The Repository Hook Echo

Kilauea 2018: The Volcano That Fell In On Itself

October 7, 2026 · Runtime 15:27 · Watch on YouTube

In 2018 the summit of Kilauea, Hawaii, collapsed 62 times in three months while lava poured out of Leilani Estates, 40 km away. USGS footage shows both ends. We downloaded the raw records -- the USGS earthquake catalogue, a seismometer on the caldera rim and a GPS station that sank almost two metres -- and measured how the volcano emptied itself.

Chapters

  1. 0:00 Leilani Estates, May 2018
  2. 0:19 The summit, 40 km uphill
  3. 0:55 Before 30 April
  4. 1:47 The first fissure
  5. 2:04 The magnitude 6.9
  6. 3:12 The lava lake drains
  7. 3:30 Twelve explosions
  8. 4:41 62 events in the catalogue
  9. 5:06 Fissure 8
  10. 6:32 Kapoho Bay
  11. 7:25 The summit starts to fall
  12. 8:24 The seismometer on the rim
  13. 9:30 The collapse of 24 July
  14. 10:10 The GPS that sank
  15. 11:36 The summit as a piston
  16. 11:56 Where the summit went
  17. 12:19 62 collapses vs one earthquake
  18. 13:06 The end, and after
  19. 14:14 What the record supports
  20. 15:01 Next: Hurricane Michael

Transcript

Leilani Estates, May 2018 0:00

This is Makamae Street, in Leilani Estates, on the Big Island of Hawaii, at 9:32 in the morning on May 6th, 2018. The first crack had opened three days earlier, inside the same subdivision. Two days later, another crack opened across Leilani Street, between the houses.

The summit, 40 km uphill 0:19

Forty kilometres uphill, at the summit of Kilauea, something else was happening. This is the crater on the morning of July 24th, filmed by the observatory's own camera. In the next three months, the top of this volcano fell in on itself sixty-two times. And the instruments around the summit recorded every one of them.

So here is the question this episode answers. Why would the summit of a volcano collapse, again and again, while the lava was coming out forty kilometres away? And how could the instruments tell, almost to the hour, when the next collapse was coming?

Before 30 April 0:55

Before April 30th, Kilauea had two open vents. A lava lake inside Halema'uma'u crater, at the summit, active since 2008. And Pu'u 'O'o, a cone on the East Rift Zone about twenty kilometres away, erupting since 1983. In mid-March, the tiltmeters at Pu'u 'O'o began to record the ground swelling. On April 21st, the summit lake rose so high that it spilled over onto the crater floor.

Then, at 2:15 in the afternoon on April 30th, the floor of Pu'u 'O'o crater collapsed. The earthquakes started moving east, down the rift, away from the summit. That is what a dike looks like to a seismic network: magma forcing a crack open underground, kilometre by kilometre. On May 1st, the observatory warned the communities downrift that an eruption was possible.

The first fissure 1:47

The ground deformation and the satellite radar showed the intrusion heading for Leilani Estates, about twenty kilometres past Pu'u 'O'o. Just before five in the afternoon on May 3rd, the first fissure opened inside the subdivision. There would be twenty-four of them.

The magnitude 6.9 2:04

The next day, at 12:32 in the afternoon, the south flank of the volcano slipped. Magnitude 6.9, about six kilometres down. It was the largest earthquake on the island in forty-three years. The fault moved up to five metres, and GPS stations on the surface jumped seaward by as much as seventy centimetres.

The two events were probably connected. The earthquake sits on the flat fault between the volcano and the old sea floor beneath it. Stress models show that when a rift zone is pried open, it pushes on that fault. And when the flank slips, the rift can open further. Kilauea had gone almost sixty years without an intrusion in this stretch of the rift, so the strain was already there.

The first lava was slow and sticky. Its chemistry matched lava that erupted in the same area in 1955, which means it had been sitting in the rift for decades. Fissures opened, spattered for minutes or hours, and went quiet. On May 12th a new one opened a mile and a half downrift, exactly where the earthquakes had been clustering for two days.

The lava lake drains 3:12

Meanwhile, at the summit, the lava lake was draining. By May 10th, it had dropped more than three hundred metres and could no longer be seen from the crater rim. The observatory had a precedent for that. In 1924, after the lake drained, the summit exploded.

Twelve explosions 3:30

So on May 9th, the observatory warned of possible explosive eruptions. Two days later, most of Hawaii Volcanoes National Park closed. On May 16th, the first explosion came. Over ten days there were twelve, eight to forty-five hours apart. The largest, on May 17th, sent ash about eight kilometres into the sky.

To the seismometers, these explosions did not look like ordinary earthquakes. Each one came with a slow, very-long-period signal of magnitude 4.7 to 5.1, and a step in the summit tilt. When the scientists worked out the source, it was not a fault slipping. It was a change in volume, deep under the crater.

In 1924, the explanation was groundwater: water seeping into a hot, empty conduit and flashing to steam. The 2018 instruments disagreed. During the explosions, the summit's sulphur dioxide went up two to three times. That gas comes from magma, not from rain water. The paper that reports it says the old hypothesis is now in question, for 1924 as well.

62 events in the catalogue 4:41

Each of those explosions also shows up in the earthquake catalogue of the U.S. Geological Survey. We downloaded it. Within eight kilometres of the summit, between April and August, it lists sixty-two events classified as volcanic eruptions, each with a moment magnitude. The first twelve, from May 16th to May 26th, are the explosions. The other fifty are something else.

Fissure 8 5:06

Down in Leilani, on May 18th, the lava changed. It came out hotter and more fluid, and five days later it reached the ocean. That was fresh magma, from the summit, arriving at the far end of the rift. Late on May 27th, a vent called fissure 8 came back to life, and within twelve hours the whole eruption was focused there.

Fissure 8 threw fountains up to eighty metres high. It fed a channel of lava that ran all the way to the sea, at fifty to two hundred cubic metres every second. The sulphur dioxide climbed past fifty thousand tonnes a day, and the haze reached Guam, more than six thousand kilometres away.

Two hundred cubic metres a second is hard to picture, so here is one way. An Olympic swimming pool holds about two and a half thousand cubic metres. At the high end, fissure 8 was filling one with molten rock every twelve and a half seconds. Pu'u 'O'o had erupted about four point four cubic kilometres, but over thirty-five years. No Kilauea eruption ever observed had kept a rate this high for this long.

This is the channel from the air on June 29th. The observatory flew drones along it again and again, measuring its speed, its gas and the shape of the ground around it. Every new overflow went into a computer model that drew where the lava would go next, along the steepest path downhill.

Kapoho Bay 6:32

On June 3rd, the lava reached Kapoho Bay. It filled the bay and kept going, building new coastline. By the end, the eruption had added eight hundred and seventy-five acres of new land to the island.

This is the same flow at six in the morning on June 14th, seen from a helicopter. Near the vent, standing waves rise in the channel. Closer to the sea, a crust forms on the surface as the lava spreads out. At the coast, the ocean entry is active along about a mile, with small explosions where lava meets water, and plumes of laze: acid steam.

Lava covered thirty-five and a half square kilometres. More than seven hundred structures were destroyed, and about thirty miles of road. There was no way to stop it. What the forecasts could do was say where it was heading, and how soon.

The summit starts to fall 7:25

Back at the summit, near the end of May, the ground itself started to move. The floor of the caldera around Halema'uma'u began to sink, and the crater walls slumped inward. The vent filled with rubble, and the ash plume faded. And the earthquakes settled into a pattern.

This is the summit from the air on June 5th. Where the lava lake used to be, there is now a deep hole piled with rubble from its own walls. The western part of Halema'uma'u has dropped and slid toward the centre of the crater. New cracks are opening on the caldera floor to the west. The magma underneath is still leaving, toward the rift.

Those are the other fifty events in the catalogue. From May 29th to August 2nd, a collapse of magnitude 5.2 to 5.4 roughly once a day. We measured the gaps between them. The middle value is thirty-two hours. The shortest was twenty, the longest fifty-three and a half.

The seismometer on the rim 8:24

Between collapses, the earthquakes did not stop. They built up. We took the raw record from a seismometer called RIMD, on the rim of the caldera, and counted the shocks above a fixed level, hour by hour. In the hours after the collapse of July 22nd, it counted five to eleven an hour.

By the next evening, it was one hundred and eighty an hour. The swarm kept going all night. The paper describes more than seven hundred small earthquakes a day at this stage. At 6:41 the following morning, the summit collapsed again. In the next hour, the seismometer counted ten.

We also counted the whole summer from the catalogue, day by day, within eight kilometres of the summit. In late April, between one and twenty-five earthquakes a day. On the day of the big flank earthquake, a hundred and forty-seven. By late June, more than seven hundred, with a peak of seven hundred and eighty-five on July 7th. From April 23rd to August 18th, thirty-nine thousand five hundred and four.

The collapse of 24 July 9:30

This is that collapse. July 24th, 6:41 in the morning. At 6:41 and 8 seconds, a small tree on the right begins to sway. Two seconds later, a pressure wave passes through the steam in the crater. One second after that, rock begins to fall from a cliff on the far side. The catalogue gives it magnitude 5.3.

The observatory sat on the rim of that caldera. This is the observatory during the collapse of July 5th. The scientists had already moved out and were running the response from somewhere else. The shaking went on damaging the empty building.

The GPS that sank 10:10

There was also a GPS receiver on the rim, called CRIM. We downloaded its daily positions. Between April and the end of September, it sank one point nine three metres. And it moved one point three metres north and almost a metre west. Toward the crater.

The GPS also shows when the sinking started. On May 1st, the rim was where it had been all April. By May 16th, the day of the first explosion, it was already forty-five centimetres lower. By June 1st, almost a metre. By August 2nd, the day of the last collapse, one point seven eight. After that, it barely moved.

Each large collapse dropped the caldera floor by several metres. Fifty of them, one after another. In places, the floor ended up more than five hundred metres deeper than it was in April. This is the crater rim from a drone on June 24th, halfway through.

A caldera collapsing is something scientists had almost never watched closely. At Bardarbunga, in Iceland, in 2014 and 2015, it happened under a glacier. At Fernandina in 1968, Miyakejima in 2000 and Piton de la Fournaise in 2007, it was over in hours or days. At Kilauea it took three months, in the middle of one of the densest monitoring networks on Earth.

The summit as a piston 11:36

And here is the part that links the two ends of the volcano. Starting in July, instruments down the rift picked up a pulse of pressure after each summit collapse. Some of those pulses were followed by more lava pouring out at fissure 8. The summit was not just sinking. Every time it fell, it pushed.

Where the summit went 11:56

So where did the summit go? Measured from the change in the ground's shape, the collapse removed about 0.825 cubic kilometres. The lava erupted downrift comes to about 0.8 cubic kilometres. That is roughly three hundred and thirty thousand Olympic pools. The hole at the top and the lava at the bottom are, within the error, the same size.

62 collapses vs one earthquake 12:19

One more number from the catalogue. A magnitude 5.3 collapse releases a lot of energy. But magnitude is a logarithmic scale, so we added up the energy, not the numbers. All sixty-two events together come to a single magnitude 6.45. The earthquake of May 4th, on its own, released about four point seven times that.

Each step of one on the magnitude scale means about thirty-two times more energy. So one collapse, at magnitude 5.3, is about five and a half trillion joules, roughly thirteen hundred tonnes of TNT. The summit released that about every thirty-two hours, for two months. And it still never caught up with a single slip of the volcano's flank.

The end, and after 13:06

The last collapse came on August 2nd, at 11:55 in the morning. Two days later, the flow from fissure 8 stopped almost at once. The summit had stopped sinking at about the same time. On September 5th, the observatory saw active lava inside fissure 8 for the last time.

So what comes after a collapse like that? The observatory looked at 1924 again. After the summit collapsed that year, there were only a few small eruptions inside the crater for ten years, and then none anywhere on the volcano for eighteen more. The paper's forecast was several quiet years, with one warning: it could also come back sooner than expected.

On July 25th, 2019, the observatory spotted a small pond of water at the bottom of the new pit. It was the first lasting body of surface water at Kilauea's summit in more than two hundred years. For seventeen months it rose, until it was about fifty metres deep. On December 20th, 2020, a fissure opened in the crater wall, and the lava boiled the lake away in about an hour and a half.

What the record supports 14:14

Here is what the record supports. The summit and the rift were one connected system, and the instruments watched the connection work. The seismometer counted down to each collapse. The GPS on the rim measured the fall. And the volume that left the top is the volume that came out at the bottom.

And the forecasts held up. On May 1st, a warning that an eruption downrift was possible. Two days later, the first fissure. On May 9th, a warning of explosions at the summit. A week later, the first one. On May 10th, a warning that more outbreaks were likely. On May 12th, a new fissure, where the earthquakes had gathered. And the first lava path drawn for fissure 8 was close to the route the channel took.

Next: Hurricane Michael 15:01

Next, the wind. In October 2018, Hurricane Michael came ashore at Mexico Beach, Florida, and was announced as a Category 4. Months later, after the aircraft and radar data were analysed, it was upgraded to Category 5. That is the next episode. Michael, 2018: the hurricane the data upgraded to Category 5.

Description and sources

The eruption began on 3 May 2018 with the first of 24 fissures in Leilani Estates; a magnitude 6.9 earthquake followed on 4 May. Fissure 8 fed a lava channel to the sea at 50 to 200 cubic metres a second, covering 35.5 square kilometres and destroying more than 700 structures. At the summit, the USGS catalogue lists 62 eruption events between 16 May and 2 August: 12 explosions, then 50 collapses of magnitude 5.2 to 5.4, a median of 32 hours apart. On the rim seismometer RIMD, shocks rose from about 5 an hour after a collapse to 180 an hour before the next one. GPS station CRIM sank 1.93 m. The collapse volume, 0.825 cubic kilometres, matches the lava erupted downrift.

PRIMARY SOURCES

Hook Echo reads a disaster through the instruments that recorded it: radar, seismometer, survey and report, with the primary sources above.

About the material. Video: U.S. Geological Survey, Hawaiian Volcano Observatory, from the USGS multimedia gallery, public domain. The earthquake counts, the 62-event series, the hour-by-hour count on seismometer HV.RIMD, the energy sum and the GPS heights are our own measurements from public data: the USGS ComCat catalogue, EarthScope's archive of the Hawaiian Volcano Observatory network, and the Nevada Geodetic Laboratory's daily solutions for station CRIM. Every diagram is drawn in code from the sources listed. Nothing in this video is AI-generated imagery.

#HookEcho #Kilauea #Volcano