Earth

One earthquake, heard by 113 seismometers

On 19 August 2018 a magnitude 8.2 broke 600 km beneath Fiji — deep enough that nobody was hurt, big enough that the whole planet rang. Here is the hour that followed at — seismometers, one trace each, stacked at the station’s true angular distance from the epicentre. Press play. The wiggles do not start at random: the ground first moves at —, — away, — after the rupture, and last at —, — away, — after. In between, the onsets lie on a curve that flattens — the first 45° of arc cost — minutes, the next 45° cost —, the 45° after that only — — because a wave crossing a longer arc dives deeper, and the rock down there carries it faster. Every trace is scaled to its own largest swing, so the nearest station and the furthest are drawn the same height: this picture is timing, not size. Nothing is modelled over the data. Tap or hover any trace.

One station’s vertical ground motion, scaled to its own largest swing

First motion — five times that station’s own pre-event noise

—

Data & source

Data: EarthScope (SAGE) FDSN dataselect waveforms and FDSN station metadata — vertical broadband (BHZ) channels of networks IU (Global Seismographic Network, USGS/NSF) and II (IRIS/IDA), one hour from the origin time. Event parameters from the USGS (us1000gcii): 2018-08-19 00:19:40.67 UTC, M_ww 8.2, 18.1125°S 178.153°W, 600 km deep. Fetched —. EarthScope distributes GSN and IRIS/IDA data openly and applies CC-BY-4.0 where the network owner declares no licence; USGS event data are public domain.

Scope and caveats, plainly. This is one earthquake at the stations that happened to be listening, not a census of the world’s seismometers: — stations of two global networks returned an hour that was at least 90% complete, and the picture is a slice chosen because those two networks are spread evenly enough over the planet to draw a curve. Thousands of other instruments recorded this quake. Coverage is uneven — the Pacific is thinly instrumented, so there are few traces inside 20° — and no station sits beyond 161°, because the point opposite Fiji is in the Atlantic. Amplitude is deliberately destroyed. Each trace is divided by its own robust peak (the 99.9th percentile of its absolute value, so one telemetry spike cannot flatten a whole trace), which is why a station next to the rupture and one on the far side of the world are the same height; the few samples above that peak are clipped. Raw 20–40 Hz samples are averaged into 0.5 s bins, the mean and linear trend of the hour are removed, and gaps shorter than the hour are drawn straight across. The first-motion marks are a threshold crossing, not phase picks. Nothing can arrive anywhere in the first 60 s — the source is 600 km down — so that window measures each station’s own noise, and the onset is the first sample after it reaching five times that noise. It is not identified as P, or as any named phase; — station(s) whose largest swing was under eight times their own noise are shown but left untimed. A quiet station therefore catches a weak arrival that a noisier neighbour at the same distance misses, which is why the onsets past about 135° fall into two groups rather than one line. No model is drawn over the data. EarthScope has retired both its TauP travel-time service and its ASCII waveform service, so there are no predicted arrival curves here at all — every curve you can see is the traces themselves. The inset is an azimuthal equidistant projection centred on the epicentre, so distance from its middle is the same scale as the vertical axis of the wall; the ring is drawn at the distance of the furthest station that has moved, not at a computed wave speed.