Earth
Volcanoes and earthquakes heard as one resonating body built from the ground's own modes, starting with the Svartsengi eruption sequence on the Reykjanes peninsula, 2023–2026.
IMO's lava-flow and effusion data is not part of this build.
The timeline zooms through five levels, from the whole sequence to a few minutes. What you hear is one resonating body. Its 36 modes are built from the ground's own movement in each eruption cycle: the three GNSS stations' motion, transposed up into hearing. A comb, tuned from the cycle, is the room it rings in. The magma's inflow bows it, and each eruption strikes it. The earthquakes of IMO's catalogue are the Swarm. At the hours and minutes levels you also hear the MELH seismometer itself, the Ground, played 250 times faster than real time. The sound starts at your first click or key press. Press L for the legend: hold a chip to hear one quantity on its own.
Credits
- Data: the Icelandic Meteorological Office (IMO, Veðurstofa Íslands): its earthquake catalogue, the GNSS stations SKSH, SENG and THOB, and the MELH seismometer. Compiled and shared by Leif Karlstrom (Department of Earth Sciences, University of Oregon; Volcano Listening Project), with whom this design is made.
- Built with DarkSonification. Designed by Miguel Angel Crozzoli; implemented with Claude AI agents.
Data
- Earthquakes. IMO earthquake catalogue (Svartsengi polygon), located + magnitude-estimated by IMO.
- GNSS. daily N/E/U, IMO stations SKSH, SENG, THOB (files *_060423_092426_I.csv).
- Seismometer. RK.MELH.10 HH* 100 Hz + LM* 1 Hz miniSEED, 2024-01-14 00:00-10:23 UTC, raw counts.
- Eruption sites. Landmark coordinates: © OpenStreetMap contributors, available under the Open Database License (ODbL).
- Lava fountains. Observed fountain heights from Rhodes et al. (2026), Bulletin of Volcanology 88:89, doi:10.1007/s00445-026-02006-3 (supplement ESM2); the other eruptions' heights are simulated with Leif Karlstrom's model. Fissure positions are digitised from Rhodes et al. (2026), Fig. 3, or placed from IMO's reports.
- Eruption dates and places. From IMO's published reports (en.vedur.is).
- Fissure segment positions. The fountains' segment centroids by Leif Karlstrom (2026-11-04), approximate (150–1000 m): the observed eruptions' segment positions are Leif Karlstrom's reconstruction from Rhodes et al. (2026): the fissure trace digitised from Fig. 3, each segment placed along it with the webcam-to-segment bearings in the supplement (the webcam located by aligning Fig. 6 with Fig. 3); the observed heights are from the supplement ESM2; the simulated eruptions' segments are his estimates.
- MELH seismometer position. Approximate, from Leif Karlstrom (2026-11-04), rounded to 3 decimals (about ±100 m).
- GNSS network. IMO GNSS time series (via Leif Karlstrom); station positions: IMO GNSS browser, brunnur.vedur.is.
- Binaural decoder (the ambisonic mode). Its four first-order filters are derived from the MIT KEMAR HRTF measurements (compact set) by Bill Gardner and Keith Martin, MIT Media Laboratory, 1994 (sound.media.mit.edu): “This data is Copyright 1994 by the MIT Media Laboratory. It is provided free with no restrictions on use, provided the authors are cited when the data is used in any research or commercial application.” W. G. Gardner and K. D. Martin, “HRTF measurements of a KEMAR,” Journal of the Acoustical Society of America 97(6), 3907–3908 (1995). The decoder: weighted mode-matching over all 710 measured directions, max-rE weighted, diffuse-field equalised.
- Coastline. Natural Earth 1:10m land, public domain.
The full list, with every licence: CREDITS.txt.
What the data can and cannot say (export of 2026-11-17T11:44:50.855Z, contract 0.13):
- MELH's position is Leif Karlstrom's approximate one (email 2026-11-04; pos_status approx, +/-100 m): no public service lists RK.MELH; SKSH, SENG and THOB are from IMO's GNSS browser station list (measured)
- waveforms are raw counts (no instrument response yet)
- MELH covers 2024-01-14 00:00-10:23 UTC only; the station was overrun by lava
- GNSS days without an IMO solution are gaps (null) and flagged <ST>_interp; all three stations have a gap across the 14 Jan 2024 eruption (the sound reads the straight line across a gap, veiled)
- GNSS 1-2 day excursions that return, in N, E or U, are flagged <ST>_spike
- the rooms and waits are bounded by IMO's eruption timeline (events/eruptions.json); deflation_events is W1's GNSS detection, kept for the comparison (gnss_match)
- fc_hz assumes a 3 MPa stress drop and treats ML as Mw
- eq_b uses one whole-catalogue Mc (max curvature)
Sound engine
The sound is made by SuperCollider running inside this page: its synthesis server (scsynth) and
its language (sclang), compiled to WebAssembly. They run the design's own SuperCollider files, the same
ones the desktop version plays: the SynthDefs, the cooking and the message parser. Two things differ from the desktop,
both in how data reaches the engine. The engine cannot read files, so those files are included in the code it runs.
The seismometer's samples are sent to it from the page, block by block, as the desktop's bridge does.
Software and licences
- SuperCollider, WebAssembly
build from the development branch at commit
5d47207(unreleased). scsynth and sclang: GPL-3.0; the WebAssembly bindings: AGPL-3.0. Unmodified; the licence text is inengine/LICENSE. - D3 v7.9.0 (ISC licence).
- Fonts of the paper look: Karla and Rubik (SIL Open Font License 1.1), served by Google Fonts.
Built 2026-11-18 10:37:55. ← back to Earth