Earthquakes that occurred from 2022 to 2025 in the Campi Flegrei caldera about 7 miles west of central Naples overlaid in red on Google Earth satellite imagery. Credit: Xing Tan
Researchers have used artificial intelligence to assess the earthquake risk of Campi Flegrei, a volcanic area in Italy home to 500,000 people, to help predict and mitigate the impacts of future seismic activity.
The technology revealed there were 4 times as many earthquakes in the area as had been previously detected. The researchers suggest that this more accurate data will help shape public responses to future earthquakes, keeping residents safer.
The Campi Flegrei caldera is an active volcanic field located west of Naples. The volcano is responsible for producing 2 of the largest eruptions in Europe, although according to this new study, an eruption might not be the main concern of residents.
“One of the biggest concerns in the short term in Campi Flegrei is not an eruption, but a moderate earthquake at shallow depth,” says Greg Beroza, a geophysics professor at Stanford University in the US and the study’s co-author.
The region is currently experiencing a period of significantly increased seismic activity, with 5 earthquakes above a magnitude 4 recorded in the first 8 months of this year.
“Seismicity could change at any time, and that may be the most important thing about this study: this capability of getting a clear view is now operational,” says Beroza.
When scientists have access to more accurate earthquake data, such as the location and length of the fault, it can help researchers determine the range of magnitudes of future earthquakes. Providing this information to city planners and residents is crucial to mitigate any potential upcoming risks that may occur in Campi Flegrei.
The team developed a model that uses machine learning in a similar way to how a camera uses a lens to adjust individual objects into focus. The Stanford-designed AI can identify precise earthquake location and magnitude data which previous tools could not pick out from massive collections of seismic data.
The researchers analysed recent seismic data from January 2022 to March this year. Existing tools had tracked around 12,000 earthquakes during this period. With help from AI, the researchers expanded this index to over 54,000 earthquakes.
(A) Map of Campi Flegrei caldera showing the seismic events (circles) from January 1, 2005, to March 20, 2025. Red stars mark events greater than magnitude 4.0. (B) Vertical ground displacement recorded from 1905 to 2023 at Pozzuoli port. (C) Seismicity evolution from 2005 to 2025. Credit: Tan et al 2025, Science https://doi.org/10.1126/science.adw9038
In particular, the data uncovered 2 faults converging under a town west of Naples called Pozzuoli.
“We’ve known that this is a risky place for a long time, since the ’80s when part of the city was evacuated, and now we’re seeing for the first time the geologic structures that are responsible,” says Bill Ellsworth, co-director of the Stanford Centre for Induced and Triggered Seismicity.
“These long faults suggest that an earthquake in the magnitude 5 range is not out of the question.”
The researchers are hopeful that this model could also be used to pinpoint the exact location of seismic activity in other earthquake-prone areas like Santorini, Greece, which could allow officials to execute evacuations earlier, potentially saving lives.
While the researchers did not observe any evidence of upward magma migration, reducing their concern of an eruption in the short term, the AI model revealed a clearer picture of the ring fault system on the edge of the caldera – the hollow caldron-like depression created after a volcano erupts.
The caldera was formed by major volcanic eruptions about 39,000 and 15,000 years ago.
“Previously, the structure of seismicity in the caldera was indistinct,” says Beroza.
“Now we’ve seen a very thin and well-marked ring fault that is consistent with the surface features, especially offshore, and also with the area that’s being uplifted.”
The land near the caldera can experience a rising and sinking known as bradyseism.
“Our Italian colleagues were surprised to see the ring so clearly,” added lead study author Xing Tan, a geophysics PhD student in Beroza’s lab.
“They expected to see something in the south where previous data had revealed scattered seismicity, but in the north, they’d never seen it so clearly.”
The research, published in Science, was a collaboration between Stanford University, Italy’s National Institute of Geophysics and Volcanology and the University of Naples Federico II.
“INGV is now running the tool by themselves as needed, so it should be helpful for scientific response and ultimately public response if something changes,” says Beroza.
The research has been published in the journal Science.