机器学习揭示玄武岩构造背景与源区性质  

Tectonic setting and source characteristics of basaltic rocks revealed by the machine learning

作　　者：郭鹏 GUO Peng(Department of Earth and Space Sciences,Southern University of Science and Technology,Shenzhen,Guangdong 518055,China)

机构地区：[1]南方科技大学地球与空间科学系,广东深圳518055

出　　处：《矿物岩石地球化学通报》2023年第1期26-33,共8页Bulletin of Mineralogy, Petrology and Geochemistry

基　　金：国家自然科学基金资助项目(42002046)。

摘　　要：玄武岩作为地幔的衍生物,是研究地幔物质组成与演化、地壳物质再循环和多圈层相互作用的重要介质。玄武岩的微量元素和同位素特征常被用于制约玄武岩形成的构造背景和地幔源区性质,但单一地球化学指标或二维图解方法常给出模棱两可或者互相矛盾的结果。与传统方法相比,机器学习方法能更全面和深入地分析数据,在多维空间上挖掘散点数据之间的内在联系和规律。本文简述了近年来机器学习方法在判别玄武岩构造背景、划分地幔组分与揭示玄武岩源区性质等方面取得的一系列成果,以期通过这些方面的应用实例为地幔地球化学研究带来新的认识。机器学习有望成为研究地幔深部过程与宜居地球形成机制的重要手段。As a derivative of the mantle, basalt is an important research object for studying the composition and evolution of mantle, the recycling of crustal materials, and the interactions between different spheres. The characteristics of trace elements and isotopes of basaltic rocks are commonly used to constrain their tectonic setting and source characteristics. However, previous geochemical proxies and two-dimensional discrimination diagrams of geochemical data of basaltic rocks often give us some ambiguous or contradictory results. Comparing to traditional methods, the machine learning method can be used to more in-depth and comprehensively analyze geochemical data and to mine the internal connections and regulations among scattered data in multidimensional spaces. This article briefly reviews the current achievements of the machine learning method in identifying the tectonic setting, discerning the mantle components, and revealing the lithology of the basalt source. These application achievements have brought some new understandings to the study of mantle geochemistry. The machine learning is expected to be an important way for effectively studying the deep mantle processes and the formation mechanism of a habitable Earth.

关 键 词：机器学习 玄武岩 地幔地球化学 

分 类 号：P591[天文地球—地球化学] P595[天文地球—地质学] P599