In this week's space roundup, I caught a story about the NISAR spacecraft: a NASA and ISRO satellite compiled a time-lapse of the eruption of Kamchatka's Krasheninnikov volcano, which began after an 8.8-magnitude earthquake. NASA's wording was almost a verdict: the earthquake "apparently roused" one of two volcanoes, and for the first time in nearly five centuries it began erupting.
I didn't like that phrasing right away. In volcanology, the question "can a major earthquake trigger an eruption" is a forty-year dispute with directly opposing conclusions in peer-reviewed journals, where one of the most cited papers was later carefully dismantled using the same data. And then a second thing emerged: the eruption they wrote about as breaking news continues to this day — KVERT's July 12, 2026 report records effusion on the eastern slope of the Northern cone and a thermal anomaly. So what lay before me wasn't news, but a proving ground that's been operating for over a year to test how science distinguishes cause from coincidence.
The earthquake occurred on July 30, 2025 at 11:24 local time (July 29, 23:24:52 UTC). Magnitude 8.8, thrust fault on the Kuril-Kamchatka Trench, focal depth around 35 km, epicenter 119 km east-southeast of Petropavlovsk-Kamchatsky. The rupture grew for about 270 seconds. This is the strongest earthquake in the world since 2011, when Tohoku shook, and sixth on the list of most powerful in the entire history of instrumental observations.
It had a long prelude: 425 foreshocks from magnitude 4.0 and above, the strongest — 7.4 ten days before the main shock. After — more than four thousand aftershocks in the same category, and on September 18, 2025 there was a magnitude 7.8 aftershock. The southern part of the peninsula shifted southeast by almost two meters.
The tsunami turned out strange. Around the entire Pacific, waves were about a meter or less, in Japan 21 people were injured during evacuation and one died indirectly — from evacuation consequences, not from water. But nearby in a narrow river valley south of Vestnik Bay they recorded a local run-up of 33.1 meters. On Shumshu, six-meter waves came 650 meters inland and left a maximum run-up of 10.71 meters at Baykovo. In Severo-Kurilsk the water traveled about 600 meters. One and the same process gave a meter in the open ocean and thirty-three meters in a specific gully — shoreline geometry matters more than source energy.
Krasheninnikov is a complex of two merged stratovolcanoes inside an ancient caldera nine kilometers in diameter, in the very center of Kronotsky Nature Reserve, about 200 km from Petropavlovsk and 13 km south of Kronotskoye Lake. The southern cone rises to 1856 meters, the northern to 1760. Named after Stepan Petrovich Krasheninnikov, a naturalist who in the 1730s described Kamchatka and published the first scientific monograph in Russia's history in Russian — "Description of the Land of Kamchatka," 1755. He himself didn't live to see the eruption of "his" volcano by exactly 270 years.
The chronology of prior activity varies among different authors, and that's a separate story. By tephrochronology the last explosive eruption was about 1100 years ago, and lava flow extrusion and formation of a cinder-lava cone in the Northern cone's crater was 400–600 years ago. The Smithsonian's Global Volcanism Program dates the last eruption to 1550. In the press "600 years" and "last flow in 1463" flashed by. Same volcano, four different answers. Formally everyone's right: we're talking about different types of events and different dating methods, but the headline reader gets only one number, and it's almost always the catchiest.
The eruption began on August 2, 2025 around 16:50 UTC, meaning August 3 at 04:50 local time. Explosions from the northern crater lifted ash initially to 3–4 km, then to 8–10 km; simultaneously a fissure opened on the northwestern flank and lava came out. By August 12 three flows stretched about 3 km each with a total area of 2.7 km², by the end of September the area was estimated at 2.5 km² and volume no less than 0.025 km³. Composition — moderately potassic dacites, SiO₂ 64.9–65.7%. On sulfur: by 1 AM on August 3 about 14 kilotons of sulfur dioxide went into the atmosphere, then plumes from different volcanoes merged and there was nothing left to count.
In 1998 Linde and Sacks published a paper in Nature that still lives in popular retellings. They took global catalogs from 1500 onward and formulated a rule: an eruption with explosivity index 2 and above, occurring within five days after a magnitude 8 and above earthquake within a radius of 800 km, is triggered. By their count such pairs occurred about four times more often than pure chance would give. Of eleven eruptions seven fell in Chile, and three of those on one volcano — Villarrica with an almost permanently open vent.
Twenty years later Sawi and Manga rechecked this with more complete data. They expanded the window to magnitude 6, five days and 800 km and removed everything that happened before 1964 — before the launch of the Worldwide Standardized Seismograph Network catalogs are incomplete, and any statistics based on them are biased. They found 30 volcanoes that fell into the "triggered" category at least once. And then — demolition: by structure, tectonic setting and volcano type these 30 differ in no way from the same set with randomized dates. No "higher magnitude — farther effect" dependence, which must exist for a physical mechanism. Eruptions within several days of a major earthquake occur at the frequency given by chance.
One tail remained that Sawi and Manga didn't hide: two months to two years after major earthquakes the number of explosive eruptions does grow by 5–12%. Not by multiples. Percentages.
Then others joined in. Bebbington and Marzocchi in 2011 estimated that directly, within several days after a shock, about 0.4% of eruptions are triggered. For Indonesia as a whole they found no statistically significant connection at all, and of 35 volcanoes with sufficient history seven showed signs of triggering — and only after adding the volcano's own internal state to the model. From there came the conclusion that seems to me the main one in the whole story: triggering is a property not of the earthquake, but of the magmatic "plumbing" of a specific volcano. The earthquake can only turn the clock forward.
Jenkins, Rust and Biggs in 2024 went through modern catalogs with Monte Carlo simulations and got that the eruption rate within 750 km for a year after a magnitude 7 earthquake is about 1.25 times background, and the excess can hold for two-four years. Before the earthquake — 0.9 times background, meaning a bit quieter than average. Separate finding: deep earthquakes, from 70 km, affect eruptions more noticeably than shallow ones, which fits poorly with the intuition "shook closer — responded stronger."
And Eggert and Walter back in 2009 showed that the entire global statistics rests on several regions. In Japan, Indonesia and Melanesia the clustering of eruptions and earthquakes is significant, in other places — no. And their formulation: individual bright observations distort the perceived weight of the connection more strongly than the entire table.
The physics here is calculated honestly, and it's also what sets the trap. There are static stress changes — they work nearby, within tens of kilometers. There are dynamic ones, from passing seismic waves — they fly hundreds of kilometers and are determined by peak ground velocity. Typical estimates for "triggered" cases — less than 1 MPa, up to 3 MPa in the toughest ones. And the pressure needed to rupture a magmatic reservoir is usually estimated at 1–10 MPa. The ranges barely overlap, and from here comes the standard caveat: an earthquake triggers nothing if the system isn't on the brink.
The preprint by James Hickey's group from Exeter, posted on EarthArXiv on September 2, 2026, calculated this for Krasheninnikov with concrete numbers. Static change in Coulomb stress — 0.01–0.05 MPa, tensile stress on the plane of the future dike — 0.002–0.006 MPa, volumetric compressive strain about minus two ten-millionths. Peak dynamic stress — 1.0 ± 0.6 MPa, and more than 2 MPa is only obtained by taking upper boundaries of ground velocity and shear modulus estimates. Statics is two-three orders weaker than needed, dynamics hits the lower boundary of the threshold, and also works in a narrow intersection where probability is low.
But here's the problem: in 2021 Takeshi Nishimura in Scientific Reports got exactly the opposite picture. He took 35–55 years of data, divided volcanoes by type of impact — and discovered that the probability of a new eruption grows two-three times within 5–10 years after a major earthquake, if accumulated static tensile strain exceeded 0.5 microstrain. And where the volcano only got strong shaking without extension, there's no probability growth at all. This is exactly the range that Hickey in the preprint calls mechanically ambiguous: for Nishimura statics works and triggers gas bubble growth, for Hickey for a specific volcano it's two-three orders weaker than the threshold. Both calculated from catalogs. Both are right in their assumptions. No one's managed to combine these two results in one head yet, and I won't pretend I understand who's in charge here.
And there's a chronology that breaks the simple explanation. The dike — the crack through which magma goes up — intruded on August 1: that day the regional network recorded twelve earthquakes around magnitude 4 in the interval from 12:37 to 17:40 UTC, typical companion of dike intrusion. The eruption began August 2. Between the shaking and the dike — about 1.5–2.5 days. If the earthquake had directly ruptured the reservoir, the dike would have gone immediately, not after two days of deliberation.
What remains? The hypothesis the authors put forward: prolonged shaking (and the rupture lasted almost five minutes) triggered processes with delay — volatile diffusion in already gas-saturated magma, bubble growth, density decrease, pressure increase. Against this version speaks only that it's unprovable directly. For it — two things: the calculated magma compressibility points to gas-saturated composition, and the dike volume 3.2×10⁷ m³ with reservoir loss 1.3×10⁷ m³ gives a ratio around 2.46, which for ordinary incompressible magma is impossible. The reservoir sat at a depth around 6.3 km.
Here begins the most interesting part, because the preprint authors themselves break their own strongest picture.
The logic of "hidden critical state" sounds elegant: the volcano stood charged, only looked dormant, the earthquake became the trigger. The problem is there's no evidence of being charged at all. Satellite radar measurements from 2016 through July 2025 showed no inflation — instead nine years of slow subsidence around 4 mm per year, and horizontal displacements pointed not to a cooling reservoir but to surface flank creep, possibly from permafrost thaw. The TROPOMI instrument found not a trace of sulfur dioxide over the volcano before the earthquake. KVERT, which monitors Kamchatka, reported no precursors.
And in the preprint text stands a phrase you rarely meet in a scientific article: we have no evidence of critical state existence. Then the authors honestly throw up their hands — either the system wasn't in critical state, or such a state in principle is poorly caught by remote methods. The second, by the way, drags along an unpleasant consequence for all volcanology: if a "dormant-looking but charged" volcano is indistinguishable from just dormant, then satellite-based forecasting hits a ceiling.
I'll separately add skepticism from practitioners. Mike Krawczynski from Washington University just in those days published work on water content in Klyuchevskoy magmas and noted that magma under that volcano rises from mantle depths very quickly, so tying a shallow earthquake to an eruption directly is difficult, though he didn't exclude shaking's influence. Ruslan Zhostkov from the Institute of Earth Physics RAS in a comment for Komsomolskaya Pravda reminded that Klyuchevskoy has been erupting since April and is active almost annually, so the earthquake could have amplified an already ongoing process.
Alexey Ozerov, director of the Institute of Volcanology and Seismology FEB RAS, told journalists on August 4–5, 2025 that the earthquake activated magmatic chambers and "pumped" additional energy into them. The formula "parade of volcanic eruptions" came precisely from his comments: according to institute data, seven Kamchatka volcanoes were erupting simultaneously, something not seen in nearly 300 years.
And then he gave Rossiyskaya Gazeta a breakdown that slightly cools this drama. On average in Kamchatka two to five volcanoes erupt daily. At that moment six were erupting — "a small deviation." In 2014 eight erupted simultaneously. Kamchatka is ten percent of global volcanism: 41 volcanoes were erupting on the planet, six of them Kamchatkan, according to general geological monitoring data the academician cited in an interview. "Exceeding the quota," the academician joked.
Both statements are simultaneously true, and this is a perfect example of how a number in a headline works. "First time in 300 years" and "in 2014 there were more" describe the same event with different rulers. The first ruler — simultaneous eruption of seven specific volcanoes, a rare configuration. The second — total number of active centers, which fluctuates from two to five on an ordinary day. Both are truthful, and both say nothing about causality.
A separate detail, for which I read all this in the first place. NISAR, that very mission whose time-lapse caught my eye, lifted off from Sriharikota cosmodrome on July 30, 2025 at 17:40 India time, meaning just over twelve hours after the main shock. The spacecraft entered operational work only by December 2025, made the first image of Krasheninnikov on December 25, and every twelve days returns to the same point twice — south to north and back. 17 frames through mid-August 2026 were compiled into an animation where you can see lava filling the small inner caldera, overflowing the rim and spreading in a fan. Pixel — 10 meters, meaning about half a tennis court.
The irony is that the physics of this eruption were dissected using data from another spacecraft. The calculation of stresses and strains in Hickey's preprint is built on C-band images from European Sentinel-1, tracks 38 and 60, from 2016. NISAR with its L-band gave cinema shot from the moment when the event was already in full swing.
L-band, however, is important for another reason. Radar with wavelength around 24 cm sees through foliage and weak vegetation cover, where C-band loses coherence. The promised horizon — almost all 1300 active land volcanoes on the planet, but this particular story shows that coverage and understanding are different things. We've learned to look at a volcano every twelve days. Answering the question "why did it wake up right now" — not yet.
The practical conclusion from this whole mass of articles is fairly dry. To the question "after a megaquake should we raise the alert level at neighboring volcanoes" statistics answers: yes, but not much and with a caveat. Excess above background exists, but it wobbles around 1.25 per year, and instant triggering is about 0.4% of cases. Round-the-clock watch "just in case" across the whole peninsula after every M8 burns service resources on false alarms.
At the same time it makes sense to watch the state of the magmatic system itself, not the amplitude of yesterday's shaking. Sawi and Manga's criterion is simply volcano activity: "sensitive" volcanoes turned out sensitive because they erupt more often. Bebbington and Marzocchi's criterion is internal state, primarily the presence of gas-saturated magma in fragile equilibrium. Both formulations lead to one thing: the duty officer watches the volcano, not the seismogram.