A Framework for Integrating XRF and Magnetics in Sedimentary Provenance Assessment

Thesis Abstract

Sedimentary provenance assessment requires robust, integrated geochemical and geophysical approaches to resolve the sources, transport pathways, and tectono-kinematic controls that shape sedimentary records. This study addresses the need for a coherent framework that combines X-ray fluorescence (XRF) spectroscopy with magnetic susceptibility and related magnetics data to enhance provenance discrimination and sediment transport interpretation in complex sedimentary basins. The aim is to develop and validate an integrative framework that links compositional signatures from XRF with magnetic properties to improve source-to-sink attribution. Specific objectives are (i) to compile a representative sedimentary sample set (n = 1,200 hand specimens and n = 300 polished slabs) across stratigraphic intervals spanning deltaic to paralic environments; (ii) to generate high-resolution XRF bulk geochemical data for major, trace, and rare earth elements alongside magnetic susceptibility, frequency-dependent susceptibility, isothermal remanent magnetization (IRM), and hysteresis parameters; (iii) to develop a data fusion workflow that integrates XRF geochemistry with magnetics through multivariate statistical methods and machine learning algorithms; (iv) to formulate a hierarchical provenance model that calibrates XRF–magnetics signatures against known source terranes using modern rivers and outcrop analogs; and (v) to evaluate the framework’s performance against conventional provenance indicators (including petrography, detrital zircon U-Pb age spectra, and clay mineralogy). The methodology adopts an explanatory mixed-methods design grounded in Bayesian inference and geostatistical principles. The study area comprises a tectonically active frontier basin with well-characterized uplifted source regions and stratigraphic records spanning mid-Cretaceous to late Eocene. Population and sampling involve sedimentary units selected to maximize provenance heterogeneity, with stratigraphic control guiding sampling density to 100 units per lithofacies. Instrumentation includes XRF spectroscopy for bulk elemental concentrations (SiO2, Al2O3, Fe2O3, TiO2, K2O, Na2O, P2O5, MnO, Cr2O3, Ni, Sr, Zr, and REEs), in-situ logging of magnetic susceptibility (MS) and frequency-dependent susceptibility (FD), as well as rock magnetic experiments (ARM, IRM acquisition curves, coercivity spectra) conducted on representative aliquots. Data collection will follow standardized preparation protocols to ensure inter-laboratory comparability, with duplicate analyses (n = 10%) to assess analytical precision. The analytical framework comprises three interconnected components. First, multivariate statistical analyses (principal component analysis, hierarchical cluster analysis, and canonical correlation) will identify covariation between geochemical and magnetic features and their linkage to potential source rocks. Second, a supervised machine learning module (random forest, support vector machines, and gradient boosting) will classify provenance signals by source terrane using calibrated training sets from modern river sediments and outcrop suites; model performance will be evaluated by cross-validation, receiver operating characteristic (ROC) curves, and confusion matrices. Third, a Bayesian hierarchical model will integrate probabilistic provenance assignments across stratigraphic levels, quantifying uncertainty in source attribution and sediment-routing pathways. Theoretical grounding draws on tectono-sedimentary frameworks and the rock magnetism theory of compositionally driven magnetic mineralogy, supplemented by the detrital balance concept and provenance theory of sedimentary dispersal. Expected findings include (i) a reproducible XRF–magnetics signature set for proximal and distal provenance scenarios, (ii) a robust data fusion schema that improves discrimination between crystalline and sedimentary source rocks, and (iii) a probabilistic provenance model with quantified uncertainty at multiple stratigraphic scales. It is anticipated that the framework will reveal systematic covariations between certain geochemical proxies (e.g., high SiO2 and low Fe2O3/MnO where siliceous carbonates predominate) and magnetic parameters (elevated low-coercivity minerals in juvenile sediments) that discriminate between basement-dominated versus sedimentary-cover sources. The study will contribute to knowledge by demonstrating a transferable, scalable integration approach that reduces reliance on costly detrital zircon dating or detailed petrography in preliminary provenance assessments. The study’s contribution to knowledge lies in operationalizing an integrative XRF–magnetics framework for sedimentary provenance that is transparent, reproducible, and adaptable to other basins worldwide. Practical implications include improved expeditious ranking of potential source areas for hydrocarbon exploration, mineral resource appraisal, and paleoenvironmental reconstruction, with explicit guidance on sampling density, analytical protocols, and data-sharing standards. Recommendations include expanding the framework to incorporate additional geophysical proxies (e.g., paleomagnetic orientation data), validating the approach in cratonic versus orogenic settings, and developing open-access software tools to facilitate broader adoption by geologists and basin analysts. The main conclusion is that XRF and magnetics, when integrated within a probabilistic, theory-grounded framework, provide a powerful, cost-efficient means to enhance sedimentary provenance assessments across diverse tectonosedimentary settings.