Statistical Techniques for Leveraging Geochemical Data in Ore and Non-Ore Characterization for Mining and Environmental Stewardship

The timeline from exploration through to extraction of a mineral deposit often spans decades, resulting in multi-generational geochemical data collected utilizing a variety of digestion and analytical methods. To extract value from these diverse datasets is challenging. This is due to lack of comparability in elemental concentrations produced by different digestion and analysis methods. Relationships between these multi-generational, variable digest geochemical datasets are typically non-linear, requiring a more sophisticated approach to data integration. Two case studies are presented to address this integration problem using simple machine learning workflows. Case study 1 outlines a workflow to derive a common molar element ratio used in porphyry deposit exploration and alteration quantification (2Ca-Na-K/Al) from 4-acid digestion data as a proxy for the degree of feldspar destruction caused by hydrothermal metasomatism. It further illustrates the prediction of this ratio (derived from 4-acid digestion geochemistry) using aqua regia digestion geochemical data as an input. Case study 2 illustrates the use of aqua regia-derived Ca as a proxy for neutralization potential in mineralogical systems dominated by carbonate dissolution in aqua regia digestion, and presents a workflow to predict neutralization potential from 4-acid data, trained to aqua regia Ca. Both case studies showcase the integration of aqua regia and 4-acid datasets via non-linear machine learning algorithms, which exploit the mineralogical and elemental controls governing differences between digestion methods.