The Earth rang like a bell for weeks

What happened on 26 December 2004

At 00:58 UTC, 7:58 in the morning on Sumatra in Indonesia, the fault beneath the ocean off the island broke along about 1,300 kilometres, as far as the Andaman Islands. The rupture lasted between eight and ten minutes.

The earthquake had a magnitude of 9.1, according to the US Geological Survey (USGS); estimates in the first months ranged from 9.0 to 9.3. The earthquake and the tsunami it set off killed or left missing 227,898 people in 14 countries, from Indonesia and Sri Lanka to East Africa.

The earthquake's waves moved the ground by more than a centimetre everywhere on the planet's surface. After they had passed, the Earth kept vibrating like a struck bell, in its own tones. Nobody could feel this vibration: it is too slow and too small. Seismometers recorded it for weeks.

How hard the Earth breathed

The tone that lasted longest is the one in which the whole planet swells and shrinks at once. Seismologists call it the “breathing mode”.

Every 20 min 28 s, the ground everywhere rose and fell by about 0.05 millimetres either side of its rest position at first: 0.1 mm from top to bottom, about the thickness of a sheet of paper. The 12 stations in the table, from Târgușor in Romania to Japan, Zambia and Chile, read between 0.038 and 0.055 mm, 13% below to 25% above the median. The real motion of this tone should differ from place to place by only about 1% (Rosat and colleagues, 2007), so the spread comes mostly from the instruments: Park and colleagues found 5–10% errors in the response of a substantial minority of seismometers. In the first days the stations kept the same beat: they were at most a tenth of a cycle apart, about two minutes out of 20.

The team that studied records from more than 400 stations in 2005 arrived at the same number: 49 micrometres, or 0.049 millimetres (Park and colleagues, Science).

The motion halved roughly every 17 and a half days. After 30 days it still had 30% of its starting size. It could be followed for almost three months, until 28 March 2005, when another great earthquake, near the island of Nias, drowned it out (Okal and Stein, 2009).

Left out: Pallekele, Wellington and South Pole, where the tone's day-by-day fall does not follow a straight line (the rule in the method).

The ground under Romania rang too

The station at Târgușor, in Constanța county in south-eastern Romania, run by the German geoscience centre GFZ with Romania's National Institute for Earth Physics, recorded the same breath: 0.044 mm on the second day, right at the median of the 12 stations. Between 1 and 3 January 2005 the station sent no data for 45 hours; in the recording on this page, those hours are silent.

Hear the breath recorded at TârgușorThe whole month from Târgușor

Why the Earth rings like a bell

Anything that is struck vibrates in its own tones: a bell, a glass, a string. The tones depend on the object's size and what it is made of. The Earth is 12,742 kilometres across and made of rock and iron, so its tones are very low. The lowest takes almost an hour for a single vibration.

Our ears hear from about 20 vibrations a second upwards. The Earth's breath makes one every 1,228 seconds, a tone nearly 25,000 times lower than the deepest sound we can hear. So on this page the recording is sped up 500,000 times: a month fits into five seconds, and the breath sounds as a clear tone of 407 hertz.

Only very large earthquakes strike the planet hard enough for these tones to rise above the background. They were first measured after the great earthquakes of the mid-20th century, above all the magnitude 9.5 Chilean earthquake of 1960, from paper records digitised by hand. In 2004, more than 400 digital stations caught them more clearly than ever before.

The tones tell what is inside. Their pitch depends on how dense and stiff the deep layers are, so seismologists use them to weigh the planet's interior: the standard reference model of the Earth's interior (PREM, 1981) was built partly from them. In 2005 they also gave one of the larger estimates of the earthquake's size, magnitude 9.3 (Stein and Okal, Nature).

The Earth vibrates very faintly even without earthquakes, in the same kind of free oscillations, kept going by the atmosphere and the oceans instead of one blow: a “hum” first reported in 1998.

How the sound was made, and what is transformed

What is real

The sound is made from the seismometer's record (vertical component, one reading per second, 30 days from the earthquake) after removing the instrument response and filtering; every 20th reading is kept. The frequencies and the ratios between the tones are as measured. The fading is real but made gentler in the sound (see Loudness), and each single tone is played at the same loudness, so their strengths are not compared. The stations on offer are Albuquerque (the Global Seismographic Network, GSN, USA), Târgușor (GEOFON and Romania's National Institute for Earth Physics) and Charters Towers (GSN, Australia).

What is transformed

• Speed: everything plays 500,000 times faster, so every tone is 500,000 times higher. A month lasts five seconds.
• Band: only vibrations from 0.25 to 1.35 millihertz are kept (on the page, 125–675 Hz), where the tones are. Tides and most of the noise are left out.
• Loudness: the difference between the very loud start and the faint end is halved, so the end can be heard too. Silence in the Târgușor recording means hours without data.
• Single tones: each is the same recording passed through a very narrow filter around the tone, then played at the same speed.
• The globe: each tone is drawn with a largest swing of 3% of the Earth's radius, so each has its own magnification; for the breath (0.05 mm) it is nearly four billion times. The motion runs 800 times faster than in reality so that it can be seen: the breath comes every 1.5 seconds. The tones' strengths are not compared; the way each one rises and fades follows the median of the stations.

How the breath was measured

Each station's record was converted from instrument units into ground motion, using the instrument response published by the network. On two-day windows a sine wave at the breath's frequency was fitted, with the neighbouring 20-minute tone fitted beside it, and the day-by-day fall gave the half-life. A seismometer also feels gravity weaken as the ground rises, which inflates the apparent motion by 12%; the numbers in the table are corrected. Stations were chosen by a rule set before reading the amplitudes: less than 10% of the data missing, and a fall that follows a straight line on a logarithmic scale.

The median of the stations gives a half-life of 17.5 days, a quality factor Q of about 5,600 (between 5,000 and 6,000 at single stations); Okal and Stein measured 5,579 ± 140 at four stations (CTA, MAJO, PPT and YSS), with 85-day windows.

Sources

1. Park, J. et al. (2005). Earth's free oscillations excited by the 26 December 2004 Sumatra-Andaman earthquake. Science 308, 1139–1144. doi.org/10.1126/science.1112305
2. Okal, E. A. & Stein, S. (2009). Observations of ultra-long period normal modes from the 2004 Sumatra-Andaman earthquake. Physics of the Earth and Planetary Interiors 175, 53–62. doi.org/10.1016/j.pepi.2009.03.002
3. Stein, S. & Okal, E. A. (2005). Speed and size of the Sumatra earthquake. Nature 434, 581–582. doi.org/10.1038/434581a
4. Lay, T. et al. (2005). The Great Sumatra-Andaman Earthquake of 26 December 2004. Science 308, 1127–1133. doi.org/10.1126/science.1112250
5. USGS. M 9.1 – 2004 Sumatra – Andaman Islands Earthquake (event page). earthquake.usgs.gov/earthquakes/eventpage/official20041226005853450_30
6. USGS. Today in Earthquake History, 26 December (227,898 dead and missing, 14 countries). earthquake.usgs.gov/learn/today/index.php?month=12&day=26
7. Masters, G. & Laske, G. REM, Reference Earth Model: tables of mode frequencies and attenuation (UCSD). igppweb.ucsd.edu/~gabi/rem.html
8. Dziewonski, A. M. & Anderson, D. L. (1981). Preliminary reference Earth model. Physics of the Earth and Planetary Interiors 25, 297–356 (Table V). doi.org/10.1016/0031-9201(81)90046-7
9. Suda, N., Nawa, K. & Fukao, Y. (1998). Earth's background free oscillations. Science 279, 2089–2091. doi.org/10.1126/science.279.5359.2089
10. Rhie, J. & Romanowicz, B. (2004). Excitation of Earth's continuous free oscillations by atmosphere–ocean–seafloor coupling. Nature 431, 552–556. doi.org/10.1038/nature02942
11. Albuquerque Seismological Laboratory/USGS (1988). Global Seismograph Network (GSN – IRIS/USGS), network IU. doi.org/10.7914/SN/IU
12. EarthScope Consortium (1986). Global Seismograph Network – II, network II. doi.org/10.7914/SN/II
13. GEOFON Data Centre (1993). GEOFON Seismic Network, network GE. doi.org/10.14470/TR560404
14. MedNet Project Partner Institutions (1990). Mediterranean Very Broadband Seismographic Network (MedNet), network MN. doi.org/10.13127/SD/fBBBtDtd6q
15. EarthScope Consortium, FDSN data and metadata services (the LHZ records and instrument responses). service.iris.edu/fdsnws/
16. GEOFON, GFZ Potsdam, FDSN services (station TIRR, Târgușor). geofon.gfz.de/fdsnws/
17. Rosat, S., Watada, S. & Sato, T. (2007). Geographical variations of the 0S0 normal mode amplitude: predictions and observations after the Sumatra-Andaman earthquake. Earth, Planets and Space 59, 307–311. doi.org/10.1186/BF03353109