Abstract Native hydrogen and helium have been considered important resources in assisting the energy transition. Hydrogen and helium seeps have been reported worldwide, which may indicate large reserves within the subsurface. However, generation of hydrogen and helium is complex; poorly understood and constrained for both generation processes and migration. One source of native hydrogen is ultramafic rocks, which have experienced serpentinization together with water radiolysis. In contrast, helium generation occurs as the result of the radioactive decay of uranium and thorium present within radiogenically enriched basement. An exploration tool, dedicated to identifying areas with the geological settings and conditions favourable for native hydrogen and helium generation, has been developed and tested. Several databases have been created and integrated as part of this study (geological and geochemical generation models) to support and focus the search for both hydrogen and helium. Machine learning algorithms which extract value from geospatial data types for detecting various accumulations have been implemented. The first machine learning results demonstrate the significant value in integrating data and machine learning for high grading areas more conducive to accumulating hydrogen and helium.

Abstract Knowledge of subsurface stresses is critical to management and prediction of fluid behaviour in the Earth's crust. However, uncertainties associated with the estimation of vertical stress magnitudes are rarely explored. In settings where the principal stresses are close, small changes of only a few MPa can have drastic effects on, stress regime prediction, fault reactivation and expected fracture behaviour. Using petroleum datasets from the Moomba Gas Field in the Cooper Basin, Australia, we assess the effects of interpolation between gaps in density wireline logs that have been filtered to remove sections of poor density tool contact in coal-bearing sequences. In addition, we compare the sonic velocity–density Nafe–Drake and Gardner transforms by comparing estimates of density from sonic velocity against density from wireline logs. The results indicate that vertical stress could be underestimated by c. 3 MPa at c. 3 km in well Moomba 61, using traditional methods. The Gardner transform shows a stronger correlation between sonic and density values for rocks in the Cooper Basin. However, once calibrated to the regional sonic velocity–density trend, the Nafe–Drake transform better matches the density logging tool in Moomba 61. This study delivers a more robust workflow for estimating the vertical stress in basins with deep coal-bearing strata.