Identification of the Original Tectonic Setting for Oceanic Andesite Using Discrimination Diagrams: an Approach Based on Global Geochemical Data Synthesis

Systematical analyses of data from GEOROC and PetDB database show that large amount of Cenozoic andesites occurred in the various oceanic environments such as mid-oceanic ridge, plume-related island and oceanic arc. In this study, we employed the geochemical data of 351 mid-ocean ridge andesites (MORA), 2 539 plume-related andesites (PRA) and 3 488 oceanic arc andesites (OAA) from the database to discuss the relationship between andesite tectonic settings and their geochemical features, thereby making an attempt to construct tectonic discrimination diagrams. Based on the data-driven pattern, all available elements were employed to derive logratios for the possible coordinates, and the overlap-rate calculation was adopted to evaluate the discrimination effect of more than 330 000 prospective diagrams. Finally, four tectonic discrimination diagrams have been successfully established to identify MORA, PRA and OAA, which can be utilized to identify the original settings of andesite with an age range from Cenozoic to Archean a certain extent. Of these diagrams, PRA is mainly distinguished by high LREE/HREE ratio due to enriched mantle source. Whereas, OAA is mainly characterized by high LILE/HFSE ratio, which reveals that fluids derived from subducted slab play an important role in forming oceanic arc andesites. Consequently, the petrogenesis of andesites is closely related to their tectonic settings. However, it should be noted that those andesites formed in both continental and oceanic environments cannot be effectively distinguished using these diagrams. We strongly recommend integrating the discrimination diagrams result with other geological information to reach a comprehensive interpretation of evolution history with those ancient andesites. This paper presents a case study which suggests that data-driven method is a powerful tool for solving geological problems in this ‘big data” era.