Wouldn’t it be nice to be able to forecast major eruptions? So far, the opportunity has been lacking. VEI-7 eruptions are rare: we have not seen one since Tambora in April 1815. To make a right forecast, the event needs to happen!

Part of the forecast is recognising that the mountain is a volcano. Tambora had not shown activity until around 5 years before the climactic explosion and it had not been considered as an active volcano – or even an inactive one. And it had not attracted much interest: although reportedly the second highest mountain of Indonesia, we have no drawings of it prior to the eruption, nor did anyone even bother enough to measure the height. In contrast, Santorini’s major explosion (perhaps not quite a VEI-7) came with warnings, sufficient for the island to have been evacuated well before it exploded. And that was more than 3000 years ago!

Has volcano forecasting made any progress since? That is hard to tell. It requires that there is another VEI-7, which we either managed to forecast or completely failed to do. The signs so far are mixed. Among the smaller (but still major) events, Pinatubo can be considered a success: the signs of an impending eruption from this inactive volcano were picked up, and evacuations prevented a major disaster. Krakatau too seems to have given some advance signs. Verbeek was worried about Krakatau 5 years before it made history, but he did not write down what made him worried. On the other hand, even in recent years Hunga Tonga came as a complete surprise.

The VEI scale

It may be useful here to explain the VEI scale. It stands for ‘Volcanic Explosive Intensity’ and it is meant as a measure of the power of an eruption. It may be compared to a waterfall. The flow rate would tell how much water passes the waterfall per second, but the power would come from how far the waterfall falls. It makes a lot of difference whether the height is 1 meter or a Niagara-like 100 meter!

The power of an eruption is not trivial to measure. The original VEI classification used a range of criteria, of which only two have survived: the height of the eruption column and the ejecta volume. The height ranges from >10 km for a VEI-4 to >40 km for a VEI-7, while the volume goes from >0.1km3 for a VEI-4 to >100km3 for a VEI-7, in steps of a factor of 10.

Neither measurement is fully precise. The height of an eruption column depends on the amount of heat, as this is what drives the column upward, but it also depends on atmospheric conditions and on water content in the eruption. From the column height, Hunga Tonga would have been classified as a VEI-8! And of course, this number is only available for recent eruptions. For Tambora we can make a guess based on how far the ejecta traveled, but we have no other record.

The ejected volume is measured from what falls on the ground. It is not what the volume was of the rock and magma before the explosion. It is like snow and water. Tephra is low density, like snow, so takes up more volume than the original rock. The difference becomes extreme once pumice is considered, which is so airy it floats on water. Over time, tephra compresses and the measured volume may go down. There are other problems. Part of the ejecta may be lost to measurement, ending up in water (as it did in Hunga Tonga) or falling back in to the caldera. Different papers may quote significantly different values for the ejecta volume of the same eruption, even when based on the same measurements.

Where a crater is found after the eruption, that can be used as an alternative to the ejecta volume. When VC assigned a VEI-5.9 to Hunga Tonga shortly after the eruption, that was based on the apparent size of the crater, not on either of the two official indicators. But crater size also has its problems. Santorini is often assigned VEI-7 based on the size of the caldera, but part of this caldera already existed before the eruption.

The ejected volume does not take into account how long the eruption lasted. That makes a difference! Effusive eruptions can result in voluminous lava flows, but at much less power than an explosive eruption of that size.To use our previous example of the water, it forgets about the height of the waterfall, which was the whole original purpose! (That is why tephra volume is used rather than lava volume: it takes power to rip rock or lava apart.) The original scale did use the duration of the eruption: explosive eruptions take place in hours, but effusive ones may last weeks to months. However, duration is not normally known for ancient eruptions. Often, the volume is the only number we have.

Still, a VEI-7 is vastly larger than a VEI-6, so the confusion mainly applies to eruptions where the ejecta volumes are within a factor of 2 or so of the boundary between classes.

Make it a seven

What turns a decent eruption into an indecent one? I think Verbeek in the 19th century was the first one to mention this problem, when he noted that Krakatau’s eruption in the weeks and months before the big explosions was notable, but not significant – nowadays we would have called that phase VEI-5. What caused the sudden shift to the whole mountain blowing up? He did not know, and it is still being argued.

The formation of a caldera is now seen as a collapse event. If you can still remember those ancient vacuum tubes (younger readers: don’t worry. One of those things older people feel embarrassed about): when their structural integrity failed, the tube collapsed – and exploded. In recent years, the Titan submarine recreated that mistake when the owner felt that damage to the hull would not matter if it was ignored. (A not uncommon attitude. I noticed when moving to the UK that train companies would only top up the antifreeze after the frost arrived. The first day of heavy frost, train cancellations were a certainty.)

Back to the present. There are two main version of collapse models for VEI-7 eruptions. Both start with the same requirements: a large magma chamber with sufficient volume to go for a VEI 7, and something that sets off the chain of events, i.e. a trigger.

One model begins with a normal eruption which runs down the magma chamber. At some point, the internal pressure in the magma chamber become too low to support the roof of the magma chamber. The magma chamber collapses, and, like the Titan vacuum tube, the collapse causes a sudden major explosion which rips the mountain apart. Krakatau may be an example of this. The second model is quite similar, but does away with the initial eruption. Some event weakens or fractures the roof (such as, an earthquake, as in the case of Pinatubo) and the collapse comes from that, after some delay. There are many variations of this, of course. Flank collapses can be significant, as in the case of St Helens (albeit a much smaller eruption). And since Hunga Tonga, the role of water as an amplifying mechanism is being looked at. This, of course, is mainly an issue in volcanoes with access to sea water, but these form a large group which are (or were) overlooked.

It has become clear that a large eruption is not just a scaled-up version of a smaller one. This is even more the case for super-eruptions (VEI-8) which seem to be very different beasts to the run-of-the-mill VEI-7’s.

Recent events

How often do VEI-7 eruptions happen? That can be estimated from the list of known large eruptions (see below). Based on these, and allowing for the fact that the list will be incomplete further back in time, a rate of 1 every 500 years seems reasonable. The main uncertainty comes from assigning which eruptions reached the VEI-7 threshold. The list below gives the likely VEI value, but it includes high VEI-6 values: because of the uncertainties, some of these may in fact have been VEI-sevens. Santorini is in the list, but this is unlikely to have reached near the threshold. This list is from 2018 and therefore lacks, for instance, Okmok’s eruption in 43 BC which was likely around VEI 6.8, whilst for Ilopango, the date has changed to 431 and the volume is now above the VEI-7 threshold. The eruptions of 536 and 540 are not in the list as no culprit is yet known for either and without a culprit no volume can be determined. Given that there is incompleteness even for eruptions as recent as the early middle ages, we may be underestimating the frequency, especially when including eruptions close to the threshold.

In the last two millennia, the large eruptions have been in Indonesia (twice), Papua New Guinea, the Pacific basin (Kuwae), Central America (Ilopango) and New Zealand (Taupo). Go back a bit further and the northern Pacific comes into view, with candidates in Japan, Kamchatka and the Aleutian islands. VEI-7 eruptions can and have happened in most volcanic arcs!

Back to the future

Can we say something about which volcanoes may be candidates for future VEI-7 eruptions? Where should we be spending our monitoring money? (Science funding may not be a priority for governments, especially for problems that may be left until after the next elections.) Sadly, we don’t know much about what a volcano looks like just before such an eruption. Neither Tambora nor Rinjani has left us much of a historical record, probably in part because so few people within range survived. There are no written or oral records of similar eruptions further back. Kuwae, Tianchi/Paekdu, nor Ilopango have left us eyewitness accounts. (Taupo’s eruption is excused – it predated any human presence in the region.) Even depictions of volcanoes prior to eruptions are very sparse. For Vesuvius’s eruption in 79 that destroyed the Naples region, the only depiction we have comes from one rough wall painting which clearly used considerable artistic license and which contradicts written documents describing fields. Inferences can be made: Tambora can’t have been too impressive or passing sailors would have made drawings and used it for navigation. We are not looking for winners of a beauty contest!

If not its appearance, and the size of the magma wallet is not immediately obvious, what should we been looking for? Newhall et al. (2018) lists 6 criteria for a volcano to become a candidate for a future VEI-7.

1. A history of such eruptions. Volcanoes doing one such eruption are likely to do it again. And again. They are repeat offenders that should not be trusted. In these volcanoes, VEI-7 eruptions may occur between 10,000 and 100,000 years apart. That fits with typical magma resupply rates of 0.01 km³/yr, which would allow the magma chamber to refill after some 10,000 years.

2. The volcano has produced silicic magma within the past 100,000 years from different vents. The silicic magma can show as rhyolite or rhyodacite, evidence that the magma has time to evolve between eruptions. The different vents should show indications for an arc or circle around the magma body, not just follow a straight fault line.

3. The volcano should be a mature stratovolcano (not a shield volcano) and may be part of a cluster of such volcanoes, which has been relatively quiet in recent millennia. Clustering of volcanoes is not uncommon. Samalas is an example of a VEI-7 volcano which left the remnants of another stratovolcano (Rinjani) on its caldera rim.

4. The volcano should leak only a limited amount of gas compared to the supply rate of new magma. This indicates that the conduit is blocked, and the magma chamber will accumulate volatiles which eventually will form gas bubbles in the magma. Evidence for this is found in eruptions from El Chicon and Pinatubo, both of which produced considerably more sulphate (a volatile) than should have been in the magma. This excess gas may drive the eruption towards a VEI-7, while an eruption with less volatiles would stall earlier.

5. Volcanoes with a higher supply rate of magma from the lower to the upper crust are better candidates. The magma arrives in the lower crust as mafic (basaltic), becomes more silicic as it ages there, and is transferred to the upper crust as silicic magma. These higher rates can happen where there is a hot spot, but in general occurs where the crust shows some local extension, as for instance in Taupo.

6. Recent unrest may increase the chance of a VEI-7 eruption. This may actually be seen as in contradiction to criterium 3! It asks for a volcano that is both quiet and restless. But Tambora showed increased activity from around 1810, 5 years before the event when it acquired a summit cloud which would not disperse. The eruption did not come entirely out of the blue.

Based on these criteria, Newham et al present a list of candidates worldwide which could have such an eruption in the future. It is a long-range forecast: there is no timescale associated with it, although if it takes more than a million years the volcano will likely be fully extinct and no longer at risk. Neither is it a prediction: it is a list of volcanoes that have a chance of doing a VEI-7, not a certainty. The authors neither predict nor schedule, just warn of the risk. Neither do they go for short-term forecasting, other than mentioning that after a period of unrest or small eruptions, the run-up to a big event may only take hours to a day.

No doubt the readers will want to peruse the list of volcanoes-at-danger. The paper can be found at
https://pubs.geoscienceworld.org/gsa/geosphere/article/14/2/572/529016/Anticipating-future-Volcanic-Explosivity-Index-VEI It is an open-access journal, you will be happy to hear. The list is in Table 3, which can be found from the list of data and figures next to the paper, or much more conveniently in the pdf version of the paper. For the curious but more time-constrained, I have appended a full list of names at the end of this post, but this does not include the other information on each volcano in the original paper.

Indonesian calderas

Two of the most recent VEI-7 events occurred in Indonesia. They also came from the same region of Indonesia: the islands east of Java. From the summit of Rinjani, it is possible to see Tambora: the two events occurred only 100 miles apart. Does that mean anything? Or was it just a coincidence? What is the risk of a third event from this region? The Newhall paper lists 18 potential future VEI-7 candidates in Indonesia, but only a few of them are in this particular region. Is the area around Rinjani and Tambora at higher risk? Have there been more such eruptions here?

Large eruptions cause large holes. A VEI-7 should create a caldera of 5 km or more across. That should be easy to find! So one way to look for likelihood of such eruptions is to look for the calderas of past events.

But that is not as easy as it sounds. Indonesia can change fast. In the tropical climate, erosion can be rapid and holes get filled in with water and sediment. If several large eruptions have occurred, then the caldera may be composite and badly distorted for its original round shape. Tropical vegetation can hide much. And of course, calderas are volcanic and new volcanoes may grow inside, obliterating the evidence of past misdemeanors.

A quick trawl through the landscape reveals some calderas of different appearances in the region, where I start from East Java and end in Sumbawa. This list is not complete! Additions are welcome as long as no volcano gets damaged in the process. Note that I did not look at the western half of Java and Sumatra, or the islands from Flores eastward.

Tengger volcanic complex, Java

The Tengger complex in eastern Java consists of a set five volcanoes, including the oft-active Mount Bromo. The group is located inside an old caldera. The caldera wall is still some 500 meters tall; the caldera itself is filled with a sea of sand – literally so as it is called Tengger Sand Sea or Sand Sea Caldera (Segara Wedi) – which surrounds the set of volcanoes. (I do not know what a set of volcanoes is called: a ‘team’, a ‘giggle’, or an ‘eruption’?) The sand derives from eroded lava from Mount Bromo. Outside of the caldera is another well-known, active volcano: Mount Semeru.

The Sand Sea Caldera measures 8.5 by 10 km. It formed through a series of caldera-creating eruptions, starting 150,000 years ago. The crater created in totum measures 16 km across. The first event was caused by the collapse of the oldest, Ngadisari volcano. A second volcano in the region, Tengger volcano, was next to go, more than 45,000 years ago. Other volcanoes grew in the caldera, each collapsing in turn: one ignimbrite is dated to 33,000 years ago. After the most recent of the major events, the current set of five cones grew, which remain active. One VEI-4 eruption is dated to 3000 years ago but more recent activity has been limited to frequent VEI-2.

Just outside the caldera lies Semeru, Java’s highest mountain and a very active volcano. It is closely related to Tengger and is seen as a high-risk volcano because of frequent eruptions, lahars (it can be very wet here) and major populations on the low plains to the south. In view of its frequent eruptions, it is not on the VEI-7 danger list, but Tengger is.

There is another caldera in the region, Jambangan caldera located north of Semeru. It is little studied but the mountain Gunung Ayek Ayek seems to be part of the northern caldera wall. The caldera appears to be some 4 km in diameter (north to south) and is younger than 45,000 years.

Further to the east lies Lamorgan volcano, 1600 meter tall. It has no caldera associated with it. Notably, this volcano was a frequent eruptor in the past but has been quiet since 1900. It is not a VEI-7 risk but worth keeping an eye on.

Ijen caldera, Java

This caldera lies near the eastern edge of Java and measures an impressive 15 by 18 km. Kawah Ijen volcano lies in the caldera: it is known from its highly acidic crater lake and the blue sulphur fire. The most recent eruption was in 1817, apart from some phreatic explosions and geyser activity. Gunung Raung is nearby and is much more frequently active but is not related to Ijen. A large number of other volcanoes lie along the caldera rim, all now inactive. This is an example of a volcano cluster, where one or more members had a large blow-out.

The caldera is in some ways the opposite of Tengger: without internal volcanic destruction, it is a fertile region, used for growing coffee.

The age of Ijen caldera is not known, surprisingly. Post-eruption deposits from the rim volcanoes have dates up to 50,000 years ago, so the caldera must be older than that. An upper limit is 300,000 years. That leaves a large range! The eruption is attributed to a stratovolcano; the volume of the explosion is estimated at 80 km3 (DRE, presumably, given the size of the caldera), a major VEI-7.

Buyan-Bratan, Bali

In the centre of Bali is a caldera measuring 6 by 11 km, where the shorter (north-south) size is not well known because the southern part of the caldera has been obliterated by younger volcanoes. There are 6 of these, ranging from Tapak in the centre to Batukaru furthest south. There are three lakes along the caldera rim. The caldera is also known as Bedegul, the name of a geothermal field.

The age of the caldera is not known, apart from it being older than 30,000 years since the Batur eruption next door deposited debris inside this caldera. The shape suggests that this may be a composite caldera, created by two explosions a few kilometers apart. Studies of the region have focussed on the geothermal potential, not the ancient origins, so we do not know whether there were two explosions and what their sizes were, but at least one is likely to have reached VEI-7.

The region has not seen volcanic activity in recorded history; in practice, this means the last 200 years. However even the new volcanoes show no signs of damage to the vegetation, nor do they exhibit recent lava flows. This suggests the quiescence has been going on for much longer than 200 years, although in the tropics it is hard to tell for how much longer. Buyan-Bratan is on the list of possible future VEI-7 eruptions. And it is also covered in our NDVP series, where Bali was number 6 on the list.

Batur caldera, Bali

The caldera of Batur is 14 by 10 km. It is beautiful area, with a large lake nestled against the caldera rim. It is Bali’s most visited volcano, being the most accessible of them. However, unlike Buyan-Bratan, it remains volcanically active. A new volcano has grown up at the centre and it has erupted many times since 1800. The last activity was in 2010, but limited to the lake.

There are two collapse structures. A smaller caldera is located inside the outer one, with the same centre but about half the diameter of the main caldera. Thus, there have been two large eruptions here. The first one has been dated to 29,000 years. It erupted 84 km³ of ignimbrite and left the 1-km deep caldera. The second explosion was smaller but still substantial at 19 km³. It occurred 20,000 years ago and formed the lake located on one side of the caldera. A third eruption 5000 years ago produced 0.1 km³ of andesitic ignimbrite. For a large eruption, we may guess that the ignimbrite contains half the total erupted volume. This would make the oldest eruption a VEI-7 and the second one VEI-6.

These are the two known large calderas of Bali. The map above shows indication for two more, one just northeast of Buyan-Bratan and one on the southeast coast – the latter is more tentative. Both appear to be much older than the others.

The next island to the east is Lombok. It is known for one main tourist attraction: climbing Mount Rinjani. If they had done so 1000 years ago, the view would have been very different.

Rinjani caldera, Lombok

Rinjani remains the second highest volcano of Indonesia, but it is half the mountain it used to be. Where the missing half once was is now a caldera, measuring 8 by 6 km. Ash from this caldera has been found in Greenland and Antarctica. There are Indonesian documents that describe the eruption, albeit based on oral histories which were written down centuries later. The eruption occurred in 1257.

The mountain that blew up has been argued to have been over 4 km tall. However, a smaller size more similar to Rinjani seems just as likely. It is now called Mount Samalas – but named so only after it ceased to exist. A small volcanic peak inside the caldera is occasionally active. It has filled in part of the caldera lake.

There are two further apparent collapse features on Lombok. One is Gunung Punikan, a much smaller mountain west of Rinjani and which sits on the rim of a caldera a few km across. Nothing seems to be known about it. The second one is Sembalun caldera, east of Rinjani. It is half enclosed by a mountain ridge (including Gunung Sempana, peaking at 2300 meters) but is open to the east. The size is similar to the Rinjani caldera. The age of this feature is not known

The next island along is Sumbawa, the last one on our exploration. It is known for hosting Tambora.

Sangenges caldera, Sumbawa

The Sangenges caldera, in the west of Sumbawa, lies well hidden in a dense forest. The peak of Sangenges volcano lies on its southern edge. It is considered extinct but the age is not known.

The caldera lies to the north of the Sangenges peak. The circular rim is clear visible and is some 6 km across. To the north the caldera opens into a valley.

Tambora caldera, Sumbawa

The most famous caldera of Sumbawa is that of Tambora. The explosion of 1815 left a deep caldera measuring 7 km across. It is beautifully circular. We know very little about what the mountain looked like before the eruption: an estimate for its original height is 3700 meters, but values well over 4 km have also been claimed. The lack of knowledge suggests that it wasn’t exceptionally tall, giving more credence to the smaller value. This is the best studied VEI-7 eruption we have and we still know remarkably little about it. There is more background on Tambora’s history elsewhere in VC.

Ntoke, Sumbawa

Maps show a third large caldera on the island, around the small village of Ntoke. The mountainous rim around it has Doro Pundunence as its highest point. The ridge is open to the north. The caldera has about the same size as the Tambora caldera but is deeply eroded.

The area

This gives 8 calderas in the general area, with ages ranging from 200 years to (probably) more than 100,000 years. That indicates around one such eruption every 10,000 to 20,000 years or so, which is pretty much consistent with what ages we have. If this is correct, then the chance of having two of them within a 1000 year period is slim at best. There may be more calderas hiding underneath younger volcanoes: there may have been more large eruptions from these locations. In fact, the 6 rules listed above suggests as much.

The Newhall list contains 7 volcanoes-at-risk in this region which is about 5% of the full list. The rate of one per 10,000 years for a region containing 5% of the list gives a rate for the full list of one per 500 years, pretty much the actual value (to the best of our knowledge!). That suggests that this region is not at an enhanced risk: the rate is as would be expected from the number of candidates.

The chance of a new VEI-7 eruption from this region in Indonesia is not particularly high. There is only a chance of 5% that the next VEI-7 will come from here. We should look wider.

It is notable that Samalas/Rinjani and Tambora were both tall mountains. It is not known how tall there were, but values around 3500 meters seem plausible. When looking elsewhere in Indonesia, it may be worth paying attention to volcanoes around this size, e.g. Semeru, Kerinci. Especially if they seem somewhat sleepy. However, both of these two are active. That would make Rinjani the volcano most at risk in Indonesia!

Still, the chances are that the next VEI-7 will come from elsewhere in the world. You heard it here first.

Albert, February 2026

Newhall et al. 2018 Skip Nav Destination
Research Article| February 28, 2018
Anticipating future Volcanic Explosivity Index (VEI) 7 eruptions and their chilling impacts

The Newhall list of volcanoes meeting their VEI-7 risk criteria is linked below. You are referred to their paper for more information on these volcanoes: this extract has only the full list of names.

List of volcanoes with VEI-7 potential