AI Detects Thousands of Hidden Earthquakes and Faults Beneath Italy’s Volcanic Campi Flegrei

The GIST AI Model Exposes Hidden Earthquake Swarms and Faults in Italy's Campi Flegrei

Scientists are applying artificial intelligence to analyze increasing seismic unrest in Italy's Campi Flegrei, a volcanic caldera housing hundreds of thousands of residents. This AI approach functions like adjusting a camera lens, turning once blurry seismic data into a sharp, clear picture that reveals earthquakes previously undetectable by conventional methods.

A collaboration between Stanford University, Italy's National Institute of Geophysics and Volcanology (INGV)—Osservatorio Vesuviano, and the University of Naples Federico II has quadrupled the number of detected earthquakes in the region. The study also identified previously unknown faults. Pinpointing fault locations and lengths is essential to estimate the potential magnitudes of future earthquakes, offering valuable insights for risk assessment and urban planning.

Published in Science, the research leverages a Stanford-developed AI model that delivers precise earthquake locations and magnitudes in near real time. The model’s success in Campi Flegrei suggests it could be adapted to other volcanic regions with seismic monitoring networks, such as Santorini, Greece, which experienced significant earthquake swarms earlier this year. Quickly locating seismic sources during intense activity is crucial to effective emergency response.

“Seismicity could change at any time, and that may be the most important thing: the ability to get a clear view is now operational,” said Greg Beroza, co-author and geophysics professor at Stanford’s Doerr School of Sustainability. INGV is already deploying the tool independently, supporting both scientific and public response efforts.

A History of Unrest

Campi Flegrei is an active volcanic caldera situated in a densely populated area near Naples. The region has seen episodes of unrest since the late 1950s. The most recent phase began in 2005, with seismic events escalating significantly in 2018. Notably, five earthquakes above magnitude 4 occurred within the first eight months of 2025. The AI model increased the number of recorded earthquakes from approximately 12,000 to over 54,000 between 2022 and 2025.

The data uncovered two faults converging beneath the town of Pozzuoli, west of Naples. This area has been monitored since the 1980s, when ground uplift exceeded 6 feet and over 16,000 earthquakes forced the evacuation of 40,000 residents. Bill Ellsworth, co-director of the Stanford Center for Induced and Triggered Seismicity, noted, “These long faults suggest that an earthquake in the magnitude 5 range is not out of the question.”

Understanding these geological structures clarifies the seismic risks in Campi Flegrei, which has a history of significant eruptions, including two of Europe's largest within the last 40,000 years. However, short-term concerns focus more on moderate, shallow earthquakes that could threaten the local population and infrastructure.

Pressure Beneath the Surface

The Campi Flegrei caldera spans approximately 8 miles and was shaped by major eruptions around 39,000 and 15,000 years ago. Besides eruptions, the area experiences bradyseism—the slow rise and fall of the ground. Previously, seismic activity patterns within the caldera were unclear. The AI model revealed a distinct, thin ring fault that aligns with surface features, particularly offshore, and corresponds with zones of uplift.

Xing Tan, lead author and geophysics Ph.D. student, shared that Italian colleagues were surprised by the clarity of the ring fault, especially in the northern section where earlier data showed scattered seismicity. The findings indicate that overall inflation of the caldera is generating pressure that triggers earthquakes.

Importantly, the study found no signs of magma moving upward beneath the caldera, which reduces immediate concerns about an imminent volcanic eruption.

• AI can improve seismic monitoring by detecting smaller or hidden earthquakes.
• Identifying faults helps estimate earthquake magnitudes and potential impacts.
• Clear, near real-time data supports faster emergency response and public safety.

For those interested in AI applications in geoscience and seismic monitoring, exploring Complete AI Training offers valuable courses to deepen knowledge in this area.