Prelude to Krakatau. III

In Part 1 and Part II we went over the current state of Krakatau and its history over the 300 years before the Big One. Now it is time to find out what caused the Big Bang, whether it is a recurrent offence, and to try our hand at the big question: why is such an offensive volcano located in the middle of an otherwise volcanically inactive region? Let’s begin at the end.


The ending began on September 1, 1880, when there was a strong earthquake in the Sunda Strait. It was felt far and wide, damaged houses, and destroyed a light house. It happened to be the day after the birth of Wilhelmina, future queen of the Netherlands, and early the following morning after the earthquake the residents of Batavia were shaken -and frightened- again when woken up by cannons being fired in celebration. Over the next three years, there were reports of smaller earthquakes around the Sunda Strait. Whether these quakes were related to the coming eruption is difficult to know. They were almost certainly tectonic and the main one was not caused by Krakatau. But Krakatau lies on the fault where the Sunda earthquakes occur. Physical damage to its plumbing by a large earthquake is not impossible.

A few months before this earthquake, in July 1880 Verbeek had make a quick visit to Krakatau and to the small island of the ‘Polish Hat’, spending only a few hours here. On Long Island he found a light grey lava with small crystals, whilst on Polish Hat there were glassy, greenish-black rocks. The collected rocks were identified as pyroxenic andesite: the rocks on Polish Hat were a dark andesitic obsidian. Analysis showed 72 to 73% of silicic acid, very different from the volcanoes on Java or Sumatra.

Rogier Verbeek is 4th from the left. This picture was taken during the investigation after the Krakatau eruption, in October 1883. It is not clear which island this is on; the desolation suggests it is one of the remaining fragments of Krakatau.

After a quiet interlude, small earthquakes were noticed again around the Sunda Strait in May 1883. On May 20, again the people in Batavia were woken up by the booms of cannon fire. But this time, the booms were continuous – and volcanic. The explosions lasted two days. There was much speculation about which volcano was to blame, but no one knew. The towns around the Sunda Strait reported by telegraph that they had heard nothing. It was only when ships that had been in the Sunda Strait reached Batavia, that they reported the source: Krakatau, 150 km from Batavia. Why had the towns around the Sunda Strait, much closer to Krakatau, not heard any explosions? That sounds strange but in fact is normal: the sound waves bend away from the Earth surface, leaving a zone where there is silence. The ships reported black clouds and white ash, and floating pumice. There were no reports from Krakatau itself: the island was not inhabited at the time and was visited only by fishermen who stayed on the coast. No one had noticed any precursor activity.

Traders quickly saw an opportunity to make money. An excursion was organised from Batavia, for a picnic at Krakatau with a view of the spectacle. 86 people paid up and went for the trip of a life time. It could easily have been the last trip of their life time, but they were very lucky. The famous picture of the exploding Krakatau, subsequently coloured by the Royal Society, comes from this trip. The ship left on Saturday after work and arrived after an overnight trip, on Sunday May 27. (Sunday had been chosen deliberately, so people could go without having to miss work.) The ship stayed until Sunday evening, allowing people to explore the eruption site. The island had been stripped of vegetation, apart from the trees on the southern side of Rakata. The tourists found that the eruption came from a crater at Perboewatan. Again, any health-and-safety manager would have vetoed the entire event! But the eruption was seen as a harmless affair, and Krakatau was too far from the main land for anyone to be inconvenienced. It was much like how Hawai’ians view Kilauea.

To see how dangerous it really was, I am including the report from Schuurman from this expedition:

Toward midnight, when we reached Anjer, we saw on the horizon, in the direction of the island of Rakata, a red intermittent glow, from the activity of the volcano. The flare appeared every 5 at 10 minutes, for a few moments only, and from which fell a rain of fire.” […]

Strange was the appearance of the island, which instead of the lush tropical vegetation, had only arid and bare soil, which raised a smoke like the smoke of a furnace. The peak [Rataka] alone still showed greenery, but the northern slope was buried under a thick layer of grey ash, from which came out here and there a few rare tree trunks stripped of branches and leaves, meagre remnants of the impenetrable forest which only recently had covered the island. Within this dark and desolate landscape, standing out on the sea as a picture of the most complete destruction, stood with a indescribable splendour a powerful column of smoke; several tens of meters wide at its base, this column was expelled with a roar of thunder, rising and swirling, up to a height of 1000 to 1200 meters, then 2 or 3000 meters higher, fading more and more and giving up in the easterly wind, the ash with which it was charged and which, falling in the form of a blackish fog, constituted the covering of the ground.

The famous photo (coloured later) taken on the excursion of May 22. The smoke rises from Perboewatan. As the wind was reported to be from the east, this picture must have been taken from north of Krakatau. The mountain on the left edge is likely Danan; The darker one on the right is probably Verlaten Island. Rakata is not visible.

“[…] Although rising without interruption, she was, every 5 to 10 minutes, considerably widened at the bottom by emitted smoke clouds, which were exceptionally powerful, and announced by a terrible noise; during these periodic eruptions, the dark-coloured smoke was expelled more quickly than usual, and dropped, at the height of about 200 meters, a shower of stones, which, viewed from the ship, seemed like black shadows moving in the air. Those dark clouds of smoke, even during the day, sometimes presented a reddish glow; at night they took the appearance of red flames, while the rain of stones, so hardly visible by day, changed into a rain of fire: phenomena that must be explained by the reflection, on the clouds of smoke, of ignition materials contained in the crater, and by the incandescent state of the ejected stones.

Along the north coast of the island, which in places is rocky and steep, we landed at a point where the coast came down to a wide beach. […] On the white sand, which constituted the proper beach, rested a bench of pumice stone a foot thick, and on these pumice had fallen the ashes, forming a 2-foot layer. […] The ash, which seemed soft, had actually packed enough to carry us; by putting the foot on it, we recognized that it only sank to the ankle, and therefore any objection to an excursion inside the island was overcome. Following the trail of the most courageous, perhaps also the most stunned, we climbed the bare mounds, which presented no obstacle other than the ash, yielding under our feet. The route followed went first to a hill, where one still saw, emerging from the ashes, some parts of trees, stripped, broken, a few meters high and showing undeniably that the branches had been violently torn. The wood was desiccated, but there was no indication for burning or charring; in addition, in the volcanic ash there were no leaves or barbs, so the deforestation must probably be attributed to a whirl of wind, such as occurs frequently during volcanic eruptions within atmospheric layers disturbed in their equilibrium by local and excessive heating.

Soon all traces of vegetation disappeared, and the ascent was continued among the dark mounds of ash that hid the sea from our eyes, without being able to hide the column of smoke, still growling and roaring. Finally, the last hill climbed, We stood on the steep edge of the eastern wall of the crater. Hollowed out from the ashes, a beautiful cavity was seen in a basin shape, with a diameter of about 1000 meters, whose circular bottom, located 40 meters lower and 150 to 250 meters wide, had sunk a few meters and was covered with a black crust, a little shiny. Through this crust, but only in a circle 50 meters across and touching the opposite edge of the crater, emerges, with a terrible crash, the powerful column of smoke.

[.. .] Among the sublimated materials, sulfur was deposited in fine dust not only in two small solfatares, but also on the ash that covered certain hills, and to which it gave a yellowish-green colour.

In the evening, around 8 o’clock, we began the journey back to Batavia, under deep and for many of us, undoubtedly ever lasting, impression of the grandiose spectacle that we had just observed.”

Map by Ferzenaar. The dashed line shows his route and landing place. Summits B. D and E were active. B had partly collapsed.

Their descriptions of Perboewatan (remember that they were the first to report from here over almost 300 years of history!) indicated that the crater was surrounded at different points by high walls of lava, and which Verbeek later suggested probably formed (at least in part) in the year 1680. It had started at the same place (or very close to) the 1680 eruption. But whereas 1680 caused extensive lava flows, the only mention of any lava being associated with the 1883 events is the black shiny crust deep in the crater. Rather than lava flows, the explosions a week before had formed a crater 1 km across.

No one mentioned this, but what had happened to the peak of Perboewatan? The peak was reported to be 120 meters tall. It seems possible that the crater was in fact the location of the previous, higer peak. A crater 1-km across can have supported a 100 to 200-meter tall edifice, making the mountain over 200 meters tall before the explosion. Such a height fits with the older drawings. The May eruption may therefore have been the demolition of the ‘barren, reddish, sulfurous spot’ of the oldest record: Krakatau had lost one of its summits. That would make the May explosion around 0.1-0.2km3 (DRE).

The eruption continued. There was another explosion on June 19, and when on June 24 the island re-appeared from the dark clouds, a second, stronger column of smoke was present, originating east of Perboewatan, at the foot of Danan – now there were two simultaneous erupting craters. There was no activity from the summit of Danan. The change in appearance of Perboewatan was finally noted: the three rock masses that once formed the summits of this mountain had disappeared. The observations were made from Anjer, from where the island was only intermittently visible because of the weather, so much of the eruption was unobserved.

The last ever visit to Krakatau was made by Captain H. Ferzenaar on 11 August, sent to map the island (a tad late, perhaps). He found three active craters, two near Danan and one at Perboewatan, with 14 bursting fumaroles on the southern slope of Danan. All active sites were on land: there was no evidence for any off-shore activity. The southern summit of Danan seemed to have partly collapsed towards the northwest. He measured this summit at 400 meters – previously it had been 450 meters. Rakata was not active. The ash on the beach was a meter thick. (This may sound a lot but extrapolating suggest that so far less than 1 km3 of ash and tephra had been ejected.) The commonly reported height for Perboewatan may come from his measurements – after it had been demolished. Ferzenaar traveled along the east and north coast of Krakatau. He did not note any change in the coast line, but we don’t know how familiar he was with the area. It is a pity he did not visit or see the western shore south of Verlaten Island where the major eruption later occurred. Neither did any of the earlier parties go there. (it appears that it was not possible to navigate the channel between Verlaten Island and Krakatau.) We don’t know what changes had occurred there before the big eruption.

Over the three months, the number of eruption centres had been increasing. Normally, when eruptions become established they focus on one point: magma always takes the easiest route. The most likely cause is that magma was so close to the surface that for different parts of the magma the easiest path became a different one. This should have been seen as a warning, and in fact after having been there for two days, Ferzenaar declared the island too dangerous to visit. Fifteen days later, the explosions began that would tear Krakatau apart; 36,000 people died from tsunamis and from (unrecognized) pyroclastic flows. It was precisely three months after the excitement of the Krakatau picnic.


After the tsunami

The eruption had been intensifying on Aug 26. The explosions became strong enough to cause a rough sea throughout the Sunda Strait. During the night, the tsunami waves began. They threw ships which were anchored at the coast on to the beach; the waves will have been several meters high. The next morning, the first of the big waves arrived at 6:30. One ship reported seeing a town destroyed before their eyes. The darkness came shortly after, lasting for 18 hours. The main event started during this darkness, at 10am, when the rain of pumice near the Java coast became a dense rain of mud, more than 20 kilometers from Krakatau. It covered ships in smelly sea floor. The sulfur smell became stifling and people reported difficulty even breathing. The wind increased and reached storm force, and the big waves followed which reached a run-up height of 30 meters. The mud rain lasted for several hours, before becoming intermittent. The ash and pumice continued until the following morning. Afterwards it was found that there had been two massive explosions, 50 minutes apart, with the last one at 10 am the biggest.

Somehow, the eruption had changed from a decent-sized event (VEI-5? Verbeek called it ‘not unimportant but insignificant’) to something very much larger, the most significant eruption of the past 200 years. The rain of mud shows that the final explosions came not from the island, but from underneath the sea. That was a change as none of the earlier ejecta showed any evidence of interaction with water. The hole left afterwards confirms this shift, as it is centred just off the west coast of old Krakatau. Almost 150 years later, we still don’t really know how this happened.

The traditional explanation for the Krakatau Big Bang is the collapse of the magma chamber. The alternative, that the island collapsed into the sea and that this caused the tsunami was already considered and dismissed in 1885. Some collapses did occur, but the main event does not fit this scenario: instead there was an undersea explosion. But if there was little or no lava, why would the magma chamber collapse? The excavations had all been from the surface, not from the magma chamber. It can’t have been because of pressure reduction, since with a volatile-driven eruption, the pressure goes back to the one before the eruption – not lower. Perhaps the heat in the chamber had melted the roof, but in that case the final event was a bit of an accident and only indirectly related to the Krakatau eruption. There is still a bit of a mystery here.

The answer may be in the lack of data from the west side. And perhaps the 2018 collapse of Anak Krakatau gives us a clue. A major cause of volcanic disasters is flank failure. Flank failure had been suggested as the cause of the tsunamis, but dismissed as it wasn’t mainly directed towards the northwest. However, a flank eruption could still have initiated the final phase. What happened on the west side of Krakatau? Ferzenaar could not visit the west because of the smoke. This may not have been just from the three craters on land: a fourth site could have been hiding itself. And this fourth site was in a dangerous place. I imagine a flank eruption building on the western slope of Danan, similar to what happened in St Helens. The new site was building at or near sea level. It over-inflated, and collapsed. The flank collapse exposed the magma chamber below sea level, which triggered the final explosion. The rest is history. This idea fits the big event, including the mud and the fact that the explosions were not primarily in one direction, as is the case if the main event had itself been the flank explosion.

If Krakatau’s Big Bang was indeed triggered by a flank collapse, it would fit the pattern that flank collapses are far and away the largest cause of casualties from volcanic eruptions. They can turn an eruption from a picnic into a disaster.

An interesting sideline: Verbeek in 1885 predicted that when Krakatau would begin to rebuild, it would do so in-between Danan and Perboewatan. In 1929, that is exactly where Anak Krakatau appeared. Verbeek was the only one ever to understand Krakatau. He later wrote that during the 1880 visit he had found indications for a rift running through the Sunda Strait. (He also complained about the lack of accurate maps of Krakatau. I sympathise!)


Rakata after the eruption

The collapse took away half of Rakata. This gave Verbeek the opportunity of a life time to see a volcano’s innards, during his visit in October. The conduit was clearly visible, and Rakata was seen to consist of many separate layers, each 10 meters or more thick. This indicates multiple sequential lava flows, well separated in time. Rakata was not an instant volcano: it had taken time to build. There is an impression that the oldest layers were build up to the east of the later summit. This must clearly have been long before the modern era, because Rakata was inactive and probably extinct by 1600. But how long?

We can get some idea from the rate at which magma is added to the island. Since 1883, all growth has been at Anak Krakatau which grew to a volume of roughly 0.2 km3 before its 2018 collapse. That gives a supply rate of order 0.002 km3 per year – this is not huge, but suffices to explain Anak. At this rate, recovery from the 1883 eruption (20 km3) will take 10,000 years. Rakata may indeed be that old. The old caldera is a few times larger and would take 50,000 years for recovery. These rough numbers suggest that the indicated age of the ancient caldera of 60,000 years is in fact very reasonable.

So it appears that Krakatau, for all its violence, does its violent eruptions only infrequently. The normal mode is of small, cone-building (or cone-destroying) eruptions, similar to how Anak Krakatau build itself up. Collapse events as in 2018 are probably common, but although highly damaging, on the volcanic scale these are small fry. The 1883 eruption dug much deeper.

But there are persistent claims on the web that the formation of the large caldera took place in either 416 or 535 AD. Where does this come from? The answer to this is two-fold.

Let’s first address the claimed 535 AD eruption. This was based on (1) the fact that there was a major eruption somewhere on Earth at this time; (2) that the 416 AD eruption was misdated. If you are waiting for a punch line, there isn’t one: this is all of the evidence. Of course our knowledge has advanced a bit since this claim was made. We now know that the 535 eruption seen in the ice cores was in the northern hemisphere, and that there was a tropical eruption in 540 but there is a good candidate for it in Ilopongo in El Salvador. The 535 Krakatau eruption can be dismissed as fairly wild speculation.

How about 416? This comes from a book called Pustaka Raja Purwa: the book of king Purwa, and was first advocated as describing Krakatau by John Dudd in 1889, Nature 40, 365-363. The Pustaka was written by Raden Ngabehi Ranggawarsita, a famous Javanese ‘pujangga’ (a combination of chronicler and poet) from the Surakarta Sultanate. The Pustaka Raja is the major achievement of a major writer: a massive collection of stories covering the history of the central Java kings, an Indonesian 1001 Nights. The Purwa is part of it, a five-volume introduction to the Pustaka Raja, covering the earliest years in a semi-historical, semi-mythical manner.

Pustaka Raja Purwa volume 2 p. 46, containing the eruption story

The particular story goes as follows:

A thundering sound was heard from the mountain Batuwara which was answered by a similar noise from Kapi, lying westward of the modern Bantam. A great glowing fire, which reached the sky, came out of the last-named mountain; the whole world was greatly shaken and violent thundering, accompanied by heavy rain and storms took place, but not only did not this heavy rain extinguish the eruption of the fire of the mountain Kapi, but augmented the fire; the noise was fearful, at last the mountain Kapi with a tremendous roar burst into pieces and sank into the deepest of the earth. The water of the sea rose and inundated the land, the country to the east of the mountain Batuwara, to the mountain Rajabasa, was inundated by the sea; the inhabitants of the northern part of the Sunda country to the mountain Rajabasa were drowned and swept away with all property. The water subsided but the land on which Kapi stood became sea, and Java and Sumatra were divided into two parts.

Confused? There are two versions of the story, in two different editions of the Purwa. The first edition gives no date whilst the second edition gives the date of 416. We do not know what the story is based on. (The second edition appeared a few years after Krakatau erupted and must have been prepared by someone else as Ranggawarsita had died a decade before the eruption.) It is normally quoted as an ‘ancient Javanese document‘ but the writing dates from the late 19th century, and we don’t really know whether there is any older tradition behind it. So how do you judge the authorative-ness? That is done using several criteria: (1) How reliable is the author? (2) Do the details make sense? (3) Is there independent evidence?

The reliability is not known as none of the stories in the Purwa can be tested. The dates are definitely unreliable as the book in another story gives different dates for the same story. (And for our story, the date comes from a version with uncertain authorship and is not in the ‘authorative’ (i.e. known author) version.) More ancient Javanese documents do not go back nearly this far; we do have some stone inscriptions from the 4th to 6th century but these only refer to acts of kings. As for the details in the story, the separation between Java and Sumatra makes no sense. They were last connected during the ice age and the separation came with the sea level rise 10,000 years ago, not with any volcanic activity. Kapi may or may not be Krakatau. The other locations are around the narrow, northern end of the Sunda Strait. The description of a tsunami seems accurate, suggesting there may well be a historic event behind the story, which affected the north of the Strait. As for independent evidence, there is none. There are no indications for a major eruption around the Strait at this time. Drilling off the southeast coast of Sumatra has found two dacite tephra layers which because of their composition and location can be attributed to Krakatau. One is recent, located in the upper ten centimeters and is attributed to the 1883 eruption. The other is 2.2 meters below the surface, and is tentatively dated to 60,000 years ago. Although there may have been many smaller eruptions in between, there is no indication for anything else at the level of the 1883 eruption, let alone one several times larger.

The conclusion from this is that the large caldera of Krakatau is most likely 60,000 years old. If there is a true story behind the mythology of the Purwa, the eruption that it describes was too small to form a tephra layer in the ocean, and it certainly wasn’t the formation of the Krakatau caldera. An event similar to the 2018 collapse of Anak Krakatau, which generated a very damaging tsunami, can not be excluded: a small (continuous) eruption, collapse into the sea, and a tsunami can fit the description. In addition, the date of the story is highly uncertain. It is beautiful writing – but science requires evidence. And the evidence we have does not support a large, historical eruption in the Sunda Strait other than the 1883 one.

Why Krakatau?

So Krakatau is a dangerous volcano but not a high-volume one. That may seem a surprising conclusion, seeing how frequently Anak erupted before its disappearance. But these eruptions were small ones, driven by the fact that since 1883 the magma has had a fairly easy route to the surface. Look around the Sunda Strait, and you find extinct volcanoes. There is evidence for one or more very large eruptions around 100,000 years ago which formed ignimbrite layers, but not much since. All other island volcanoes in the Strait are inactive. Krakatau is a peculiar exception.

To get a greip grip, we need to look at Indonesia as a whole. Java is straightforward, a volcanic chain perpendicular to the subduction front. But Sumatra lies at an angle of 40 degrees to the subduction direction. The Great Sumatra Fault runs along the length of the island and takes up the strike motion that come from this angle. It is a highly segmented fault, and where segments are offset, a small basin forms. The Sumatran volcanoes lie along this fault; the basins feed the massive calderas that are characteristic of Sumatra (including Toba). There are also two segments underneath the Sunda Strait, and these have formed a deep subsidence basin. This basin is aligned with the Sumatra Fault close to Sumatra, but bends north-south towards Java. The basin is divided into a west graben and an east graben, with the Tabuan Ridge in the centre. Both grabens are bounded by multiple faults.

The Sumatra fault in the Sunda Strait. Source: Mukti, Ris.Geo.Tam Vol. 28, 115 (2018).

There is a ridge on the east side of the east graben, called the Krakatau ridge which runs north-south, and passes to the west of Krakatau. Beyond that, the floor of the Sunda Strait looks fairly featureless. That is surprising, as the Strait formed by Sumatra and Java pulling apart, so a pull-apart basin should run along the length of the Strait. Where is it?

The seismic reflection data, 25 km south of Krakatau. The trace runs SSW-NNE

A seismic reflection study found a hidden deep graben, roughly where the letter ‘K’ is of “UKF’ on the figure above. It is 3-6 km deep, less than 20 km wide, and bounded by steep normal faults. But nothing is visible on the surface: it is completely filled with sediment. Somewhat optimistically, the discoverers (Susilohadi et al., 2009 Tectonophysics, 467, 55–71) called it the the ‘Krakatau graben’, a name that has not yet caught on. The study also saw some indication of this graben some 10 km south of Krakatau where they had another trace. Here it was not as well defined. From the two detections, it seems this graben may run north-south, and if it extends as far as Krakatau, the volcano will either be in it or close to its edge. The graben may line up with Lampung Bay on Sumatra but there is currently no evidence it extends that far.

The next part of the puzzle is the line of volcanoes that includes Krakatau. The earthquakes define it as a fracture zone, and the depth of these quakes shows that it is in the bedrock – it is an ancient fault, as already indicated by the fact that its alignment, NNE-SSW is different from that of any recent faults. The volcanic activity is at least a million year old; its presence suggests that the fault was re-activated at some point, but the fact that only Krakatau remains active shows that the fault is being pushed back into retirement.

What happened? The rule of thumb in Asia is that everything is India’s fault. Its crash into Asia deformed the entire continent, and through a chain reaction involving the opening of the Andaman Sea and the rotation of Borneo, caused Sumatra to move northwest and to rotate. The strike-slip motion is currently accommodated by the Sumatra Fault.

From Pramumijoyo and Sebrier, Journal of Southeast Asian Earth Sciences, 1991,
Volume 6, Page 37: Neogene and quaternary fault kinematics around the Sunda Strait area, Indonesia (

The Sumatra fault reaches the Sunda Strait at the Semangka bay. The Lampung Bay further east looks quite similar, and is also bounded by a fault on its northern side. This fault has been argued to continue a short distance across the Strait, in Java, but it has not been traced underneath the Strait. The Lampung bay fault shows evidence for slip-strike movement before around 5 million years ago, but only normal motion since (i.e. before that time there was sideways motion of both sides, since that time the relative motion is up/down). It may have acted as a branch of the Sumatra fault at the time. Movement along this fault could have re-activated the volcanic fracture zone, which begins at this fault.

This makes the history of movements inside the Sunda Strait extraordinary complex. Before 5 million years ago, the area between Lampung Bay and Semangka Bay pulled away from Java, leaving behind a basin with subsidence and thinned crust. After that time, the slip movement here ceased, and the pulling motion now happened southwest of Semangka Bay as the Sumatra Fault took over. The pull-away motion could be origin of the Krakatau graben, present mainly at the southern end because the movement continued there for longer. The sediment-fill is because this was a long time ago.

Now look at Krakatau, It sits where the volcanic fracture zone reaches the Krakatau graben, and just before it hits the Krakatau ridge where it may terminate. In a graben, the crust is thinned and the stress is extensional. This may have given decompression melt, and the low stress made it easy to store the magma and move it up. The composition of the lavas suggest that the magma may be stored a long time in the chambers. The location allowed Krakatau to flourish, while the other volcanoes withered and died as the re-activated fault re-lost its vigour, after the Sumatra Fault stole the northwest movement. There are magmatic no-go areas in all directions from Krakatau: go either way on the fault or on the graben and you find a volcanic dead-end: four volcanic stop signs.

This picture of Krakatau’s cause and origin remains speculative. Most geology research in the region focusses on the oil deposits on Sumatra, and little is done on the Sunda Strait. We don’t know the ages of the faults or volcanoes, the location of the grabens, or even if the Sunda Strait was opened by extension or by rotation (nor where the pivot would be). After all these years, Krakatau’s world still hasn’t been properly mapped.


Krakatau spent centuries or even millennia building up to its explosion. The signs were fumarole activity, drying and moving hot springs, and possibly inflation. After Verbeek briefly visited the island in 1880, he pointed out its peculiar location, where different lines of volcanoes intersect. Had he seen something that had made him uneasy? It is not recorded, and he does not mention any sign of activity apart from the 1680 lava flows, but he later said that, if people had visited more often, signs of an impending eruption might have been spotted. Still, without knowledge of its ancient history it would have been hard to foresee the catastrophe that came from the collapse. It is a typical case of lurking danger not being recognized for want of history.

The main lesson for us is not to rely on apparent inactivity of a volcano. Krakatau may only do its big explosions once every 50,000 years. But with over 1000 active volcanoes in the world, that still means we may expect something this big every 50 years. And Krakatau had a second method to cause disaster, which it probably has done far more frequently. In the volcanic lottery, someone will draw the losing ticket. Only the past can tell us what risk each ticket carries.

And there is one volcano with a behaviour similar to that of Krakatau before its eruption. Iwo Jima has shown centuries of inflation, with fumaroles and hot springs. Nothing may happen – but if it does, we should not call it unexpected. It too is condemned to repeat its past.

Albert Zijlstra, May 2019

Part I and Part II

There have been previous posts on Krakatau:

  • Krakatoa: a blast from the past
  • Krakatoa skies: when the Sun turned blue
  • The rise and fall of Anak Krakatau
  • After this heavy post, here is some light Escher

    153 thoughts on “Prelude to Krakatau. III

    1. Thanks Albert for the wonderful article and the even better series as a whole 🙂

      “The main lesson for us is not to rely on apparent inactivity of a volcano. Krakatau may only do its big explosions once every 50,000 years. But with over 1000 active volcanoes in the world, that still means we may expect something this big every 50 years.”

      I think this is an important comment, and reflecting on some of the discussion in the previous post on the Taupo Volcanic Zone, I think it’s important to realize that the bigger (VEI5+) eruptions of tomorrow likely won’t come from the “usual suspects”. I find it interesting how the most active volcanoes get the most monitoring (often for good reason), yet from a bigger picture risk perspective, we likely aren’t even doing much in the way of actively monitoring volcanoes that can create very large eruptions.

      Iwo Jima / Ioto is a good example of this, but there are many other wonderful examples. Ata in Japan has always struck me as a volcanic region that can potentially catch a lot of people off guard one day as everyone in that area is heavily focused on the Sakurajima / Aira caldera system. Many other good examples of volcanoes like these, some of which are monitored to a decent level, but may not have the instrumentation we see in places like Iceland of course.

      • In response,, it’s worth noting that Krakatau was exceptional in one respect; of the VEI 6+ eruptions of the last 200 years or thereabouts, it’s the only one with any historical record (however fragmentary the evidence) of previous activity. Tambora, Santa Maria, Katmai and Pinatubo had shown zilch in historic time (admittedly for Katmai that was less than a century, but still…)

    2. A very satisfying conclusion to a great series! Many thanks!

    3. FINALLY!! Some closure. I first read about this beast as a teen in the local library. I quickly ran out of information to read. Thank you VERY VERY much for filling in the gaps!


    4. It is very interesting the comparison between iwo jima and krakatau, and the evidence of similar processes between the two. On that note too it is unlikely iwo jima will have the second part of its name in the next century if the events leading up to 1883 are analogous to the uplift at ioto since ww2. It would be quite incredible if it blows up in 2083, to continue the line of the ’83 years and volcanism (1783 (skaftar fires), 1883 (krakatau) and 1983 (pu’u o’o) 🙂 )

      • 2083 is reserved for Grimsvotn Saksunarvatn, round 2. 😀

        Another in the line of great posts in this Krakatau series, putting a roof on the very strong walls build by previous two posts.

        • That woud be alot of fun….Saksunarvatn 2.0 year 2022 ( So I will be alive )
          First it begins with 2011 pherato- glacial plinian on steroids ( 10 times larger ) with many souch columns around in something like a ring row. Whole Iceland gets darkened under basalt mammatus and extremely heavy lapilli fallout in east Iceland many meters. And North hemisphere airspace are of course turned off. heavy water flooding too.
          The ice melting woud be severe dwarfing Katlas 1918 floods.
          Whle caldera complex is blown away

          And later many many kilometers tall glowing fountains merges with higher ashclouds shoots up ….streaked with lighting and hot as a blast furnace.
          Eruptive rates as gas is so high that the fluid basalt is blown up
          But its much much hotter too than a sillic plinian.
          Well into its phase the columns base glows ferocuisly
          The yellow white hot plinian fountain column from Glacier is like a huge fiery pillar
          From Faore island its an impressive sight glowing.
          Faroe Islands woud get quite alot of ashfall from a X100km3 basalt plinian

          In the end Grimsvötn forms a new caldera and ebcomes dormant for 100 s of years.

        • Another even more lavaly scenaro is IF the 800 km3 reosvair eruopts in a Laki like Show
          That woud yeild at least 300km3 Laki like event ( but much larger and more intense )
          But also very very very very very unlikley to happen.

          This is how Grimsvötn Saksunarvatn 2.0 woud look like….
          But now we imagines the glowing hot ashclouds and pumice fountains to be 20 km high
          And that feeds a 40 to 50 km high Grimsvötn ashcloud depening eruptive rates….
          Calbuco was impressive enough and imagine Grimsvötn 100 times larger
          and much much hotter since its thoelite or picrite from plume

    5. Great article!

      I wonder if there is a more subtle variable for very large explosions in eruptions. I am currently trying to figure out if a100,000 megaton explosion for a mid range VEI 8 eruption is possible mainly because i want to write an article on the possibility of mesospheric injection with VEI 8 eruptions(and i am really bored).

      • Look for the wikipedia article on megaton yields. There is a graph in the article that gives mushroom cloud height vs yield. That should be a good approximation to eruption plume height since the underlying mechanics at that point are the same (heat energy).

      • Always bearing in mind that a nuclear explosion releases all its energy in a pico-second, whereas a volcanic event would take hours or days to reach the same total energy release

        • True. But the energy of the initial eruption onset should scale quite well with nuclear weapon yields. This is what determines if it reaches the stratosphere or not. Other than that M. Don is dead on accurate. Total eruptive energy is spread out over time and using that value is very likely to yield bogus results.

      • Just be aware that moving from one air mass to another may change the dynamics of how the plume interacts with it. I’m not sure if the mesosphere has a different dynamic than the stratosphere. Also note that the further away from a fitted curve you get, the greater the uncertainty that the fitted curve I’d still valid. Also be aware that what LOOKS to be an exponential curve could just be a sigmoid curve before the inflection point. Sigmoids show up a lot in natural science. Sparks even noted sigmoid like behavior in bubble nucleation rates in magma bodies of some volcanoes in an preface to a book on explosive eruptions. (I’m not sure of the title)

        • At the risk of bordering on tediousor monotonous…

          Fitted curves tend to have a basic underlying function that is adjusted to fit the available data. These underlying functions have general characteristics that all functions like them share. For example. A function that contains the term x squared. Will always take off in the positive or negative direction on each end of the curve the further away from the minima or maxima you go. An x cubed function will do the same. But each end will proceed in the opposite direction of the other end. Higher orders of “x” will show similar behavior, each order having its own characteristic. This is one of the reasons that many excel manuals will warn about using “polynomial fitted” curves. As a general rule, the usefulness of a fitted curve, no matter what type, is within the boundaries of the fitted (modeled) data. The sound way of approaching this is to have less reliance in the model output the further you get from the fitted region. Albert may know a few tricks in determining how this assumed error adds up, but I just wanted to warn anyone trying to construct a model based on the plume height data above. The important thing is to realize where it can give you unverifiable and possibly horrible results.

          Note: Both Albert and Carl can give you further guidance if necessary. They are far better at it than I.

          • Note: This “further guidance” thing is completely up to them. I am in no way volunteering their services. So. If you need help, you had better butter them up and be ultra nice about it. I’m just pointing you at people that I know for a fact know what they are doing.

          • …and, as s former instructor, I try to make things understandable. Sometimes too simple. For example. Though what I have stated is not calculus, its based on principles from it. Calculus really isn’t that hard, I’ve passed Calc 101 twice, just as many times as I’ve dropped out. First drop out was due to human nature (Homo Stultus), second time was due to job requirements. I wound up taking it a second time because my transcripts arrived late. Too bad also. My second time through was much much easier though it was 20 years between courses. Ya see. In calculus you learn how to actually DO interesting stuff with math.

      • A 100 gigaton bomb is possible with a few shipping containers and a few hundred kg of lithium deuteride… and some other stuff too 🙂

        Yes it is the timing of these processes, a VEI 8 is a days-weeks long release of energy, a nuclear bomb is so fast it is outside the scope of the physics we encounter and can see. Supervolcanoes are far larger than anything we have made as a bomb, but the tsar bomba was probably bigger than any single explosion in a VEI 8. Tsar bomba could have been made 2 times more powerful without changing the design at all only using uranium as a damper instead of lead, and it was only as big as it was so it coukd ve transported, built it on a ship and you can make over 1000 of them and still get them somewhere. That is your 100 gigaton bomb 🙂

        No volcano comes close, you need an asteroid for that, asteroid belt is nice but they are all orbiting tbe same direction as earth, but an oort comet hitting earth head on retrograde at its maximum velocity would be 10 times worse, there isnt enough uranium on earth to compete with the potential apocalypse if hale-bopp hit the earth…

        Also intergalactic comets are a thing and galaxy escape velocities are hundreds of km second or more which is 10 times faster than hitting the sun after falling from the oort cloud… the chance is a number containing a long line of zeroes and a 1 at the end somewhere but its still there.

        Or a neutron star could hit the earth, watch the entire planet compress onto the surface of thus tiny glowing speck. Then it would hit the sun (or probably the other way round actually) and you get a supernova

        Yeah supervolcanoes arent that bad really…

        • I think one of the limiting factors in nuke sizes, is that past a certain point, you are just releasing energy into space and not doing much more surface damage, so its it’s its a “waste” of effort in getting bigger boom boom.

        • @Turtlebirdman. Thank you for pointing out that though an ultra low probability it is still there. That’s an important distinction that many people just gloss over. Just knowing there is a non-zero probability is important.

        • Can you tell me how much energy a VEI 8 eruption would have because I keep getting different answers from different studies. ranging from 865.000 megatonnes for a low end to 150,000 megatons for an eruption of the same scale.

          • No, I can’t. But what I would do is take the most referenced numbers and average them out. And remember that it was a range and run a few percent above and below that. Remember, you’re just trying to ball-park it to get an idea of what is possible, so you don’t need to beat yourself up over un needed accuracy. In your dialogue, make sure you point out the uncertainty.

          • And if you want to make it ‘seem’ more accurate, find the 95% confidence interval and use that as your high and low range. Sure, it might be B/S if it has no real basis, but it will look pretty and sell lots of shampoo in a commercial. Most people have no clue what the numbers mean anyway. That’s why the news uses them all of the time. Be careful if you run across someone that understands them though. They will eat you alive.

          • If you do get called to task on your numbers. Take the time to try to understand where you made your mistakes. You’ll get better at it. And don’t be afraid to fail. That’s how you learn.

          • You can start with the thermal energy, and then scale the kinetic energy to the thermal. If we assume that a VEI 8 eruption has a Magnitude of greater than or equal to 8 and less then or equal to 9, the total mass is 10^15 Kg to 10^16 Kg. With a temperature difference (from ambient) of 700-1000 K and a heat capacity of about 1000 Joules per Kg/K, the total thermal energy will be of the order of 10^21 – 10^22 Joules. Eruptions rates of these events are not well known but values of 10^9 – 10^11 kg/s seem likely to explain the dispersal and area of large ignimbrites and associated co-ignimbrite ash falls in the geological record.

            • …and do note that Mastin et al based their work (mass ejection rate vs plume height) on Sparks similar work. Both were based on available heat energy of the ejected rock.

              So… taking this approach seems like a sound way to approach the problem. Several experts in the field used this method.

            • (I use the Mastin et al formula to estimate ejection rates, then use linear interpolation to guesstimate the prevailing rate between the known sample points on a second by second basis, then integrate that curve to get a quick estimate of the VEI of ongoing eruptions.)

              And you can find the Mastin et al paper here.

              Note that the resultant volumetric output is in Dense Rock Equivalent. (DRE)

          • Nah, that is not the correct numbers.
            You have to remove a few zeros on that one.
            People never get the “lofting, crushing, milling” potential of the Tsar Bomba for instance.
            But this is not the time and place to go into nuclear debates. I did that once, and that is enough.

    6. Bravo, Albert! Many thanks for all of your work, it is very much appreciated. A wonderful series of articles.
      Go pour yourself a large one and put your feet up. Cheers!

      Take care, now.

      • Thanks. This was quite exhausting – I feel like i have now written a book on this thing, over 6 posts!

      • I absolutely concur. As detailed and as reasoned as available information sources permitted. These three articles were utterly riveting.

        Thank you Albert.

    7. A little action from Japan…

      The Japan Meteorological Agency (JMA) raised Alert Level for Hakoneyama volcano from Level 1 to Level 2 at 17:15 UTC on May 18 (02:15 JST on May 19). The last magmatic eruptive activity at this volcano took place about 2 900 years ago. The volcano is located in Kanagawa Prefecture, south of capital Tokyo.
      JMA said the number of volcanic earthquakes has increased since the morning of May 18 (LT), along with fumarolic activity in Owakudani and the surrounding areas.
      Large volcanic rocks could fall at Owakudani, JMA said, prohibiting access to the popular tourist spot.
      According to The Asahi Shimbun, Hakone Ropeway Co. suspended operations between Sounzan and Togendai stations, the starting to final stops, and started to provide bus service between these stations.

          • Perhaps now an unpleasant but sub-catastrophic eruption can be referred to as a DCD! My contribution to sheepy culture. 🐑

        • It did have a phreatic eruption in 2015 so the new unrest isn’t that surprising.

          • I remember following and watching this one. Hakone is an interesting volcano, and any time you get a volcano that hasn’t erupted in 2900 years potentially waking up, it’s worth paying attention to.

            With that said, probability would say that this is unlikely to lead to an eruption. But *if* we got a true eruption, there is definitely some potential for something of the larger sort here compared to our usual volcanoes of discussion.

    8. I have a question on a different topic (after a few weeks living with Krakatau it is hard to move on!). The plot below shows the atmospheric transparency at Mount Montezuma, measured by the solar observatory there. There is an obvious dip in 1932 which seems to be dust from the eruption of Azul. But there is also a dip in 1953 which has no clear cause. It was attributed to smelters in the area but that does not seem to correlate well, and the dip is also seen (weakly) in the northern hemisphere. Nuclear explosions have been suggested but they correlate too well either. A volcanic eruption seems most likely, in the southern hemisphere or equatorial. But there doesn’t seem much on record. Does anyone have any ideas?

      The small dip in 1952 is attributed to Mount Lamington, but that is too early for 1953.

      • An eruption of that magnitude would not go unnoticed in the modern times, so its likely not volcanic. It must be some other source. Have you checked gkobally if this signal comes up?

        • That was my first feeling but the authors of the paper are quite adamant in ruling out alternatives, and it seen (a little) n the northern hemisphere as well. Are you sure that a VEI-5 somewhere in the south could not have been or missed or underestimated?

          • A VEI-5 should not be hard to miss, even weaker booms at earlier times got noticed. Its just too big.


        Groningen gas output should be cut more quickly, says minister

        AMSTERDAM (Reuters) – Production at the Groningen gas field in the Netherlands should be reduced at a faster pace following a 3.4 magnitude earthquake in the northern Dutch region on Wednesday, Economic Affairs Minister Eric Wiebes said.

        …Wednesday’s earthquake was among the most severe to ever hit the region. There were no reports of casualties, but local authorities said around 90 homeowners reported damage in the first hours after the quake.

        After last year’s quake, SodM said production should be cut to less than 12 bcm as soon as possible to minimize seismic risks.

        • There are two problems, first not specific for Groningen.

          First one is the lowering of the groundwatertables. Going on in NL since agricultural intensive cropping became dominant in the 50’s 60’s. Getting a low watertable makes it easier to mow and crop more early in season (higher soil temperature). In sandy soils not much of a problem (well, talking about buidings damaged here only). In clay and peat soils, containing lot of water, in the northern and western parts of the country that certainly is. Older buildings have no sufficient structure to hold against settling of the soil caused by the lowering of the groundwatertable. It is common in many areas and seen as major problem for future decades. Lowering tables is no long term solution for agricultural improvements, because the soil settles near the new low watertable, so a new even lower table is necessary … and so on. Foundating structures of buidings (often wood) become unstable in the dryer condition, wood starts to rot.

          Second is the soil subsidence caused by gas field production (up to 32 cm in 2013). Fact is the porous layers from which the gas is extracted, collapses bit by bit resulting in quakes.
          Special for areas like Groningen, is the extra adjustment (lowering) of the watertables, because of subsidence of soil caused by gas field production.
          Walls are cracking, older buidings may become and have become unstable. The (not that strong) quakes have worsening effect on the problem of instability. Some houses even should be teared down and/or rebuild completely.

          Actual subsidence in the Netherlands mm/year. Groningen is clearly present in the north (Nederlands Centrum voor Geo­desie en Geo-informatica (NCG))

          Future subsidence 2002 -2050 without correcting actions (Deltares, TNO-GDN and Wageningen Environmental Research).

          • Interesting! I know about the quakes and the problems the people in Groningen are facing, but the influence of water level management is new to me. Glad to live on the sandy soils, though (although drought is a bit of an issue here) 😉

            • Every person I know of the Netherlands persuasion is trying to persuade me that there is draught. I found my trip there to horribly cold, windy and wet this Tuesday.
              Germany was even worse, they had reinvented the Autobahn 1 into the Autozee 1. My trusty Merc was transformed into a stately canal-ship.

        • Its unclear to me whether this is actually a problem. Presumably the gas company will pay to remedy any damage, some of which will be pre-existing (but maybe made worse) and everyone is happy?

          It would be unlikely that substantial gas could be taken without some settlement.

          • Welcome to the Netherlands 😉 We, as a country, have become mighty rich from the gas, but recompensation?

          • Yes, compensation has been promised to the Groningers, so there should not be a financial problem.
            It takes a long time though. The oil company (NAM) has been marking its own paper, at first. Gouvernment has taken over now, but the complicated relation oilcompany – Dutch state (NAM is 50% owned by British / Dutch Shell which have large tax benefits in NL) might have a slowing effect on the settlement/payment of damage. Not easy to solve, damage that gets worse every earthquake… It is unclear when the soilsettlement will stop finally. Some gas is taken from the field still.

            The other problem (causing far more financial damage), overall subsidence of the soil in West and North, only can be stopped when lowering of the watertables stops. This would cause a strong diminish of the Netherlands exporting agricultural stuff, quite a lot at the moment. Large impact I think. But if not solved it will cause billions damage to buildings/infrastructure. And with the rising sealevel lowering tables aggravates salinization of the water in coastal areas too. It’s time for new methods. 🙂

      • 3.4 according to KNMI. ‘Heavy’ earthquake say newsagencies.
        Groningen is blown up once more.

    9. This is quite off topic to krakatau but it is relevant to one of the discussions on the last article.

      Found this just now, seems last years eruption was apparently not as big as I thought, 0.19 km3, and nowhere near as big as holuhraun or fissure 8, and neither was the 1979 eruption, which was about half the volume of fissure 8 and holuhraun, 0.65 km3. Still an impressive size for a basaltic eruption that was not fed by gravity though.

      I am somewhat sceptical of the lower 2018 flow being only 10 meters thick however, lava flows near the ocean are often a lot more than that anyway even on eruptions that are not considered ‘big’, and this eruption was also directly fed by a vent on the coastal plain so it would be expected the lava would accumulate around the vent a lot more than an eruption higher up. The eruption also formed a 200 meter wide cone, which is about as big as the cone formed by fissure 8, and while this isnt a great size indicator it is probable given the similarity that this vent erupted quite a bit more lava than the study shows, the video I linked shows an icelandic-style curtain of fire that is 1.5 km long and the tallest fountain is at least 100 meters high, which requires a pretty serious eruption rate, several hundred m3/s on the whole thing.

      • There is a vent in the E flank which is about 185 m tall and 1.5 km wide, that one called my attention because it is an oversized reticulite cone, similar to Kilauea Iki but much bigger, it would have been quite a spectacular eruption to watch.

        • Yes I saw that, last years eruption very briefly had the lower flank vent fountain to 500 meters when it first opened, so maybe not surprising either. Last year seems to have been that the eruption sent for a day including the low flank vent (fissure 4), then stopped, then new fissures opened near the lowest original vent a few days later which is what that video on the last article was and these lasted until mid August. Really hard to tell though without continuous observation.

    10. An M2.5 at Grimsvotn. And a special one. I assume Carl might have something to say about it. 🙂

      Here is the low frequency component. Anyone sees something special? 🙂

    11. Thanks Albert for the awesome series, the first two parts made me really looking forward to the last one!
      Is the chemistry of the lava any different from the next active volcanos aorund? The description of Krakatoa not being active for a very long time before its eruptions got me thinking: How about the nearby “extinct” volcanos of Rajabassa/Sebesi/Sebuku/Panaitan Island? Especially Sebesi and Sebuku seem to be in a quite similar geological setting. Are they monitored?

      Thanks again for the time you put into this, happy to know more about one of the most interesting and volcanos around the world!

      • Not happy though about more OT flaring up again.. Maybe shouldn´t say this, but most of the off topic is not respectful at all in regards to Albert and the time he put into this.. cautiously spoken!

        • It has been noticed but I’ve lacked the time to fully review everything.

          Guys, can we please keep things mostly on topic. A lot of time is invested in these posts, please respect that by not beating your own drum excessively at the expense of other people’s questions and interest in the article.

          • The trouble is its TOO good. The arguments are well documented and well argued and its really hard to comment other than “me too” to congratulations on a job well done.

            I think its called “the last word” on a topic….

      • I had the same thought about Sebesi and Sebuku.

        Also while looking at Google Earth trying to find Pinatubo I noticed a similar looking mountain to the north of it. Does anyone know the name of it? I guessed Pinatubo is the one with visible sediment drainage.

    12. Maybe we should worry most about round lakes that show seismic activity not seen before….

    13. I do have 1 question that is actually about this post, though it also applies elsewhere too. It has been said quite often that the volume of an eruption can be found going off the amount of missing material on a volcano if it is a really powerful explosion. However shouldnt the term eruption only refer to the new magma and not the material blown off the existing cone? This is something ive been wondering, because a lot of very powerful explosive eruptions especially if they are phreatomagmatic tend to include a lot of old material so the volume of new magma might not be that big compared to the volume of tephra, kilaueas 1790 eruption was about 50% wall rock except at the start and end which were all magma, and a lot of maar volcanoes can sometimes even be made entirely out of the base rock and have no new ash except a thin veneer on top and some lava in the crater tgat often gets buried by sediment later.

      This is more a question about the part where it is attributed that the missing volume between two of the expeditions to krakatau is attributed to a VEI 4 eruption, that missing cone could have been lost to a landslide and then partly filled in with a much smaller volume, very similar to last year. Support of this is from last year actually. Currently if you look at anak krakatau you wouldnt actually immediately think it collapsed and was formerly a small stratovolcano, it just looks like a normal tuff cone because of subsequent eruptions. We know better because it was observed but that was not the case in older times. Last years eruptions after the collapse are labeled as VEI 3 but including the volume of the collapse as part of the eruption makes it a solid VEI 4.

      Obviously this has way bigger implications for the enormous collapses that happen on oceanic shield volcanoes, if mauna loas southwest rift falls into the ocean again would it be included in the eruption that is (probably) required to initiate it? Because if that is the case you get 10,000 km3 of rock… technically VEI 9…

      Really all this is about is whether eruption induced collapses should be considered in the eruption volume.

      • That is a good question (and glad someone asked a question..) In effusive eruptions, only magma can erupt so the output is close to the DRE input. Fountaining does not change that as those happen inside the melt on the surface: the pressure is not enough to break rock. In an explosive eruption, it really is an explosion and it fragments the cold rock. The mountain is blown to ‘smithereens’. The magma provides the energy and the volatiles (pressure) but how much of the ejecta were molten can vary a lot between eruptions. At Krakatau 1883, initially it was a minor part, I think, as little lava was reported. Once the top has blown off, the heat of the magma is exposed and now you get this smoke column with ash, but it doesn’t add much volume. You can also get further explosions (a lot of them, at Krakatau) and these are now (probably) more magma and less cold rock. In Krakatau, the new islands that were formed from the final explosion steamed for weeks. So they must have been hot. But after the May 1883 eruption, people walked on the ash just fine so it must have been cold.

        A landslide is not really a VEI event. I would not call Anak Krakatau a VEI-4 just because of the collapse volume. Although the collapse happened during a minor eruption, I would consider it as a separate thing. If I am right about Krakatau 1883, a flank failure triggered an underwater explosion, so that is again a VEI event. In Anak, there wasn’t much indication of water-magma interaction: the magma chamber was never exposed. Neither did the chamber respond to the change in weight which I find interesting. It just died (before coming back to life last week). In Krakatau 1883, the big explosion was also the last one so it removed all trapped volatiles. Afterwards there were some collapses but no more volcanic activity.

        • Interesting. Regarding labdslides, it could be different in some cases though, regarding massive collapses. The interiors of the really huge volcanoes will be hot, but especially so for hawaii. The rift zones are magmatic temperature, and the parts further out are glowing to some extent until within 1 km of the surface at least on kilauea, the same is probably true of mauna loa maybe even more so, and removing the flank of the volcano will expose this to the ocean, which will also be racing back as a massive wave. Technically its not new magma but there must gave been some almighty steam explosions associated with the alika collapse events, same thing on a slightly smaller scale with the collapses in the canary islands, and reunion. I guess that is a different circumstance to what happened last year at anak though, but even tgen there was a massive thunderstorm above anak while it was erupting so there is some real heat there.

          The collapse at anak last year was probably not very deep, only the visible part and a bit underwater falling down, already the area towards the caldera has built up a lot above sea level and far offshore of the vent, so its probably not very steep anymore on tge offshore slope which means future collapses are less likely? Anak is really not very big still, maybe 1 km3 at most, it probably isnt really big enough to cause decompression eruptions from collapses. I expect it will carry on as normal soon enough.

          • The collapse volume (0.2km3) is not enough to make a dent in the hole in the sea! The crater has re-sealed itself but I think that the shallow bit coincides with where the island used to be. The new volcano may be just as unstable as the old one was.

            The Kilauea rift zone is very hot, and if this becomes exposed to trapped sea water, you will get a big explosion with a decent VEI rank. If the water is not trapped you get big geysers but not much VEI, I think.

            Not sure that the thunderstorm was related to the collapse. There are a lot of thunderstorms there at most times.

            • If Kilaueas seaward flank collapsed Into the sea it woud cause an enormous tsunami

              And pretty much a major VEI 7 pheratosurtseyan plinian or short simply massive explosion
              Lots of hot ( above 1300 C ) rubble and magma woud get quenced by the sea resulting in the most violent holocene sea eruption ever
              A very very very unlikley scenario

            • It would be interesting to test this idea of massive collapse induced eruptions with prehistoric collapses at other islands to see if such large scale eruptions happened afterwards. We already know seemingly harmless volcanoes have often rather devastating eruptions in the presence of water. Krakatau is probably one of the best cases, and the best observed, though in this case it is not basaltic. I dont know if it was intended this way but it sounds almost like the caldera was actually made by the explosions rather than collapse, most VEI 6 volcanoes dont tend to erupt soon after a caldera event but krakatau was probably erupting again before 1900 and has been steady since, and the original vents are still sort of these as you say.

              While on the topic of massive collapse eruptions, east molokai volcano, the biggest volcano on earth at that time, was basically beheaded by its collapse. Unlike the similar sized collapse of koolau volcano on oahu around the same time, east molokai collapsed when it was still active, when it was similar to mauna loa now. Most of its rift and a large part of its summit was removed in the event. That must have been a monumental eruption, even if the volcano wasnt hyperactive at the time its interior would have still been hot. Kilaueas magma system is potentially hundreds of km3 including the now generally accepted deep rift, and kilauea today is a lot smaller than east molokai was at its peak. Normally these deep rifts are non-eruptive and left over from when the volcano was smaller but it is still magmatic temperature olivine crystal mush. Exposing all that in a collapse would be like having 400 km3 of basalt magma turned to tephra, this is potentially a VEI 8… Even without water, instantly decompressing 400 km3 of basalt magma wouldnt be a gentle event.

              No idea how viable this is at all but it is an interesting parallel to the points in the article, and its only human to take an idea to its limits… 🙂

            • The original extension of the Wailau slide is not known so it is not clear wether it affected the summit and rift zones or not, but the main problem with that idea would be that the resulting explosive eruption should have mantled Oahu in a thick tephra layer and such deposit doesn’t exist. Also the slide will remove maybe the 2 or 3 upper kms of lava pile but that won’t be enough to reach the deep rift which is also a mush of crystals that probably doesn’t contain much volatiles so it is not going to blow up as easily as a plug of fresh gassy magma in a stratovolcano.

              In the Canary Islands some landslides managed to dismantle the rift zones (the Cumbre Nueva rift zone of Taburiente volcano for example was partly destroyed during the Aridane slide), this is easier in the Canary Islands because of the steeper slopes and smaller size of the volcanoes, In Hawaii I doubt any lanslides were able to destroy the summit area of the volcanoes, later erosion did it. After Aridane part of the deep rift of Cumbre Nueva came out effusively as picrite basalts probably remelted through descompression but I am not aware of any evidence indicating that a large explosive event took place. This is also similar to what happened after the double Micheque-Orotava collapse in Tenerife that dismantled the NE rift zone, again no explosion seems to have happened there.

            • I think we still have a lot to learn about how the flank collapses happen, what causes them and what the effects are. They are the most dangerous volcanic events. To predict them better, we need to understand them better.

            • I agree with that, flank failures are very frequent, most stratovolcanoes and oceanic shield volcanoes ongo cycles of rebuilding and lateral collapse multiple times throughout their lifetime, with a much higher frequency in stratovolcanoes. There are probably different possible triggers, intrusions of magma into the system, adding weight to the flank (possibly what caused the collapse of Anak Krakatau?), large earthquakes (the Hilina Slump had a small rupture caused by the 1975 M 7.7 earthquake).

              Regarding large oceanic shield volcanoes the collapse I think is usually dry, at least in the Canary Islands that seems to be the case, in Hawaii this is still open for discussion but certainly explosive eruptions on the VEI 6-8 range should have left some trace on the older islands, that suggests that any explosive interaction with water was “small”.

            • Looking at molokai it does look like at least some of the caldera was taken by the collapse, as well as some of its east rift, but not the main part. It is very likely however that there was some magma exposed in the process. Such eruptions might happen under significant water, which would prevent large plumes so the eruption stays somewhat localised despite its size.

              Regardless of this, there is one slide in hawaii that we know definitely took out a rift zone, the alika slides on mauna loa. There is nothing downwind of there, so if anything happened it would be at the bottom of the sea now, This was long enough ago that any evidence of it on land would be buried anyway, the only parts of the big island older than those slides are on kohala, way upwind. The current part of mauna loas deep rift that is within the estimated area of the alika 2 slide is about 150 km2, and if the rift is as deep as the one at kilauea (7 km) that is a volume of about 1050 km3, likely the rift is much deeper at mauna loa on account of its size and age, not that it really matters much. As you say it probably wont be exposed for the most part but I disagree with it being degassed, at those depths there will still be some volatiles that cant escape. Fissure 8 was probably related to the deep rift and it was very rich in SO2.

            • Turtlebirdman 😉

              Do you think Halumi’alumi’namnam will erupt again, I sooo much loved it 🙂

            • At the time of Alika 2 the SWRZ of Mauna Loa was located more to the south and east (the so called Ninole Rift Zone), I am not so convinced then that the landslide touched the rift.

            • The landslide scarp also cuts back through beyond the crest of the current rift, it is likely it exposed a lot of the ninole rift, it was enough to get the rift to relocate, which is not trivial and in hawaii is much more difficult than it appears, hilina pali is close to a rift but radial eruptions inside it will never happen because it goes against the grain, so to speak. Same thing would have applied back then but eruptions did start happening in the slide scarp.

            • Turtles, I so agree!

              I love the Hilina, especially to go swim outside with all the honus.

          • Turtlebirdman if the African Superplume was out in the pacific ocean
            Woud it do a Hawaii, Tamu Massif or an huge oceanic LIP ? Ontong Java Plateau 2.0

    14. After reading all three articles, I have to say: What a monumental feat, Albert. Best I’ve read on Krakatau by far!

    15. After a couple of days relative quiet on the Big Island

      3.0 10km SSE of Pahala, Hawaii 2019-05-23 06:03:58 (UTC) 47.7 km
      2.3 9km SE of Pahala, Hawaii 2019-05-23 05:58:15 (UTC) 48.5 km

      As seen at Mauna Loa

    16. Flicker is down so I cant share the image. More deep tremor on the big Island associated with these quakes.

      2019-05-23 06:03:58 3 47.7
      2019-05-23 05:58:15 2.3 48.5

      This can be seen on many of the seismographs. If someone could look at HUAD (Hualalai’s seismograph) and explain if this instrument is recording more of the tremor that looks to start at around 04:28 and continue through the two above until 07:00 ish. Times are UTC. Still looking for any other instrument that shows the same signal.

        • Squonk

          It almost looks like due to its location or sensitivity it picked up more of the event?
          Thanks for the image

          • The signal at HUAD starts more than an hour before the tremor picked at the stations in Pahala so I guess it must be some sort of noise that later overlapped with the tremor from Pahala.

      • More tremor


        SWRD – Mauna Loa

        2.3 10km SSE of Pahala, Hawaii 2019-05-24 11:26:59 (UTC) 38.8 km
        1.9 13km NE of Pahala, Hawaii 2019-05-24 11:19:13 (UTC) 26.9 km

        • 2.3 25km E of Honaunau-Napoopoo, Hawaii 2019-05-24 17:22:42 (UTC) 0.0 km

          Close to RCOD and SLC

          SLC Tilt


          • And as often recently with Mauna Loa quakes the manual review isn’t far behind.

            M 2.1 – 26km E of Honaunau-Napoopoo, Hawaii
            2019-05-24 17:22:42 (UTC) 19.412°N 155.620°W -1.1 km depth

    17. Your best series to date Albert, quite stunning!

      Just a tinsy winsy correction. Unless Ilopongo has uprooted and hollered Groot lately while galumphing across borders, it is still in El Salvador and not in Guatemala. I think we have enough volcanoes as it is.

      • Oops. Thanks for the fact check. I have now moved the volcano back to its origin.

        • Good to know that it has not joined you in the quest to run marathons. It would have been a strange sight seeing Ilopongo running New York Marathon.

    18. I was asked about a connection to the Mysterious Island”of Jules Verne. That wasn’t Krakatau, but it may have been based on Graham Island which appeared and disappeared in 1831/32.

      • One day the world really needs an expose of Vernian volcanism. He really loved his volcanoes, and was fairly well versed in it (for his time). I have a nagging feeling that I read about it being a disappearing Island in the mediterranean being his example.

        But, I may have the Island thing backwards.

        Edit/ I had it correctly, as did Albert. I never knew that Isola di Ferdinandea had been named Graham Island. So, if both of us point to the same culprit in the Meds, then it is probably it.

        • Got a decent program on PBS America on the eruption on kilauea in 2018. . On Nova. Its calked Kilauea on fire. Enjoy 😁😁

          • And this had what to do with what? Ferdinandea is not the same as Kilauea.

            This is why we will soon have to clean house again.

            • Fernandina Is remakabely diffrent from Kilauea in structure……
              La Cumbre ( Fernandina ) is much and completly lacks a rift zone. Lavas erupts trough caldera ring faults and flows down the cone.
              This makes the Galapgos shields very very rounded.
              Many short lives eruptions makes the lava stop up at the slopes
              and that explains the rather very steep slopes of some of these shield voclanoes.Cerro Azul is one of the steepest and tallest in galapagos
              Overtuned dinner soup plate shields are galapgos.
              And the calderas can be enromous and seems to mostly result from intrusive drainage and the enromous drained volumes never seems erupt most of it.
              The closest analouge is the marsian and venusian shields that also lacks hawaiis rift zones.
              Galapgos is a quite prolif hotspot with the huge galapagos submarine lava plateau.

            • Galapagos Wolf 2015 is by far one of the most impressive lava fountain events I ever seen filmed and around 500 million cubic meters erupted.
              It completely dwarf Holuhraun and Leilani during the first days
              These wolf curtain of fire si impressive as hell!
              Making it one of the largest lava at once every caught in one single frame

              Here is Wolf 2015 in 1080 P!
              There is no words of how impressive this is!
              Curtian of thoelite fountains thats 100 s of meters tall

            • Carl and Albert are talking about Ferdinandea, not Fernandina, a volcanic island that popped up south of Sicily claimed by a few countries and then rapidly beaten down by the waves, to the spanish the name of Ferdinandea is Isla Julia.

            • Then I am wrong Julia was the french name. I guess there must also be an spanish name for the island since they also claimed it, not that I have been able to find it though.

          • I think it was a mistake in which ‘reply’ button to use. It happens.

    19. Gunung Agung Bali/Indonesia has become more restless again past months. About every 5,5 days (average) new eruption takes place (Jacky Pomeroy, facebook). Todays:

      I wonder how the pattern of eruptions will develop. Last year we all held our breath for an expected big go, based on the volcanoes history. But things turned out very different.

      • I do not really think that was the main event. Could be that this is still just a build up phase.
        That being said, Agung has a bit a too good PR agency compared to actual eruptive oompf.

        • Hmm yes, it’s Bali…
          Nevertheless, some are following intensive and have knowledgable contacts. Fun.

        • I think agung is potentially very hazardous. Whilst the majority of the populstion seems to be safely out of the way when you drive from denpasar inland its rather clear that this is all built on a rather scary pyroclastic flow.

          Hopefully the coming eruption is modest but agung does rather dominate the skyline.

      • I dont think agung is at the stage in its life to do a really big eruption yet. The eruption in 1963 was somewhere on the high VEI 4 low VEI 5 part of the scale, and the lava erupted is not that viscous, it formed a flat lava pancake i stead of a rubbly solid dome. 1963 was the first eruption in over 100 years so it had not erupted in living memory and took people by surprise especially after already erupting for a long time before blowing up, but that almost certainly wont happen now, if an eruption as big as 1963 happens now it wont be nearly as bad either, we know its capabilities and it has been giving ample warning. What could be more of a problem is a flank collapse or an effusive flank vent erupting at low altitude. The lava is hot andesite and basaltic andesite at about 1000 C similar to fissure 17 or hekla so it would very probably flow as thick looking a’a flows as opposed to making a viscous dome like sinabung, but this is flank vent is probably an unlikely scenario anyway.

        Its been a while since there was a decent VEI 4 though, last one was in 2015. We got kilauea to do something that is technically way bigger than any VEI 4, but sometimes only a nice big stratospheric mushroom cloud is the way to go 🙂

        • I hope Mauna Loa will produce tephra, I want to try new tephra in my pineapple-tephra facial masks that I sell in my boutique.

          • If it’s the chemistry you’re after, try sand. Tephra is exploded rhyolite, and one of the governing minerals is silica…

            Ground pumice might make a good analog.

            • Sand is becoming scarce in places. Sumatra has lost 24 islands which were there recently and are now gone, dug up (illegally) to provide sand for China and Singapore.

            • Beach sand is not abrasive enough to exfoliate the skin. And it contains salt and horrible poisons from skin lotions here in Hawaii.
              I used ash from the March 24 2008 eruption of Halema’Uma’U, but then the bad government fined me. Whopsies!
              Now I use imported ash from Japan that is legal.

        • 1963 was a VEI 5 and bigger than St. Helens.

          Not Pinatubo though.

            • I used 0.28 km3 DRE but that is for the March 1963 eruption only. Adding the May eruption brings it up to 0.4 km3 DRE. (In fact the paper you quote also states 0.4 km3.) A year later, a plane still reported ash at 20 km high. We can safely assume that the ash cloud reached at least this high. And Agung had a worldwide climate effect, unlike st Helens. Location in the tropics helped, and it may have had more sulfur, so this is not only dependent on eruption size. But it seems likely that Agung was underestimated rather than overstated.

    20. Agung seems to be erupting weekly now.

      Here is another dated 24 May 2019:

    21. Here’s another view. He is quite the rock-chucker.

      • Thanks for that video – it’s much clearer that the other one I saw!

      • Turtles, what do you think would give the best exfoliation in a Hawaiian facial mask, Strawberry, Kiwi or Guava? I am trying out a new recipe inspired by Leilani. Toodles!

      • Amazing! thanks for sharing! I had been looking for a video of the Mauna Ulu eruption for quite some time.

        Also, I knew some lava devils/lavaspouts had been spotted during Mauna Ulu but the size of the second one shown was incredible, that, combined with the next clip of the cascades into Aloi crater either shows how light the surface of a lava lake/open channel is or how strong the updrafts generated by the lava are, probably both.

        • Yes I was also very surprised to see the lava tornadoes, and the lava blowing off the lava falls. Baring a large ring fault eruption near Bardarbunga it will be a long time before something like this happens again.

    22. Murray Gell-Mann has died. The last (and perhaps the first) of the great particle physicists. Much of the theory of quarks (including the name) is his work. His big achievements were in the 1960’s, and he build it up so far that there has been little progress since.

      He is also know for working and clashing with Feynman. Even in physics, personalities matter

    23. Pretty good-sized eruption of Popocatepetl yesterday:

    24. Here is Anak Krakatau, on May18, imaged by Sentinel

      And here it is on May 24, after the minor eruption. This is also Sentinel but radar (VV polarization to be precise). It looks to me like the crater lake is partly (mostly) filled in. Anak Krakatau may be close to resurfacing.

    25. [meta] I’m not sure what you’ve changed, but lately Ctrl+End does not work correctly on this site to scroll instantly to the bottom of the page. Instead it scrolls partway down, but often only a short distance, a few PgDns worth perhaps, though sometimes a larger chunk.

      The problem, whatever its cause, is getting steadily worse as well. When this first started a week or two ago it sometimes took two Ctrl+End to jump to the bottom and sometimes still only one. Now it’s typically six to ten Ctrl+End keypresses needed and never just one.

      The likely cause is some Javascript change. Why there’d be any scripts that mess with basic browser UI behavior in the first place, I don’t know, but one of the scripts intercepts the Ctrl+End keypress now and changes the scroll distance that it produces.

      Please put it back to normal so that Ctrl+End does what it did before, and what it still does on literally every other website and in every other window with a vertical scrollbar in existence: scroll all the way to the bottom immediately in one fell swoop.

      • I tried this now, but for me ‘CMD down’ (Mac) does go to the end, and the reverse to the top, even n this mammoth post.

      • There’s nothing we’ve changed on the site, unless a wordpress update changed something. If it’s working on other PCs I would try with clearing your browser cache or trying a different browser. It could also have been a Windows update at your end that changed something?

    Comments are closed.