Name Surname
Advancing petro-volcanological hypothesis formulation and thermodynamic insights through Machine Learning
Main Supervisor: Andras Lukacs (Eötvös Loránd Tudományegyetem)
Co-Supervisors: Diego Perugini (University of Perugia), Szabolcs Harangi (Eötvös Loránd Tudományegyetem)
Objectives
Trace elements are pivotal in understanding and constraining pre-eruptive dynamics, especially in silicic systems that are prone to behave explosively, posing a significant risk to society. The project aims at developing new data-driven and physics-informed predictive models to elucidate trace element partitioning, as function of pressure, temperature, and magma composition. The investigation will complement literature data with new determinations on different volcanoes, including samples the Ciomadul Volcanic Complex, a dormant volcano with evidence of a persistent magma reservoir, where past eruptions pointed to the potential occurrence of fast crystal mush reactivation. The extraction of new parametrizations in the context of the lattice strain model by symbolic regression and additional ML techniques will be also investigated.
Epected Results
- The development of new predictive models to constrain trace element partitioning combining experimental and natural samples for silicic systems, often posing significant hazard;
- Improve our knowledge on pre-eruptive dynamics of the Ciomadul Volcanic Dome Complex, used as a proxy for in silicic systems.