In Switzerland, the Sectoral Plan for Deep Geological Repositories regulates the search for identifying suitable siting regions for the safe disposal of radioactive waste. To obtain a quantitative basis of fundamental rock properties such as mineralogy and porosity for numerical modelling of repository performance assessments and radiological consequence analysis, nine scientific boreholes were drilled by Nagra in northern Switzerland between 2018 and 2022. The aim was to assess the suitability of the Mid-Jurassic Opalinus Clay and its confining units for hosting a safe deep geological repository for radioactive waste. Over six kilometres of core were recovered and extensively analysed for porosity and mineralogy, alongside geomechanical, sedimentological, geochemical and hydrogeological analyses, supported by an extensive wireline logging program. A novel approach combining supervised Machine Learning and stochastic multimineral log analysis enabled continuous quantification of mineralogy and water content (total porosity), calibrated against core data. The machine learning improved the individual quantification of matrix silicates, carbonates and four major clay minerals: smectite, illite, kaolinite, and chlorite, which could not be satisfactorily resolved by conventional multimineral analysis alone. Total clay and porosity estimates correlated well with core measurements (R 2 = 0.91–0.94), with minor discrepancies in thinly laminated or heterogeneous intervals such as the upper Opalinus Clay, below the 1.5 ft log resolution. Uncertainties were quantified using Monte Carlo simulations incorporating random errors from wireline log repeatability and endpoint variability. A novel approach for quantifying the log random error standard deviation, based on the difference between the Repeat and Main logging passes, was successfully implemented in our study and proved its robustness. The combined Machine Learning-Multimineral workflow provided a reliable, core-calibrated representation of rock properties across the nine boreholes. The resulting continuous datasets form a robust quantitative basis for characterizing the Opalinus Clay and its confining units in support of Switzerland's deep geological repository programme.
Marnat et al. (Mon,) studied this question.