTw) able to mirror or predict the performance of its corresponding physical twins. Therefore, the main target of the Topical Collection is exploring how the development of DTw can benefit the development of safe, efficient solutions for the pre-disposal and disposal of radioactive waste. A particular challenge for DTw in radioactive waste management is the combination of concepts from geological modelling and underground construction which will be addressed by linking structural and multi-physics/chemistry process models to building or tunnel information models. As for technical systems, engineered structures a variety of DTw approaches already exist, the development of DTw concepts for geological systems poses a particular challenge when taking the complexities (structures and processes) and uncertainties at extremely varying time and spatial scales of subsurface environments into account.

The assessment of the long-term performance of the engineered barrier systems of high-level radioactive waste (HLW) repositories requires the use of reactive transport models. Here a non-isothermal reactive transport model of the long-term geochemical evolution of a HLW disposal cell in a granitic host rock is presented. The model includes the vitrified waste (40 cm in diameter), the carbon-steel canister (5 cm thick), the saturated FEBEX bentonite buffer (75 cm thick) and the reference granitic rock. The model accounts for the thermal transient stage and assumes generalized steel corrosion under anaerobic conditions with a corrosion rate equal to 1.41 m/y. Canister failure is assumed to occur when the remaining canister thickness is equal to 3.5 cm at t = 25,000 years. Canister corrosion caused an increase in pH. The computed pH in the canister just before canister failure (t = 25,000 years) was equal to 9.25 and ranged from 7.82 to 9.25 in the bentonite. Magnetite, the main corrosion product, precipitated in the bentonite and especially in the canister. The thickness of magnetite precipitation band in the bentonite was & AP;1 cm. Siderite precipitated at both sides of the canister/bentonite interface. The precipitation front penetrated >1 cm into the bentonite. Nuclear glass started dissolving after canister failure (t > 25,000 years). The concentration of dissolved silica increased in the inner part of the glass until t = 30,000 years and decreased in the outer part of the glass due to the out diffusion of dissolved silica into the canister and the bentonite. This diffusive flux caused the precipitation of greenalite at the glass/canister and canister/ bentonite interfaces. The pH at the end of the simulation (t = 50,000 years) ranged from 7.93 to 7.89 in the glass, from 7.89 to 8.66 in the canister and from 7.87 to 8.6 in the bentonite. Magnetite precipitated in the canister while there was carbon steel to corrode. Once the canister was fully corroded, magnetite redissolved near the glass/canister interface. Greenalite precipitated in the canister and the bentonite, especially at the glass/canister interface and siderite precipitated at the canister/bentonite interface. The simulation results should be useful for the performance assessment of engineered barriers of radioactive waste repositories in granitic host rocks.

Lindane (a harmful contaminant) was produced in Sabiñánigo (Huesca, Spain) and deposited at the Sardas landfill. This site contains a large mass of pollutants, which have an extremely large contamination potential of the Ebre River. The site has undergone numerous human interventions that have modified the natural conditions. The site exhibits complex hydrogeological patterns and has been monitored systematically for a long period of time, and a large amount of geological, hydrological, and hydrogeological data are available. Here, a 2D finite element groundwater flow model along a vertical profile heading east–west along the thalweg of the former gully is presented. The main goal is modelling groundwater flow through the landfill, the Gállego River alluvial aquifer, and its interactions with the Sabiñánigo reservoir. The numerical model confirms the prevailing conceptual hydrogeological model of the site. The main results include: (1) Groundwater flows into the landfill mainly along perimeter ditches, which do not properly drain the surface and subsurface runoff (13.84 m3/d) and from the underlying marly rock (8.84 m3/d); (2) The total landfill leachate outflow towards the alluvial floodplain underneath the front slurry wall through a shallow marl layer is equal to 17 m3/d; (3) The oscillations of the Sabiñánigo reservoir water level produce a tidal effect that results in periodic changes of the hydraulic gradient between the alluvial gravels and the reservoir; (4) Groundwater flows generally from the alluvial aquifer towards the reservoir in an average E–W direction with an average Darcy velocity equal to 5 cm/d. The flow direction, however, changes to W–E when the reservoir level rises suddenly and; (5) The hydrodynamic parameters of the alluvial silts and reservoir silting sediments are crucial in determining the influence radius of the inversion of groundwater flow direction when the reservoir level rises suddenly. Model results enhance the confidence of the conceptual model, provide the basis for detailed specific models of the landfill and the alluvial aquifer, and highlight the importance of considering the tidal effect of the reservoir level oscillations. They also provide valuable information for managing the landfill and its impact on the surrounding groundwater system.

Digitalization and Artificial Intelligence is the fastest emerging paradigms in engineering and natural science applications. As an example data-driven and physics-inspired machine learning methods have been developed and evaluated to accelerate numerical simulations; evaluating their usability for applications related to the radioactive waste management cycle is therefore of high relevance. Under the umbrella of the European Joint Programme on Radioactive Waste Management (EURAD) and on Pre-disposal Management of Radioactive Waste (PREDIS), different initiatives have been established to facilitate evaluation and implementation of digitalization technologies (and specifically digital twins mirroring their corresponding physical assets) for long-term radioactive waste management. Previous studies indicate that digitalization is an important tool, to improve processes throughout the whole waste management cycle including pre-disposal activities like waste treatment, conditioning, storage and aspects related to the operation and long-term performance of disposal systems. However, before implementation, several challenges related to different facets of digital twins and digitalization need further research. At this stage, the specific potential and role of different aspects of digital transformation for different topics of waste management is still somewhat vague. In this contribution, an overview will be given addressing selected research developments s based on activities that are ongoing in EURAD and PREDIS or in closely related activities.

Carbon steel and compacted bentonite have been proposed as candidate materials for the overpack and buffer, respectively, of the multi-barrier system of a geological high-level radioactive waste repository. Carbon steel corrosion may impair bentonite properties. The interactions of corrosion products and bentonite are analyzed with laboratory corrosion tests. Here coupled thermo-hydro-chemical-mechanical (THCM) models of two types of heating and hydration tests performed on compacted bentonite in contact with Fe powder are presented to study the iron-bentonite interactions at representative repository conditions. Tests on small cells (SC) were performed under unsaturated non-isothermal conditions in 25 mm long columns containing 21 mm of bentonite and 4 mm of Fe powder. Tests on medium-size cells (FB) were performed under unsaturated non-isothermal conditions in 99.8 mm long columns containing 86.8 mm of bentonite and 13 mm of Fe powder. Model results for the SC tests showed that magnetite and Fe(OH)2(s) were the main corrosion products which compete for Fe2+ precipitation. Computed corrosion products precipitate mainly in the Fe powder, penetrate a few mm into the bentonite and reproduce the measured iron weight data. Model results of the FB tests showed that magnetite precipitates throughout the Fe powder interface and reproduce the main trends of the corrosion products. Model results of these corrosion tests will be of great relevance for the performance assessment of engineered barriers of radioactive waste repositories.

Modeling radionuclide reactive transport in host rocks is one of the major challenges with respect to the performance and safety assessment of high-level radioactive repositories. The rock heterogeneity has a significant impact on the scale effects of flow and solute-transport parameters such as the sorption coefficient. The scaling of the radionuclide sorption coefficient is related to the reactive mineral facies (RMF) distributions in fractured rocks, which can be simulated with transition probability-based geostatistical methods. Geostatistical and geochemical analyses of the RMF distributions are performed on granite samples taken from the Beishan site in northwest China. This paper presents an approach to identifying RMF by using a deep-learning-based mineral identification model. The volume proportions and mean lengths of the RMF are accurately estimated in this way compared to the results of X-ray diffraction analysis. This approach overcomes the limitations of traditional methods which are subjective and have lower accuracy. The results show that the composite covariance model of sorption coefficients is related to the volume proportion and mean length of the RMF. The effective sorption coefficient obtained by the upscaling model is greater than the geometric average and smaller than the arithmetic average. Furthermore, through global sensitivity analysis, it is found that mean retardation factors have the most significant effect on the effective sorption coefficient. The results of this study provide substantial information that contributes to improving understanding of radionuclide transport in fractured granite.

One of the most common solutions for more than 50 years for the recovery of mine holes generated during mining activities is their transformation into lakes. This paper presents the activities carried out from the closure of the As Pontes mine to the total filling of the mining lake. The environmental solution adopted for the recovery of the mining hole and its transformation into a mining lake is described, as well as the methodology followed for its filling, defining the water sources used, its volume and its chemical quality. Perfect mix and stratified lake models were performed to simulate the final chemical quality of the lake. The details of the studies carried out on waves on the shores of the lake and the stability of the slopes of the mine are also presented. The hydraulic works carried out to reintegrate into the lake the watercourses that were diverted during the mining exploitation are also indicated. Water chemical data of the As Pontes lake collected during and after the filling of the lake show the presence of two well-differentiated zones separated by a chemocline. The uppermost layer has a slightly acid pH and oxic conditions, while the deeper layer has very low pH values and anoxic conditions. The methodology used was successful and the forecasts made are in line with the results obtained both in relation to the filling time of the lake and the quality of the overflow waters, which comply with all the established limits.

Generating reasonable heterogeneous aquifer structures is essential for understanding the physicochemical processes controlling groundwater flow and solute transport better. The inversion process of aquifer structure identification is usually time-consuming. This study develops an integrated inversion framework, which combines the geological single-sample generative adversarial network (GeoSinGAN), the deep octave convolution dense residual network (DOCRN), and the iterative local updating ensemble smoother (ILUES), named GeoSinGAN-DOCRN-ILUES, for more efficiently generating heterogeneous aquifer structures. The performance of the integrated framework is illustrated by two synthetic contaminant experiments. We show that GeoSinGAN can generate heterogeneous aquifer structures with geostatistical characteristics similar to those of the training sample, while its training time is at least 10 times faster than that of typical approaches (e.g., multi-sample-based GAN). The octave convolution layer and multi-residual connection enable the DOCRN to map the heterogeneity structures to the state variable fields (e.g., hydraulic head, concentration distributions) while reducing the computational cost. The results show that the integrated inversion framework of GeoSinGAN and DOCRN can effectively and reasonably generate the heterogeneous aquifer structures.