Chapter 13: Who Lives on the Flank

Nobody asks why people live in Bangladesh.

That is not quite true — people do ask, and the asking is usually just as condescending. But the question gets asked with far more energy about volcanoes than about floodplains, which is odd, because floods kill vastly more people. It gets asked about volcanoes and not about the coasts, where a majority of humanity lives within reach of a storm surge, or about the seismic zones of Tokyo, Istanbul and Los Angeles, or about the parts of the world that are running out of water.

Why would anyone live there is a question we reserve for hazards that look dramatic, and the reservation tells you more about the questioner than about the resident.

So let me put the number down and then do it the courtesy of taking it seriously. Depending on how you define proximity — and the definition matters enormously, so I will state it — something in the region of eight hundred million people live within a hundred kilometres of a volcano that has erupted within the last ten thousand years. Tighten the radius to thirty kilometres and the figure is a couple of hundred million. Loosen the criteria and you get to a billion.

These people are not there through ignorance of geology. Many of them have a considerably more direct relationship with the mountain than any reader of this book. They are there for reasons, and this chapter is about what the reasons are, who gets to weigh them, and who does not.

Register: the soil science and the exposure figures are empirical. The historical and ethnographic material is interpretive. The argument about political economy in the second half is an argument, and I will make it as an argument.

Why the soil is worth it

Start with the reason that is not cultural or historical but chemical, because it is the one that holds everywhere.

Volcanic ash weathers fast. That is the whole thing, and everything else follows from it. Basaltic and andesitic glass is thermodynamically unstable at the surface, and in a warm wet climate it breaks down over decades to centuries rather than the millennia that crystalline rock requires. Weathering releases the elements plants need — phosphorus, potassium, calcium, magnesium — on a timescale that matters to a farm.

The soils that result are classified as andisols, and they have an unusual mineralogy. Instead of the well-ordered clays that develop on most parent materials, they form short-range-order minerals with enormous specific surface area. The consequences are agriculturally excellent: high water-holding capacity, high capacity to retain nutrient cations against leaching, low bulk density, and a loose friable structure that is easy to work and resists compaction.

If you were designing a soil, you would design something close to this.

An honest complication, because I have promised them throughout: andisols are not uniformly wonderful. That same high-surface-area mineralogy binds phosphate very strongly, and phosphorus fixation is a real and well-documented limitation on volcanic soils — the nutrient is present and the plant cannot get at it. Volcanic soil is not a free lunch; it is an unusually good lunch with one specific and manageable problem.

Then look at where people actually are.

Java carries somewhere north of a hundred and fifty million people on an island roughly the size of England, at among the highest rural population densities anywhere on Earth, and it does so on ash-derived soils sustaining multiple rice harvests a year. The island has around forty-five volcanoes considered active. The correspondence between the population map and the volcano map is not a coincidence and it is not a tragedy. It is a system that works.

Campania was known to the Romans as Campania Felix — fortunate Campania — for the productivity of the ground around Vesuvius, which they farmed intensively and wrote about admiringly, without any apparent sense that they had made an error.

The Ethiopian highlands, built of flood basalt from the province that split the Red Sea open, carry the majority of that country's population and its agriculture, in dramatic contrast to the lowlands on either side.

And a disproportionate share of the world's coffee, cacao and wine comes off volcanic ground, from Java and Sumatra to the Ethiopian highlands, Guatemala, Costa Rica, Colombia, the slopes of Etna and the Canaries.

Now the arithmetic that governs the decision, and it is worth stating precisely because it explains almost everything in this chapter.

The benefit is annual, reliable, and compounding. The cost is generational, probabilistic, and catastrophic.

A farmer on volcanic soil gets a better harvest this year and next year and every year. The eruption may come in her lifetime, or her grandchildren's, or not for four centuries. Any reasonable discounting of future risk against present return favours staying, and it favours staying by a wide margin. Human institutions handle that shape of problem badly, and so do human intuitions — but the underlying calculation is not irrational. It is arguably the same calculation every society makes about every low-probability catastrophe, and the volcano only makes it visible.

The rest of the ledger

Soil is the largest entry but not the only one.

Building material. Volcanic tuff is light, workable, and strong enough — it built a great deal of Rome. More consequentially, Roman concrete used pozzolana, volcanic ash named for the town of Pozzuoli, which sits inside the Campi Flegrei caldera we met in Chapter 10. The ash reacts with lime to produce a cement that sets underwater and that has proven extraordinarily durable in seawater — Roman harbour structures are still standing after two thousand years, and the reason involves mineral phases that continue to grow within the concrete as seawater percolates through it. Some of the best-performing construction material in human history is a volcanic product, and it came out of an active caldera.

Obsidian, which for most of prehistory was the sharpest material available and one of the most widely traded. Because obsidian from different volcanoes carries distinct trace-element signatures, archaeologists can source individual flakes to individual eruptions, and have used this to map trade networks across the Mediterranean, Anatolia and Mesoamerica in remarkable detail. Volcanic glass is one of the best tracers of early human exchange we possess.

Geothermal energy and mineral deposits, both of which are Chapter 14's business.

Tourism, which is now a major economic sector for Iceland, Hawai'i, Bali, Sicily and elsewhere, and which is worth noting because it inverts the hazard relationship — people travelling toward the volcano on purpose, and occasionally, as at Whakaari in 2019, dying there.

And one entry that belongs on the ledger precisely because it is grim. At Kawah Ijen in East Java, men carry baskets of solidified sulfur out of an active crater, breathing acid gas, for wages that are poor by any standard and good by local ones. It is one of the more punishing forms of work anywhere. It is also a volcanic resource being converted into a livelihood by people with limited alternatives, and any honest ledger of what volcanoes provide has to include the entries that are provided on bad terms.

Goma

And now the case that breaks the tidy version of everything I have just written.

Nyiragongo sits above the city of Goma, in the eastern Democratic Republic of the Congo, on the northern shore of Lake Kivu. It has a persistent lava lake, and it produces the most dangerous lava on Earth.

The reason is chemistry, and it takes us back to Chapter 2. Nyiragongo's magma is extraordinarily low in silica — a foidite, alkali-rich and silica-poor — which means it is very poorly polymerized, which means its viscosity is exceptionally low. Chapter 2 established that basaltic lava typically moves at a walking pace and takes everything except your life. Nyiragongo's lava has been recorded moving at speeds that would be respectable on a motorway.

It is the one volcano where the cinematic image — a wave of lava outrunning people — is approximately true.

In 1977 the lava lake drained through fissures on the flank, and the flows moved so fast that people were caught in the open. In 2002, fissures opened toward the city; lava entered Goma, destroyed a substantial fraction of the built area, and displaced several hundred thousand people. In May 2021 it happened again, with very little warning, killing dozens and displacing tens of thousands more.

Here is the detail from 2021 that belongs in this chapter rather than in Chapter 5.

The Goma Volcano Observatory — the institution responsible for monitoring Nyiragongo — had, some months before the eruption, lost its principal external funding amid a dispute over financial management. Its staff were going unpaid. Its internet connection had been cut, which meant it could not receive data from its own remote seismic stations. Its capacity to see what the volcano was doing was substantially degraded at the moment the volcano did it.

I want to be careful and not turn this into a simple morality tale, because the funding suspension had its own stated reasons and I am not in a position to adjudicate them. But the sequence is not in dispute: the monitoring capacity was compromised, and then the mountain erupted, and people who might have been warned were not.

And there is a second thing about Goma that dismantles the framing I offered at the start of this chapter.

Goma's population has grown enormously over recent decades — to something around two million — and the growth was driven overwhelmingly by conflict. People came to Goma because the alternative was worse: displacement from war in the surrounding region, refugee flows from Rwanda, the collapse of security in the eastern provinces. They did not weigh volcanic risk against agricultural yield and decide the trade was favourable.

They were not making a bargain. They were fleeing something else, and this is where they arrived.

The soil argument, the discounting argument, the rational-choice framing — all of it assumes an agent with options. A great many of the people most exposed to volcanic hazard worldwide do not have options in any meaningful sense. They live where displacement, poverty, land tenure and violence have put them.

Which means the sentence I want this chapter to be built around is this: volcanic risk is distributed by political economy, not by geology. The mountain does not choose who lives at the bottom of it. We do.

Armero, revisited

Which brings me back to the chapter this book turns on.

Chapter 5 established that at Nevado del Ruiz the science was correct, the hazard map was produced in time, and the failure was downstream. I stand by that account, and it is the account the discipline has largely accepted. But it answers the question what went wrong and not the question for whom, and the second question is this chapter's.

So ask a different set of things about the same night.

Who lived on the alluvial fan at the mouth of the Lagunillas? Armero was a prosperous agricultural town — cotton, rice, coffee — and prosperous agricultural towns are inhabited disproportionately by agricultural workers, which is to say by people whose livelihoods were literally the ground and who had the least capacity to abandon it on a probability.

Who received the hazard map? It went to national, regional and local authorities and to the press. It did not go, in any form designed for the purpose, to the households inside the shaded polygon. The people whose lives it described were not its audience.

Whose interests did evacuation threaten? The harvest was in progress. The cost of moving a town falls immediately and locally and visibly on identifiable people; the benefit is a catastrophe that may not happen and for which nobody receives credit.

And what else was occupying the Colombian state in November 1985? A week before the eruption, the Palace of Justice in Bogotá was seized by an armed group and retaken in an assault that killed around a hundred people, including much of the Supreme Court. The government's attention, capacity and political bandwidth in the days before Armero were consumed by a national crisis of a different kind. That is not an excuse. It is the ordinary condition of the institutions that volcanic warnings have to travel through, and any system that assumes an undistracted state is a system that will fail.

At a mass funeral in Ibagué afterward, mourners carried a banner reading: The volcano didn't kill 22,000 people. The government killed them. I do not think that is the whole truth, for the reasons Chapter 5 set out at length. I do think it is a more accurate summary than most of what was written at the time, and it was produced by the people who had buried the dead.

The lesson I draw is a modification of Chapter 5's. That chapter concluded a warning is a social product rather than a scientific one. This one adds: it is also a political product, and it moves along channels that are shaped by who has standing. Getting the science right and delivering it to officials is not the end of the job. It is not even obviously the middle.

The knowledge that is not in the model

Communities that live with volcanoes develop their own hazard knowledge, and the relationship between that knowledge and the instrumental kind is more interesting than either romanticization or dismissal allows.

At Merapi, in central Java, there is a traditional office of spiritual gatekeeper, appointed by the Sultan of Yogyakarta, responsible for the relationship between the community and the mountain. For decades that role was held by a man who became widely known and widely respected.

In 2006, during a period of unrest, he declined to evacuate and urged others to stay. The eruption spared his village. His authority — already considerable — grew, and the episode was widely read as vindication.

In 2010, during a much larger crisis, he again declined to leave. Pyroclastic flows reached his village. He was killed, along with others who had remained with him.

I tell it that way because the two outcomes together are the honest version, and either one alone is a different and misleading story. Local authority was right the first time and catastrophically wrong the second, and it was competing for legitimacy with the scientific warning throughout.

It is also worth saying that the 2010 crisis was, overall, a success: Indonesian volcanologists raised the alert progressively as the eruption escalated beyond anything in the modern record for that volcano, and several hundred thousand people were evacuated. Several hundred died. Without the evacuation the number would have been very much higher.

The general point stands, though, and it cuts in a direction that is uncomfortable for a book like this one. Traditional and local knowledge encodes real information, particularly about rare events, because oral tradition and ritual can preserve the memory of an event that occurred outside the instrumental record entirely. The clearest demonstration comes from a different hazard: on one Indonesian island in 2004, an oral tradition about the sea withdrawing before a great wave — preserved from an earthquake a century earlier — sent the population uphill immediately, and almost everyone survived while neighbouring populations did not.

What local knowledge does badly is generalize from single events, and it can anchor hard on the most recent experience, which is exactly the failure mode at Merapi in 2010: a volcano behaving outside its own historical envelope, interpreted by an authority whose framework was built from that envelope.

The instrumental and the traditional are not competing accounts of the same thing. They fail differently, and a serious system uses both.

What the archaeology shows

The long record is more encouraging than the individual disasters suggest, and worth a paragraph as a corrective.

The consistent finding across volcanic regions is that people come back. Vesuvius has destroyed settlements repeatedly and the slopes have been resettled within a generation each time, right up to the present, where a large population lives inside the zone that would have to be evacuated. Iceland has cycles of settlement, abandonment and return. Java has been continuously farmed through eruption after eruption.

At Cerén, in El Salvador, a farming village was buried by ash around fourteen centuries ago and preserved in extraordinary detail — buildings, fields, painted vessels, meals abandoned mid-preparation. It is often called the Pompeii of the Americas, and the most interesting thing about it is what has not been found: bodies. The evidence suggests the residents had warning and left.

And the case that gets cited most, Thera and the Minoans, has been substantially qualified. The eruption was enormous, the effects on Crete were real, and Minoan civilization nonetheless continued for a considerable time afterward. The simple story in which a volcano ends a civilization does not survive close dating.

The pattern across all of it is that resilience is the normal outcome and collapse is rare, and that where collapse does occur the eruption is usually one stressor among several rather than a sufficient cause on its own.

Why people do not leave, without condescension

Finally, the practical question that all of this bears on. Evacuation orders are issued and are frequently not complied with. Why?

The literature is consistent and it is not flattering to the way the question is usually framed.

Livestock. This is the single most cited reason and it is enormously underweighted in planning. A family whose entire capital is four cows will not abandon four cows on a probability, and if evacuated will return to feed them. People die going back for animals. Evacuation schemes that make no provision for livestock are, predictably, less effective than schemes that do.

No destination and no savings. Evacuation costs money — transport, lost wages, food, somewhere to sleep. For a household with no buffer, a two-week evacuation that turns out to be unnecessary is a genuine catastrophe of its own kind.

Property and looting. Empty houses get robbed, and people know it.

Prior false alarms, which are inevitable given everything in Chapters 2 and 5, and which erode compliance in a way that is entirely rational on the individual's part.

Distrust of the state, which in many of the relevant places has an extensive evidentiary basis, and which no amount of technical accuracy in a hazard map will repair.

Place attachment, which is the one usually treated as sentimental and which I think is the least sentimental of all. For a great many people, leaving a place means the end of a livelihood, a community, a set of relationships and a way of being that does not reconstitute elsewhere. That is not irrational attachment to scenery. It is an accurate assessment of what is being asked.

Set these out together and the picture inverts. Non-compliance is usually the rational response of someone facing constraints that the person issuing the order does not share. The mayor ordering the evacuation has a salary, a car and somewhere to go. That asymmetry, more than any deficit of information, is what the last mile actually consists of.

Which yields something practical, and it is the finding that most improves outcomes: evacuation works when the people evacuating were involved in designing the plan, when the destination is known in advance, when livestock and property are provided for, and when the relationship between the community and the observatory was built before the crisis rather than during it.

That is not a scientific insight. It is a social one, and it saves more lives than any instrument in this book.

The bargain, restated

So: a billion people, and none of them are stupid.

For many of them the arrangement is a genuine bargain, and a good one — fertile ground, reliable water, a livelihood, in exchange for a risk that is real and rare. That trade has been made deliberately for thousands of years by people who understood the terms, and it has, on the whole, paid.

For many others there is no bargain at all. There is a city that grew because a war pushed people into it, and an observatory that lost its funding, and a mountain that did not consult anyone.

Both of those are true at once, and a book that offered only the first would be a comfortable book about resilience, and a book that offered only the second would be a lecture. What the two together establish is the thing I want to carry into the final chapter: the hazard is geological and the risk is political, and the difference between them is where all the available leverage lies.

Which raises the obvious next question. We have spent this chapter on what it costs to live beside the engine and what it pays.

What if we used it on purpose?

Draft notes — verification status

Standing convention. This chapter makes claims about identifiable institutions and about people who died, including a named traditional office-holder at Merapi. Those passages need the most careful checking in the book after Chapter 5.

Structural note. The outline calls for Part V to open with a field vignette at the Goma market. Not written, for the reason given in the Chapter 1 and Chapter 10 notes.

Pending verification — high priority:

• Global exposure figures: ~800 million within 100 km of a Holocene-active volcano; ~200 million within 30 km. Source these to the Global Volcano Model / VOGRIPA population exposure work (Brown et al.) and state the definition used, since the draft leans on definitional precision.
• Goma Volcano Observatory funding suspension preceding the May 2021 eruption: the funder, the stated reasons, the timing relative to the eruption, and the specific operational consequences (unpaid staff, loss of internet, inability to access remote seismic data). This is a serious claim about an institution and a funder and must be sourced precisely and characterized fairly.
• Nyiragongo eruptions: 1977 (flank fissures, lava lake drainage, death toll estimates — these vary widely), 2002 (proportion of Goma destroyed, deaths including the fuel station explosion, displacement figures), 2021 (deaths, displacement, warning time). Lava flow velocities and the foidite composition.
• Goma population (~2 million) and the conflict-driven character of its growth.
• Merapi: the office of juru kunci; the 2006 and 2010 decisions; the 2010 eruption's scale relative to the modern record, evacuation numbers (~350,000) and death toll (~350). Name the office-holder only after verifying details of both episodes — the draft deliberately does not name him.
• Palace of Justice siege, Bogotá, 6 November 1985: date, casualties (~100, including Supreme Court justices), and its proximity to the 13 November eruption. The draft uses this to argue about state capacity and the connection should be defensible rather than merely suggestive.

Pending verification — standard:

• Andisol mineralogy: short-range-order minerals (allophane, imogolite), high specific surface area, water retention, cation exchange capacity, low bulk density. And the phosphorus fixation limitation, which the draft raises as the honest complication.
• Java: population (~150 million+), island area comparison, number of active volcanoes (~45), rural population density claims.
• Campania Felix and Roman agricultural writing on the Vesuvian slopes.
• Ethiopian highlands: flood basalt origin and the population/agriculture concentration relative to lowlands.
• Roman concrete and pozzolana from Pozzuoli; the seawater durability mechanism and the mineral phases involved (aluminous tobermorite). Confirm the current understanding, which has been refined recently.
• Obsidian trace-element sourcing and its use in mapping prehistoric trade.
• Kawah Ijen sulfur mining: working conditions and wages.
• Whakaari 2019 as a tourism-related fatality case — already flagged in Chapter 5 notes; keep consistent.
• Cerén: date (~600 CE), the Loma Caldera eruption, and the absence of human remains implying evacuation.
• Thera/Minoan: the qualification that Minoan civilization persisted well after the eruption.
• Simeulue and the smong oral tradition in 2004; the 1907 earthquake it derives from and the survival statistics.
• Evacuation compliance literature: livestock as the dominant return factor, looting, cost, false-alarm fatigue, and place attachment. Cite the specific studies rather than generalizing.