Mineral prospectivity mapping of orogenic gold deposits using a mineral systems approach in the Huangling region, western Hubei

Abstract

The Huangling region hosts the only documented Precambrian gold metallogenic event within the Yangtze craton. However, dense vegetation significantly hinders direct-detection exploration techniques, necessitating a predictive targeting approach. This study proposes a mineral systems model for the gold deposits in the Huangling region, comprising four key components: geodynamics, fertility, architecture, and preservation. Geological, geophysical, and geochemical proxies corresponding to these components were identified. Regional tectonics associated with collisional processes are crucial geodynamic factors, elucidated through geomagnetic data analysis. Mantle-derived materials, consistent with the prevalence of mafic dykes, are considered the primary gold source, with coeval mafic dykes serving as a spatial proxy for architecture. Mineralization is structurally controlled by secondary NW- to NNW-trending faults, interpreted as fluid conduits and employed as a spatial proxy for architecture. While deposit preservation is reflected in post-formation geochemical signatures, the transported nature of stream sediment geochemical anomalies presents interpretational challenges. To address this challenge, the weighted drainage catchment basin method was implemented to refine anomaly patterns. For predictive modeling, both random forests (RF) and convolutional neural networks (CNN) were employed to integrate the four spatial proxies. The CNN model demonstrated superior performance, achieving an area under the curve (AUC) of 0.855, marginally outperforming the RF model (AUC = 0.797). Analysis of the success rate curve further revealed that the CNN model successfully predicted all known mineral occurrences within the top 50 % of the highest-probability zones. The application of deep learning methodologies exhibited remarkable efficacy in forecasting orogenic gold deposits. CNN-based mapping identified NW- to NNW-trending high-mineralization zones, which align with the known spatial distribution of gold deposits in the region. These results not only validate the model but also suggest promising exploration targets in adjacent and deeper areas. The majority of known deposits are situated within zones exceeding 80 % probability, while the model also identifies previously unexplored high-probability areas. These findings underscore the mineral systems approach as a robust and effective tool for prospectivity mapping of orogenic gold deposits.