Integrating Petrophysical Logging and XRF Data for Mineral Fraction Estimation of Lower Crustal Rocks from the ICDP-DIVE Project using a Bayesian Inversion Framework

The mineral abundance, their properties and geometrical arrangement on small spatial scales directly affect the physical characteristics of the continental crust at large scales. Consequently, the mineral assemblages determine to a large extent how geophysical methods respond to these rocks. Determining the mineral volume fractions is an essential first step for modelling and interpreting geophysical data, constraining crustal structure, and understanding the evolution of the Earth’s lithosphere. In this study, we develop a Bayesian inversion framework that integrates petrophysical information from downhole well logs and multi-sensor core logging data with X-ray fluorescence (XRF) data to estimate continuous mineral fraction profiles along two ICDP-DIVE boreholes (Greenwood et al. 2026) drilled through the exhumed lower continental crust of the Ivrea–Verbano Zone (IVZ) with almost 100% core recovery. The framework involves two schemes: (1) an overdetermined inversion of relative sparse XRF oxide weight fraction data from powdered rock samples combined with core density logs, and (2) a severely underdetermined inversion of potassium, magnetic susceptibility, and core density logs, conducted by groups derived from a cluster analysis of these logs. The latter scheme is constrained by the first scheme, which allows to retrieve a continuous mineral fraction estimates along both boreholes from the limited number of 3 petrophysical logs. An ensemble Markov Chain Monte Carlo algorithm (Cheng et al. 2022) is adapted to recover the posterior mineral fraction distributions while quantifying uncertainties. An essential input is the prior knowledge of the minerals present and their chemical formula, which may require supplementary measurements, especially for minerals such as amphibole, whose chemical formula is difficult to determine. The results show that the XRF Oxide–density inversion approach provides robust mineralogical estimates that are consistent with independently obtained modal estimates from section observations. The constrained inversion of the petrophysical logging data successfully captures mineral fractions across most lithologies despite the underdetermined nature of the problem. The study demonstrates that combining XRF-derived oxide fractions with continuous downhole and core logging data within a Bayesian framework provides a powerful approach for obtaining quantitative, mineral fractions in a range of lower crustal lithologies.