4:15 PM - 4:30 PM
[MZZ45-13] Geochemical discrimination and feature Extraction global volcanic rocks using data-driven statistical analysis
Keywords:tectono-magmatic setting, machine-learning, geochemical data, volcanic rock
Geochemical discrimination and characterization of magmas formed in different tectono-magmatic settings are classical and fundamental issues of solid Earth science. This approach allows us to identify the tectonic origin of unknown volcanic rocks and to characterize and discuss the differences and similarities of various magmas. Discrimination and feature extraction of magmas formed in various tectono-magmatic settings require analysis of high-dimensional and large geochemical data sets consisting of multiple elements. Data-driven statistical analysis based on machine learning (ML) method is a powerful approach for conducting a high-precision analysis between multiple classes based on high-dimensional data.
We proposed an ML-based statistical approach for geochemical discrimination and feature extraction of magmas formed in different tectono-magmatic settings (Ueki et al., 2018, G-cubed). We have further developed an ML-based approach to perform feature extraction, and have identified a small number of fundamental geochemical features of different tectono-magmatic settings (Ueki et al., 2022, Frontier Earth Sci.).
We have newly constructed an ML-based geochemical discriminator for easy use that runs in the Excel spreadsheet (Ueki et al., 2024, G-cubed). The newly constructed system (Sparse Geochemical Tectono-magmatic setting Probabilistic membershiP discriminatoR: SGTPPR) allows us to conduct a high-precision geochemical analysis using the eight major (SiO2, TiO2, Al2O3, Fe2O3, MgO, CaO, K2O, and Na2O) and six trace elements (Rb, Sr, Y, Zr, Nb, and Ba) that can be analyzed by conventional methods such as X-ray fluorescence spectroscopy (XRF). The system outputs the probability of membership for eight different tectono-magmatic settings (mid-ocean ridge, oceanic island, oceanic plateau, continental flood basalt province, intra-oceanic arc, continental arc, island arc, and back-arc basin) for a given volcanic rock sample. We will present an overview of these series of machine learning-based analyses of global magmas. The application of the new geochemical discriminator to several volcanic rocks around the Japan arc will also be presented.
We proposed an ML-based statistical approach for geochemical discrimination and feature extraction of magmas formed in different tectono-magmatic settings (Ueki et al., 2018, G-cubed). We have further developed an ML-based approach to perform feature extraction, and have identified a small number of fundamental geochemical features of different tectono-magmatic settings (Ueki et al., 2022, Frontier Earth Sci.).
We have newly constructed an ML-based geochemical discriminator for easy use that runs in the Excel spreadsheet (Ueki et al., 2024, G-cubed). The newly constructed system (Sparse Geochemical Tectono-magmatic setting Probabilistic membershiP discriminatoR: SGTPPR) allows us to conduct a high-precision geochemical analysis using the eight major (SiO2, TiO2, Al2O3, Fe2O3, MgO, CaO, K2O, and Na2O) and six trace elements (Rb, Sr, Y, Zr, Nb, and Ba) that can be analyzed by conventional methods such as X-ray fluorescence spectroscopy (XRF). The system outputs the probability of membership for eight different tectono-magmatic settings (mid-ocean ridge, oceanic island, oceanic plateau, continental flood basalt province, intra-oceanic arc, continental arc, island arc, and back-arc basin) for a given volcanic rock sample. We will present an overview of these series of machine learning-based analyses of global magmas. The application of the new geochemical discriminator to several volcanic rocks around the Japan arc will also be presented.