Introduction
On 15 January 2022, 04h:15 UTC, the volcano Hunga Tonga-Hunga Ha'apai in the south-west Pacific Ocean (20°32'32.37''S, 175°23'38.67''W) erupted in what proved to be the most powerful such event since Krakatau in 1883. Among the many impacts of the eruption, a substantial tsunami propagated throughout the south-west Pacific Ocean. The signatures of the eruption were recorded at a wide range of recording stations globally, including the atmospheric pressure wave, the tsunami itself and, in addition, higher-order responses, such as a tsunami associated with the pressure wave (Carvajal et al. Reference Carvajal, Sepúlveda, Gubler and Garreaud2022).
Here we describe what is probably the most southerly oceanic measurement of the tsunami arrival, as the event was detected in a sub-ice mooring at the grounding line of the Ross Ice Shelf at 82.47°S. It was recorded on the margin of the Ross Ice Shelf along the Siple Coast at the Kamb Ice Stream 2 (KIS2) field camp by instruments deployed to record ice-ocean interactions in order to improve understanding of ocean effects on ice-sheet stability (Whiteford et al. Reference Whiteford, Horgan, Leong and Forbes2022).
Analyses of tsunamis and ice shelves tend to focus on the observable stability of the terminal face (e.g. Brunt et al. Reference Brunt, Okal and MacAyeal2011) as well as implications for research infrastructure (Burbidge et al. Reference Burbidge, Power, Gusman, Wang, Lukovic and Black2020). However, there is evidence that the tsunami propagates into the ice shelf cavity, as demonstrated by the analysis of the 2011 Honshu (To¯hoku) event's impact on the Sulzberger Ice Shelf (Papathanasiou et al. Reference Papathanasiou, Karperaki, Theotokoglou and Belibassakis2015). Extending this, the effects of a 2015 Chilean earthquake-driven tsunami were observed with a broadband seismic array deployed in a north-south line through the middle of the Ross Ice Shelf (Bromirski et al. Reference Bromirski, Chen, Stephen, Gerstoft, Arcas and Diez2017), revealing the wave propagation. The Tongan wavefront event described here followed a more direct oceanic route to the cavity than the 2015 Chilean event. The Tongan tsunami is closer in path to a modelled response due to a 2011 event in the Kermadec region (Bromirski et al. Reference Bromirski, Chen, Stephen, Gerstoft, Arcas and Diez2017).
Results
The data were recorded at the KIS2 camp (latitude -82.470442°; longitude -152.291643°; see Fig. 1). The sampling was measuring ocean processes in a 6 km-long sub-ice channel that extends landward from the edge of the ice-shelf cavity. The channel is ~240 m high and 150 m wide beneath 410 m of ice and snow. In early 2022, a hot-water borehole was drilled through the ice shelf to access the ocean at the head of the channel. Hydrographic instruments were suspended in the ocean cavity late on 11 January 2022 local time, and 4 days later and 7000 km to the north the Hunga Tonga-Hunga Ha'apai eruption occurred.
a. Ross Ice Shelf and the location of the Kamb Ice Stream 2 (KIS2) channel (inset shows location within Antarctica) with b. an outline of the known extent of the channel (see Whiteford et al. Reference Whiteford, Horgan, Leong and Forbes2022) and the KIS2 sample station near the channel head. The solid black line is the estimated edge of the ice-shelf cavity. ASAID = Antarctic Surface Accumulation and Ice Discharge project; wgs84 = World Geodetic System 1984 coordinate reference frame.
The instruments providing the data were three RBR Concerto conductivity-temperature-depth (CTD) instruments. These instruments included a pressure sensor rated to 1000 (dbar) with an initial accuracy of ± 0.05% full scale and a resolution of < 0.001% full scale, equating to 1 cm. Sensors were sampled at 60 s intervals with a response time of < 0.01s. The three instruments were at pressures of 425, 495 and 648 dbar, respectively, and all three gave consistent readings, so only information from the 495 dbar unit is described here (i.e. ~85 m beneath the ice underside and 155 m above the sea floor).
Considering the central instrument as representative, the timeseries of the pressure, after high-pass filtering to remove the 0.75 m-amplitude tide (primarily diurnal and including the spring-neap cycle), clearly shows the arrival of the tsunami (Fig. 2a). The record (Fig. 2b) shows an initial perturbation ~6.4 h after the eruption. This was followed 7 h later (i.e. 13.4 h after the eruption) by a sudden and sustained burst of variability due to the tsunami's arrival. The variability in the high-passed signal persisted for ~4 days and did not return to background variability seen prior to the arrival until approximately day 20.
The ice-cavity channel pressure timeseries showing a. the raw (blue) and high-pass-filtered (black) pressure perturbation timeseries (note that for clarity time here starts at 1 so that noon on 1 January = 1.5). b. An expanded version of the high-pass-filtered pressure record for 15 January 2022, showing the arrival of the initial atmospheric front and the start of the tsunami-driven response (1 dbar ≈ 1 m).
Discussion
The initial perturbation 6.4 h after the eruption is the air pressure wave, and the 7000 km separation implies a propagation speed of 300 m s-1. This compares favourably to the pulse speed of 307 m s-1 described by Carvajal et al. (Reference Carvajal, Sepúlveda, Gubler and Garreaud2022). The tsunami arrival 13.4 h after initiation is ~4 h longer than the direct travel time implied by a deep-water propagation speed. This reflects the slower wave travel on the continental shelf and the propagation response to varying topography.
Closer examination of the timeseries over the days following the wave's arrival shows short bursts of oscillating events of periods of ~3–6 h (Fig. 2a). Assuming that the channel is a rigid lid cavity, we expect pressure variations at the channel mouth to propagate into the mooring site as acoustic waves with a velocity of ~1445 m s-1 and a transit time of <5 s. In addition, the channel is ~150 m wide and the sampling location is ~500 m from the end of the channel, so the effects of these boundaries will be felt very quickly at the sample location. Current meters were included in the sampling but in a lower-frequency regime (30-min interval). The near-bed unit shows a perturbation of ~2 cm s-1 at approximately the time of the tsunami's arrival, but there is no clearly distinguishable associated response in the surface-mounted precision GPS data.
The 2015 Chilean event produced elevated high-frequency perturbations of the Ross Ice Shelf surface for < 2 days (Bromirski et al. Reference Bromirski, Chen, Stephen, Gerstoft, Arcas and Diez2017). The implication then is that the 3–6 h variations and the 4 day decay period seen at KIS2 are associated with the wider cavity embayment and not the 5 km-long sub-ice channel. Such post-event ‘ringing’ could be associated with either the local coast or the regional continental shelf shape. Being ~750 km from the continental shelf break, this would tend to rule this out. However, local to the channel are a number of features of the Siple Coast that would enable a complex wave response, including the concave-curved coastline and the presence of some islands and headlands (Fig. 1).
Notably, in terms of our normal sampling protocols for ice-cavity instruments, in the case of the 2022 deployment we were working on ice fixed to the bedrock and so not moving with the tides. In addition, we were trialling a sampling protocol that provided a faster sampling rate (60 s). If we had been sampling as per earlier experiments, our observations of the tsunami would have been far less clear. This serendipitous dataset has implications for understanding high-latitude shallow-water wave processes beneath ice shelves.
Acknowledgements
The work was funded by the Antarctic Science Platform (ANTA1801) and Antarctica New Zealand. We thank the Victoria University of Wellington Hot Water Drilling Team led by Darcy Mandeno, the NIWA capex programme and the IceFin team led by Britney Schmidt. Andy Mullan, Regine Morgenstern and the camp team assisted with the field deployment. We thank the traverse team, the KIS2 camp team and Paul ‘Woody’ Woodgate. Yusef Karim and Nicolas Zuluaga at RBR Systems provided timely technical support. We thank two anonymous reviewers for their comments.
Competing interests
The authors declare no known competing interests.
Data availability
Data are available at SeaNoe (https://doi.org/10.17882/96263).
Author contributions
Craig Stewart developed the instrumentation, led the deployment and first identified the signal. Huw Horgan led the overall experiment and mapped the channel. Craig Stevens conducted the analysis and led the write-up. All authors contributed to the manuscript.
Open access
