In the bustling heart of Italy’s Campi Flegrei, a region fraught with geological uncertainty, scientists are harnessing the power of artificial intelligence to unveil a clearer picture of the volcanic unrest that has long challenged researchers. This caldera, situated just west of Naples and home to over half a million residents, has historically been a source of seismic anxiety. Yet traditional seismic detection methods have struggled to decode the sheer volume and intricacy of earthquake activity beneath its surface. Now, with cutting-edge computational tools, researchers have penetrated this veil of complexity, revealing previously hidden seismic events and fault structures with unprecedented precision.
This breakthrough leverages an AI model developed by Stanford University, which meticulously sifts through vast seismic datasets to identify earthquake occurrences that conventional algorithms miss. The result is a seismic catalog that is fourfold larger than earlier records, expanding from approximately 12,000 detected earthquakes between 2022 and 2025 to over 54,000. This leap in detection sensitivity not only quantifies the dynamic seismic background more accurately but also enables geophysicists to precisely map fault lines that underlie the region. The newfound clarity in seismic patterns enriches our understanding of the fault mechanics at Campi Flegrei—a crucial advancement given the potential risks posed by future quakes.
Faults, which are fractures along which rock masses slip during earthquakes, are central to the region’s seismic hazard assessment. Before the advent of AI-enhanced seismicity analysis, the spatial arrangement and extent of these faults remained vaguely understood due to the overlapping signals and noise typical of dense seismic swarms in volcanic regions. With better fault delineation, scientists can refine earthquake magnitude models and probabilistic hazard forecasts, crucial for urban disaster preparedness. Recognizing fault geometries with enhanced accuracy also sheds light on the stress accumulation and release mechanisms within the caldera’s complex crustal environment.
Published in the prestigious journal Science on September 4, 2025, this research represents a collaboration among Stanford University, the National Institute of Geophysics and Volcanology (INGV) – Osservatorio Vesuviano, and the University of Naples Federico II. The study illustrates how real-time earthquake monitoring systems augmented by AI can transform our ability to manage volcanic risks. Campi Flegrei, with its continuous seismic activity, serves as an ideal natural laboratory. The success here hints at broader applications for other volcanic and tectonically active regions worldwide, such as Santorini in Greece, which recently experienced an earthquake swarm of similar complexity.
Greg Beroza, a geophysics professor at Stanford’s Doerr School of Sustainability and a co-author of the study, emphasized the operational readiness of this AI-based approach. “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,” he remarked. The INGV is already integrating the tool into their workflows, allowing for near real-time seismicity analysis that can profoundly enhance scientific and public responses to emergent geophysical changes.
Campi Flegrei’s history of unrest is both rich and sobering. The caldera has experienced episodic ground deformation and seismic crises since the late 1950s, including a particularly intense period beginning in 2005. More recently, in the first eight months of 2025 alone, the region has recorded five earthquakes exceeding magnitude 4, raising alarms among local authorities. The new AI-assisted seismic catalog reveals that the ongoing earthquake activity is far more extensive than previously known, offering a granular view of seismicity that extends well beyond what conventional monitoring systems documented.
A closer inspection of the seismic data reveals two previously underappreciated fault systems converging beneath the densely populated town of Pozzuoli, a locality historically vulnerable to seismic hazards. This area has been under scientific watch since the 1980s, when episodes of ground uplift reached over 6 feet, and earthquake swarms triggered evacuations affecting up to 40,000 residents. The recognition of these long fault structures, which span several kilometers, aligns with their geological potential to produce moderate earthquakes in the magnitude 5 range, heightening awareness about the magnitude of seismic risk faced by local communities.
Bill Ellsworth of Stanford’s Center for Induced and Triggered Seismicity highlighted the significance of these findings: “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.” This enhanced understanding marks a monumental stride forward in volcano-seismology, bridging knowledge gaps that have persisted for decades.
Campi Flegrei is a vast caldera measuring roughly eight miles in diameter, formed by cataclysmic volcanic eruptions approximately 39,000 and 15,000 years ago. The caldera’s cyclic behavior is characterized by episodes of ground inflation and deflation, phenomena collectively known as bradyseism, driven by complex subsurface processes including magma movement and hydrothermal activity. Prior to this study, the internal structure of seismicity within the caldera was diffusely mapped, with scattered earthquake events providing little coherent pattern.
The AI model unveiled a strikingly well-defined ring fault encircling the caldera’s rim, aligning closely with surface morphological features, particularly offshore. This discovery offers compelling evidence linking deep subterranean processes with surface deformation patterns recorded by geodetic instruments. Xing Tan, a geophysics doctoral student at Stanford and lead author of the study, noted the reaction of the Italian colleagues: “Our Italian colleagues were surprised to see the ring so clearly. 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.”
Intriguingly, the researchers found no seismic signatures indicative of magma ascending towards the surface during the study period, a finding which modestly reduces the immediate concern for a volcanic eruption. Instead, the study suggests that the primary driver of seismic activity is the general inflation of the caldera due to pressurization within the subsurface. This inflation creates stress that activates existing faults, causing the detected earthquakes. Disentangling these processes helps clarify the nature of unrest and improve risk assessment models.
Despite the diminished eruption risk implied by the absence of magma migration, the study highlights that “a moderate earthquake at shallow depth” remains a significant short-term hazard in Campi Flegrei, capable of damaging structures and jeopardizing residents. The advancement of AI-assisted seismic monitoring thus stands as an essential tool for hazard mitigation, enabling authorities to better anticipate seismic threats and respond swiftly.
The study’s success is a testament to the interdisciplinary collaboration spanning geophysics, data science, and volcanology, supported by diverse funding sources including the Dipartimental Project LOVE-CF, Pianeta Dinamico project Nemesis, the Stanford University Doerr School of Sustainability Discovery Grant, the RETURN Extended Partnership, and the European Union Next-GenerationEU program. As AI continues to permeate earth sciences, these developments foreshadow a new era in which seismic hazards can be tracked, understood, and managed with unprecedented accuracy and timeliness.
This transformative research not only elevates our understanding of Campi Flegrei’s current unrest but also establishes a framework that can be adapted to other seismically active volcanic regions, thereby enhancing global preparedness against natural disasters driven by the restless Earth.
Subject of Research: Volcanic seismicity analysis using artificial intelligence at Campi Flegrei Caldera
Article Title: A clearer view of the current phase of unrest at Campi Flegrei Caldera
News Publication Date: 4-Sep-2025
Web References:
https://sustainability.stanford.edu/news/ai-detects-hidden-earthquakes
http://dx.doi.org/10.1126/science.adw9038
Image Credits: Xing Tan
Keywords: Earthquakes, Geophysics, Artificial intelligence, Machine learning