Semi-automated identification of regolith/basement interfaces using hyperspectral and geochemical data

Regolith processes can mask, alter, and limit the expression of underlying mineralisation and related alteration halos by restricting the migration of most trace elements to the surface. Characterising cover sequences may permit geochemical and mineralogical features and boundaries in this zone to be used to enhance rather than hinder mineral exploration. Mapping and characterising regolith interfaces is a key theme of the MinEx CRC program, with the objective of evaluating a range of in situ data that permit near real-time decision-making in drilling programs linked to the National Drilling Initiative. Multivariate analysis of geochemical and mineralogical data, derived from techniques such as pXRF, and hyperspectral data acquired by portable spectrometer or HyLoggerTM, using Data MosaicTM and optimised machine learning approaches are being used to improve identification of significant interfaces within regolith and basement profiles. A semi-automated workflow based on both data-driven and knowledge-driven approaches has been developed for cover characterisation of drillholes in the Cobar Basin. An optimised subset of geochemical and spectral variables available from scanning a test drillhole through portable instruments has boosted the boundary detection and classification accuracy in cover characterisation up to 96%. This workflow is being extended to include other data types (e.g., petrophysical parameters) to assist in classification and characterisation of drill materials as part of the process of improving the effectiveness of exploration in regions containing transported or deeply weathered cover.