It’s amazing the scale of what has happened at Kilauea volcano over a little more than a year and a half. Since the current eruption started in December 2024, the caldera has been filled with 0.31 km3 of new lava.  A very respectable amount. In a series of 54 fountain episodes, the eruption has raised the caldera floor by about 100 m and built a tephra cone 50 m tall on top of the western caldera rim. The eruption style has been one of spectacular beauty. Roaring fire fountains have towered to heights of 540 m, showering the summit area in fist-sized clasts. Among the tephra fell spongy pieces of reticulite, glimmering gold and lighter than pumice itself. The strongest episodes fed plumes all the way into the jet stream, which led to ash falling to the east and north of Kilauea, opposite to the trade winds, as far as Hilo, Pahoa, or Kalapana. Fountains were dual during many of the episodes (from the south and north vents) and shot vertically or sideways, at times unleashing into a giant glowing V, often associated with black clouds of tephra or dancing dust devils. Episode 41 took meteorological action beyond that of common dust devils when the spinning vortex became fixated under the plume, which itself showed rotation in the radar, and became a large, black, swirling mass almost reminiscent of a mesocyclonic tornado.

But everything has its time, and everything ends. The fountains, as has happened during other Kilauea eruptions, continue to grow in height due to the conduit widening and thermally eroding the walls, causing friction to decrease (allowing magma to rise faster). Eventually, due to the increasing conduit width, the fountain ends, and the conduit enters what seems to be a convective mode of an incipient lava lake or pond; in this state, the convection releases gas in the conduit, which blocks the ability of it to flash into bubbles, the ability to shoot into a towering jet of gas and lava. In other words, the convective mode kills the fuel needed for high fountain activity. The south vent was the first to reach this tipping point and shut its fountains with episode 43, in March this year. This vent has, however, continued to host a permanent lava pond, and has even developed some satellite vents. North was always weaker in its fountains, possibly due to being a branch of the south vent, and continued to produce fountains long after south, but eventually conduit widening may have catched up to south, and it looks like episode 54 a bit over a month ago may have been its last. Overflows, spattering, and captivating hydrogen flames have continued to happen after the fountains at both of the vents, but they seem to be struggling.

In the aftermath of the fountains, a volatile situation has emerged. Kilauea started inflating rapidly right after episode 54 as magma rose into the shallow Halema’uma’u chamber under the caldera to resupply the magma erupted in the episode. The climb was bumpy due to a series of DI events (deflation-inflation events) that triggered ups and downs in the pressure of Kilauea’s system. By about September 7-8, pressure had reached the threshold that, at other times, initiated fountain episodes. Nothing happened. The vents were not capable of flashing into fountains, and inflation largely halted. What seems to have happened is that the fountains struggled to start due to the efficient degassing that the vents were now producing, with both vents likely featuring stable convection in their conduits by now, magma rising to the surface, releasing their gas and sinking back down, creating a dense, heavy column of melt that can neither expand into a gassy foam nor even overflow steadily given the large weight the degassed magma column now exerts on the magma system. As it has now become clear, this change has redirected the magma flow in other directions, two of them actually.

Around September 10, shortly after inflation hit a plateau, earthquake swarming started in the northern corner of the caldera; it seems that as early as this day magma started to leak slowly into an intrusion along the caldera fault zone. On September 14 and 15, without any immediate precursors, a fissure opened along the base of the western wall of the 2018 caldera, signalling that the magma intrusion of a few days earlier had reached the surface. Deformation maps published by HVO show that during this time a line matching an incipient caldera fault formed towards the end of the 2018 eruption (that never fully collapsed) had been shifting slightly. After a lull, activity picked up dramatically at 2 PM on September 30, with a swarm under the active vents, showing a magma intrusion had started at the SW corner of the caldera. In a matter of hours, the tiltmeter at Uwekahuna Bluff (UWD) was up by 40 microradians. Both the earthquake locations and tilt direction changes showed that magma was propagating from southwest to northeast, likely parallel to the caldera wall and along the uncollapsed caldera fault that was developing near the end of the 2018 collapse.

The caldera edge intrusion is not the only thing going on. While the UWD tiltmeter at Uwekahuna Bluff stopped rising, the cross-caldera distance instead continued rising almost as fast as before. This indicates the summit is widening horizontally but without rising much. Related to this development is likely the emergence of seismic activity in the uppermost rift zones, the so-called connectors. Earthquakes in this area indicate strain due to extension of Kilauea’s deep system, strain that can lead to dike intrusions. The caldera itself lies over this deep dike-like body (the deep rift) and is also subject to its effects, though it doesn’t show as much earthquake activity due to its aseismic nature caused by the magma chamber overlying the deep rift. The cross-caldera distance increase shows magmatic-driven spreading is currently strong at the summit, with the uppermost bits of the rift zones also feeling the effects (possibly to a lesser degree, just being more noisy). This means that a summit dike intrusion can start at any given time now, inside Halema’uma’u or in the South Caldera area.

Now, why do I think the current situation is volatile? Added to the enormous supply that Kilauea is getting lately, there are some conditions being met that carry certain parallels to two major and extraordinary eruptions of Kilauea, those of 1790 and 1823.

The first is that magma is seeping into the caldera ring fault on the northern side of the caldera. This I find worth consideration given that ring dikes are more or less what we seem to be seeing at Kilauea now. Ring dike intrusions fill a caldera fault while a caldera collapses, and they have a unique geometry that makes them quite troublesome: they are outward dipping, like a truncated cone, or truncated cone segment, with the wider base on top of the chamber. This specific geometry means the intrusion opens in width as the caldera collapses, and it also helps the caldera collapse by slicing off the support the caldera block has from surrounding rock. I have strong suspicions that I will explore below that explosive eruptions of Kilauea were related to ring dike intrusions, sometimes with water/groundwater interaction.

I still think at present an explosive outcome is VERY unlikely given that so far activity is very limited in extent to one side of the caldera. The faulting would have to extend all around the caldera block to initiate full collapse, and it’s unlikely current forces are capable of accomplishing such a feat. I do not know of any shield volcanoes outside volcanic arcs having experienced caldera collapse through summit activity alone. Caldera collapses in ocean island and ridge volcanoes seem to be due to lateral draining with voluminous dike intrusions associated with major rifting or eruptions at low elevations that reduce pressure at the summit enough for the caldera roof block to collapse and overcome the resistance from rock friction. Whether the ring dike can push itself farther around the caldera is also unclear, given the northern caldera floor is at the same elevation or slightly higher than the active vents. So it’s a very remote scenario right now, but one I can’t fully rule out. And it’s not as unlikely that a rift zone eruption at low elevations may occur in the next several years while the current caldera fault intrusion (or part of it) is still molten, so that the lateral draining by the rift eruption leads more easily into caldera collapse and ring dike widening occurs right away, triggering explosive activity in the ring fault.

The second “interesting” situation that is being met is that there’s a huge volume of liquid rock filling the 2018 caldera that is stored at high elevations above most of the rest of the volcano. The summit fissure eruption of December 2020 created a ponded lava lake inside Kilauea that later eruptions have kept pouring lava into. This ponded lava lake contained a volume of 0.19 km3 at the end of the September 2023 eruption, point up to which it had been growing endogenously. Now, the current eruption seems to have been erupting on top of this lava lake, but has it contributed volume to it? As far as I’ve been able to observe, the lava emplacement mode has been mixed, including flows that have cooled on top of the lava lake crust but also lava flowing under the crust and lifting it up from below. Up to the end of September 2025, or even in some parts up to December 2025, there were some surfaces with white sulfur spots and bands that could be seen to be pushed up from below in between episodes due to endogenous growth of the caldera lava fill. Some fumarolic features on the lake crust lasted several months without being flowed over. Most of the 0.18 km3 of lava erupted into the lake during these first several months of eruption must have been added to the molten lake. The remarkably tall fountains towards the end of 2025 and the start of 2026 created thick clastogenic flows (flows fed by falling tephra) with a deep, lightweight scoriaceous crust that flowed on top of the lava fill and seems to have stayed “floating” there, building a platform on top of which successive layers of lava have been piling up. Parts of the caldera fill do continue to be thinly crusted and seem to founder when lava comes over them. In these thinly crusted sections, some of the lava has probably continued to seep into the underlying molten lake during 2026, while the rest of the volume has been emplaced in a semisolid stack on top of the molten lake.

It’s evident that right now the summit caldera holds an enormous volume of liquid lava, maybe over 0.4 km3. There is a realistic risk that this lava will break out catastrophically at lower elevations along the Southwest Rift (this rift can be intruded directly from the summit with a dike, unlike the ERZ, which uses the connector to supply rift storage and, once there, create dike intrusions). Such a lake drain-out has happened before, in 1823.

Because of this developing situation, I wanted to talk about the 1790 and 1823 eruptions of Kilauea. These two events happened close in time and were of extraordinary character for Kilauea volcano. Both eruptions are unlike anything the volcano has produced since and were potentially life-threatening events. 1790 is well known; it was an explosive eruption that caused an unknown number of casualties on the order of hundreds of people. The following 1823 Keaiwa eruption was an effusive eruption with unusual characteristics, including 10-meter lava run-ups over obstacles; thankfully, it happened in an uninhabited area. A recent article (Andrea Tonato et al. 2024) looked into this eruption, and with geophysical modelling they obtained 11,200 cubic meters per second as the minimum effusion rate required to have produced the observed run-ups, potentially making this eruption the most intense historical lava flow known of any volcano, being superior to that of Nyiragongo in 1977. Unfortunately, these two events happened just before the start of regular monitoring of the volcano, and the only information is that which was provided by the native Hawaiians, and is almost entirely thanks to William Ellis questioning the Hawaiians when he visited Kilauea in 1823 shortly after the Keaiwa eruption had taken place. Had regular observation started just a few decades earlier, there would be priceless observations on the behaviour of Kilauea, and volcanoes in general.

As it happens, however this article has snowballed like some sort of humongous accretionary lapilli into a size that is far beyond my original size projections. Given that an explosive outcome is at present very remote, if possible at all, I will just focus on the Keaiwa eruption now, and deal with the 1790 eruption in a future article that is already prepared.

The 1823 Keaiwa eruption

The 2024 article by Andrea Tonato et al. lists many of the strange aspects and key characteristics that make the 1823 Keaiwa flow one of Kilauea’s most unusual and impressive eruptions. First of all, the lava had a glass composition of 6.4 wt% MgO, which is normal for Kilauea, and thus puts this eruption within the typical range of viscosity and temperature of the volcano, so that the strange aspects can’t be blamed on this. Kipukas (older stretches of land enclosed by younger lava) of the 1823 flow often show lava run-ups on the side facing upslope, showing that the lava flowed fast enough to flow upslope into obstacles it came across, and in some cases it even overcame them completely. Run-ups of the flow over obstacles range 4-10 m in height, with the most notable case being the so-called Lava Plastered Cones, which were draped in lava on the side facing upslope towards the fissure, and lava pushed 10.5 m up and even down the other side of the cones in some places. Another aspect is the small average thickness at the flow edges, only 40 cm thick. Added to this, the flow never became properly channelized, with flow levees being completely absent, which bears similarity to other Kilauea flows that erupted rapidly, within several hours, not giving time for the flow to “organize”. Keaiwa erupted from a fissure system known as the “Great Crack” that has unusual aspects to it. Most remarkably, the fissure is one single winding but continuous line running for a length of 10 km; this makes it basically unique when compared to other fissures on the planet, which are made of separate offset segments, if only slightly. I attribute this characteristic to maybe the shallow depth in which the dike was emplaced, which I think would have run close to the surface, while in a typical situation the main dike flow is at depth and the surface fissures are distinct lobes that rise from it. Another aspect of the Great Crack is the total absence of pyroclastic material, except close to the sea where it had some phreatomagmatic action. It can be discerned that the melt had been extensively degassed before being effused. The article also mentions the unusual lava balls, which are boulders from the collapsing walls of the Great Crack that were coated in lava, and some of them were ejected out along with the flow. It was confirmed that the Great Crack formed during the 1823 eruption because it cuts through the solid outer crust of the sheet pahoehoe flows, but melt from inside these flows flowed back into the fissure; in places, this coated the walls of the Great Crack in a few cms of lava.

There are some additional aspects of the Great Crack that I also consider worth mentioning. One of them is that the Great Crack, rather than a typical fissure, is a massive gaping chasm, in places over 25 m wide. Uprift of the eruptive portion, the Great Crack can possibly still be followed as first a series of intermittent collapse pits connected by crack systems and then networks of large cracks up to several meters wide that zigzag all the way up to the SW end of the Kilauea caldera (though it’s not clear all of these were formed in 1823). Speleologists have also managed to descend 158 meters down into the Great Crack, finding multiple levels of intact passages that resemble tall canyons and reach over 10 meters wide with vertical walls ascending to darkness. Such widths are well beyond those of ordinary dikes and likely helped transport the magma at high speeds. I’d argue, though, that the width is ultimately a consequence of this particular style and occurs due to intense, rapid erosion that creates trench-like features that seem characteristic of lava lake drain-outs, when the very enlarged dikes flowing close to the surface have their roofs collapse into long chasms.

Regarding the flow morphology, I’d like to note that some publications misleadingly treat Keaiwa as a flow that turns to aa lava at a distance of about 1 km from the vent. This is technically correct: the flow surface beyond this distance is aa, but it only became so at the end of the effusion. When the effusion ends, the lava flow naturally becomes more sluggish and rapidly slows down; in this stage the still-flowing lava has crusted over, and this crust is torn apart. However, as far as the coast, the flow edges and the parts where lava flowed over old flow fronts and obstacles are sheet pahoehoe lava: lava with a glassy surface that did not have a solid crust when flowing. This means at the time of the eruption was at its peak the lava didn’t form a solid crust anywhere; it was a flood of silver pahoehoe lava as far as the coast, 4 km away from the fissures where it entered the ocean.

What actual historical evidence is there on the Keaiwa eruption? Information comes from British missionary and writer William Ellis, who visited the volcano on August 1, 1823, shortly after the Keaiwa eruption and questioned around. He mentioned two different dates of the eruption: one five months before his visit, told by the people of Kealakomo, and another three weeks before, from a person from Kapapala. I think the second one is more reliable given Kapapala is right next to the fissure. According to this later source, the events that led to the Keaiwa eruption may have also unfolded gradually, starting with the first cracks opening 11 months before his visit. 2 months prior, a large earthquake estimated at M 7 caused damage over the southern flank of Kilauea and collapsed the cracks near Kapapala. Then, at 3 weeks before, the outburst would have happened (or five months earlier by the alternate chronology); only some fishermen were apparently in the area who lost five canoes to the flow. According to the Hawaiians, the flow carried a large boulder nearly “30 m” across, broken off from the ground into the sea, producing a tall island at some distance from the coast. The speleologists do mention large boulders the size of houses suspended in between the walls of passages inside the Great Crack, so it’s possible one of these was dragged to the sea.

At the summit of Kilauea, at the time of Ellis visit, the caldera consisted of two craters nested within each other, separated by a flat ledge floored in pristine lava and running along the whole circumference of the caldera, called the Black Ledge. The inner pit was filled with tens of lava cones, many of them floating pyroclastic islands on two vast extents of lava in fluid motion at the S and N ends of the inner crater. Ellis deduced the inner pit had been filled with lava up to the ledge and then drained through underground fissures to the location of the Keaiwa eruption. This indeed seems to be what happened. In fact, the Inner Pit would fill completely drain again in a nearly identical manner in 1832 and 1840, returning the caldera to the almost exact same appearance it had in 1823. After rising to the Black Ledge and well above in elevation, pushing the formed Inner Pit area upwards like a piston from below, the caldera had another, perhaps less voluminous and less extensive, draining in 1868. Observations during this time seem to point to the entire area of the Inner Pit, 3 kilometers in length, being a vast lava lake. Mainly, the observation made in 1838, that massive portions of the Inner Pit would suddenly founder and overturn, swallowing floating cones in the process. Also, that, throughout the 1823-1868 period, the volcano had no trouble forming cones and transient lava lakes and ponds anywhere within the extent of the Inner Pit. While the vast lava lake filling the caldera was a “ponded” lake formed by passive accumulation of melt within the depression, there was also a permanent, vigorously active lava lake at Halema’uma’u, a proper convecting lava lake connected in a wide shaft to the summit chamber of Kilauea. This convecting lava lake likely provided an input of gas into the broader one from below, dynamizing activity across the whole extent of the Inner Pit.

This major lava lake thus drained four times. The 1823 event, though in literature hasn’t been recognized as a lava lake drainout, unless Ellis’ statements are taken as such, here in Volcanocafe well known to a few of us Kilauea watchers as very clearly one such drain out. Most evidently because the perfect degassing of the Keaiwa eruption means the lava had been stored at surface level for some time beforehand before coming out of the Great Crack. But also due to several similarities to Nyiragongo eruptions in 1977 and 2002, which drained large lakes that had been filling its summit crater. One of the most interesting features are trench-like features that seem characteristic of these drain-outs. They likely form as the ground collapses into the very wide expanding dikes that run close to the surface. Both the 1977 and 2002 eruptions of Nyragongo have images of trench-like structures very reminiscent of the Great Crack. Another such feature also formed at Kilauea when Mauna Ulu lava that had ponded inside the pit crater Alae drained spectacularly on August at a flow rate of 5500 m3/s. Below are images comparing the aforementioned structures:

The only time something similar to Keaiwa has been at Nyiragongo was in 1977 and, to some degree, in 2002. I’m not sure if there are good observations before 1977 to establish what kind of lava lake was present, though my guess is major ponded lava lake. The one before 2002 was quite definitely a ponded lava lake formed in fissure eruptions during 1982, and 1994–1996 that had flooded the crater interior. At the time of the 2021 flank eruption, there was instead the narrow shaft of a convecting lava lake which only occupied a small part of the crater, with the rest of the floor rising through successive overflows, so that the ensuing eruption was not as intense as its predecessors.

The 1977 Nyiragongo eruption in particular had some terrifying effects. The lava lake contained within the summit crater burst through fissures on the flanks, unleashing a flood of 0.02 km3 within an hour. Its volume may not seem like much, but it was the speed that was fearsome, given the 6000 m3/s effusion rate. The lava front reached the outskirts of Goma in only 20 min, an average speed of 30 km/h, which is faster than the sprinting speed of an average adult and initially the lava is estimated to have advanced at 100 km/h. Some villages were wiped out, with somewhere between 60 and 300 direct casualties from the lava. Sudden exposure of large surfaces of crust-less lava to the air created enormous heat-driven wind currents that tore off entire banana groves. Lava left behind strange sights, like banana leaves with 1-cm thick coatings of lava, which are reminiscent of descriptions made by William Ellis regarding the 1823 flow where he mentions that he could see the grass the lava had flowed through and that there were lava stalactites hanging from tree branches (near the fissure).

The slopes of the cone and fluidity of Nyiragongo’s magma are superior to those of Kilauea, so it’s likely the flow front speeds were not as fast during the Keaiwa eruption, but still it’s likely that they were enough to have overrun people and be potentially life-threatening. The area was uninhabited back in 1823, and it still is today (as is the entire SWRZ and the areas downslope of it), which means a tragic result is unlikely to happen in the case of a repeat, but could be dangerous if people are trekking through the area.

Will we see a repeat of the Keaiwa eruption, and how will we know if so?

At present, the amount of molten lava contained at the summit, and probably the elevation of the lava fill, seem similar to that of before the 1823, 1832, and 1840 drainouts. Based on the reported depths and extent of the Inner Pit (its silhouette can be seen to this day as a circular line of faint fumarolic deposits across the northern part of the caldera) the volume of each of these three drain events was somewhere around 0.3-0.5 km3. This means the 1823 eruption may have actually been gigantic and poured a large volume of lava into the sea or continued to open eruptive fissures underwater down the submarine SWRZ. I used to think a large 0.2 km3 lava delta at Kapaoo Point had been formed at the time of the eruption but upon closer inspection I’ve realized it’s actually an older delta draped in 1823 lava, possibly the delta from the original eruption of the Lava Plastered Cones. Due to this, 1823 doesn’t have any apparent delta deposits yet entered the coast along a large stretch of coastline, maybe the flow rate was so fast that it effortlessly passed beyond the water surface and flowed down the submarine slopes of the volcano.

The similar situation, or apparently similar situation now to 1823, and also that the fountain events have stopped and pressure is rising, with magma trying to find new ways out, means a similar scenario is definitely not impossible. Though it may still be an unlikely outcome. The Inner Pit lava lake of the early 19th century drained four times, but only one of them resulted in a Keaiwa-style eruption. The draining in 1832 is enigmatic and it’s not known where the magma went- whether a gigantic intrusion or a submarine eruption; at this point, I don’t think the mystery will ever be clarified. 1840 drainout went into both an exceptionally long system of en-echelon dikes and an eruption at the Lower East Rift Zone (because this eruption fed from East Rift storage, it did not directly tap into the summit lava lake and only indirectly deflated the summit chamber, which drove back the summit lava lake). The event of 1868 was mostly intrusive, caused by the large amount of spreading the M 7.9 earthquake caused, which was immediately taken up by an intrusion.

Something like 1840 can probably not happen right now however, because the current summit lava lake is truly rootless, unlike the 18th-century one, which was connected to the wide shaft of Halema’uma’u down the summit chamber. Any drainout of the current lake, as long as conditions do not change, will have to be a direct dike connecting the lake with a lower elevation area. A large enough intrusion is also unlikely to happen at present that can drain the summit ponded lava. As such, if it does drain, it will have to be in a dike-fed eruption somewhere in the SWRZ, though it could be underwater. Alternatively, it might not drain at all if activity for example, shifts into the East Rift Zone.

Currently, the strain is being built at the summit of Kilauea, so it seems a dike intrusion is eventually going to take place and erupt within Halema’uma’u. If it stays into Halema’uma’u, then it’s all fine and normal. Yet if this intrusion propagates into the SWRZ, it could result in a sudden outburst there. It could also happen that the dike stays within the caldera and start pouring lava into the lake, and the rise in lava level triggers an intrusion from the lake itself down the SWRZ. This intrusion would likely start slowly and grow in spurts, maybe over weeks and months, and eventually culminate is a climactic event like Keaiwa in 1823.

But of course there are other potential scenarios. One has magma continuing to intrude stubbornly into the caldera ring faults, as has happened these past weeks, until it somehow happens to bring the caldera block down and puts us on a completely different but equally or even more exceptional course of events. Another could have the lava lake just sitting there while activity shifts elsewhere, for example to the East Rift Zone, or maybe the active vents on the SW part of the caldera grow their own shallow dike into the SWRZ, not connected to the lake, and supply a satellite shield there. In these cases, the lava lake would take many decades to solidify, which might indeed happen eventually if no more attempts at rising the caldera fill happen.

As can be seen, the current situation at Kilauea is getting quite unpredictable, as it usually happens whenever there’s a change in the mode of activity. However, a scenario similar to the 1823 Keaiwa eruption or related is currently a very possible outcome, though perhaps still an unlikely one.

In my next article I will be looking at the 1790 eruption of Kilauea, a somehow even more crazy period of Kilauea than that of 1823-1868, which it too would be rather timely topic given Kilauea has developed a liking to intruding magma into caldera faults as of late.

 

Main references

Tazieff, H. (1977). An exceptional eruption: Mt. Niragongo, Jan. 10th, 1977. Bulletin of Volcanology, 40(3), 189–200. https://doi.org/10.1007/bf02596999

Tonato A*, Shea T, Downs DT, Kelfoun K (2025) Rapid emplacement of the Keaīwa Lava Flow of 1823 from the Great Crack in the Southwest Rift Zone of Kīlauea volcano. Journal of Volcanology and Geothermal Research 466:108391. https://doi.org/10.1016/j.jvolgeores.2025.108391 download pdf

Don Coons. CAVE EXPLORATION WITHIN THE GREAT CRACK OF KILAUEA VOLCANO. http://www.cavepics.com/IVS17/COONSI.pdf