Modulating the Impacts of Stochastic Uncertainties Linked to Deposit Locations in Data-Driven Predictive Mapping of Mineral Prospectivity

The operation of large-scale ore-forming processes triggers the development of neighboring mineral deposits of the same or related types in a metallogenic province. While these deposits often bear striking similarities, variations in local geological settings cause differences in many deposit features. Therefore, in a metallogenic province, geochemical, geophysical, and geological signatures of local areas mineralized with a certain deposit type can show considerable inherent differences. The application of deposit-type locations as training sites, thus, introduces a type of stochastic uncertainty into data-driven mineral prospectivity mapping (MPM), impairing the predictive capability of this activity. This study delves into this type of uncertainty and applies an ensemble technique combining bootstrapping and naive Bayes classifiers to measure this uncertainty and lessen its impact on the MPM-generated exploration targets. Two components, one representing the quantified uncertainty and the other a modulated predictive model, are retained by the proposed framework. This framework was applied to a suite of mineral-systems derived targeting criteria of skarn-type Cu mineralization in the Alborz-Azerbaijan magmatic belt of northern Iran. The predictive results derived by the proposed technique outperformed those derived using a single classifier, showcasing its efficacy. In addition, a novel approach is described and applied to demarcating exploration targets marked by low uncertainty.

Automated summary

The operation of large-scale ore-forming processes in a metallogenic province triggers the development of neighboring mineral deposits, which bear similarities but also differences due to variations in local geological settings. This inherent differences in geochemical, geophysical, and geological signatures of mineralized areas with certain deposit types introduces stochastic uncertainty into data-driven mineral prospectivity mapping (MPM), impairing its predictive capability.