ROUTINISED DATA FUSION AND CLASSIFICATION WORKFLOW TO IMPROVE OREBODY KNOWLEDGE AT THE ONTO Cu-Au DEPOSIT (Invited Presentation)

This is a presentation on the application of routinised data fusion and classification workflow on multi-element geochemical, spectral mineralogy and structural logging data collected from drill core at the Onto lithocap-hosted high sulphidation epithermal Cu-Au – porphyry deposit in the Huu district, NTB province, Indonesia. Data were collected from diamond drillholes that were selected to represent the range in the geology from distal to proximal parts of the deposit.

Data engineering workflows were undertaken to align the source data. Exploratory data analysis (EDA) approaches utilising deposit-specific geochemical ratios and indices can provide a rapid visual representation of alteration signatures. The spatial distribution and orientation of the interpreted hydrothermal alteration mineralogy from the geochemistry, spectral mineralogy and structural logging data enabled detailed characterisation of the extensive vertical and lateral argillic and advanced argillic alteration zones, which overprints older potassic porphyry alteration signatures.

Classical EDA approaches were augmented with several data dimension reduction and multivariate machine learning (ML) techniques, including Self-Organising Map (SOM), Uniform Manifold Approximation and Projection (UMAP), and Density-Based Scanning (DBSCAN) clustering technique. A combination of classical EDA and ML techniques applied at the deposit were found to be robust in repeatability across the selected drillholes. These were used to create training models for refining of existing lithological and alteration domains in the deposit. Supervised machine learning methods such as decision trees and random forest were utilised to identify parameters that can distinguish between domains, and to be used for classifying data from other drillholes outside of the training model.

Routinising the data analysis and interpretation process allows resource companies to productionise the creation of orebody knowledge. This approach allows geologists to 1) inform, augment and validate lithological and alteration core logging on-site, 2) Improve the reliability of the 3D geological model with follow-on benefits to the resource and geometallurgical models and 3) Incorporate the improved orebody knowledge as the basis for a conceptual system to identify other deposits in the vicinity of the Onto deposit.