Mantle to Mine: An integrated Machine Learning, Minerals Systems and Geomechanical Approach to Copper and Gold Exploration

Many mineral deposits lie under thick accumulations of post-mineral cover, making exploration expensive and risky. Exploration targeting is thus often focussed mainly on various combinations of gravity and magnetic anomalies. In the case of IOCG style deposits the fluid pathway footprint is generally expressed as district-scale alteration and geochemical haloes. However, it is generally difficult to focus successful exploration due to the extensive geochemical 'smoke'. In addition, many seemingly attractive accumulations of iron oxide alteration are often barren or at best host low metal grades, and it is rarely obvious which iron oxide masses might host economic Cu and Au deposits. Our solution for successfully identifying the 'fire' in IOCG systems of the Gawler Craton of South Australia is to apply a range of methodologies that combine the strengths of both empirical and conceptual, process-based targeting. Our integrated 'Mantle to Mine' approach minimizes interpretation bias and reduces the reliance on specific deposit models. Specifically, we have utilized a data-driven methodology (machine learning) that systematically tests for the relative importance of a large number of geological factors on known deposits, which is additionally guided by an integrated mineral systems approach that analyses the processes operative in the passage of ore fluids and metals to their eventual deposition sites. Finally, we have utilized an analogue-driven methodology to test geological processes (geomechanical modelling) that simulates rock failure and ore fluid flow at potential metal trap sites in the upper crust to further hone district-scale targeting. As an independent test of their accuracy, the results of the machine learning and geomechanical modelling were compared with the location of known mineral occurrences and geochemical anomalies. The hit rate was very high in the case of both methodologies and this test has given us confidence in the predictions made for less well-known zones.Open-Access Online Publication: May 29, 202