Quantitative Mineral Mapping of Drill Core Surfaces II: Long-Wave Infrared Mineral Characterization Using<i>μ</i>XRF and Machine Learning

Abstract Long-wave infrared (LWIR) spectra can be interpreted using a Random Forest machine learning approach to predict mineral species and abundances. In this study, hydrothermally altered carbonate rock core samples from the Fourmile Carlin-type Au discovery, Nevada, were analyzed by LWIR micro-X-ray fluorescence (μXRF). Linear programming-derived abundances quantified μXRF data used as training construct series of regression models. The models produced proportion estimates with root mean square errors 1.17 6.75% (model predictions) 1.06 6.19% (compared quantitative X-ray diffraction data) for calcite, dolomite, kaolinite, white mica, phlogopite, K-feldspar, quartz. These results are comparable error linear spectral deconvolution (±7–15%), commonly unmixing technique. Having mineralogical chemical set makes it possible identify quantify mineralogy provides more robust meaningful interpretation than current methods utilizing library or end-member extraction. Using method presented here, spectroscopy overcome limitations inherent use short-wave (SWIR) in fine-grained, low reflectance rocks. This new applied any deposit type, improving accuracy speed interpretation.