Compositional estimation of protolith and metamorphic mass transfer in metabasaltic rocks: A Machine-Learning-Based Approach
Abstract
Geochemical data are essentially multidimensional and contain valuable information on geochemical processes in the Earth's interior. To evaluate the mass transfer due to geochemical processes (i.e., metamorphic mass transfer), it is necessary to quantitatively estimate the composition of protolith from samples that are directly available to us; however, in previous studies, only discrimination diagrams have been used to distinguish protolith types, it was not possible to quantitatively estimate the composition of the protolith.
Metamorphic mass transfer is usually evaluated by comparing the rock sample compositions with assumed protolith compositions. By contrast, at the regional scale, heterogeneities of the protolith compositions often obscure the effect of mass transfer; quantitative evaluations of mass transfer in regional metamorphism has been hindered. In this study, we developed a machine learning model for estimating the protolith composition from metabasalt and calculated the mass transfer due to metamorphism /alteration in metabasaltic rocks. The elements in the rock are divided into two groups: mobile elements and immobile elements. In other words, it is assumed that the concentration of immobile elements does not change even if they are affected by metamorphism and alteration. Accordingly, if immobile elements are set as input elements of the machine learning model and mobile elements are set as the output, it can be applied to metabasalt. Therefore, the developed machine learning model can estimate the composition of protolith from the metabasalt. The machine learning model was constructed by using fresh basalt composition from PetDB as the training data. The machine learning model was applied to the data of metamorphic and fluid-rock reaction experiments with known protolith, and it was found that the model was able to estimate the composition of the protolith correctly. Applying this machine learning model to altered/metamorphic rock, the chemical composition of the protolith can be estimated quantitatively. Furthermore, this model can be applied to the regional metamorphic terrains where their protoliths are heterogeneous; we can estimate regional mass transfer.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMV041...11M
- Keywords:
-
- 3613 Subduction zone processes;
- MINERALOGY AND PETROLOGY;
- 3660 Metamorphic petrology;
- MINERALOGY AND PETROLOGY;
- 8104 Continental margins: convergent;
- TECTONOPHYSICS;
- 8170 Subduction zone processes;
- TECTONOPHYSICS