Imagine. You are being put in charge of planning for a future major volcanic eruption. All you have is an incomplete list of volcanoes in the world, some experience of past eruptions, and a few volcanologists as consultants. The politician who gave you the task has little idea of what might be involved and although you have a remit, the instructions are rather vague. What would you do?

People have gone here before. There are organisations dedicated to providing assistance to countries before, during and after volcanic eruptions; the US VDAP program comes to mind. Volcano risk assessment schemes exist. They tend to focus on several aspects:

Hazards: What are the dangers associated with volcanic eruptions? This is quite a diverse set, including CO2 and SO2 emissions, volcanic ash, pyroclastics and extending even to tsunamis. Volcanoes can worsen their environment by many different means.

Exposure: Who are exposed to the dangers? How many people, what are their circumstances? Why are they there?

Vulnerability: How badly would an eruption affect them? That may depend strongly on their personal volcano. At Anak Krakatau, the primary vulnerability turned out to to come from a tsunami.

Resilience: Can the local people become more resilient to eruptions? Of course, there are limits to this. In the end, it is better not to live on a volcano. But there are things that can be done, ranging from houses that don’t collapse under ash and have secure water supplies, to the presence of warning systems and signposted evacuation routes.

An example of such a risk assessment is shown below. It lists as main goals in the long term protecting people and assets, and in the short term saving lives. It is hard to argue with this!

This is a good start for your task. But this is designed for smaller eruptions which affect one specific area only. Many eruptions, of course, are like that. But should you focus on those, or are in the end the large eruptions indeed more important?

This problem also exists for earthquakes. Magnitude-7 events which catastrophically damage one or more cities happen regularly, while the largest events, reaching magnitude 9, do much more damage but also occur much less frequently. They require different risk assessment. For the magnitude-7 events, enforcing building standards in cities at risk is probably the most effective way to build resilience. For the magnitude-9 events, tsunamis can cause major damage far from the earthquake. Therefore, a tsunami warning system is required.

We cannot predict earthquakes, but we do know which cities and regions are most prone to them. So the effort can be focussed. Still, in countries such as Turkey and Myanmar, the risks are well known with adequate building standards in place, but even some of the newest buildings do not adhere to them. Standards without enforcement have limited effect. It is Amazon’s Law: few people complain when they are sold something which is a lot worse but a bit less expensive.

Volcanoes are not earthquakes. They are both less and more predictable, have a much wider variety of impacts and can affect people at very large distances. So again, how do you go about this?

Experience

Many countries already have had to deal with these issues. Systems have been developed and they have often been successful. Indonesia is of course a prime example. It has many volcanoes, including dangerous ones and frequently erupting ones (sometimes the same volcanoes), large populations living close to those volcanoes and with a tendency to creep ever closer. The number of volcanic casualties is amazingly small (though not zero), showing that the systems developed in Indonesia are effective. But it can still run into unexpected events. A example of this was the unexpected (though foreseen) collapse of Anak Krakatau in 2018. And the system has never been tested against eruptions exceeding VEI-5: the last such eruption in Indonesia was that of the elder Krakatau in 1883, long before volcano mitigation was seen as a social responsibility.

The VEI-6 eruption of Pinatubo in the Philippines is the ‘before versus after’ event of volcano risk management. The systems were developed on-the-fly during the build-up, and they saved the day with flying colours, although there was some amount of luck involved. The Philippine Volcano Observatory was on the case before the eruption, so the first hurdle (don’t ignore) was successfully passed. Later on, the combination of the local scientists and the USGS prevented a catastrophe. Neither of the two could have done it on their own. But it worked only because of personal risks taken by the people in charge. The ‘system’ without these personal interventions would likely have worked neither adequately nor in time. The reasons for this include the unknown nature of the volcano, the lack of experience with large eruptions, and the fact that although the program of international cooperation was a priority for the people involved, it was seen as a distraction by the USGS as an institution. It was a tremendous success but it could easily have ended very differently.

Lessons were learned from this. But lessons learned have shown that lessons learned are quickly forgotten again.

The current situation is that almost all countries are able to handle their usual volcanic eruptions. There may still be cases where the local resources just do not exist. An example is the Democratic Republic of Congo. The city of Goma has over 2 million people, is located next to a dangerous volcano (Nyiragongo) with people spreading over the fertile slopes, and it is a war zone with little government control of any sort – or any government. The Nyiragongo eruption of 2021 could easily have become a major disaster.

Warnings

The most effective way to reduce the risk associated with eruptions is to have an early warning system in place. This is an advantage volcanoes have over earthquakes: they may be less predictable, but they are also much less likely to occur without pre-warning.

The numerical or colour-coded warning levels that were devised during and since Pinatubo are brilliantly designed. The first phase of a build-up to an eruption is slow and can take months. This phase is used to warn people but does not require any emergency actions which would be difficult to maintain over long periods. As the build-up continues, the warning levels increase while the timescales become shorter. Evacuations are hard to organise over a day, but if lasting over a week people are likely to go back. This timescale is captured very well by the warning levels.

The usefulness of early warnings depend on the trust put in them by both the authorities and by the local population. This trust can be hard to win and is easy to loose. It is important not to make mistakes and not to ring the bells too early. But the usefulness also depends on the cooperation of the volcano – never taken for granted! Pinatubo did oblige. Nyiragongo might not.

Monitoring

So what are the early warning signals that volcanoes themselves send out? The predominant ones are rising surface heat (leading to steam clouds and later phreatic explosions), slow inflation (which may affect the water table) and earthquake swarms. None of these guarantee an eruption! The early warning systems are monitored through seismographs close to the action, satellite radar and cameras. The first and last need to have capability to send the data via mobile signal or something similar. Finally, there is a need for people with time and knowledge to analyze the data.

In a place like Goma, the seismographs and cameras may get stolen, the mobile links may not work, and the scientists aren’t safe. Instead, observations by people visiting or living on the mountain become important.

Other places may lack monitoring for different reasons. At Hunga Tonga, there was a lack of land on which to install the instruments. Some places may be too glaciated or politically inaccessible. In such cases, satellite monitoring is crucial but this may in itself not give adequate early warning.

Prediction and response

After the early warning, the next requirement is that of prediction. What kind of eruption may be expected, and what would be the impacts? This is a bit easier. Knowledge of past behaviour of the volcano is helpful. This is where the roaming scientists come in. There may be clues in the landscape, such as old tephra layers or the hummocky landscapes that indicate a past flank collapse.

Next comes the response. The data may be unambiguous and the models clear, but how do you make people take notice? How do you overcome disinterest, distrust and disbelief? These are all very human ‘qualities’. This was done successfully at Pinatubo but it took a lot of effort and the evacuations were late and in neighbouring cities, left incomplete. At St Helens, foresters were being send back into the danger zone by their employers – luckily the explosion came on the Sunday before the working Monday. In some countries, the current situation with backlash against science might make it harder for people to take their volcano seriously. Links with communities need to be build using people trusted by both camps. Thus is not something that science is traditionally good at.

Failure

In 2018, Anak Krakatau suddenly collapsed – the entire mountain wasn’t there anymore, having slid into the sea unobserved. The collapse is often said to have been caused by an eruption, but this is not really correct. The mountain had been erupting before (as it commonly did) but it was quiet when the collapse came. Anak Krakatau had just grown a bit too much and had become unstable. The ensuing tsunami came out of nowhere and killed 400 people.

It turns out that this event had been predicted some years before, and the models had both the collapse and the tsunami right. But people had not taken notice. Publishing the warning behind a paywall may have hindered the message. Aatellites had observed the slow sliding of them mountain in the months before the collapse – but the satellite data was not analyzed until afterwards.

The scientists had not communicated well, other scientists had not paid enough attention (and in fact, it wasn’t clear who should have looked at the data) and the authorities had ignored the danger. There had been a lack of a coherent strategy. No one was to blame – the danger had not been covered in the system.

Fatalities and survival

The total number of fatalities due to volcanic eruptions is not accurately known. Indirect casualties from long-term, long-distance effects are very hard to quantify. For direct casualties (which may include local after-effects such as famines), around 270,000 fatalities have been counted since 1600. The total is dominated by the largest 5 disasters. Two of those are also among the largest eruptions during this time: Tambora (VEI 7) and Krakatau (VEI 6). Pinatubo might well have been in this list had it not been for the actions of the volcanologists. But the large 1809 eruption is not in the list – we still do not know where this occurred. And the other three are small eruptions with large consequences. The most recent major event was in 1985, when mudflows covered a city at the foot of Nevado de Ruiz.

The numbers listed above are conservative. For instance, for Tambora a total of 92,000 has also been quoted.

Typically, there are 2 to 3 eruptions per year which cause fatalities. Often, these involve people being too close to the eruption. Often, they are caught out by the fact that phreatic explosions do not have early warning signs and can be quite unpredictable.

The fraction of eruptions which cause fatal incidents has increased over the past 50 years. They account for 10% to 15% of all eruptions. It depends on the size of the eruptions. For VEI 2 or less, the fraction is small (though non-zero!) but for VEI 4 or larger, it is over half and for VEI 6 and above it is 100% (though this does become small number statistics, given how infrequent these are!) Notably, the VEI-6 eruption of Novarupta caused 2 fatalities and Hunga Tonga, which was potentially a VEI 6, caused 6: four in Tonga and two in Peru. (Large eruptions can have long-range impacts.) In contrast, Pinatubo caused around 300 direct fatalities. (The full death toll of Pinatubo of around 850 included victims of diseases in evacuation camps and mudflows which continued for years afterwards.) Location is important: both Novarupta and Hunga Tonga occurred well away from settlements.

For large eruptions, the dominant cause of fatalities is not the explosion itself. For Krakatau, it was the tsunami. For Tambora, around 10,000 died directly from the explosion (the number is not well known – the entire local culture was wiped out and there are only estimates of the size of the population within 20 km where there were no survivors). Many more more died of starvation. For smaller eruptions, causes remain varied but important are (again) tsunamis, lahars and pyroclastic flows.

Apart from Tambora, for each eruption the fatalities came predominantly from a single cause, which differs between eruptions. In all these cases (again with the possible exception of Tambora) the majority of fatalities, could have been prevented by a timely evacuation. The most important aspects of volcanic safety are the early warning systems, detailed understanding of the local dangers, an effective emergency plan involving large-scale evacuations, and a trusted system of communication.

People

There will always be people who are attracted to the dangers of an active volcano. Both at Reykjanes and at Kilauea we have seen people walk up to a cone during a quiet interlude, and almost getting caught out by the resumption of the eruption. There is little that can be done about this.

Volcano tourism events have a tendency to continue the tours even when the activity begins to increase. White Island is a good example, and there have also been recent cases in Japan. Volcanoes are well worth visiting and tourism there is a good thing. But it can be difficult to stop an activity that is bringing in much needed money for the local economy.

How about people who live in the vicinity of a potentially active volcano? No matter how well organised a country is, volcanoes remain dangerous. Half of all eruptions that cause fatalities are in Indonesia. The country has many volcanoes, and has many people living near to those – a difficult combination. And not all eruptions give sufficient warnings, and not all people pay heed to urgent warnings.

The most dangerous distance at which to live from a volcano is 8 to 10 km. This is the distance of the largest (by far) number of fatalities. So where are the danger zones in the modern world?

The figure shows the distance between various cities and their volcano. Auckland and Naples are well known for their proximity, although note that for Naples the volcano is Campi Flegrei, not Vesuvius. Quito is an interesting case as the nearby volcano that is listed is not its only neighbour: Cotopaxi and Reventador are also nearby. For typical eruptions, all these cities must be considered to be at risk, although the risk does depend on how eruptive the volcano is.

A common feature for many of these places is that the volcano has long quiet periods, or where there are long periods where the eruptions are too small to affect the city. It gives a sense of security based on living memory. This can be deceptive. To a volcano, a repose period of 500 years is nothing. To us, it seems safety. The recent past may not be a good marker for what a volcano can do. Each one needs to be studied individually.

Handbook

Around 15 years ago, a project called ‘Miavita’ developed the “Handbook for Volcanic Risk Management: prevention, crisis management, resilience” It was funded by the European Union and adopted by the United Nations. The acronym stands for ”Mitigate and Assess Risk from Volcanic Impact on Terrain and Human Activities”. (Scientists and committees may lack eloquency.)

The handbook defines risk as R = H x V x M / C

If this doesn’t mean much, R stands for risk, H stands for hazard, V stands for vulnerability, M stands for the value at risk (I used ‘M’ for money) and C stands for the local ability to cope. Of course, these still need to have numbers attached! But it gives an idea of what is important.

In order for this risk analysis to be made, the first requirement is to understand the volcano. That requires studying the landscape, creating geological maps (yes – these still exist), determining the eruption history, and finding evidence for structures such as collapse features, a caldera, or a river bed with old lahar layers. This gives the hazards.

The distribution of people needs to be documented as well as their living conditions, road networks etc. This provides the vulnerabilities as well information on their ability to cope.

Monitoring systems are needed where volcanic hazards are present, and people are needed to run these. Having a local volcano observatory is helpful. At Pinatubo, both were set up within the months before the eruption. This method of monitoring-on-demand worked well there, because the country already had a national volcano observatory.

Once the volcano and its environment are well understood, risk reduction becomes the priority. This can be short-term (panic mode) or long-term (for volcanoes with no immediate threat). A reasonable forecast for the size and nature of any eruption is needed. Pinatubo seemed to do only large events, so it worked well there. But for a volcano such as Taal with a variety of eruption sizes, this can be a challenging task! But even for a small eruption at Taal, it was clear that people should not be living on the island: their relocation was both a short-term and a long-term requirement. A large eruption would have threatened the region around the Taal lake, but in that case, evacuation rather than relocation would have been appropriate.

Different risks (lahars, tephra falls, toxic gasses) for the same volcano affect different regions, so this can become quite complicated but that is the nature of volcanoes.

Hazards are classified on a scale running from ‘slight inconvenience’ via ‘major financial losses’ to ‘major loss of life’. Worst case scenarios need to be created, although once available these are best used sparingly. For long-term planning, the frequency of the various volcanic events should be determined.

The handbook defined a threat matrix based on the intensity of the hazard and their frequency. At the high end of hazard and frequency (red and beyond), human settlements should not be present. Taal island falls in this category. At orange, people can be present with specific precautions. At the lowest frequencies, it appears that even the highest hazard levels become acceptable. That may reflect what happens in reality, since events that happen once per century leave little awareness in the local population. Still, if there are 100 volcanoes in this group, it still gives a major disaster every year.

Basically, to manage a volcano requires awareness (what it can do), prevention (stop settlements in the most at-risk places), prediction (eruption warning) and protection. But eruptions will happen, and action will be needed afterwards. This ranges from immediate rescue, to the later revivification of the land.

Evacuation requires pre-planning and is best done before any eruption, as transport is likely difficult after a significant eruption. At Pinatubo, one third of the fatalities occurred in the evacuation camps, caused by crowding and poor hygiene.

A major problem after an eruption is ash covering fields and houses. Fields covered by more than 30cm of ash will not recover easily, and become largely sterile. 10Cm can be ploughed in but should not be just left a it severely restricts growth. The ash may also be toxic and can sometimes kill plants (and animals) even with a limited cover. Food aid may be required for an extended period. At Tambora, this was the largest cause of fatalities. Water supply may also be critical.

Road networks are vulnerable especially to lahars. Escape routes should avoid crossing rivers where possible. Lahars can occur long after an eruption, causing a long term risk.

There are many more aspects that could be covered. The risks associated with volcanoes are as diverse as the volcanoes themselves!

Volcanoes are changeable. An eruption can change the mountain sufficiently that the hazard map for future events needs to be redone. St Helens is an example, although this is also a case where the hazards before its main eruption were significantly underestimated. Even the most developed country may have gaps in its volcanic knowledge.

What is the role of the scientists in all this? The general rule is that scientists advise but do not decide. Governments and local leaders are responsible for any actions (or lack thereof). That requires building trust, from both sides. A scientist who attacks governments, local leaders or communities will not achieve much.

The rare events

These schemes may work well, as they largely did (with much improvisation and some luck) at Pinatubo. That was a VEI-6 eruption. But how would you handle a VEI-7 eruption? There hasn’t been one for 200 years, we have no living memory of such events and we know rather little about the run-up. Would we see it coming? And afterwards, could we cope when an entire nation might lie buried and impacts could spread world-wide?

The first step in VEI-7 management should be prediction. That already gioves some problems. Very large eruptions may be more likely in volcanoes that haven’t erupted for a long time. Pinatubo was unrecognized and unmonitored. A VEI-7 capable volcano may not look like a volcano or even like a mountain! For Tambora, we have no information regarding what it looked like before the 1815 eruption. It is often assumed to have been the second highest mountain of Indonesia, based on extrapolation from the remaining slopes, but this seems unlikely given the lack of prior drawings used for navigation, and general lack of attention. Perhaps it had a series of smaller cones, or was deeply eroded like may old volcanoes are. And it wasn’t known to be volcanic, until five year before the main event. This raises the question whether we would be looking in the right places? Or would such an eruption come from a hidden volcano? Did Ilopango once look like Taal?

A VEI-7 eruption may dispose of 50 km³ of magma. Even at a supply rate of 0.01 km³/yr, that takes 5000 years to accumulate. In most cases, the rebuilding may take much longer. Some large volcanoes show evidence of an ancestral volcano that self-destructed perhaps 20,000 years or longer in the past. But those are the volcanoes which may be most at risk of an imminent repeat.

The run-up to the eruption would likely take years (and perhaps decades) and would certainly be noticed. But the location might not be interpreted as at risk of such a large event. Both Tambora (VEI 7) and Pinatubo (VEI 6) showed only relatively minor explosions until days or weeks before the climatic event. Krakatau had done a VEI-5 early on, but until weeks before its explosion (VEI 6) it was seen as a nice day out for a picnic. Hunga Tonga (VEI close to 6 ) had precursor eruptions which were not seen as significant until the day before the island went up. Will we know whether an eruption is likely to progress to a destructive size?

If the ultimate event is driven by caldera collapse, i.e. depletion of the magma chamber leaving its roof unsupported, then there may not be an obvious sign at the surface that this will happen. At Pinatubo, really the only sign of this danger came from the remnants of a large ancient caldera and of course the thick ignimbrite deposits in the region. Luckily these signs were recognized during the build-up.

Afterwards, a massive rescue operation may be needed. There is experience with this within the UN and various aid organisations, mainly in the context of war or famine. Months or a year later, there may be issues with worldwide food supply. A temperature drop of 1C would put as back to the climate of a century ago and although this would come as a big shock to us, that was a climate we coped with ion the past. But agriculture would also be affected by the lack of sunshine due to the sulphate haze, and the lack of rain caused by low evaporation over the oceans.

Tambora gave us the last major subsistence crisis. Could we cope with the next one?

This is the state of volcano management plans. Some parts we know well, some parts are doable if resources are available and in many cases the countries most likely to be affected meet the existing guidelines. The uncertainty lies in the once-in-500-years events, frequent enough that they should be included in risk analysis, damaging enough that world-wide action will be required, but barely covered but the existing plans.

Volcanoes will do what they want: we cannot prevent the next eruption. But are we ready?

Albert, July 2026