We know very little about the processes that lead to a reeruption of supervolcanoes such as the mostly underwater Kikai caldera in Japan (pictured) and are therefore ill-equipped to make predictions. (Credit: SEAMA Nobukazu)
In A Nutshell
- Japan’s Kikai Caldera, which produced probably the largest eruption of the past 10,000 years, appears to be slowly refilling with magma from below.
- Seismic imaging revealed a large magma reservoir sitting 2.5 to 6 kilometers beneath the seafloor, in roughly the same location as the reservoir that powered the original eruption 7,300 years ago.
- A similar recharging pattern has been observed at Yellowstone, Toba, and Santorini, leading researchers to suggest this may be a common behavior among giant caldera volcanoes.
- The refilling model is inferred from seismic data rather than directly observed, and no eruption timeline or risk level has been established.
Most Americans know Yellowstone as the supervolcano lurking beneath a national park, a magma-filled giant that scientists monitor obsessively for any sign of unrest. Yellowstone, it turns out, may not be the only one worth watching. Researchers in Japan have found evidence suggesting that Kikai Caldera, a submarine volcano south of Kyushu Island, is slowly refilling with fresh magma after producing what was probably the largest eruption of the past 11,700 years. More striking still, the two volcanoes may be following a similar pattern.
About 7,300 years ago, Kikai unleashed roughly 160 cubic kilometers of material in a single catastrophic blast, collapsing a crater wide enough to swallow a mid-sized city into the seafloor. Now, research published in Communications Earth & Environment proposes that the same underground reservoir thought to have powered that ancient cataclysm is still there, in roughly the same location, and likely receiving new magma from below.
Japan’s Sleeping Supervolcano and the Yellowstone Connection
Kikai Caldera does not carry the cultural weight of Yellowstone in the Western imagination, but the scale is hard to fathom. Its Kikai-Akahoya eruption is probably the largest eruption anywhere on Earth during the Holocene, the epoch spanning roughly the entirety of human civilization. For perspective: the 1980 eruption of Mount St. Helens ejected less than one cubic kilometer of material. Pinatubo in 1991, whose ash cloud measurably cooled global temperatures, ejected around 10. Kikai ejected 160.
Since that ancient blast, the volcano has been anything but idle. Over the past 3,900 years, fresh magma has been pushing up through the caldera floor and accumulating into a lava dome now recognized as the largest of its kind anywhere in the world, exceeding 32 cubic kilometers in volume. Mapping the reservoir feeding that dome was a central goal of the new study.
Yellowstone offers a useful point of comparison. A magma reservoir sits at depths of 3 to 8 kilometers beneath that caldera, roughly the same depth range where pre-eruption magma is thought to have been stored before Yellowstone’s last giant eruption, more than 600,000 years ago. Kikai, the new data suggest, may fit a similar mold.
How Scientists Mapped the Supervolcano’s Interior
Imaging the inside of a submerged volcano is not straightforward. A research team from Kobe University and the Japan Agency for Marine-Earth Science and Technology strung 39 underwater sensors along a 175-kilometer line across the caldera and used a ship-mounted array of airguns to fire sound pulses into the seafloor. Molten rock slows seismic waves relative to solid rock, so where waves slowed down, the team knew something hot and partially liquid was lurking. After processing more than 12,000 individual wave recordings, they had a detailed picture of what lies beneath.
A Reservoir Slowly Refilling Beneath Kikai Caldera
What the data revealed was a large zone of sluggish seismic waves starting at about 2.5 kilometers below the seafloor and reaching down to 6 kilometers. In cross-section, the reservoir is roughly trapezoidal, narrow at the top and wider at the base, with a total volume estimated from a 2D model at around 220 cubic kilometers. Between 3 and 6 percent of that volume appears to be liquid magma, with 10 percent as a reasonable upper ceiling.
That depth range aligns closely with rock chemistry studies of both the original Kikai-Akahoya eruption and the more recent lava dome activity, pointing to the same reservoir doing the same job across millennia. At an average estimated refilling rate of more than 8.2 cubic kilometers per thousand years, a meaningful amount of material has been accumulating. As the paper states, “these melt re-injection processes in a magma reservoir at a shallow depth just beneath the caldera could be a step towards the next giant caldera eruption.”
Kikai carries less liquid magma than Yellowstone at present, but the architecture is familiar: a large, shallow reservoir directly beneath the caldera, at depths matching those of previous eruptions. Toba in Indonesia, widely regarded as the site of the largest terrestrial eruption of the Pleistocene epoch, and Santorini in Greece show a similar configuration. A pattern may be emerging.
Watching for Warning Signs at the World’s Giant Caldera Volcanoes
If refilling a shallow reservoir after a catastrophic eruption may be what giant calderas commonly do, it reframes how scientists should be monitoring them. Tracking changes in seismic wave speeds beneath these systems over time could serve as an early gauge of whether a reservoir is changing in size or melt content, well before any surface activity makes the news. Right now, that kind of long-term seismic monitoring does not exist at Kikai on the scale it does at Yellowstone, making the new baseline measurements from this study especially valuable for future comparisons.
Nobody can say yet how much magma needs to accumulate before Kikai, or any of its counterparts, reaches a tipping point. But for a volcano that probably produced one of the largest eruptions of the past 10,000 years, sitting in one of the most seismically restless corners of the planet, the fact that it appears to be quietly refilling is reason enough to keep watching.
Paper Notes
Limitations
Melt fraction estimates rely on several assumed values, including the background rate of heat increase with depth, the temperature at which rock begins to melt, and a factor describing how seismic wave speed responds to temperature. The first two had little effect on results, but the third introduced meaningful uncertainty, meaning the exact size and total melt volume of the reservoir cannot be pinned down precisely. Data were also collected along a single two-dimensional line, so the reservoir’s full three-dimensional shape is inferred rather than directly imaged. Modeling tests suggest the actual reservoir may be larger and contain a higher proportion of melt than the reported estimates. The melt re-injection model is also inferred from seismic imaging and prior geological studies rather than directly observed, and whether this behavior is common across giant caldera systems remains a hypothesis rather than an established rule.
Funding and Disclosures
This study was supported by Japan’s Ministry of Education, Culture, Sports, Science and Technology under its Third Earthquake and Volcano Hazards Observation and Research Program, as well as a JSPS Grant-in-Aid (grant number 20H00199). Fieldwork was conducted aboard the research vessels Kaimei, Kaiyo Maru No. 2, and Fukae-maru. Authors declare no competing interests.
Publication Details
Authors: Akihiro Nagaya, Nobukazu Seama, Gou Fujie, Satoru Tanaka, Hiroko Sugioka, and Shuichi Kodaira, affiliated with the Department of Planetology and Kobe Ocean-Bottom Exploration Center at Kobe University, Kobe, Japan, and the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokohama, Japan. Paper title: “Melt re-injection into large magma reservoir after giant caldera eruption at Kikai Caldera Volcano.” Published in Communications Earth & Environment (a Nature Portfolio journal), vol. 7, article 237, 2026. DOI: https://doi.org/10.1038/s43247-026-03347-9. Received June 5, 2025; accepted February 17, 2026; published online March 27, 2026.







