Unsupervised chemofacies classification from core HHXRF data in the Ohio Trenton–Utica succession: a reproducible multivariate workflow

Organic-rich mixed carbonate–siliciclastic unconventional successions commonly exhibit centimeter- to decimeter-scale mineralogical and geochemical heterogeneity that complicates stratigraphic interpretation and reservoir characterization. Although handheld X-ray fluorescence (HHXRF) resolves this variability at high resolution, multielement profiles are high-dimensional and strongly correlated, limiting reproducible translation of coupled elemental behavior into stratigraphically coherent chemofacies. Here, we present a reproducible informatics workflow for chemofacies delineation and interpretation using 20-element core HHXRF data (interpreted in terms of relative downcore variability) from two subsurface cores spanning the Upper Ordovician Trenton/Lexington, Point Pleasant, and Utica interval in Ohio, USA. Principal component analysis (PCA) reduces redundancy and provides a stable feature space for distance-based classification. K-means clustering of retained PCA scores delineates five chemofacies: (i) Detrital Dolomitic, (ii) Siliceous Argillaceous, (iii) Mixed Silica–Argillaceous, (iv) Carbonate-Dominated, and (v) a localized Mo-enriched anoxic end-member that marks extreme conditions over a short stratigraphic thickness. Exploratory factor analysis (EFA) extracts three interpretable latent factors representing terrigenous/siliciclastic influence, redox-sensitive trace-metal enrichment, and anoxia/sulfidic conditions. Median factor scores and Kruskal–Wallis tests ( p < 0.00001) provide statistical support for overall differentiation among chemofacies in factor-expression space. The workflow is transferable to other core-based geochemical datasets and reduces subjectivity in facies-based chemostratigraphic interpretation.