Chapter 12: Volcanic Winter, and the Temptation to Borrow It

In the summer of 1783 a haze settled over Europe.

It was not ordinary weather and the people who recorded it knew that much. In Hampshire, the naturalist Gilbert White wrote of a summer filled with a peculiar dry fog, of a sun that rose and set the colour of blood and could be looked at with the naked eye, of an oppressive heat, and of an atmosphere that struck him as portentous. Similar accounts came from across the continent. The haze had a sulfurous smell. It did not disperse with rain. It lasted for weeks.

Benjamin Franklin, then in Paris, wondered in print whether the cause might be volcanic — whether something erupting somewhere had put enough material into the air to dim the sun over a continent. It is often cited as the first attempt by anyone to connect an eruption to a change in the weather at a distance, and he was right, though he could not have known why.

Fifteen hundred kilometres to the north, in the interior of Iceland, a fissure had opened.

Register: the eruptions, their dates, their sulfur loads and the climate signals are empirical, and for the more recent cases very precisely so. The attribution of specific historical consequences to specific eruptions is interpretive and I will be explicit about how far I think it can be pushed. The second half of the chapter shifts into engineering assessment and ethics, and I will announce the shift when we reach it.

Laki

The eruption that began in June 1783 came from a fissure system some twenty-seven kilometres long, opening in stages, eventually with a hundred and thirty or so vents along it. It ran for eight months. It produced on the order of fourteen or fifteen cubic kilometres of basaltic lava — the largest lava flow in recorded human history.

By the standards of the last two chapters, that is not a large eruption. It is a thousandth of a flood basalt province. It did not form a caldera. Its Explosivity Index was unremarkable, because it was effusive rather than explosive.

It was, nonetheless, one of the deadliest volcanic events in European history, and it killed almost nobody with lava.

What it produced was gas — an immense quantity of sulfur dioxide, and, critically for Iceland, fluorine. In Iceland the consequences are remembered as the Móðuharðindin, the Mist Hardships. Fluorine settled on grass; livestock that ate the grass developed skeletal fluorosis and died. The great majority of the island's sheep, and a large fraction of its cattle and horses, were lost. What followed was famine, and the famine killed something like a fifth of the population of Iceland.

In continental Europe, the sulfur produced the haze that Gilbert White described, and — this is the part that is argued — a spike in mortality. There is statistical work identifying substantial excess deaths in England in the late summer of 1783 and the following winter, with a plausible mechanism in respiratory illness from a sulfur-rich atmosphere. The magnitude of that excess is contested and the attribution is not airtight, but something happened, and the timing is hard to dismiss.

The winter that followed was severe across Europe and North America.

The claim I decline to make

There is a story about Laki that appears regularly and that I am not going to endorse.

The story runs: Laki disrupted European climate, the disruption caused crop failures, the crop failures raised bread prices, and expensive bread produced the French Revolution in 1789.

I want to be careful here, because it is not absurd. Bread prices were high in 1789. Food insecurity was genuinely a factor in the unrest. Climate does affect harvests and harvests do affect politics.

But the chain is six years long and every link is loose. The most proximate agricultural catastrophe in France was the failed harvest of 1788, following a devastating hailstorm in July of that year — an event with no established volcanic connection. The Revolution had causes that were fiscal, institutional, intellectual and dynastic, and had been building for decades. Historians who work on the period are, as far as I can tell, generally unimpressed by the volcanic explanation.

What Laki demonstrates is not that a volcano caused a revolution. It is something more useful and less dramatic: that a modest, non-explosive eruption in a thinly populated place can degrade agriculture and public health across a continent, and that it can do so through gas alone, without any of the phenomena people associate with volcanic danger.

The detective story of 1257

The ice sheets of Greenland and Antarctica preserve annual layers, and those layers record atmospheric sulfate. Large eruptions leave spikes. If a spike appears at both poles, the eruption was tropical, because only a tropical injection distributes sulfate to both hemispheres.

Somewhere around 1258 there is a spike, at both poles, and it is the largest of the last two thousand years — larger than Tambora, larger than Krakatoa, larger than anything else in the era of written records.

For decades nobody knew what had caused it.

This is a genuinely strange situation. Here was the biggest volcanic event of the last two millennia, and the volcano responsible was unidentified. Candidates were proposed and eliminated. The search went on for a long time.

The answer, published in 2013, was Samalas — a volcano in the Rinjani complex on the island of Lombok, in Indonesia, which had destroyed itself in the eruption and left a caldera now occupied by a lake.

The case was made by assembling evidence of several completely different kinds. The chemistry of volcanic glass shards recovered from the ice cores was matched against the composition of tephra from Lombok. Radiocarbon dates were obtained from material charred by the eruption. The stratigraphy and the deposit volume were mapped and were consistent with an event of the required magnitude.

And there was a written source. A Sasak chronicle from Lombok, recorded on palm leaves, describes the destruction of a kingdom called Pamatan by a catastrophic eruption of Samalas — a text that had existed all along, in a language and an archive that the geochemists proposing the identification had not initially been reading.

I find this a satisfying piece of work, and it illustrates something about what the discipline has become. Reconstructing this event required ice core chemistry, electron microprobe analysis of glass, radiocarbon dating, field stratigraphy, and philology. Five methods, four disciplines, one afternoon in the thirteenth century.

The climatic aftermath included cold years and crop failure in the late 1250s across the Northern Hemisphere, recorded in European chronicles. Some early accounts linked the eruption to mass burials excavated in London; the dating of those burials has since been questioned, and I would not lean on the connection.

536, and the worst year to be alive

Procopius, writing in the sixth century, recorded that the sun gave its light without brightness, like the moon, throughout a whole year, and that it seemed a great deal like an eclipse that would not end. Other chroniclers in the Mediterranean and the Near East describe the same thing: a dimmed sun, a failure of the seasons, frost and famine.

Tree rings confirm it. The mid-530s register as one of the sharpest and most sustained cold intervals in the Northern Hemisphere record of the last two thousand years, with summer temperatures depressed by degrees rather than fractions.

Ice cores identify the cause, and it turns out to be not one eruption but a sequence: a large eruption in 536, probably at high northern latitude; a second, larger and tropical, around 540; and a third within the following decade. Three major injections within about fifteen years, arriving before the climate system had recovered from the previous one.

The revised alignment of the ice core and tree ring chronologies, published in 2015, tightened this considerably and is one of the more consequential papers in the field. A follow-up analysis proposed that the sequence initiated a prolonged cold period through the sixth and seventh centuries — the Late Antique Little Ice Age.

And the plague arrived in 541.

A historian's description of 536 as the worst year to be alive has circulated widely, and I understand why. But I want to flag the interpretive difficulty rather than ride it. Attributing societal transformation — the fortunes of empires, the spread of an epidemic — to a climate signal is exactly the sort of inference that this book has spent eleven chapters being careful about. The cooling is measured. The famines are recorded. The causal chain to political outcomes is an argument, and a contested one.

What is not contested is the mechanism, and it is the same mechanism throughout: sulfur reaching the stratosphere, forming aerosol, reflecting sunlight, and cooling the surface for a few years until it falls out.

Tambora, and the year without a summer

In April 1815, Tambora, on the island of Sumbawa, produced the largest eruption in recorded history. Direct deaths on Sumbawa and Lombok — from pyroclastic flows, ashfall, and the famine and disease that followed the destruction of agriculture — ran into the tens of thousands.

The following year, 1816, is remembered in the Northern Hemisphere as the year without a summer. There were frosts in June, July and August in New England. Harvests failed across Europe, food prices rose sharply, there were riots, and a typhus epidemic followed the malnutrition. It was, in the accounting of subsequent historians of climate, the coldest summer of the millennium in some regions.

Tambora matters for this book for a specific reason: it is the best-documented pre-modern instance of a volcanic climate forcing, with contemporaneous instrumental records, agricultural accounts, and a global spread of written observation. It is the closest thing we have to a controlled experiment in what a large stratospheric sulfur injection does to human societies.

It also produced, incidentally, some of the more durable cultural consequences of any eruption. A wet, dark, cold summer beside Lake Geneva kept a small party of writers indoors, and the ghost stories they wrote to pass the time included the first draft of Frankenstein. The failure of the oat harvest, and the resulting shortage of horses, is credited with prompting the invention of a wheeled running machine that became the ancestor of the bicycle.

I include those not as trivia but because they are a reminder that the effects of these events propagate into places nobody would think to look for them.

The eruption that did the opposite

Everything above supports a generalization: large eruptions inject sulfur, sulfur forms reflective aerosol, the planet cools for a year or three.

In January 2022 an eruption came along that did not fit.

Hunga Tonga–Hunga Haʻapai, which we met in Chapter 4 as the most densely instrumented eruption in history, was a shallow submarine event. Seawater had direct access to the magma, and the resulting interaction was extraordinarily violent — the eruption column reached something like fifty-seven kilometres, into the mesosphere, the highest ever measured and the largest since Krakatoa in 1883.

But what it put into the stratosphere was unusual. Its sulfur output was modest for an eruption of that violence. What it injected in enormous quantity was water, vaporized seawater, in a mass that increased the total water content of the stratosphere measurably — by something on the order of a tenth.

Water vapour in the stratosphere is a greenhouse gas. Sulfate aerosol cools. Hunga Tonga delivered a large dose of the first and a modest dose of the second, which meant that for the first time in the instrumental era we had a major eruption whose net radiative effect was not obviously negative.

Early analyses concluded that the warming effect dominated. More recent work has argued for a slight net cooling. The sign is genuinely disputed, and I am not going to resolve it for you; what matters for this chapter is that it can be disputed at all.

Because the lesson is methodological. Volcanoes cool the planet is a generalization drawn from a set of eruptions that happened to be subaerial and sulfur-rich. It is not a law. What an eruption does to the climate depends on what it injects, how high, at what latitude, and in what proportions — and a different kind of eruption produces a different answer.

Hold on to that, because in about two pages I am going to describe a proposal to deliberately reproduce a volcanic climate effect, and it rests on the assumption that we know what volcanic climate effects are.

The hazard has grown while the volcanoes have not

One more observation before the register changes, and it is the one I think is most underweighted in public discussion of volcanic risk.

In April 2010, Eyjafjallajökull erupted in Iceland. It was a small eruption. On the scale that has occupied this chapter it barely registers — a fraction of Laki, a rounding error against Tambora.

It closed European airspace for the better part of a week. Something like a hundred thousand flights were cancelled and millions of passengers were stranded, at a cost to the airline industry alone in the billions. The reason is that volcanic ash is abrasive silicate glass, and jet engines ingest it, and it melts in the combustion chamber and re-solidifies on the turbine blades and the engines stop — a hazard established memorably in two incidents in the 1980s in which passenger aircraft lost all engines in flight and restarted them, barely, on the way down.

And Hunga Tonga, in addition to everything else, severed the single submarine fibre-optic cable connecting Tonga to the rest of the world. The country was substantially cut off for weeks, in the immediate aftermath of a disaster, at exactly the moment communication mattered most.

Neither of those failure modes existed in 1815. Tambora could not have grounded a fleet of aircraft or severed a data cable, because there were none.

So here is the argument. The volcanoes have not become more dangerous. We have become more exposed, in new dimensions, and some of the new exposures are triggered by much smaller events. A modern society is more efficient, more interconnected, and considerably less redundant than a nineteenth-century one — with concentrated supply chains, just-in-time inventories, grain exports dominated by a handful of regions, and infrastructure that fails in cascades.

A Tambora-scale event today would not produce a nineteenth-century outcome. It would produce a twenty-first-century one, and we have no worked example of what that looks like.

The turn

Register shift, announced. Everything from here to the end of the chapter is engineering assessment and ethics rather than geology. The empirical content is what Pinatubo did; the rest is argument about what to do with that knowledge, and I will be explicit about where I decline to make a recommendation.

Pinatubo as a design document

Pinatubo, in June 1991, injected roughly twenty million tonnes of sulfur dioxide into the stratosphere. Over the following months it converted to sulfate aerosol and spread globally, and for about two years the planet was measurably cooler — by something in the region of half a degree.

That is the most important number in the entire literature on deliberate climate intervention, because it is the only full-scale test anyone has of the central proposition.

The proposal known as stratospheric aerosol injection is, in essence, Pinatubo on purpose and continuously: deliver sulfate or another reflective aerosol into the stratosphere, maintain it, and offset some fraction of greenhouse warming by reducing the sunlight reaching the surface. It is technically not far beyond current capability and it is, by the standards of climate interventions, remarkably cheap — which is itself part of the problem, as I will come to.

Everything anyone knows about whether it would work comes from eruptions, and mostly from that one.

What Pinatubo established, on the positive side: that a stratospheric sulfate layer does produce global cooling of a predictable magnitude per unit of sulfur; that it spreads globally within months; that it persists for one to two years before falling out; and that the climate response is broadly what models predicted, which was a significant validation of those models.

What Pinatubo also established, and this is generally given less prominence by advocates:

Ozone. Sulfate aerosol particles provide surfaces on which chlorine chemistry proceeds much faster. In the years following Pinatubo, stratospheric ozone depletion was significantly worse than it would otherwise have been. With atmospheric chlorine now declining, the effect of a future injection would differ — but the mechanism is real and it is a direct consequence of the intervention.

Precipitation. The global hydrological cycle is driven by surface heating. Reduce sunlight and you reduce evaporation, and there is evidence of a measurable decline in global precipitation and continental river discharge in the period after Pinatubo. This is the finding that most complicates the proposal: sunlight reduction and greenhouse warming do not have equal and opposite effects on the water cycle, so an intervention that restores the global mean temperature does not restore the global climate. Some regions get drier than they would have been under either scenario alone.

The same experiment underwrites the proposal and the strongest objections to it. That symmetry is worth sitting with.

Where the volcanic analogy fails the engineers

And now the structural problem, which I think is underappreciated even among people who follow this debate.

An eruption is a pulse. A deployment would be a plateau. Pinatubo delivered its sulfur in a day and the aerosol decayed away over two years. Any real intervention would require continuous replenishment, sustained for decades or centuries, because stopping restores the warming.

Those are not the same physical system. At sustained high loading, aerosol microphysics changes: particles coagulate, and larger particles scatter less efficiently per unit mass and fall out faster. So the cooling achieved does not scale linearly with the sulfur delivered, and the efficiency declines as you push harder. Pinatubo cannot test this, because Pinatubo was a spike. The best evidence we have is structurally the wrong shape for the question being asked.

Termination shock. A sustained deployment masking a large amount of greenhouse warming creates a commitment. Stop abruptly — through war, economic collapse, political change, or simple loss of interest — and the masked warming arrives over a few years rather than a century, at a rate far beyond anything ecosystems or agriculture could track. The intervention would have to be maintained by every successor institution for as long as the underlying carbon remained in the atmosphere. No eruption tests that either.

Altitude and latitude. Where you inject determines where the aerosol goes and how long it stays. A volcano injects where it injects. A deployment would choose — and the regional distribution of consequences would follow from that choice, which is the point at which this stops being an engineering question.

And the proposal to bleed a supervolcano

A related idea deserves a short and unsympathetic hearing, because it recurs and because working through why it fails is instructive.

The proposal is to defuse a large caldera system — Yellowstone is the usual example — by extracting heat from it. Drill into the periphery of the magmatic system, circulate water, bring the heat to the surface, generate electricity, and progressively cool the reservoir until it can no longer produce a large eruption. A version of this received attention some years ago with a suggested cost in the billions and the attractive framing that it would pay for itself in power.

It fails on the arithmetic, and the arithmetic is not close.

The thermal mass of a caldera-scale magmatic system is enormous. At any plausible rate of heat extraction, cooling it appreciably would take on the order of tens of thousands of years — longer than recorded history by a wide margin, and requiring uninterrupted operation and maintenance across that entire span. This is not an engineering challenge. It is a category error about timescales.

And there is a more serious worry, which the proposal's own authors recognized. The integrity of the roof is one of the things restraining the system, as Chapter 10 explained. Drilling into the upper part of a pressurized magmatic system in order to make it safer is a proposition that requires enormous confidence in a subsurface model that Chapters 3 and 4 established we do not have.

By way of contrast, here is an intervention that worked.

In January 1973, a fissure opened on the Icelandic island of Heimaey, close to the town and threatening to close the harbour on which the community's fishing economy depended. Over the following months, seawater was pumped onto the advancing lava front in enormous quantity — millions of cubic metres of it — chilling the flow margins, slowing the advance and steering it. The harbour survived. It is the most successful deliberate intervention in an active eruption anyone has achieved.

The contrast tells you where the ceiling is. At the scale of one lava flow and one harbour, humans can intervene in a volcanic process and win. At the scale of a caldera, we cannot, and the proposals which claim otherwise are not engineering but arithmetic that has not been done.

Who decides

I said I would set out the decision structure rather than make a recommendation, and I want to explain why I am declining before I do it.

The question of whether to deploy solar radiation management is not primarily a scientific question. The science can characterize the effects, their uncertainties, and their distribution. What it cannot do is determine how to weigh a reduction in global mean temperature against a shift in monsoon rainfall affecting a billion people, or how to trade a present benefit against a maintained obligation on every future generation. Those are questions about values and about who bears which risks, and a volcanologist has no special standing to answer them — the same division of labour I argued for in Chapter 5, applied at planetary scale.

The features of the problem that any governance arrangement would have to confront:

It is cheap. Cheap enough that a single moderately wealthy state, or conceivably a private actor, could attempt it unilaterally. This inverts the usual structure of climate politics: the problem is not persuading everyone to act but preventing anyone from acting alone. There is no existing international instrument that clearly prohibits it.

Effects are global and unequal. An injection in one hemisphere affects the other. The distribution of consequences would not match the distribution of the decision, and it would not match the distribution of historical responsibility for the underlying problem either.

Attribution is very hard. After deployment, a drought will occur somewhere. Determining whether that drought was caused by the intervention, by the greenhouse warming it was offsetting, or by neither, is a statistical problem that may be unresolvable in any individual case — which makes liability nearly impossible to assign and makes accusation nearly impossible to refute.

It creates a commitment that must be honoured indefinitely, by institutions that do not yet exist, under the penalty described above.

And there is a moral hazard argument, which is that the availability of a cheap partial offset reduces the pressure to address the cause. Whether this effect is real is empirically contested; that it is a reasonable worry is not.

I do not think anyone should read this chapter and conclude that I have told them what to think about this. I have tried to give an accurate account of what one eruption in 1991 demonstrated, of the specific ways in which it is not an adequate model for a sustained intervention, and of the shape of the decision. The decision is not mine to make and it is not the discipline's.

What geology contributes is narrower and worth stating plainly: we know this works, in the crude sense that we have watched a volcano do it. We do not know what happens when you do it on purpose and do not stop.

Part IV has been about the engine running at rates nothing can buffer — the calderas, the great dyings, and the winters that arrived inside human memory.

Part V returns to the ordinary case: not the catastrophe but the mountain that has been there all along, with a market at the bottom of it and something like a billion people living within reach.

Draft notes — verification status

Standing convention. Hunga Tonga's core facts were resolved during the thesis-paper verification pass. This chapter has the highest density of historical and quantitative claims in the book and the second half enters contested policy territory.

Already verified — do not re-check:

Hunga Tonga–Hunga Haʻapai, 15 January 2022; column ~57 km reaching the mesosphere; highest recorded since Krakatoa 1883; large stratospheric water vapour injection; radiative sign disputed — earlier studies reported net warming, more recent work slight cooling.

Pending verification — high priority:

Laki: fissure length (~27 km), number of vents (~130), duration (June 1783 – February 1784), lava volume (~14–15 km³), SO₂ release (~120 Mt), and the claim that it is the largest lava flow in recorded history. Icelandic livestock losses and the ~20–25% population mortality figure.
English excess mortality in 1783–84 attributed to Laki (Grattan, Brayshay and successors). The draft calls the magnitude contested — verify that characterization is fair and that the effect is real at all.
Samalas/1257: Lavigne et al. (2013, PNAS) identification; the Babad Lombok palm-leaf chronicle and the kingdom of Pamatan; tephra geochemistry matched to ice core glass; eruption volume (~40 km³ DRE); the bipolar sulfate spike as the largest of the last 2,000 years. Also verify that the London mass-burial link has indeed been undermined by revised dating — the draft distances itself from it and should be right to.
536 sequence: Sigl et al. (2015) ice core chronology revision; Büntgen et al. (2016) Late Antique Little Ice Age; the 536 / ~540 / ~547 eruption sequence and the inferred latitudes; Procopius's description (verify the quotation's sense and source).
Pinatubo SRM numbers: ~20 Mt SO₂; ~0.5 °C cooling; ~1–2 year aerosol residence. Post-Pinatubo ozone depletion, and the precipitation/river-discharge decline (Trenberth & Dai 2007).
Eyjafjallajökull 2010: flights cancelled (~100,000), passengers affected, industry cost. The 1982 British Airways Flight 9 and 1989 KLM 867 engine-failure incidents.
Yellowstone heat-extraction proposal: the NASA-associated study, its cost estimate, its own acknowledgement of roof-integrity risk, and the timescale for appreciable cooling. The draft's dismissal is confident and must be arithmetically defensible.
Heimaey 1973: seawater volume pumped (~6 million m³?), duration, and the extent to which the intervention actually determined the outcome versus the eruption ending on its own. This is the chapter's one success story and it should not be overstated.

Pending verification — standard:

Gilbert White's 1783 account in The Natural History of Selborne; Franklin's 1784 speculation on a volcanic cause.
The historiographical status of the Laki–French Revolution claim. The draft says historians of the period are "generally unimpressed" — verify rather than assume.
Tambora: April 1815; direct and total death tolls; erupted volume; the 1816 Northern Hemisphere anomalies; European harvest failure and typhus.
Villa Diodati and Frankenstein; Karl Drais's running machine and the oat-shortage explanation. Both are widely repeated and the Drais causal link in particular should be checked rather than assumed.
Hunga Tonga stratospheric water increase (~10%) and SO₂ output for comparison.
Aerosol microphysics at sustained loading — coagulation, reduced scattering efficiency per unit mass, faster fallout — and the resulting sublinear scaling.
Termination shock literature and the rate of warming on abrupt cessation.
SRM cost estimates and the "free driver" framing; the current status of international governance instruments.
Moral hazard evidence — the draft says the effect is empirically contested; confirm.

The ordinary case, and the billion people downslope