Data-driven expeditious mapping and identifying granites in covered areas via deep machine learning: A case study on the implications for geodynamics and mineralization of Eastern Tianshan

Abstract

Mapping granites and identifying their genetic types are critically important for understanding the tectonic evolution and patterns of metallogenesis of a defined region. This is because granite usually serves as a significant geological probe enabling us to track and examine the process of crustal accretion and recycling in various geodynamic contexts. However, even geological mapping of granite, especially at large regional scales, in areas covered by regolith, vegetation, water, or ice, remains a considerably challenging task due to the coverage, resulting in incomplete surveys originating from inaccessible measurements. Therefore, we proposed a data-driven framework that utilizes the deep learning method to integrate multi-source survey data, including geophysical, geochemical, and remote sensing data, to map granites and identify their genetic types in covered areas. For demonstration purposes, the Gobi-desert covered area of the Eastern Tianshan orogenic belt, which is an important mineralization belt and exhibits a complex tectonic evolution, with extensive exposures of various genetic granites primarily formed during the Late Paleozoic, has been used as a case study. Using Bouguer gravity, aeromagnetic, geochemical data for stream sediments, and remote sensing data as predictors, we built two fully-connected deep neural networks for mapping undiscovered granites and identifying their genetic types in the study area, respectively. We applied the synthetic minority over-sampling technique to balance the data related to genetic types of granite. This study demonstrates that the proposed deep machine learning framework is a powerful data-driven tool for laborless and expeditious mapping of granite lithology and identifying genetic types in covered areas. It develops a novel model for the tectonic evolution history of the Eastern Tianshan from the perspective of the spatiotemporal distribution of granites with different origins. This model reveals the metallogenic patterns and exploration potential of magmatic-hydrothermal deposits related to granite in the Eastern Tianshan region.