TRANSFER

TRANSFER –  Training and Research Academy for Nuclear Safety and Geological Repository Systems

TRANSFER FSU Jena: Upscaling geological heterogeneity at the pore scale using deep learning (DL)

Funding: Federal Ministry for Research, Technology and Space (BMFTR) and Federal Company for Radioactive Waste Disposal (BGE)

Duration: 01.04.2026 – 30.04.2029

Project leader: Prof. Thorsten Schäfer and Dr Saeid Sadeghnejad

Person in charge: Denis Muturi

Description

The safe disposal of high-level radioactive waste requires reliable predictions of how radionuclides migrate through geological host rocks over timescales of hundreds of thousands of years. Opalinus clay are the clay rocks that show promise for host formation, but their pore structure is heterogeneous across many scales, ranging from individual clay matrixes at the nanometer level to sandy and carbonate lenses visible at the centimeter scale. This heterogeneity significantly impacts fluid flow and radionuclide transport, which existing homogeneous safety models cannot account for.

The Jena subproject within the TRANSFER graduate school explores novel digital rock physics and deep learning approaches to bridge this scale gap by upscaling pore-scale structures and processes. By combining multi-scale imaging (e.g., X-ray CT, SEM/FIB-SEM) with machine learning models, the complete pore architecture of laboratory-scale plugs may be reconstructed at pore-scale resolution of digital twins. Numerical simulations on these digital twins enable spatially explicit predictions of permeability and diffusivity across the entire core sample, capturing the rock's inherent heterogeneity for the first time. These permits the prediction of flow and transport properties at various scales, improving the understanding of radionuclide migration in complex subsurface systems. Results are validated against laboratory measurements and delivered to consortium partners for integration into repository-scale safety models.

The findings contribute to reducing uncertainties in long-term safety evaluation of geological repositories and support the development of reliable, data-driven models for nuclear waste management.

Project overview

TRANSFER is a national interdisciplinary graduate school dedicated to advancing the scientific foundations for the safe disposal of high-level radioactive waste in geological repositories. The program brings together five Helmholtz Centers and five universities:

TRANSFER consortium: GFZ Helmholtz-Zentrum für Geoforschung (project coordination), Freie Universität Berlin (FUB), Forschungszentrum Jülich (FZJ), Helmholtz-Zentrum Dresden Rossendorf (HZDR), Karlsruher Institut für Technologie (KIT), RWTH Aachen, FSU Jena, TU Bergakademie Freiberg (TUBAF), Helmholtz-Zentrum für Umweltforschung (UFZ), Universität Greifswald (UG).

Ten doctoral researchers work on complementary projects spanning the entire repository system from molecular-scale radionuclide sorption on clay mineral surfaces to repository-scale two-phase gas transport modeling. The program emphasizes interdisciplinary training and structured knowledge transfer, preparing graduates for positions in science, industry, regulatory authorities, and policy.

Transferability

The methodology will be extended beyond Opalinus Clay.

Model transferability is assessed without retraining, and deviations are regarded as mineralogical differences and resolved using formation-specific scaling relationships.

 Scientific Impact

The project delivers:

• First validated and uncertainty-quantified multiscale upscaling approach for clay-rich formations.
• Enhanced prediction of flow and transport in heterogeneous mediums.
• Reduced uncertainty in radionuclide migration and long-term safety assessment processes.

 This methodology is directly applicable to:

•  CO₂ geological storage
•  Subsurface hydrogen storage.
•  Environmental engineering and groundwater contamination.
• Characterisation of geothermal reservoirs.