Laukamp, Dr Carsten1, Francis, Neil1, Hauser, Dr Juerg1, Gopalakrishnan, Dr Suraj2, Mule, Shane1
1CSIRO, Perth, Australia, 2Geologial Survey of Queensland, Brisbane, Australia
The combination of magnetic susceptibility and density allows identification of iron oxide copper-gold (IOCG) mineralisation by estimating proportions of magnetite, sulphide and hematite alteration which can be indicative of IOCGs. In the frame of the National Virtual Core Library project, hyperspectral reflectance spectra acquired from drill core of the Osborne Cu-Au deposit, Mount Isa Inlier, Queensland using a HyLogger3 at GSQ’s Exploration Data Centre were compared with magnetic susceptibility and density measurements. In this study we explore the feasibility of inferring the petrophysical data from the 1) visible-near (VNIR), shortwave (SWIR) and 2) thermal (TIR) infrared wavelength regions. Specifically, we seek to predict magnetic susceptibility and density values in drill core sections where petrophysical data are not available and potentially extrapolate these to other hyperspectral data sets, such as those acquired by field or Earth Observation instruments.
Using The Spectral Geologist (TSGTM) software, partial least squares (PLS) was employed to derive predictive models using 23 unique magnetic susceptibility and density measurements, that were assigned to all nearby spectral measurements (+/- ~10cm). The values of the input magnetic sustainability and density values ranged from 0 to 2.3 K (Si) and 2.7 to 4.9 g/cm3, respectively. The hyperspectral data were not spatially re-sampled to fit with the drill core interval measured for petrophysical data. Instead, the original 1 cm spatial resolution was used to evaluate the variability of hyperspectral data within the petrophysical sample intervals. The correlation between the 23 measured and corresponding modelled magnetic susceptibility (n = 157) for the same 23 depth intervals was high for the VNIR-SWIR (r2 = 0.95) and the TIR (r2 = 0.969), but the PLS-modelled magnetic susceptibility values showed a large variance (± 0.8 and ± 0.5, respectively). Similarly, the correlation between the measured and modelled density was high for the VNIR-SWIR (r2 = 0.958) and the TIR (r2 = 0.989), with the PLS-modelled density values showing a large variance (± 0.4 g/cm3 for both wavelength ranges). However, HyLogger3 high-resolution RGB imagery showed that the predicted value ranges were sufficiently different to discriminate drill core intervals dominated by magnetite-rich rocks, from magnetite-rich breccia and least-altered (non-mineralised) rocks. PLS models based on the VNIR-SWIR wavelength ranges were mainly driven by depth changes of electronic transition absorption features related to iron and copper in the VNIR, which are most intense in the highly altered, magnetite- and/or sulphide rich rocks. PLS models based on the TIR wavelength ranges were highly influenced by the thermal background typically associated with iron oxides and sulphides. Density values modelled from VNIR-SWIR compared to those modelled from TIR showed a good correlation (r2 = 0.729), whereas the correlation between magnetic susceptibility modelled from VNIR-SWIR and TIR was comparably low (r2 = 0.518). While the small amount of data used to infer the models discussed here means that their predictive power needs to be assessed comprehensively, our results nevertheless indicate a high potential for successfully inferring petrophysical from hyperspectral data and cost-effective mapping of IOCG-related alteration.
Carsten Laukamp is a senior research geoscientist at CSIRO Mineral Resources, based in Perth, Australia and is project leader of the National Virtual Core Library. Carsten explores the potential for combined use of reflectance spectroscopy, geochemistry and geophysics for tracing hydrothermal alteration signatures through cover and advancing ore body knowledge.