Six deep-ocean buoys recorded the Pacific oscillating for more than ten hours after the Kamchatka tsunami

Nothing in the simulation was tuned to the buoy records ^[7]. The team started from the rupture model the U.S. Geological Survey built from seismic waves recorded around the world: a fault about 690 kilometers long, cut into 345 patches, with slip above 10 meters in places ^[7]. They let the seafloor rise and sink over the roughly 200 seconds the rupture took, instead of all at once ^[8]. Long-wave equations then carried the tsunami across the ocean. In water about 4 kilometers deep, those equations move it at about 700 kilometers per hour ^[9].

The buoys were never used to fit anything, so they are an independent check. Each DART station has a pressure sensor on the seafloor that registers the height of the sea surface. A surface buoy relays the readings by satellite, and the station samples every minute or faster once a tsunami passes ^[4].

An arrival error of about a minute ^[10] is small against travel times of 27 to 144 minutes, roughly 0.7 to 3.7 percent ^[1]. The height match needs more care. An average miss of 6 centimeters ^[10] is about 7 percent of the 85-centimeter crest at the nearest station ^[5]. It is also three-quarters of the 8-centimeter peak at the farthest one ^{[6] [2]}. An average across six stations can hide a weak fit at the small end. On land, at Baikovo on Shumshu, the model put a flow 6.2 meters deep about 600 meters inland. Reported figures were 6 meters and 650 meters ^[12], an inland miss of about 8 percent ^[4].

I think this is the firmer result. For this event, a rupture model built from seismic data alone explains what the deep ocean recorded, with no underwater landslide or second fault required ^[13].

The ringing comes from the buoy records themselves. For the first hour, seismic waves shook the sensors, and the traces jitter at periods of a few minutes. Then the tsunami's energy settles into periods between about 30 and 80 minutes ^[14]. More than ten hours after the first waves, every buoy was still oscillating in that band, far above its quiet level ^[15]. At those periods, ten hours is somewhere between about 7 and 20 full cycles ^[3]. "Physicists would say that the Pacific behaved like a bell with little damping: Struck once, it rang for a long time," one of the co-authors wrote ^[17].

The authors did not report whether the earthquake-only model reproduces that tail. The account reports the model's match on arrival times and peak heights ^[10]. It describes the ten-hour oscillation from the observations ^[15]. The headline says the rhythm "could help tsunami warnings," and the explanation turns to bays and harbors ^[18]. The published text does not set out a rule a warning center could apply to decide the danger is over.

Any such rule would have to be local. According to the authors, the dominant period differed from buoy to buoy, most likely because each station hears the tsunami after its own journey over ridges, trenches and islands ^[16].