Gold Prospectivity Mapping in the Sonakhan Greenstone Belt, Central India: A Knowledge-Driven Guide for Target Delineation in a Region of Low Exploration Maturity.

The Sonakhan greenstone belt in Central India is under-explored with respect to gold in spite of its similarity to auriferous greenstone belts in general, which prompted a prospectivity analysis. The workflow involved adoption of a conceptual mineral systems model, recognizing indicative spatial proxies, processing exploration datasets, generating evidence maps and integrating into GIS-based mineral prospectivity mapping. Available geological information such as key lithologic units and their contacts was combined with geochemical anomalies of selected pathfinder elements, geophysical data (aeromagnetic anomaly and K/Th ratio map) and satellite digital image data (ASTER and Landsat 7 ETM +), leading to generation of 17 evidential layers. The lack of a significant number of known mineral occurrences in the study area precludes the use of data-driven prospectivity modeling techniques. Therefore, knowledge-driven approaches such as binary and multiclass index overlay, fuzzy logic and fuzzy AHP (analytic hierarchy process) were adopted to integrate the evidential layers resulting in four prospectivity maps. The variation of cumulative prospectivity with respect to cumulative area in each model was used to determine threshold to produce binary prospectivity maps separating high and low prospectivity zones. An approach based on the unique conditions of the binary prospectivity maps was used to illustrate the combined results of different models. In order to quantify the intuitive uncertainty in exploration targeting that arose due to different model outputs, a modulated predictive model was generated taking the mean prospectivity values at each pixel. The pixels having mean values above 95th percentile were grouped and the area delineated as potential exploration targets for gold that comprises merely 5% of the study area. The estimated uncertainty and confidence values for each pixel were used in the risk analysis that returned 1.95% and 3.05% of the study area as low- and high-risk exploration targets, respectively. [ABSTRACT FROM AUTHOR]