Skip to content

A Cutting-Edge AI Approach for Ground Deformation Modelling

Martina Allegra 1,2, Flavio Cannavò1, Gilda Currenti1, Simone Palazzo2, Concetto Spampinato2

  • Affiliations: 1Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo, Catania, Italy; 2Department of Electrical, Electronic and Computer Engineering, University of Catania, Catania, Italy

  • Presentation type: Talk

  • Presentation time: Monday 10:30 - 10:45, Room S150

  • Programme No: 3.1.7

  • Theme 3 > Session 1

Abstract

The dynamics of the plumbing system directly results in ground deformation in volcanic regions reflecting the complex shape of magma intrusions, the volume of magma entering or rising, and the mechanisms of emplacement. Accurate interpretation of ground deformation data, critical for monitoring a volcano, has prompted the development of several methodologies and models to understand the underlying causative sources and mechanisms. However, traditional methods for interpreting ground deformation data usually rely on simplified mathematical models and manual analyses, which can be time-consuming and less accurate due to the complex nature of volcanic processes and subjective choices of the analyst. On the other hand, more complex models such as numerical approaches (e.g. FEM) can provide more realistic solutions, but at the higher computational and temporal cost; therefore, they cannot be considered for real-time applications. Recent advances in Artificial Intelligence may prove useful by potentially offering an automatic and more accurate tool for encoding complex patterns in deformation data and leading to better predictions of volcanic activity.  By taking advantage of the automation of deep learning, a novel inversion methodology has been approached with a combination of 3D deep neural networks to reconstruct the distribution of forces -- the 3D source -- inside the volcano that can generate the measured ground deformation. The AI model shows good ability to do instantaneous imaging of the volcanic source from station-based streamed deformation data, providing a useful tool for both monitoring and research purposes.