Santorini’s famous caldera may hide a different tsunami threat: not an eruption, but a collapse of the volcano’s flank. Simulations suggest even a localized failure could send waves reaching thirty-nine meters. Tsunamis generated by volcanic island sector collapses present a recognized threat to coastal communities and infrastructure.

In historical times, eighteen documented sector collapse events caused more than fifteen thousand casualties. Even much smaller events can trigger impactful tsunamis: the twenty-eighteen Anak Krakatau collapse, with a volume of zero point two to zero point three cubic kilometers, produced run-up heights of up to thirteen meters and caused more than four hundred casualties.

Santorini is one of the most intensively visited islands in the Mediterranean, with approximately two million visitors annually. From twenty-sixteen to twenty-twenty-three, tourism pressure increased markedly through rising visitor numbers, hotel capacity, and expanded built-up areas, and that trend is projected to continue.

The island’s permanent population is officially recorded at fifteen thousand five hundred fifty, and summer tourism substantially adds temporary workers employed in the tourism sector. This highly dynamic population exposure coincides with a complex multi-hazard volcanic setting, where hazard assessment is important for communicating risks to stakeholders.

Hazard assessments for volcanic eruptions are widely used, but non-eruptive volcanic hazards, especially those associated with volcanic unrest, remain comparatively underrepresented. Periods of unrest, including the twenty-eleven to twenty-twelve Santorini episode and the twenty-twenty-five earthquake swarm, can increase cascading hazards such as tsunamis triggered by submarine landslides.

Such unrest episodes occur annually at approximately twenty large calderas worldwide and may persist for years, with considerable socioeconomic consequences. Susceptibility maps and numerical tsunami modeling are available within the Santorini volcanic group, but focused assessments of present-day tsunami hazard remain limited.

A worst-case ten-meter tsunami run-up scenario for Kamari suggested that more than four hundred structures would fall within the inundation zone, with over half classified as highly to very highly vulnerable. Seismic studies also indicate that submarine slope failures associated with volcanic or volcano-triggered seismic activity could generate tsunamis near Santorini’s eastern coastline.

Recent results from IODP Expedition three hundred ninety-eight point to a previously overlooked hazard scenario for Santorini. Between December twenty-twenty-two and February twenty-twenty-three, the expedition recovered several tens of meters of unconsolidated pumice and ash at four drill sites within Santorini Caldera.

These volcaniclastic deposits were emplaced during a submarine explosive eruption of Kameni Volcano in seven hundred twenty-six CE. The seven hundred twenty-six CE eruption produced pumice and ash with a bulk volume of up to three point one cubic kilometers centered around Kameni Volcano.

Seven subsequent eruptions formed thick lava flows on top of slopes reaching twenty degrees and composed of underconsolidated pumice and ash deposits. Because Kameni Volcano sits on a topographic step coincident with the volcano-tectonic Kameni Line, previous research identified the edifice as potentially susceptible to flank instability.

Despite these indications, this specific mechanism had remained unquantified and had not yet been integrated into existing disaster risk frameworks. High-resolution seismic reflection data and IODP Expedition three hundred ninety-eight results show that much of the Kameni edifice, especially its submarine flanks, is built on thick accumulations of unconsolidated pumice and ash deposited during the seven hundred twenty-six CE eruption.

These deposits form slopes reaching about twenty degrees along a topographic step separating the northern caldera basin, with a maximum depth of three hundred eighty-five meters, from the southern basin, with a maximum depth of two hundred ninety meters. The step is aligned with the Kameni Line, an active volcano-tectonic lineament.

Subsequent lava flows since fifteen seventy to fifteen seventy-three CE were emplaced directly onto these weak, poorly consolidated volcaniclastic units, creating mechanically unstable conditions within the growing edifice. The rapid post-caldera growth of the Kameni volcanic edifice, lava flows over steep unconsolidated pyroclastic deposits, the structural influence of the Kameni Line, and recurrent volcanic unrest establish preconditions for potential flank instability.

A future flank failure of Kameni Volcano could generate tsunamis within the Santorini caldera and across the broader region. These mechanisms therefore warrant formal integration into regional geohazard and risk-mitigation frameworks. The study considers three flank-collapse scenarios at Kameni Volcano, called slides A, B, and C.

Slide A includes multiple lava flows from seventeen eleven to nineteen twenty-five, with a total volume of zero point zero five seven cubic kilometers; slide B consists of zero point zero seven seven cubic kilometers of the fifteen seventy to fifteen seventy-three lava flow; and slide C corresponds to the seventeen oh seven to seventeen eleven lava flow, with a volume of zero point one three cubic kilometers.

The scenarios were chosen as a first-order approximation covering a reasonable range of slide volume and orientation. The depth-averaged Landslide-HySEA code was developed for numerical modeling of tsunamis generated by submarine and aerial landslides. Its finite-volume numerical model is based on the two-layer Savage-Hutter model and was fully validated for all NOAA National Tsunami Mitigation Program mandatory benchmark problems.

The numerical scheme uses an explicit first-order Intermediate Field Capturing Parabola discretization for the system equations. The discretization of the Coulomb friction law and the treatment of wet-dry fronts follow the methods described in the cited model developments.

The numerical scheme is implemented on graphics processing units using the CUDA programming language, resulting in very high efficiency. Landslide-HySEA input parameters include non-erodible topography, the initial thickness of each layer, Coulomb friction angle, ratio of densities, friction between layers, and friction between the water or slide and the non-erodible bottom.

Other input parameters include total simulated time, the time interval for data output, the Courant-Friedrichs-Lewy condition as a factor between zero and one, and maximum velocity for each layer. For every scenario, the Coulomb friction angle was varied between nine, twelve, and fifteen degrees.

For lava flows assumed to be the sliding layer, density was varied between two thousand seven hundred, two thousand five hundred, and two thousand three hundred kilograms per cubic meter. Because friction between layers had a much smaller impact on the resulting tsunami than Coulomb friction angle and sliding-layer density, it was set to zero point zero eight for all simulations.

A topography grid with a resolution of twenty meters was used for numerical modeling. Interpolated slide planes produced volumes of zero point zero five seven, zero point zero seven seven, and zero point one three cubic kilometers for slides A, B, and C, respectively.

During cruise M two fifteen, the Modular Mobility photogrammetric tow-sled system surveyed a transect across the northern front of the fifteen seventy to fifteen seventy-three flow facing the caldera floor. At an altitude between two point five and five meters above the seafloor and a vehicle speed of zero point two five knots, stereoscopic cameras collected highly overlapping seafloor imagery at one-second intervals.

The imagery was processed into a hydroacoustically georeferenced photogrammetric three-dimensional seafloor model using Agisoft Metashape Professional version two point two point one. The resulting model covered three hundred ninety-five by ninety-three meters and had one-centimeter geometric and textural resolution at a nine-meter average track width.

The modeling results strongly depend on two key input parameters: the Coulomb friction angle and the sliding-layer density. For all slides, the main text shows the parameter combinations producing the highest and lowest resulting tsunamis within the considered range: nine degrees with two thousand seven hundred kilograms per cubic meter, and fifteen degrees with two thousand three hundred kilograms per cubic meter.

Virtual tide gauges for all slides were presented at Korfou, Armeni, Firon, and Athinios. For slide A, the resulting tsunami is directed northwest, with a smaller maximum height between two and four meters near Armeni. The first arrival at Armeni is one minute after slide motion begins.

For slide B, the tsunami runs northeast, with a maximum height between seven and thirteen meters just north of Firon. Panel a shows a broad, nearly continuous offshore footprint around the caldera, with the brightest colors concentrated toward the central and northwestern basin.

In panel b, the colored area is more fragmented and less extensive, while the strongest patches remain clustered near the center rather than along the outer coast. In panel a, the warmest colors cluster around the central caldera and the coast near Firon, while much of the outer basin remains blue.

Panel b shows a more localized green-to-yellow patch near Firon, with broader surrounding areas staying in the darker blue range. For slide C, the resulting tsunami is directed northwest, with a maximum height between twenty-one and thirty-nine meters at Korfou. The first arrival at Korfou is one minute after slide motion begins.

For each scenario, the semi-confined geometry of the caldera would promote tsunami reflections, causing multiple impacts at individual coastal locations within the first minutes. The figure’s two map panels show the strongest modeled response concentrated inside the caldera and extending northwest toward Korfou, while colors fade outward offshore.

Comparing panels A and B, the high-intensity region is broader and more continuous in A; panel B shows more blue and green around the caldera margins, indicating a visibly smaller modeled extent there. Reading across the rows, the table pairs each scenario with a direction and a named location: Armeni for A, Firon for B, and Korfou for C.

The first-arrival column shows that A and C are marked as approximately sixty seconds, while B is listed as less than thirty seconds. The numerical modeling is a first-order approximation, and the resulting tsunami height for all scenarios nearly doubles across the considered parameter range.

Even so, all three potential slides can reach coastal tsunami heights exceeding two meters at several critical locations within Santorini’s caldera. Slides B and C reach coastal tsunami heights much higher than two meters, as also shown in the virtual tide gauges.

Across the four panels, the green curves show the largest oscillations, especially at Korfou and Armeni, where the first few peaks and troughs span much more of the vertical axis than the blue and red curves. The dashed vertical markers also show that the initial response arrives at about half a minute in Firon, one minute in Armeni and Korfou, and one and a half minutes in Athiniós, followed by repeated oscillations.

The figure compares each slide at two parameter settings: the left column uses a nine-degree friction angle and density of two thousand seven hundred kilograms per cubic meter, while the right uses fifteen degrees and two thousand three hundred. Across the maps, the strongest colors cluster along the inner caldera coastline, especially near Fira and Athinios, with slide C showing the broadest high-intensity coastal band in panels e and f.

A tsunami wave exceeding two meters would significantly affect all four critical locations, where infrastructure is at sea level or elevated by only about one meter. Athinios is Santorini’s sole terminal for large vessels, handling an average of fifteen to twenty-five ferry arrivals daily during the high season.

Athinios is also the exclusive gateway for one hundred percent of the island’s essential imports, including freshwater tankers, groceries, and fuel. A tsunami-induced disruption of this single terminal would immediately paralyze the supply chain, and even a forty-eight- to seventy-two-hour port closure would lead to critical shortages for residents and daily visitors.

The vulnerability of Athinios therefore represents a systemic risk to the island’s functional continuity. Because the Kameni volcanic edifice sits on a northward-deepening topographic step coincident with the volcano-tectonic Kameni Line, previous research identified susceptibility to flank instability.

Numerical simulations of multiple potential flank-collapse scenarios reveal a significant tsunamigenic hazard. A localized failure involving a volume of zero point one three cubic kilometers could generate maximum tsunami amplitudes of up to thirty-nine meters.

All modeled scenarios, with landslide volumes from zero point zero five seven to zero point one three cubic kilometers, produce wave heights exceeding two meters at several primary ports, including Korfou, Armeni, Fira, and the Athinios ferry terminal. The study finds that modeled Kameni flank collapses could produce waves above two meters at several essential coastal sites, making continuous monitoring and flank-collapse scenarios important parts of Santorini’s tsunami planning.