Reanalysis traces Hunga's largest tsunami to a caldera collapse heard 2,000 km away

Hydroacoustics is why any of this is detectable. Sound put into seawater by a violent submarine process travels as a tertiary wave, or T-wave, with far less loss than the equivalent energy travelling through rock, and the authors describe an isolated volcano rising from the seafloor as a bell that couples the noise of its own destruction into the water column ^[11]. The geometry is the trick. The nearest seismometer sat in Fiji, about 750 km away, a distance at which many volcanic seismic signals are poorly transmitted through the Earth ^[12]. Satellites do better on the eruption itself, spotting heat, gas and plumes and giving aviation useful notice, but according to the authors they cannot say whether a wave is on its way ^[13].

Note what kind of study this is. The 14 stations were seismic stations, and the work is a reanalysis of records those instruments had already written ^[9]. That matters more than the distances do. The collapse was faint in the ground and loud in the ocean ^[8][10], which means the useful observable was inside an operating network's archive rather than absent from it. The farthest station used sat about 3.5 times as far from Hunga as the nearest seismometer did ^[20].

The reason the mechanism is worth distinguishing is the size gap. The explosive phase sent 1-to-4-metre runups onto Tongatapu within minutes ^[3]; the collapse phase, more than an hour later, reached 18 to 40 metres on islands inside 100 km ^[4]. At the top of each range that is a factor of ten ^[19]. A monitoring scheme keyed to explosions would have fired for the smaller waves and stayed quiet for the one that destroyed villages.

The thing this does not tell you is how much warning the signal buys. The strong part of the T-wave lasted about five minutes ^[10], and the worst-hit coasts were within 100 km of the source ^[4]. Establishing that the collapse caused the wave required knowing precisely when the wave reached Tonga ^[18], and that evidence came from a telecommunications tower at Kanokupolu whose weather station made its last scheduled transmission at 6 pm, 28 minutes before the collapse began ^[16][21], before the wave flattened the tower and scattered it inland ^[17]. The account available here stops as the team turns to pinning that arrival time ^[22], so the lead-time number is not yet in front of me.

There is also a discrimination problem this single event only sketches. In the first hour, submarine landslides down the flanks were loud enough to be heard hundreds of kilometres away and strong enough to cut communications cables ^[14], and they were not the source of the 18-to-40-metre runups. One eruption gives one loud tsunamigenic collapse and several loud events that were not, which is a thin basis for a false-alarm rate across the hundreds of submarine volcanoes whose activity state is barely characterised ^[15].