Submarine landslide megablocks show half of Anak Krakatau island failed on December 22nd, 2018

The study addresses how volcanic island flank collapses generate tsunamis and seeks to resolve the geometry, volume, and emplacement processes of the 22 December 2018 Anak Krakatau collapse. Despite extensive observations, key parameters such as collapse volume, failure style, headwall geometry, and submarine extent remained ambiguous due to eruptive plumes and rapid post-collapse deposition obscuring the scar. By integrating subaerial and submarine datasets, the authors aim to accurately parameterize the failure to improve tsunami source characterization and modeling. The broader significance lies in improving hazard assessments for arc-volcanic islands, where such collapses, though smaller than ocean-island events, dominate historical tsunamigenic records.

Previous work on volcanic-island flank collapses (e.g., Ritter Island 1888; Oshima-Oshima 1741; Mount Unzen 1792) has provided insights into tsunamigenesis but often lacked accurate pre-collapse bathymetry, precise volumes, and detailed failure geometries, limiting model fidelity. Anak Krakatau is the best-instrumented island-arc flank collapse, yet prior reconstructions differed widely in failure plane location and volume estimates. Studies using seismic inversion and tsunami modeling suggested volumes around 0.2–0.3 km³, but constraints on subaerial vs submarine contributions and emplacement style were uncertain. Historical development of Anak Krakatau indicates earlier instability (a 1949 flank collapse) and edifice growth atop the steep 1883 caldera wall, implying long-term structural controls. This study builds on and reconciles these findings by combining high-resolution SAR, aerial imagery, multibeam bathymetry, and seismic reflection data.

• Remote sensing of subaerial changes: Near-daily SAR (Sentinel-1, ALOS-2, TerraSAR-X, RADARSAT-2, COSMO-SkyMed) spanning the collapse; high-resolution CSK Spotlight image (12/23/2018) to delineate the headwall and subaerial scar; Sentinel-2 true/false-colour imagery to map pre-collapse lava/ejecta emplacement and crater migration; aerial photographs on 12/23/2018 to validate SAR interpretations.
• Historical analysis: Compilation and georeferencing of historical topographic surveys and photographs (1941–1950) to identify a prior SW sector failure and long-term structural controls.
• Bathymetry: Pre-event (1990) narrow-beam echosounder-derived bathymetry re-digitized and gridded; post-event (Aug 2019) high-resolution multibeam (Teledyne Reson T20-P) with tidal calibration; artifacts minimized in CARIS and analyzed in ArcGIS.
• Seismic reflection: High-resolution SPARKER profiles pre-collapse (2017) and post-collapse (2019) with consistent acquisition; processing and interpretation in Petrel; depth conversion using conservative P-wave velocity (1700 m/s) for volcaniclastic sediments; seawater velocity for TWT-depth conversion of 1990 surface.
• Volumetrics: Subaerial failure volume from applying mapped scar to a 2018 pre-collapse DEM, constraining slope angles (median ~35°, up to 60°). Two end-member submarine failure geometries constructed by combining 2018 DEM with 1990 bathymetry: (1) shallow failure to −100 to −120 m; (2) deep-seated to −230 m, with lateral limits from morphological ridges. Deposit volume computed by subtracting 1990 from 2019 bathymetry over blocky deposit; corrections and uncertainties addressed for resolution differences, tidal/directional artifacts, pre/post-event sedimentation, and basal erosion inferred from seismic. Debris-flow unit volume estimated by interpolating thickness between seismic horizons. Post-collapse sediment drape thickness mapped from seismic to estimate eruptive infill volume.

• Subaerial scar and volume: A ~2 km-long NW–SE headwall cutting the active cone defines a subaerial failure area of ~0.18 km² and volume of 0.098 ± 0.019 km³.
• Submarine deposit architecture: A fresh, block-dominated landslide deposit covers ~7.2 km² SW of the island, extending up to 1.5 km into the basin, arranged in 3–4 N–S block trains. Individual blocks are 185–520 m long, 300–500 m wide, and up to ~70 m high; internal layering and frontal folding/thrusting indicate translational emplacement with limited disaggregation and modest seafloor incision (generally <5–10 m; locally 10–15 m beneath largest blocks).
• Deposit volume: Primary blocky landslide deposit volume is 0.214 ± 0.036 km³ (after accounting for data resolution differences, sedimentation between 1990–2018, tidal/directional artifacts, and erosional/incorporated material). Individual block volumes range 0.002–0.035 km³.
• Failure volume and partitioning: Shallow failure scenario yields 0.175 ± 0.015 km³, favored by observations and consistent with deposit volume after bulking; deep-seated scenario (0.313 ± 0.043 km³) exceeds mapped deposit volume. Subaerial edifice contributed ~45% (0.098 km³) and submarine flank ~55% (~0.116 ± 0.025 km³).
• Additional units: A contemporaneous to immediately post-emplacement debris-flow unit beyond the toe has a volume of 0.022 ± 0.006 km³ and likely comprises eroded caldera-floor sediment; excluded from primary landslide volume.
• Post-event deposition: A parallel-bedded post-collapse drape (mean ~14 m thick) blankets the basin, volume 0.154 ± 0.023 km³. Combined with prior estimate of subaerial deposits (~0.029 km³), the total post-collapse eruptive volume is ~0.299 ± 0.05 km³, ~90% deposited offshore, rapidly obscuring the submarine scar.
• Failure depth and dynamics: The failure plane likely intersected the submarine slope at ~−100 to −120 m. Emplacement duration ~90 s suggests velocities ≥16.5 m/s (consistent with modeled 12–45 m/s). The block-train morphology and short runout indicate en-masse translational collapse capable of efficient tsunami generation.
• Tsunami context: Observed runups >80 m within the caldera and up to 13 m on distant coasts; fatalities 437. The mapped landslide volume aligns with independent seismic inversion (~0.2 km³) and tsunami models (0.22–0.27 km³).

The integrated subaerial–submarine dataset resolves previously ambiguous parameters of the 2018 Anak Krakatau collapse, demonstrating a predominantly translational, en-masse lateral failure with a shallow submarine intersection (~−100 to −120 m). The deposit-based volume (0.214 ± 0.036 km³) and subaerial–submarine partitioning provide robust constraints for tsunami source modeling, showing that such block-rich, short-runout collapses can still generate efficient tsunamis without invoking additional sources. Preconditioning factors—rapid SW-flank loading by lavas and ejecta, cumulative deformation and fissuring, fumarolic activity, and structural controls from the pre-1960 tuff cone and caldera wall—likely localized and weakened the eventual failure plane. The findings underscore a dichotomy between translational block-dominated collapses (like Anak Krakatau) and highly disintegrative, long-runout failures (e.g., Ritter Island), emphasizing that edifice properties and deformation history govern fragmentation and transport style. Rapid, voluminous post-collapse eruptions (~0.3 km³) quickly buried the scar and deposits, highlighting that collapse events can trigger intense eruptive phases that modify and obscure geologic evidence within weeks, complicating retrospective identification of prehistoric collapses.

This study delivers the first high-resolution, deposit-constrained volume and geometry of the 2018 Anak Krakatau flank collapse, quantifying a 0.214 ± 0.036 km³ blocky submarine deposit sourced ~45% from subaerial and ~55% from submarine flank failure. Evidence favors a shallow, en-masse translational collapse intersecting the slope at ~−100 to −120 m, with limited disaggregation and short runout that nonetheless generated a significant tsunami. The work resolves key ambiguities in prior reconstructions and provides validated parameters to improve tsunami modeling. It also identifies preconditioning indicators—flank loading, deformation, fissuring, fumaroles, and structural boundaries—relevant for monitoring similar volcanoes. Post-collapse eruptions (~0.299 ± 0.05 km³ total) rapidly rebuilt and buried the scar, demonstrating how swiftly evidence can be obscured. Future research should: refine constraints on submarine failure planes using even higher-resolution, time-lapse bathymetry and seismic imaging; integrate in situ geomechanical properties to predict fragmentation and rheology; expand real-time monitoring of deformation, thermal/fluid signals, and shoreline change; and test tsunami models across the translational–disintegrative spectrum for arc-volcanic settings.

• Submarine failure plane obscured by rapid post-collapse deposition, leading to uncertainty in precise boundaries, depth, and basal gradient; only end-member geometries (shallow vs deep-seated) could be bracketed.
• Resolution differences between pre-event (1990) and post-event (2019) bathymetry introduce vertical uncertainties; residual tidal and directionality artifacts remained despite corrections.
• Depth conversion for seismic profiles depends on assumed P-wave velocities in heterogeneous volcaniclastic and block materials, adding thickness/volume uncertainty.
• Data gaps on the NE extent of the submarine scar and limited lateral coverage affect precise mapping of failure margins.
• Debris-flow extent and any associated turbidites beyond seismic vertical resolution remain uncertain; interpolation of erosional depths across the slide area introduces additional, unquantified uncertainty.
• Subaerial scar slope angle and DEM-based reconstructions carry inherent errors due to plume cover and rapid morphological change.