Large seafloor rupture caused by the 1956 Amorgos tsunamigenic earthquake, Greece

Surface ruptures from large onshore earthquakes are routinely mapped to constrain fault source, rupture extent, and slip. Offshore, equivalent analyses are challenging yet crucial for understanding tsunami generation and seismic hazard. The 1956 Amorgos earthquake (Mw 7.2–7.8) in the South Aegean Sea produced severe damage and the largest recorded Mediterranean tsunami in the last two centuries, with run-ups up to 20 m along south Amorgos and ~10 m on Astypalaea. Focal mechanisms suggest NE–SW striking normal faulting with a SE-dipping plane, but the causative undersea fault remained unconfirmed. Variability in tsunami run-up led to hypotheses invoking both fault slip and submarine landslides, amid poorly constrained epicentral locations and coastal bathymetry. This study aims to identify and characterize the submarine seafloor rupture responsible for the 1956 earthquake and evaluate its role in tsunami generation to improve regional seismic and tsunami hazard assessment.

Prior work provided multiple epicentral solutions and focal mechanisms indicating NE–SW striking normal faulting consistent with regional tectonics. Historical accounts and limited instrumental data documented significant tsunami run-ups, while damaged tide gauges hindered near-field records. Tsunami modeling had invoked combined sources (fault slip plus mass wasting), but was limited by poorly known basin geometry and coastal bathymetry. Marine geophysical surveys over the last decade mapped the main regional faults: the Amorgos fault (main segment ~45 km, up to ~75 km with secondary structures), the Santorini–Amorgos fault (~55 km), and the Anafi–Astypalaea system (~65 km). Seismic reflection indicated substantial cumulative offsets and suggested a steep scarp at the base of the Amorgos fault, hinting at a preserved fault mirror. Onshore indicators (uplift on the southern coast and long-term subsidence on the northern coast) provided mixed signals, underscoring the need for direct submarine observations.

The team surveyed three candidate faults (Amorgos, Santorini–Amorgos, and Anafi–Astypalaea) using an AUV and an HROV during the AMORGOS-22 and AMORGOS-23 cruises. Shipboard multibeam (Kongsberg ME70) bathymetry was acquired at low speeds (2–5 kt) to increase resolution over steep scarps, producing 10 m DEMs. Near-bottom high-resolution bathymetry was collected by AUV Idef (Kongsberg Reson SMF EM2040) flying ~70 m above the seafloor, yielding 1 m DEMs after processing with IFREMER’s GLOBE software. Survey strategy targeted the steepest, simplest scarp sections away from mass-wasting features. The HROV Ariane (IFREMER) collected 4K video imagery along horizontal and vertical tracks at ~2–5 m standoff and <0.5 m/s. Video frames were preprocessed (illumination correction) and processed with structure-from-motion using MATISSE 3D (method A), and cross-checked with 3DF Zephyr pipelines (methods B and C), to build georeferenced, scaled digital outcrop models (DOMs) for seven vertical transects. On DOMs, the team mapped striae, gouge, and identified the present fault–wedge contact and the uplifted paleo-wedge contact. On-fault displacement was measured along the dip direction (parallel to subvertical striae) as the distance between the present and paleo contacts. DOM orientation issues prevented precise dip measurement on the models; instead, scarp dips (and uncertainties) were measured on the 1 m AUV DEM to compute corresponding vertical offsets and horizontal extensions. Bathymetric and imagery analyses also characterized scarp morphology (height, slope, gullies, cones, failures) and coatings (dark Fe–Mn oxide vs. fresher light-brown surfaces). Data and navigation files were archived in public repositories.

- The Amorgos fault displays a continuous >600 m high submarine scarp with steep slopes (50–60°) and a smooth, linear base lacking significant segmentation or splays, indicating localized deformation along a single plane. - A freshly exhumed fault mirror was identified over ~500 m length, with dip-parallel striae (tectoglyphs), indurated brown gouge/breccia coatings, and clear color/roughness contrasts indicating recent exposure versus oxidized older surfaces. - Remnants of an uplifted paleo colluvial wedge atop the mirror mark the pre-earthquake fault–seafloor contact; comparison with the current contact yields on-fault offsets between 9.8 and 16.8 m across seven transects (mean ~12.7 m at the presented site). - Using scarp dips from 1 m DEMs, corresponding vertical offsets at the seafloor vary from 6.4 to 13.4 m (mean ~9 m), and horizontal extension from 5.0 to 11.6 m (mean ~9 m). - Similar fresh deformation markers occur discontinuously over ~30 km along the Amorgos fault. - Scaling relationships: treating the observed offset as maximum surface displacement gives Mw ~7.5 (±0.1), consistent with seismology (Mw 7.2–7.8); assuming it represents mean displacement would imply Mw 8.0–8.3, deemed too large, thus favoring the former. - The rupture occurred within ~1 km of the Amorgos coast, plausibly sufficient to generate the observed large tsunami without requiring landslide sources, at least near the epicentral area. - Given a regional horizontal extension rate of ~4 mm/yr, ~2250 years of loading would be needed to accumulate slip comparable to 1956, suggesting long recurrence and scarcity in historical records. - Other nearby major faults (Santorini–Amorgos, Anafi–Astypalaea, Ios system, Kinairos fault) remain unbroken and largely quiet seismically, while seismic coupling is high (~80%) regionally (except near Santorini), implying potential for future large earthquakes.

The discovery and quantification of a fresh fault mirror and large seafloor offset along the Amorgos fault directly identify it as the source of the 1956 earthquake, resolving a decades-long uncertainty about the causative structure. The magnitude and proximity of vertical displacement to the coastline provide a physically plausible mechanism to generate the extreme local tsunami run-ups, diminishing the necessity of invoking coeval landslides to explain the largest waves near the source. Comparison with empirical scaling supports a maximum surface slip consistent with Mw ~7.5, aligning with instrumental estimates. The freshness of the mirror, absence of comparable large events since 1956, and lack of historical evidence for similar nearby earthquakes make the 1956 event the most likely origin of the observed scarp exhumation. Regionally, the findings imply that significant strain is stored on other long, unbroken normal faults in a strongly coupled system, elevating seismic and tsunami hazard. The work demonstrates that submarine fault source identification is feasible decades after events, which is crucial for reconstructing historical tsunamigenic earthquakes and improving hazard models.

Using AUV and HROV data 67 years after the event, the study identifies a large, freshly exhumed seafloor rupture along the Amorgos fault as the likely source of the 1956 Amorgos earthquake and tsunami. Measured on-fault offsets up to ~16.8 m and seafloor vertical displacements averaging ~9 m, together with morphological and textural evidence (striae, gouge, color contrasts), substantiate recent coseismic slip. The inferred slip and rupture dimensions are compatible with Mw ~7.5 and can plausibly account for the extreme local tsunami without requiring landslide sources. The approach showcases how targeted submarine exploration can resolve debated fault sources of historical tsunamis (e.g., Messina, Lisbon, Showa-Sanriku, Sumba) and emphasizes the need for detailed undersea fault mapping to refine seismic and tsunami hazard assessments in the Aegean and beyond. Future work should pursue paleoseismology, improved bathymetry of coastal zones, dating of mass-wasting features, and integrated tsunami modeling using better-constrained source geometries.

- Direct dating of the exhumed fault mirror is not currently possible; temporal attribution relies on historical and instrumental earthquake catalogs. - DOMs suffered from camera pan/tilt misorientation, preventing direct dip measurement on models; dip was instead derived from AUV DEMs. - The contribution of post-seismic slip versus coseismic slip to the observed offsets cannot be distinguished with available data. - Coastal and nearshore bathymetry, critical for accurate tsunami run-up modeling, remains incompletely known. - Identified submarine landslides are undated; their temporal link to the 1956 event is unconfirmed. - Scaling-law inferences depend on assumptions about rupture geometry (dip ~60°, rupture length 45–75 km) and shear modulus. - Observations of fresh markers are spatially discontinuous and concentrated along selected scarp segments, potentially biasing offset estimates.