The isle of Vulcano is a part of the central volcanic ridge of the Aeolian archipelago on the Tyrrhenian Sea in southern Italy. Over the course of its history, Vulcano has undergone multiple explosive eruptions, with the last one thought to have occurred around 1888–1890. However, there is an active hydrothermal system under Vulcano that has shown evidence of intermittent magma and gas flows since 2021 – a sign that the volcano has been in a state of unrest.
During unrest, the volcanic risk increases significantly – and the summer months on the island currently attract a lot of tourists that might be at risk, even from minor eruptive events or episodes of increased degassing. To examine why this unrest has occurred, researchers from the University of Geneva have collaborated with the National Institute of Geophysics and Volcanology (INGV) in Italy to recreate a 3D model of the interior of the volcano on Vulcano, using a combination of nodal seismic networks and artificial intelligence (AI).
Until now, few studies have examined the deep underground details of volcanoes, instead relying on looking at the outline of their internal structure. This is because the geological domains where eruptions nucleate are often inaccessible using airborne geophysical techniques, and onshore studies don’t penetrate far enough into the volcanic plumbing system to look at how the magma and hydrothermal fluids mix. Recent studies have shown the outline of the plumbing systems, but they’ve not had sufficient resolution to distinguish the magma from the hydrothermal system.
3D modelling of the volcano
To better understand what could have caused the 2021 Vulcano unrest, the researchers deployed a nodal network of 196 seismic sensors across Vulcano and Lipari (another island in the archipelago) to measure secondary seismic waves (S-waves) using a technique called seismic ambient noise tomography. S-waves propagate slowly as they pass through fluid-rich zones, which allows magma to be identified.
The researchers captured the S-wave data using the nodal sensor network and processed it with AI – using a deep neural network. This allowed the extensive seismic dispersion data to be quickly and automatically recovered, enabling generation of a 3D S-wave velocity model. The data were captured during the volcano’s early unrest’s phase, and the sensors recorded the natural ground vibrations over a period of one month. The model revealed the high-resolution tomography of the shallow part of a volcanic system in unrest, with the approach compared to taking an “X-ray” of the volcano.
“Our study shows that our end-to-end ambient noise tomography method works with an unprecedented resolution due to using dense nodal seismic networks,” says lead author Douglas Stumpp from the University of Geneva. “The use of deep neural networks allowed us to quickly and accurately measure enormous seismic dispersion data to provide near-real time monitoring.”
The model showed that there was no new magma body between Lipari and Vulcano within the first 2 km of the Earth’s crust, but it did reveal regions that could host cooling melts at the base of the hydrothermal system. These melts were proposed to be degassing melts that could easily release gas and brines if disturbed by an Earthquake – suggesting that tectonic fault dynamics may trigger volcanic unrest. It’s thought that the volcano might have released trapped fluids at depth after being perturbed by fault activity during the 2021 unrest.
Improving risk management
While this method doesn’t enable the researchers to predict when the eruption will happen, it provides a significant understanding into how the internal dynamics of volcanoes work during periods of unrest. The use of AI enables rapid processing of large amounts of data, so in the future, the approach could be used as an early warning system by analysing the behaviour of the volcano as it unfolds.
In theory, this could help to design dynamic evacuation plans based on the direct real-time behaviour of the volcano, which would potentially save lives. The researchers state that this could take some time to develop due to the technical challenge of processing such massive volumes of data in real time – but they note that this is now more feasible thanks to machine learning and deep learning.
When asked about how the researchers plan to further develop the research, Stumpp concludes that “our study paves the ground for 4D ambient noise tomography monitoring – three dimensions of space and one dimension of time. However, I believe permanent and maintained seismic nodal networks with telemetric access to the data need to be implemented to achieve this goal”.
The research is published in Nature Communications.
The post Researchers map the unrest in the Vulcano volcano appeared first on Physics World.