Thermal conductivity evaluation using an integrated method: An example from the Horn River Basin, Canada

The Horn River Basin in northeastern British Columbia, Canada, hosts significant unconventional shale-gas resources and geothermal potential within the Devonian Horn River Group (HRG). The organic-rich HRG shale is highly heterogeneous due to complex mineralogy and matrix variability across the basin, posing challenges for accurate thermal conductivity (TC) prediction. This study presents an integrated approach to evaluate spatial TC variability by combining comprehensive petrophysical characterization with two-phase mixing models: the Voigt–Reuss–Hill (VRH) average approach and the geometric mean (GM) method. Machine learning techniques were applied to extensive mineralogical data to predict mineral compositions from well logs, while total organic carbon (TOC), porosity, and fluid saturation were quantified through core–log calibrations and statistical analysis. TC profiles were then calculated using the VRH and GM models, incorporating calculated component volumes and mineral TC values adjusted for in situ temperature and pressure conditions. Both measured and modeled results indicate that TC in the low-porosity shale is predominantly controlled by mineralogy/lithology, increasing with quartz and carbonate content and decreasing with higher clay volume. Notably, the VRH approach performs better for lower TC values (<2.5 W/m·K) associated with higher clay and lower quartz and carbonate contents, whereas the GM method is more accurate at higher TC values (≥2.5 W/m·K), corresponding to lower clay and higher quartz and carbonate contents. Given the heterogeneous mineralogy and high TOC content of the HRG shale, integrating both models effectively captures the full range of TC variability and enhances predictive accuracy.