A NASA satellite tracked a magnitude 8.8 earthquake tsunami across the Pacific Ocean with 1-centimetre precision and pinpointed exactly where the seafloor rupture was shallow enough to generate catastrophic coastal waves, solving a measurement problem that ground-based sensors have failed to crack for decades. Findings published in Science in March 2026 quantify how the Surface Water and Ocean Topography satellite, known as SWOT, captured a two-dimensional image of the tsunami wavefield generated by the 29 July 2025 Kamchatka earthquake. No existing seafloor pressure sensor or land-based seismic network could have extracted that same information from the same event. The satellite passed approximately 600 kilometres from the epicentre just 70 minutes after the rupture and recorded what no instrument had ever recorded before: a full spatial image of dispersive trailing waves tied directly to near-trench slip within 10 kilometres of the trench.
The earthquake struck at 23:24 UTC off the Kamchatka Peninsula in Russia’s Far East and sent waves across the entire Pacific basin within hours. Runups, the height a wave reaches when it climbs a shoreline, exceeded 17 metres in Severo-Kurilsk, a coastal town on the northern Kuril Islands. Five DART sensors, which are seafloor pressure recorders deployed across the North Pacific to detect and report tsunamis in real time, logged the leading wave front, with the closest unit measuring a crest-to-trough height of 1.32 metres. Large waves were recorded at transoceanic coastlines thousands of kilometres from the source, confirming that the tsunami crossed the full width of the Pacific with enough energy to be destructive at distant shores. The United States Geological Survey logged the event at moment magnitude 8.83, placing it among the strongest subduction earthquakes in the modern instrumental record.
To understand what SWOT captured and why it matters for warning systems, the physics of how a tsunami forms at a subduction trench needs to be clear. A subduction zone is a boundary where one tectonic plate dives beneath another, with the descending plate pulling downward into the Earth at a deep-sea trench. When the locked section of that boundary ruptures, it can displace the seafloor vertically by several metres in seconds, and that displacement launches a pulse of water upward across thousands of square kilometres. If the rupture reaches close to the trench itself, within roughly 10 kilometres, the seafloor displacement happens in shallower water where the geometry of the fault produces a different kind of wave. Shallow near-trench slip generates short-wavelength disturbances, around 50 kilometres from crest to crest, that sit at the boundary between deep-water and shallow-water wave physics at the 5.5-kilometre ocean depths typical of this region.
Those 50-kilometre waves behave differently from the main tsunami front because at that scale, relative to the ocean depth, different wavelength components travel at different speeds. Longer components outrun shorter ones, spreading the wave energy out into a fanned, curved train of secondary waves that trail behind the leading crest. This spreading process is called dispersion, and its presence in the SWOT image is a direct physical record of where the rupture was shallowest. A nondispersive model, one that treats all wavelength components as travelling at the same speed, was run against the same Kamchatka slip data and produced only the long leading crest, with no trailing wave structure at all. A second model that excluded all slip shallower than 10 kilometres on the fault interface also failed to reproduce the dispersive tail, despite fitting the DART pressure records and the Sentinel-1A land deformation data acceptably well. Only the model that allowed slip to reach within 10 kilometres of the trench reproduced the amplitude, spacing, curvature, and propagation direction of all seven trailing disturbances that SWOT recorded.
The DART network, which the world currently depends on for real-time tsunami detection across the Pacific, has two physical constraints that prevent it from reading this near-trench signal. At a typical DART sensor depth of 5.5 kilometres, the dynamic pressure from a 50-kilometre-wavelength wave is attenuated to approximately 80 percent of its surface value, following the standard depth-wavelength pressure response relationship. The sensors are also point measurements, single instruments spaced hundreds of kilometres apart across the ocean floor, so they record a time series at one fixed location rather than a spatial image of the wave structure. A checkerboard inversion test, a standard method used to assess which dataset can resolve which part of a source, confirmed that the SWOT imagery was the only dataset in this study capable of resolving shallow near-trench slip when placed against DART records and InSAR land-deformation data in direct comparison. The five DART sensors reproduced the tsunami’s timing and amplitude adequately, but none of the five captured the dispersive wave train that SWOT imaged 600 kilometres from the source.
SWOT operates on a completely different physical principle. It is a wide-swath radar altimeter, meaning it broadcasts radar pulses across a strip of ocean hundreds of kilometres wide and measures the time and phase of the return signal to calculate sea surface height at 1-centimetre precision. The team stripped out large-scale background ocean variability using ocean-state data from adjacent days, removing signals with wavelengths above 100 kilometres, and the residual noise floor in the processed image sat at approximately 3 centimetres. The dispersive trailing waves registered between 20 and 50 centimetres in amplitude, placing them 7 to 17 times above the noise floor and leaving no ambiguity in the measurement. A Sentinel-3A altimetry satellite also crossed the tsunami front approximately 30 minutes after the earthquake, around 570 kilometres from the source, and its nadir-only profile, a single-line measurement rather than a full two-dimensional swath, picked up both the leading crest and dispersive trailing packets consistent with the SWOT image. That independent Sentinel-3A crossing at roughly 200 kilometres from the source provided a second orbital confirmation that the dispersive wave train was real.
The joint inversion that combined SWOT, DART, and Sentinel-1A InSAR data placed the preferred slip model’s peak displacement near 50.5 degrees North, 158 degrees East, with the rupture concentrated between 49.5 and 52.5 degrees North along the Kamchatka subduction zone. The inversion calculated an earthquake magnitude of Mw 8.81, matching the independently estimated USGS value of Mw 8.83 without that number being used as a constraint during the calculation itself. Maximum slip in the preferred model reached 12 metres on the fault surface, with substantial slip extending toward the trench in the uppermost 10 kilometres of the plate interface. Sensitivity testing across three simplified slip distributions showed that dispersive wave amplitude at the SWOT footprint increased in direct proportion to how far slip reached toward the trench: zero trench-breaching slip produced negligible dispersive energy, while 100 percent trench-breaching slip at 12 metres maximum produced the strongest and most clearly separated trailing wave train.
This was not SWOT’s first encounter with a tsunami. The satellite previously captured the 19 May 2023 tsunami near the Loyalty Islands, and trailing dispersive waves were again logged following the 2 May 2025 magnitude 7.4 Drake Passage event. Those three captures across different ocean basins and different source mechanisms show that dispersive tsunami signatures are not rare geological oddities. Their scarcity in historical records before SWOT reflects the limits of nadir-only altimeters and sparse point sensors, not a genuine absence of the phenomenon in nature. SWOT revisits any given location on an approximately 11-day average cycle, with an exact 21-day ground-track repeat, which means it will not always be positioned to image a tsunami. Because tsunamis travel across ocean basins for many hours after their source, the probability of at least one SWOT swath intersecting the wave train somewhere along its path is not negligible.
The existing global tsunami warning infrastructure, built around DART pressure sensors and coastal tide gauges, currently cannot resolve near-trench rupture depth from offshore wave data alone. SWOT’s 1-centimetre height precision and wide-swath two-dimensional imaging geometry fills that gap directly, as the Kamchatka event demonstrated across a 600-kilometre separation between satellite and source. The satellite is presently operational, with processed sea surface height data publicly available through NASA Earthdata and the Physical Oceanography Distributed Active Archive Center. Monitoring teams tracking active subduction zones in the Pacific, including the Cascadia zone off the US Pacific Northwest, the Nankai Trough off Japan, and the Chilean margin, now have a satellite-based tool capable of constraining where a future rupture reaches the trench, with a precision that no seafloor or land-based sensor currently matches.
Source:
Sepúlveda, I., Nilsson, B., Yu, Y., Carvajal, M., Brandin, M., Gabriel, A-A., Sandwell, D. (2026). SWOT detects dispersive tsunami tied to a near-trench source in the 2025 Kamchatka earthquake. Science, 391(6792). https://www.science.org/doi/10.1126/science.aeb8634
