Rhian Dentelbeck, MSc

The Abitibi Greenstone Belt (AGB) is host to world-class base metal deposits. However, increasing depletion of near-surface deposits requires improved geological models to guide the identification of future exploration targets. Carbonaceous argillites, or black shales, are common ore-hosting units in the AGB and their distinct electromagnetic properties has resulted in many of these rocks being targeted during drilling campaigns. This resulted in a vast yet widely underappreciated archive that holds valuable information to locate deeper mineral deposits.

The black shales are mainly synvolcanic and formed in volcanosedimentary basins spatially correlated with magmatic complexes. This study uses over 500 samples of carbonaceous argillite collected during a regional-scale sampling program in 2011. Whole-rock geochemical analysis was performed on 564 samples from more than 75 different townships. The objectives of the present study are to develop a detailed classification of carbonaceous argillite from the AGB as a tool for mineral exploration including improved understanding of i) metal enrichment, ii) source rock domains, and iii) distinguishing between barren and potentially mineralized horizons. Black shale is visually unclassifiable, however, using geostatistical methods, we have identified groups related to source rock composition (felsic, mafic), mineralization style, and post-depositional processes.

Alteration indices highlight hydrothermally influenced units, including mineralized shales and shales that were derived from volcanic rocks already altered at their source. The europium anomaly (Eu/Eu*) and Zn/Na in particular, identify known mineralized shales in mafic-dominated assemblages including Kidd-Munro and Stoughton-Roquemaure. Two machine learning approaches identify key relationships between the different mudstone units, including spatial variations in the shale geochemistry and their relationship to the host volcanic rocks. Principal Components Analysis (PCA) and Random Forest Clustering identify the element groups that explain most of the variance in the dataset and predict the source rock composition with an accuracy of 92%. (e.g., vanadium, scandium, and chromium are particularly important for distinguishing between felsic and mafic source rocks). Blind tests show the Random Forest model developed using the AGB samples can predict source rock composition and potential links to mineralization (VMS versus orogenic gold targets) in other regions.