Name Surname
Robust multi-phase machine learning thermo-chemo-barometry of volcano plumbing systems
Main Supervisor: Olivier Namur (Katholieke Universiteit Leuven)
Co-Supervisors: Livio Fano (University of Perugia), Francois Holtz (Gottfried Wilhelm Leibniz Universität Hannover)
Objectives
Reconstructing the plumbing systems of active volcanoes is essential to understand their eruptive style and minimize the threat they pose to populations. Mildly alkaline to alkaline volcanoes forming ocean islands are amongst the most active volcanoes in the World. Some of these volcanoes are continuously monitored by geophysical methods (e.g., seismometers) but these methods have shown strong limitations in imaging small magma lenses where magma differentiation is assumed to take place. Geothermobarometric equations based on mineral or mineral-liquid phase equilibria can provide direct records of the conditions of crystallization. However, these models need experimental calibrations which are currently lacking for alkali-rich systems or are of insufficient quality. Here we propose to conduct high-temperature, low- to high-pressure experiments on a range of alkaline magmas. The composition of experimental products (e.g., pyroxene, amphibole and co-existing melts) will be combined with existing data from the literature and the resulting dataset will be processed by complementing thermodynamic rules and ML algorithms to derive new physics informed ML models on phase equilibria and phase compositions as a function of pressure, temperature and melt composition.
Epected Results
- The main result will be the development of data-driven and physics-informed models that will allow calculating accurately the pressure-temperature conditions of magma crystallization under active volcanoes;
- The second critical result will be a much stronger understanding of the phase equilibria of alkaline magmas which will allow explaining the mineralogical and compositional variability observed at ocean island volcanoes;