On the evening of 15 January 2022, the Hunga Tonga-Hunga Ha’apai volcano 1 unleashed a violent underwater eruption, blanketing the surrounding land masses in ash and debris 2 , 3 . The eruption generated tsunamis observed around the world. An event of this type last occurred in 1883 during the eruption of Krakatau 4 , and thus we have the first observations of a tsunami from a large emergent volcanic eruption captured with modern instrumentation. Here we show that the explosive eruption generated waves through multiple mechanisms, including: (1) air–sea coupling with the initial and powerful shock wave radiating out from the explosion in the immediate vicinity of the eruption; (2) collapse of the water cavity created by the underwater explosion; and (3) air–sea coupling with the air-pressure pulse that circled the Earth several times, leading to a global tsunami. In the near field, tsunami impacts are strongly controlled by the water-cavity source whereas the far-field tsunami, which was unusually persistent, can be largely described by the air-pressure pulse mechanism. Catastrophic damage in some harbours in the far field was averted by just tens of centimetres, implying that a modest sea level rise combined with a future, similar event would lead to a step-function increase in impacts on infrastructure. Piecing together the complexity of this event has broad implications for coastal hazards in similar geophysical settings, suggesting a currently neglected source of global tsunamis.

Current models used to assess earthquake and tsunami hazards are inadequate where creep dominates a subduction megathrust. Here we report geological evidence for large tsunamis, occurring on average every 300–340 years, near the source areas of the 1946 and 1957 Aleutian tsunamis. These areas bookend a postulated seismic gap over 200 km long where modern geodetic measurements indicate that the megathrust is currently creeping. At Sedanka Island, evidence for large tsunamis includes six sand sheets that blanket a lowland facing the Pacific Ocean, rise to 15 m above mean sea level, contain marine diatoms, cap terraces, adjoin evidence for scour, and date from the past 1700 years. The youngest sheet and modern drift logs found as far as 800 m inland and >18 m elevation likely record the 1957 tsunami. Previously unrecognized tsunami sources coexist with a presently creeping megathrust along this part of the Aleutian Subduction Zone.

Erosion and deposition from tsunamis record information about tsunami hydrodynamics and size that can be interpreted to improve tsunami hazard assessment. We explore sources and methods for quantifying uncertainty in tsunami sediment transport modeling. Uncertainty varies with tsunami, study site, available input data, sediment grain size, and model. Although uncertainty has the potential to be large, published case studies indicate that both forward and inverse tsunami sediment transport models perform well enough to be useful for deciphering tsunami characteristics, including size, from deposits. New techniques for quantifying uncertainty, such as Ensemble Kalman Filtering inversion, and more rigorous reporting of uncertainties will advance the science of tsunami sediment transport modeling. Uncertainty may be decreased with additional laboratory studies that increase our understanding of the semi-empirical parameters and physics of tsunami sediment transport, standardized benchmark tests to assess model performance, and development of hybrid modeling approaches to exploit the strengths of forward and inverse models.

Elevations at Driftwood Bay were mapped with a Magellan PM500 real-time kinematic Global Navigation Satellite System (GNSS) survey instrument with ±1.5 cm horizontal and ±3 cm vertical accuracy. Tidal benchmarks at Nikolski were surveyed to provide comparison to tidal datums calculated by NOAA for the Nikolski, Alaska tide gage (NOAA station ID: 9462450). We also obtained a tidal curve for Driftwood Bay from a pressure sensor deployed near the study site 31 July-8 August, 2013 that was coupled to a barometric pressure sensor to correct for atmospheric pressure changes. We tied the pressure sensor deployed at Driftwood Bay to the NOAA tide gage at Nikolski during the GNSS elevation survey. The tidal curve from Driftwood Bay was converted to a tidal datum referenced to the National Tidal Datum Epoch using the Online Tidal Datum Computation Tool developed by JOA Surveys (ref to https://www.tidaldatumtool.com). All elevations we report are referenced to mean tide level (MTL) as defined by the local tidal datum tool at Driftwood Bay, which incorporates a-0.137 m deviation from MTL at Nikolski due to variation in the geoid. 2 1.2 Dating The timing of sand sheet deposition was estimated using multiple methods, including 137 Cesium activity and radiocarbon dating. Gamma spectroscopy, using low-background, high-efficiency, high-purity Germanium detectors, yielded 137 Cs activities in 1-cm-thick sediment intervals sampled within 17 cm of the surface in core 201 (Table S1). We used peak 137 Cs activity to identify the chronostratigraphic interval of AD 1963 (Pennington et al., 1973) and its relationship to the depth of the youngest sand sheet. Samples were initially dried, homogenized by grinding, packed into standardized vessels, and sealed for at least 24 h before counting. Activities were corrected for self-absorption using a direct transmission method (Cutshall et al., 1983; Cable et al., 2001). Radiocarbon dates obtained through the National Ocean Sciences Accelerator Mass Spectrometry facility in Woods Hole, Massachusetts. We sampled plant macrofossils from above and below the sand sheets and tephras for dating, including delicate peat moss (Sphagnum spp) stems and leaves, black stems of sedges (Carex spp), and woody twigs. We computed calibrated radiocarbon ages from lab-reported ages using OxCal (version 4.2.4) (Bronk Ramsey, 2009a) with the IntCal13 data set of Reimer et al. (2013). Table 2 reports one-sigma, analytic ages in radiocarbon years before 1950 (14 C yr BP), and two-sigma, calibrated ages in solar years before 1950 (cal yr BP). Posterior ages that estimate the times of sand sheet and tephra deposition are listed in Table 3 based on a Bayesian age-depth model constructed with OxCal (Bronk Ramsey, 2008). A Bayesian age-depth model incorporated radiocarbon analyses (Table 2) of samples from multiple cores and the activity of 137 Cs in the upper 25 cm of sediment in core 201 (Table S1). We used OxCal (Bronk Ramsey, 2009a), to construct the model, which computed probability