A Sideways Earthquake, an Unexplained Wave, and a New Suspect Under the Seafloor
In July 2020, a magnitude 7.8 earthquake broke the plate boundary beneath the Shumagin Islands, off the Alaska Peninsula, and shoved the seafloor upward. That is how a tsunami is usually made. The floor of the ocean rises, the water above it rises with it, and the bulge spreads outward. Three months later, on Oct. 19, 2020, a magnitude 7.6 shock struck nearby. The waves of shaking it sent through the planet pointed to a fault that had mostly slid sideways instead, and sideways motion of the seafloor displaces very little water. A large tsunami followed it anyway.
Six years on, nobody has settled what made that wave. The latest attempt was published Oct. 6, 2026, in the journal Seismica. Ronni Grapenthin, of the Geophysical Institute at the University of Alaska Fairbanks, and colleagues at Michigan State University, Cornell University and Kunming University of Science and Technology assembled what they describe as the most complete land-based record of the earthquake available. Their own fault models, they found, cannot account for the tsunami. What can, on their reading, is a slip so slow and so quiet that nothing on land registered it.
That record puts several kinds of measurement over each other: recordings of the earthquake's waves at seismometers around the world; satellite-positioning stations in Alaska logging the ground's movement many times a second as the shaking passed; the permanent shift those same stations showed once it was over; radar images taken from orbit before and after the quake, which map how far the land moved between passes; and a recalculated catalog of aftershock locations. Against all of it the team tested two fault models: one in which a single fault slid sideways, and one that added a thrust segment. Both fit what the instruments on land recorded. Neither, the authors report, makes a tsunami resembling the one that arrived.
The source they went looking for leaves no trace
That negative result is the firmest thing in the paper. To find something that would make the wave, Grapenthin and colleagues ran a systematic test of thrust-slip models: candidate patches of the plate boundary fault, each slipping in a way that would lift the sea. The version that fits best sits on the shallow part of that boundary. It slips over a very long period rather than in a jolt, and it starts minutes after the main shock, at a place where the main shock's own strain had pushed the fault closer to failure. The authors read that as a triggered tsunami earthquake, a rupture that makes far more wave than shaking, set off by the earthquake beside it.
None of that was observed. It is the best-fitting member of a family of candidate sources, which is a weaker kind of claim than a measurement. The silence is part of the proposal, not something anyone recorded. A slip that produced no measurable movement at the surface is by definition one the instruments did not catch, which is why it has to be inferred at all.
The slow slip was not the new part
The idea that something slow and thrust-like made the Sand Point tsunami did not start here. In February 2025, in the same journal, Sean Santellanes and colleagues concluded that strike-slip models matching the earthquake's measured orientation cannot produce the observed tsunami. Relatively slow slip on the plate boundary is allowable alongside the sideways faulting, and such a rupture would radiate too little seismic energy to be noticed by ordinary monitoring. In 2023, Yefei Bai and colleagues had reported something similar in Nature Communications: fast rupture on faults inside the plate, plus an induced slow thrust slip on another fault near the shelf break. That slip took minutes rather than seconds and left no clear trace in either seismic or ground-motion records while producing the distant wave.
What the new paper adds is narrower than it may sound. It moves the slow slip onto the plate boundary itself rather than into the plate above it. It makes the slip late rather than simultaneous with the shaking. And it ties both the timing and the location to the stress the main shock itself added. That is the new claim.
Or nothing slipped at all
A third group has worked from different data and reached a different conclusion. In June 2025, in Progress in Earth and Planetary Science, Akino Naitoh, Koichiro Motoi and Toshitaka Baba worked backward from the tsunami itself, inverting records from deep-ocean sensors and tide gauges to recover the shape of the sea surface that must have started it. Assuming the source acted at the same moment as the earthquake gave them no realistic sea surface at all. Assuming it acted about five minutes late gave a clean one: the surface dropping on the landward side and rising toward the trench. That is independent support for the delay, from another group using other instruments. Their explanation for what caused it is not a slip. It is a submarine landslide, and a very large one, roughly 120 cubic kilometers of rock and sediment.
A landslide of that size is a demanding thing to require, which is why this is a genuine disagreement rather than a demolition of either account. It is also an argument that has not yet been had in print: the reference list of the new Seismica paper runs to 83 entries and does not include the landslide study.
On what any of this would mean for hazard, the paper is careful almost to the point of silence, and deliberately so. It says that splitting an earthquake into sideways slip and triggered thrust slip offers a way to understand tsunami earthquakes better, and that if such pairings are common at plate boundaries, they might affect how earthquake and tsunami hazards are assessed. It makes no recommendation to anyone who issues warnings. What the three accounts share is harder to argue with than any of them on its own: whatever lifted the water off Sand Point was not the shaking the instruments recorded, and it did not happen at the same time.