Geochemical discrimination and characteristics of magmatic tectonic settings; a machine learning-based approach

摘要

Geochemically discriminating between magmatism in different tectonic settingsremains a fundamental part of understanding the processes of magma generationwithin the Earth's mantle. Here, we present an approach where machine-learning(ML) methods are used for quantitative tectonic discrimination and featureselection using global geochemical datasets containing data for volcanic rocksgenerated in eight different tectonic settings. This study uses support vectormachine, random forest, and sparse multinomial regression (SMR) approaches. Allthese ML methods with data for 20 elements and 5 isotopic ratios allowed thesuccessful geochemical discrimination between igneous rocks formed in eightdifferent tectonic settings with a discriminant ratio better than 83% for allsettings barring oceanic plateaus and back-arc basins. SMR is a particularlypowerful and interpretable ML method because it quantitatively identifiesgeochemical signatures that characterize the tectonic settings of interest andthe characteristics of each sample as a probability of the membership of thesample for each setting. We also present the most representative basaltcomposition for each tectonic setting. The new data provide reference pointsfor future geochemical discussions. Our results indicate that at least 17elements and isotopic ratios are required to characterize each tectonicsetting, suggesting that geochemical tectonic discrimination cannot be achievedusing only a small number of elemental compositions and/or isotopic ratios. Theresults show that volcanic rocks formed in different tectonic settings haveunique geochemical signatures, indicating that both volcanic rock geochemistryand magma generation processes are closely connected to the tectonic setting.