Curiosities / Episode 02
A strange seismic signal. One pulse every 92 seconds. For nine days. What could possibly keep the Earth ringing?
September 2023. Instruments around the world hear a rhythm that does not behave like an ordinary earthquake.
Investigate the signal ↓The explanation crossed disciplines: seismology, satellite imagery, field evidence and a model of moving water. Follow the clues, then separate the measurements from the reconstruction.
Instrument record / September 2023
An ordinary earthquake usually produces a mixture of frequencies that dies away much sooner. This signal was dominated by one slow rhythm: approximately 92 seconds per cycle.
Seismometers detected it internationally for up to nine days. Something was repeatedly transferring force into the solid Earth. The instruments revealed the rhythm; they did not yet reveal the cause.
Location / Combining records
Researchers compared signals recorded at different places and brought them together with reports and images from Greenland. The trail led to Dickson Fjord, a remote inlet in the island’s east.
Satellite imagery showed a major change in the landscape. A mountain slope above the water was missing. The question became: could a landslide explain a signal that lasted days?
Reconstruction / 16 September 2023
Approximately 25 million cubic metres of rock and ice collapsed into the fjord. That is a volume, not a wave height. The huge, rapid displacement of water generated a tsunami.
The 2024 investigation linked the slope’s failure to the thinning of a glacier that had supported it. Evidence from before and after the collapse helped reconstruct an event nobody was standing there to film in full.
Field evidence / Initial tsunami
Near the landslide, the initial tsunami reached a local run-up of approximately 200 metres. Run-up is how high water reaches on land above sea level. It is not the height of a wave cruising around the ocean.
This was the violent opening event. It must be kept separate from the much smaller motion that followed. A 200-metre wave did not continue for nine days.
Numerical reconstruction / A trapped rhythm
Think of bathwater rocking after a single push. Water rises at one end as it falls at the other, then reverses. The basin gives that motion a preferred rhythm. This standing oscillation is called a seiche (pronounced “saysh”).
In the study, the fjord’s geometry trapped a much smaller oscillation, modelled at approximately 7 metres. The numerical water model and the measured seismic signal were close in period, rather than identical.
Synthesis / Water moves, Earth responds
As the mass of water moved back and forth, it exerted alternating forces on the fjord. That repeated forcing coupled into the ground, producing the unusual seismic signal.
The observed signal’s approximately 92-second rhythm and long persistence fit a trapped water oscillation. In 2025, researchers added independent support using SWOT satellite observations of the water’s surface slope.
Those satellite passes were snapshots, not a continuous film of nine days. Together, different kinds of evidence made a distant, hidden process legible.
A wider basin usually takes longer to slosh. Deeper water lets the disturbance travel faster. An outlet and friction let the motion fade. Explore those changes one at a time.
Teaching model / Not Dickson Fjord
Playback is accelerated: one model cycle takes four real seconds. No automatic playback. You can always scrub or step.
The default invented basin takes about 135 seconds per cycle. This is not a reconstruction of the observed 92-second signal.
One push can leave a long echo. Confinement limits escape; damping dissipates motion. A lightly damped basin can keep oscillating long after the initial disturbance. Try the leaky outlet, then compare how much amplitude remains.
A basin’s natural rhythm also explains resonance: repeated pushes close to that rhythm can reinforce the motion, like well-timed pushes on a swing. The Dickson Fjord explanation is a freely decaying seiche, not nine days of repeated pushes.
This is a shallow-water approximation for the lowest standing mode in a rectangular basin of uniform depth. Its period is T = 2 × width ÷ √(9.81 × depth), with metres and seconds. It neglects three-dimensional geometry, changing depth, nonlinear tsunami motion, tides and detailed coupling to the Earth. The full physical simulation in the study is substantially more complex.
For display, the standing surface follows a cosine across the basin and a cosine in time. Amplitude is normalised to 1 at the push. It is multiplied each cycle by (1 − internal damping) × (1 − escape loss). These are arbitrary teaching parameters for amplitude loss, not fitted measurements of energy loss. The outlet changes loss only; we do not solve an open-boundary basin.
The real seismic signal had a dominant period of approximately 92 seconds. The 2024 study’s numerical seiche model gave approximately 87 seconds. Neither the period nor the approximately 7-metre modelled seiche is a measurement of a 200-metre wave lasting nine days. This small demonstration does not reproduce the event’s duration or absolute water height.
Changing a condition returns this demonstration to the initial push. Animation pauses when hidden or scrolled away, and does not resume automatically. Reduced-motion visitors can use the time slider and quarter-cycle steps.
The extraordinary part is the chain: a slope loses support, rock and ice displace water, a fjord traps an oscillation, and the moving water leaves a signal in instruments far away.
It also shows why the type of evidence matters. A seismic record measures ground motion. A run-up mark records water’s reach. A numerical simulation reconstructs a mechanism. A satellite pass samples the surface. They can support one explanation without being interchangeable measurements.
As glaciers thin, slopes and coastlines can change in ways that create cascading hazards. The case makes a practical argument for looking across boundaries: mountain stability, tsunami risk and seismology belong in the same conversation.
All drawings are original schematics. No research image or recorded seismic trace has been reproduced.