Quantifying Uncertainties Linked to the Diversity of Mathematical Frameworks in Knowledge-Driven Mineral Prospectivity Mapping

Knowledge-driven mineral prospectivity mapping (MPM) has been practiced with diverse algorithms for combining GIS-based predictor layers (i.e., exploration targeting criteria) into predictive models in greenfield areas. Due to the diversity of these algorithms, there is a lack of consensus on the output of methods derived; that is, different algorithms employed generate distinct outputs, rendering the MPM uncertain. This specific type of uncertainty poses certain challenges to selecting reliable exploration targets. Measuring this type of uncertainty and linking it to the decision-making process are, therefore, two issues that merit serious consideration. This study adopted a four-part framework for tackling the above problem, namely (i) generating several predictive models using different algorithms, (ii) measuring uncertainties linked to discrepancies in predictive values derived from different algorithms, (iii) integrating the output of different algorithms into a combined predictive model, and (iv) selecting low-risk target zones based on low values of uncertainty and high combined predictive values. Three algorithms, namely AHP, TOPSIS, and fuzzy logic, were applied to generate three interim predictive models, followed by developing a combined predictive model using the self-organizing map (SOM) technique. Plots of combined predictive values vs. measured uncertainty values were used to select low-risk target zones. These zones cover ~ 5% of the area investigated in this study, predicting a considerable number of its mineral occurrences. This methodology appears to be an appropriate option for enhancing the reliability of predictive knowledge-driven models of mineral prospectivity.