Translating mineral systems criteria into a prospectivity model for IOCG deposits in the Kolari region, Finland

Abstract

The mineral system approach facilitates mapping the fingerprints of geological processes in ore formation using mineral prospectivity modeling (MPM). It aims to reduce the time and cost of exploration as a priority in the mineral exploration industry. Finland, which is covered with thick soil, dense vegetation, and snow in winter and thus has limited outcrops, serves as a suitable testing ground for this method. In this contribution, MPM was employed to predict favorable targets for Fe-Cu-Au (IOCG) deposits in the Kolari region, northwestern Finland. The mineral system components ore metal and sulfur sources, pathways, energy sources/drivers, traps, and the geological factors influencing the ore-forming processes were translated into mappable criteria and were considered for a regional-scale analysis. Knowledge-driven, fuzzy logic overlay, fuzzy inference system, and geometric average methods integrate evidential layers derived from geological (scale-free geology map), geochemical (till and bedrock sample geochemistry), and geophysical data (magnetic, radiometric, electromagnetic measurements, and gravity worms). Subsequently, the data were modeled using a combined conceptual/empirical approach, i.e. fuzzified evidential layers and logistic regression. In the prospectivity models, existing IOCG deposits are consistently placed within high-favorability areas. In addition, new exploration targets were identified. The models were validated against known IOCG deposits using Receiver Operating Characteristic (ROC) analysis and average Area Under the Curve (AUC) with values of final prospectivity models consistently reaching scores > 0.8. This indicates a favorable outcome of the application of MPM, making the approach applicable to other regions and ore deposit types. A majority voting ensemble technique was employed to combine the favorable areas of each prospectivity model. Then a confidence index was adopted reducing uncertainty linked to the models generated. These outputs consistently facilitated the identification of exploration targets more reliably.