Machine Learning Reveals Source Compositions of Intraplate Basaltic Rocks
Abstract Recycling of crustal material is thought to introduce pyroxenite to the peridotite mantle. Mapping such lithological heterogeneity within the mantle is crucial to understanding the mantle's chemical evolution but remains challenging. By sampling the mantle source, intraplate basaltic m...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2021-09-01
|
| Series: | Geochemistry, Geophysics, Geosystems |
| Online Access: | https://doi.org/10.1029/2021GC009946 |
| Summary: | Abstract Recycling of crustal material is thought to introduce pyroxenite to the peridotite mantle. Mapping such lithological heterogeneity within the mantle is crucial to understanding the mantle's chemical evolution but remains challenging. By sampling the mantle source, intraplate basaltic melts provide a unique chance to reveal lithological heterogeneity within the mantle. We train machine learning (ML) models with major oxide data of experimental peridotite and pyroxenite melts to help reveal the mantle source lithology of basaltic rocks. The ML models can predict source lithologies from major oxide information with an accuracy larger than 94%. As a case study, we predict source lithology of the Cenozoic intraplate basaltic rocks in Northeast China. Our ML models suggest that pyroxenite dominates the mantle source of basaltic rocks sitting above the stagnant Pacific slab while peridotite dominates the source of the basaltic rocks located west of the slab tip, consistent with previous studies using other approaches. Our ML models could potentially be used to infer mantle source lithologies of basaltic rocks from other regions around the world. |
|---|---|
| ISSN: | 1525-2027 |