GoKawiilSeismicity
Large-magnitude events were extracted from the GCMT catalogue32 and from the ISC bulletin33,50. All events were relocated from our hydroacoustic records (Extended Data Table 1).
Hydroacoustics
The hydroacoustic array consisted of five autonomous hydrophones (Fig. 1a) moored in the axis of the SOFAR (sound fixing and ranging) channel and anchored to the seafloor. They continuously record sounds in the 1–125-Hz frequency band (250 Hz sampling rate), which include acoustic waves generated by seismic events near their epicentres24,27 (tertiary waves or T-waves) or by hot lava–seawater interactions (short impulsive events called H-waves)16,44,45,46,47, among other sounds (icequakes, large baleen whale calls or big vessels25). Source events can be precisely located (within 1–2 km)15,24,25 by trilaterating the T-wave or H-wave arrival times at each hydrophone, all synchronized with the GPS time. Following the approach of Fox et al.24, with at least four arrival times, we can derive the uncertainties in latitude, longitude and origin time from the covariance matrix of the nonlinear least-squares inversion. In that paper, they do not exceed 2 km (major axis) versus 1 km (minor axis) at 1σ (68% confidence level) inside the array; also, for each event, we systematically performed 10,000 random Monte Carlo draws within 1 s and 1 m s−1 intervals from the arrival times and sound speeds, respectively, yielding the same figures. Automatic event detection, arrival time association and trilateration are based on newly developed methods51,52. Unfortunately, the hydrophone located northwest of Amsterdam Island (top-left in Fig. 1a) failed in early April 2024. All reported events were consequently located from the acoustic records of the four remaining hydrophones. Hydroacoustic recordings were crucial in accurately repositioning the GCMT focal mechanisms and ISC events, provided in land-based catalogues with location uncertainties of 20 km or more owing to the remoteness of the area. Between the start of seismic activity on 26 April until 2 May 2024, nearly 500 T-wave events were identified and located (versus 22 GCMT and 52 ISC events) and more than 2,000 H-wave events (Extended Data Fig. 2).
Direct-path ranging
Horizontal displacements of the seafloor were measured by acoustic direct-path ranging between Exail-manufactured Canopus beacons deployed on the seafloor. All acoustic transponders were mounted on a 3.50-m-high tripod and equipped with pressure, temperature and, for some, conductivity sensors to infer the sound speed at ranging times. During the maintenance cruise in 2025, limited datasets were downloaded with an acoustic modem to ensure that all beacons were running properly (inclination, battery level, sensors, some ranging measurements). The direct-path ranging schedule was designed to ensure that the batteries of the beacons would last 4 years. Accordingly, each station in the ridge array ranged to the neighbouring stations in acoustic line of sight every 4 h. In the TF array, ranging sessions were set 2 h apart. Hence, during a ranging session, all baselines are measured twice, that is, both ways, with a precision of 1.5 microseconds for the two-way travel times (<3 mm at 1,500 m s−1). Travel times were then converted into distances based on the sound speed inferred from the ancillary sensors, using Del Grosso’s formula53. The converted ranging uncertainties increase with the length of the baseline and with the uncertainties in the sound speed, at best derived from the harmonic mean of the sound speeds measured at both ends of a baseline. However, because, in 2025, we only recovered (and redeployed) stations TP7 and TP13 to access their whole datasets, all baselines shown are solely based on the sound speed at station TP7 or at station TP13 (Fig. 3 and Extended Data Fig. 3).
Two perpendicular inclinometers also monitor the beacon attitude. All tripods tilted during the seismic crisis, up to 12° for some stations (1° tilt of the 3.50-m-high tripod will cause 6 cm of displacement of the acoustic head). Because tripods are not oriented on the seafloor, we considered all possible orientations relative to the baseline direction. Given the size of the measured displacements from TP7 (Fig. 3 and Extended Data Fig. 3), the resulting uncertainties remained nearly two orders of magnitude smaller. However, for baselines from TP13, they obscured any signal (Extended Data Fig. 3b).
All tripods were blindly lowered with a cable to the seafloor from R/V Marion Dufresne. It was therefore very challenging to position them on bathymetric escarpments or cliff edges and ensure that they were stable and roughly vertical, and that each beacon could communicate with its neighbours, especially on a rugged basaltic seabed and with maps that had a pixel resolution of 20 m × 20 m at best. Deployments of the 15 beacons took 6 h per tripod on average (ranging from 3 h to 11 h).
Bottom-pressure recorder
Vertical displacements of the seafloor were estimated with pressure sensors. A self-calibrating (that is, drift-controlled) BPR was set on the floor of the axial valley to precisely monitor its vertical deformation (cyan hexagon in Fig. 1b). For redundancy, the BPR, manufactured by RBR Ltd., has two Paroscientific Digiquartz sensors (model 46K-313) measuring the ambient ocean pressure through a single oil capillary. They also regularly measure the pressure inside the instrument casing as a calibrating ‘zero’. These ‘zero’ measurements, corrected for air-pressure variations inside the instrument with a Paroscientific barometer (model 216B-102), are used to estimate the sensor drift, with the assumption that the drift of Paroscientific pressure sensors at low pressure (that is, atmospheric pressure) is the same as that at seafloor depth, which amounts to neglecting the scale factor. Lower-resolution pressure sensors (without drift control) were also fitted to the acoustic beacons.
... continue reading