Skip to content

Enhancing Lagrangian numerical simulations of Lava Flows Using AI-based CFD emulators

Eleonora Amato 1,2,\, Vito Zago1, Ciro Del Negro1

  • Affiliations: 1 Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo, Catania, Italy 2 Department of Mathematics and Computer Science, University of Palermo, Palermo, Italy 

  • Presentation type: Poster

  • Presentation time: Tuesday 16:30 - 18:30, Room Poster Hall

  • Poster Board Number: 38

  • Programme No: 6.5.7

  • Theme 6 > Session 5

Abstract

Lava flows exhibit complex fluid behavior characterized by non-Newtonian rheology, enabling them to overcome barriers, form tunnels, and inflict damage on affected areas. Mathematical modeling and numerical simulations are essential tools for quantitatively describing these flows and predicting potential scenarios without the risks associated with in-field studies. However, achieving an optimal balance between model accuracy and computational efficiency can be challenging. For instance, while Computational Fluid Dynamics (CFD) models provide reliable simulations of lava flows, they often come with high computational costs. Recent research has sought to address this by integrating CFD with Artificial Intelligence (AI) to enhance simulation performance. Here, we introduce a CFD emulator incorporating AI specifically designed for lava flows, capable of replicating their characteristic visco-thermal coupled behavior. This model adeptly manages various physical phenomena, including phase transitions, particle solidification, and the influence of air. We have conducted simulations under diverse physical conditions to validate the model's reliability and generalization capabilities. Furthermore, we investigated the impact of different parameters on lava flow modeling and the quantification of associated volcanic hazards. For example, we analyzed the influence of the effusion rate on eruption styles using satellite-derived estimates, facilitating a deeper understanding of varying eruptive behaviors without the risks of in-field assessments. Our results demonstrate the significant potential of integrating real-world measurements, numerical models, and AI to simulate lava flows, producing near real-time scenarios that are valuable for impact and risk assessment in complex eruption events.