5:15 PM - 6:30 PM
[MGI33-P01] Numerical modeling of mantle melting degree based on major element composition of melt
Keywords:mantle, melting, data driven
Determining the mantle melting degree is essential to understand the physical and chemical conditions of the mantle, the process of melt generation, and for estimating the chemical composition of the source mantle based on the chemical composition of volcanic rocks. Due to the low thermodynamic degrees of freedom during partial melting of mantle peridotite, the composition of the partial melt is controlled by the phase relation, including phase assemblage of residual minerals and the melt fraction. Therefore, it is expected that the melt fraction can be expressed only as a function of melt composition, although it is generally expressed as a function of temperature with a fixed bulk composition of the source rock and a fixed pressure. Based on this background, we conduct numerical modeling of mantle melting degree based on the major element composition of the melt.
We compiled the previously reported experimental results of mantle peridotites at pressures corresponding to the upper mantle. Both hydrous and anhydrous experiments are included in the compiled database. We use melting experiments in which the melting degree at a given bulk composition was reported. We conduct a regression analysis based on the data-driven method in Ueki et al. (2018 and 2020). Based on the method, a small number of essential variables are automatically selected to obtain a model of robustness and high predictive capacity. A model that calculates the degree of melt based on the melt composition without temperature and pressure information is constructed based on our analysis. This model will allow us to estimate the physical and chemical conditions of the source region of various natural primitive magmas.
We compiled the previously reported experimental results of mantle peridotites at pressures corresponding to the upper mantle. Both hydrous and anhydrous experiments are included in the compiled database. We use melting experiments in which the melting degree at a given bulk composition was reported. We conduct a regression analysis based on the data-driven method in Ueki et al. (2018 and 2020). Based on the method, a small number of essential variables are automatically selected to obtain a model of robustness and high predictive capacity. A model that calculates the degree of melt based on the melt composition without temperature and pressure information is constructed based on our analysis. This model will allow us to estimate the physical and chemical conditions of the source region of various natural primitive magmas.