A recent study published in Nature Communications has successfully constructed a high-resolution subsurface model of Italy’s Vulcano Island during its 2021 unrest period using an advanced technique known as Neural Network Nodal Ambient Noise Tomography (N²ANT). This research was conducted through an international collaboration involving the University of Geneva, the Italian National Institute of Geophysics and Volcanology (INGV), and other institutions.
Background
In September 2021, Vulcano Island entered a phase of unrest marked by high-temperature fumaroles, significant emissions of CO₂ and SO₂, ground uplift, and Very Long Period (VLP) seismic events. These phenomena are commonly associated with the migration of shallow magma and fluids, though the specific mechanisms remained poorly understood.
Methodological Innovation
The research team deployed a dense seismic array consisting of 196 SmartSolo nodal seismometers. Using a deep learning algorithm to automatically extract Rayleigh wave dispersion data, and applying a two-step inversion strategy—nonlinear travel-time tomography followed by Bayesian probabilistic inversion—they obtained a detailed 3D shear-wave velocity model of the subsurface down to 2 km depth.
Key Findings
The La Fossa caldera is surrounded by low-velocity anomalies, likely related to hydrothermal activity and fluid-saturated zones;
High-velocity anomalies at greater depths are interpreted as cooled magmatic intrusions or dense rock bodies;
The hypocenter of VLP events is located at the interface between low- and high-velocity anomalies, supporting the hypothesis that fluid migration triggers these earthquakes;
Tectonic activity—particularly along the Aeolian-Tindari-Letojanni fault system—plays a key role in initiating volcanic unrest.
Implications and Outlook
This study represents the first successful application of high-resolution, near-real-time subsurface imaging during an active volcanic unrest period. It establishes a new paradigm for volcanic monitoring and hazard early warning. In the future, such techniques could be integrated into early warning systems to dynamically assess volcanic behavior, optimize emergency response strategies, and ultimately enhance risk mitigation capabilities.
👇 Read the full paper and join the discussion:
https://www.nature.com/articles/s41467-025-62846-z
Authors: Douglas Sami Stumpp, Iván Cabrera-Pérez, Geneviève Savard, Tullio Ricci, Mimmo Palano, Salvatore Alparone, Andrea Ursino, Federica Sparacino, Anthony Finizola, Francisco Muñoz Burbano, Maria-Paz Reyes Hardy, Joël Ruch, Costanza Bonadonna, and Matteo Lupi.
VolcanoMonitoring #DeepLearning #SeismicImaging #Geohazards #AmbientNoiseTomography #Vulcano #INGV #Geoscienceing blind faults play a key role in modulating the near E-W oriented stress loading on the TLFZ.