Answers are grounded in this paper and cite their sources.
Porphyry copper systems in southern Mongolia are important targets for mineral exploration, yet data-driven mineral prospectivity mapping has limited application in this region. In this study, we developed a Random Forest model to evaluate porphyry copper prospectivity and validated the results using independent hyperspectral evidence from the Environmental Mapping and Analysis Program. Despite limited data availability, a 15-layer predictor stack at 250 m resolution was created using geological, structural, topographic, geophysical, and Sentinel-2A-derived spectral indices. The model was trained using 264 balanced samples, including 132 positive samples buffered around 33 confirmed porphyry occurrences to reduce spatial overlap, and 132 negative samples selected using spatial exclusion masks. Two model configurations were tested: Model A, which included intrusion proximity variables, and Model B, which excluded them. Model A achieved an area under the receiver operating characteristic curve of 0.920 and an overall accuracy of 0.837, outperforming Model B, which reached a lower value of 0.745 and an accuracy of 0.714, highlighting the strong metallogenic influence of intrusive bodies. Prospectivity maps and top-decile target zones revealed anomalies that align with known deposits at Tsagaan Suvarga and Bronze Fox, and with prospects at Shuteen and Mandakh, while identifying additional unexplored target areas. Independent validation using Environmental Mapping and Analysis Program-derived alteration maps further demonstrated that top-ranked pixels from Model A overlapped with 22.23% of intense Al-OH alteration, 9.91% of Mg-Fe-OH alteration, and 31.12% when combined, outperforming Model B. These results indicate that combining Random Forest prospectivity mapping with hyperspectral alteration information improves prospectivity assessments for concealed porphyry exploration.
No takes yet. Share an insight, caveat, or question.
Badrakh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: