Chapter 11: The Great Dyings

In north-central Siberia there is a plateau called Putorana, and it is made of stacked basalt. Flow on flow on flow, laid down horizontally, exposed now in cliffs and river gorges to depths of kilometres. It is a large area of a large country and there is almost nothing there.

That plateau is the eroded remnant of an eruption that killed most of the life on this planet.

And here is the thing to hold on to as we go, because everything in this chapter follows from it: the eruption was not violent. It did not blow up. It was, for the most part, lava — enormous volumes of runny basalt pouring out of fissures over an extremely long period, of the sort described in Chapter 2 as the survivable kind. Anyone standing in Siberia could have walked away from it.

The dying happened somewhere else. It happened in the air and in the ocean, over thousands of years, on the other side of the world, to organisms that never saw a volcano.

Register: the deposits, their volumes and their ages are empirical, and the dating has become extremely precise in the last fifteen years. The causal chain from eruption to extinction is theoretical — well supported in outline, argued in detail, and in one prominent case genuinely unresolved.

A category above the last one

Chapter 10 asked you to picture a thousand cubic kilometres as a cube ten kilometres on a side, and called it the floor of the super-eruption category.

A large igneous province is measured in millions of cubic kilometres.

That is not a bigger version of the last chapter. It is a different phenomenon with different mechanics and a different mode of destruction. Super-eruptions are explosive, silicic, and over in days. Flood basalt provinces are effusive, mafic, and emplaced over hundreds of thousands to a few million years — though with the important qualification, established by high-precision dating over the past two decades, that the bulk often comes out in short violent pulses within that window rather than at a steady drip.

The proposed mechanism takes us back to Chapter 1's unresolved argument. The favoured explanation is the arrival of a mantle plume head at the base of the lithosphere: a large volume of anomalously hot mantle reaching shallow depth, undergoing decompression melting on a massive scale, and erupting through whatever fractures it can open. Flood basalt provinces are, in fact, among the better arguments for deep plumes — which is worth noting, since I gave the plume sceptics a fair hearing in Chapter 1 and this is a place where the mainstream position has real evidence behind it.

The provinces themselves: the Siberian Traps, at around 252 million years. The Central Atlantic Magmatic Province, at around 201 million years, erupted as Pangaea began to split. The Deccan Traps of western India, at around 66 million years. The Karoo–Ferrar, the Emeishan, the Columbia River Basalts, the North Atlantic province — and, in the oceans, plateaus like Ontong Java, which are larger still and much harder to study.

The correlation, counted honestly

Now the claim that gives this chapter its title, and I want to state it more carefully than it is usually stated — including more carefully than I have stated it myself elsewhere.

The commonly repeated version is that four of the five great mass extinctions coincide with large igneous provinces. That is a tidying-up of a messier picture, and the tidy version is doing more work than the evidence supports.

Here is the honest ledger.

End-Permian, around 252 million years ago — the largest extinction in the record — coincides with the Siberian Traps. This association is strong, the dating is now very tight, and I do not think it is seriously disputed.

End-Triassic, around 201 million years ago, coincides with the Central Atlantic Magmatic Province. Also strong.

End-Cretaceous, around 66 million years ago, coincides with the Deccan Traps — and also with a large asteroid impact. This one is genuinely contested and gets its own section below.

Late Devonian and end-Ordovician have proposed LIP associations that are weaker, more argued, and in the Ordovician case complicated by a major glaciation that is the more usual explanation.

Beyond the canonical five there are further associations that strengthen the general pattern: a mid-Permian extinction with the Emeishan Traps, an oceanic anoxic event in the Jurassic with the Karoo–Ferrar, and the Palaeocene–Eocene Thermal Maximum — a rapid warming event rather than a mass extinction — with the North Atlantic province.

So the fair summary is: the two largest well-dated extinctions of the Phanerozoic coincide with the two largest well-dated continental flood basalt provinces, several lesser events show similar associations, and one very famous case has a competing explanation. That is a real and important pattern. It is not "four of five," and the shorthand should be retired.

Correlation is also not causation, and with a sample this small and a dating uncertainty that was until recently comparable to the duration of the events themselves, the burden is on the mechanism. Which is where the interesting part is.

It is not the lava

The single most important correction in this chapter is that flood basalts do not kill by burial.

The Siberian Traps covered an enormous area of Siberia. They did not cover Africa, or the oceans, or the shallow marine shelves where the great majority of the extinction victims lived. Lava is a local hazard, as Chapter 2 established, and even at these volumes it is local — a subcontinent, not a world.

What travels is gas.

A flood basalt province degasses on a scale that is difficult to hold in the mind. Carbon dioxide, sulfur dioxide, hydrogen chloride, hydrogen fluoride — released continuously, from fissures spread over a subcontinent, for tens or hundreds of thousands of years. And unlike an explosive eruption, which fires its gas load into the stratosphere in an afternoon and then stops, a flood basalt province delivers gas persistently, which turns out to matter enormously for which effects accumulate.

What the Siberian Traps were erupting through

And now the refinement that has, over the past twenty years or so, substantially changed how this subject is understood — and which I think is the most interesting single fact in this chapter.

The Siberian Traps were not unusually large among LIPs. They are large, but the Deccan and CAMP are in the same class, and the extinctions associated with those were less severe.

What made Siberia exceptional was what it erupted through.

The province was emplaced into the Tunguska Basin, a sedimentary basin containing vast deposits of coal, organic-rich shale, and evaporites — salt beds, including halides. And a great deal of the magma did not reach the surface. It intruded laterally as sills, sheets of magma injected between sedimentary layers, cooking enormous volumes of sediment in place.

The consequences of cooking those particular rocks are specific and severe.

Heating coal and organic-rich shale drives off carbon dioxide and methane — potentially far more carbon than the magma itself carried, because the magma is a heat source and the sediment is the fuel.

Heating evaporites in the presence of organic matter produces halocarbons: methyl chloride, methyl bromide and related compounds. These are potent destroyers of stratospheric ozone.

So the Siberian Traps may have been the deadliest volcanic event in Earth's history not because of the magma but because of the geology it happened to intrude. A comparable volume of the same basalt, erupted through crystalline basement with no coal and no salt, would have been a much less consequential event.

There is a genuine test of this, and it has largely been passed: if ozone was destroyed on a large scale, terrestrial plants should show the signature of ultraviolet damage. Malformed and mutated pollen and spores are found in end-Permian terrestrial sections, consistent with a serious ultraviolet excursion.

I find this the most sobering fact in the chapter, because of what it implies about contingency. The largest extinction in the history of life may have depended on the accident of where a plume happened to surface.

The kill mechanisms

With that established, the causal chain from eruption to extinction runs roughly as follows. Several of these operate simultaneously and on different timescales, which is a large part of why the outcome was so severe.

Long-term warming. Carbon dioxide accumulates faster than the weathering sink can remove it — the thermostat of Chapter 7, overwhelmed rather than broken. Substantial global temperature rise, sustained over tens of thousands of years.

Short-term cooling. Sulfur dioxide forms stratospheric aerosol and cools the planet on a timescale of years, superimposed on the warming trend. The result is not a smooth transition to a hotter world but violent oscillation — repeated cold snaps punctuating a long warming. Organisms adapted to neither.

Ocean acidification. Carbon dioxide dissolving into seawater lowers pH. This is measurable in the rock record through boron isotopes, and the signature is present at the end-Permian and end-Triassic boundaries. It is selectively lethal to organisms that build calcium carbonate skeletons, and the extinction pattern is consistent with that selectivity.

Anoxia. Warm water holds less dissolved oxygen, and warming weakens the circulation that ventilates the deep ocean. The result is widespread oxygen depletion, and in places euxinia — sulfidic water, in which hydrogen sulfide accumulates. There are biomarker traces of sulfur bacteria that require both light and hydrogen sulfide, which means the sulfidic zone reached the sunlit surface waters. That is an extraordinarily hostile ocean.

Ozone loss and ultraviolet damage, as above.

Acid rain, from sulfur and halide emissions, with consequences for terrestrial vegetation and soils.

There is also a useful forensic tool worth mentioning: mercury. Volcanic emissions carry mercury, which is deposited globally and preserved in sediments. Spikes in mercury concentration in boundary sections have become a widely used chemical fingerprint for volcanic activity, and they have been found at several extinction horizons. It is one of the better recent methods for tying an extinction to an eruption when the dating alone is not decisive.

The end-Permian, in sequence

The largest extinction in the history of complex life removed something in the region of eighty per cent or more of marine species and a comparable proportion of terrestrial vertebrate genera. Insects — which have otherwise been almost immune to mass extinction — suffered their only major loss.

What the high-precision dating of the last fifteen years has added is a sequence, and the sequence is the argument.

The Siberian magmatism began somewhat before the extinction — on the order of a few hundred thousand years before. The extrusive lavas, the flows that built Putorana, came first, and the extinction did not accompany them.

The main pulse of sill intrusion — the phase that cooked the sedimentary basin — coincides with the extinction interval.

And the extinction interval itself is short. Estimates from high-precision uranium-lead dating of zircons put the main phase at well under a hundred thousand years, and possibly a few tens of thousands.

That sequence is a strong argument for the contact-metamorphism mechanism specifically. It was not the lava, and the timing says so: the lava was already erupting and nothing much was dying. What changed was the magma going sideways into coal and salt.

Alongside this sits a large negative excursion in carbon isotopes — the signature of a huge injection of isotopically light carbon into the ocean-atmosphere system, which is exactly what burning buried organic matter would produce.

Recovery took a very long time. Marine ecosystems did not return to comparable complexity for several million years. There is a "coal gap" in the record — a long interval with essentially no coal deposits anywhere, because the peat-forming ecosystems were gone and did not come back. There is a reef gap of similar length.

The one that is not resolved

I want to handle the end-Cretaceous carefully, because it is the case everyone knows and the case where I have the least to offer in the way of a verdict.

Around sixty-six million years ago, an asteroid roughly ten kilometres across struck what is now the Yucatán Peninsula, leaving the Chicxulub crater. The evidence for this is overwhelming and was hard-won: a worldwide layer of iridium-enriched clay, shocked quartz, impact spherules, and eventually the crater itself.

Around the same time — spanning the boundary — the Deccan Traps were erupting in western India, at a volume comparable to the other great provinces.

Both of these things happened. The dispute is about their relative contributions, and it has been running for four decades with a great deal of heat.

The impact camp points to the precision of the timing: the extinction horizon coincides with the ejecta layer to within the resolution of the record, which is very fine indeed. That is difficult to explain as coincidence.

The volcanism camp points to evidence of ecological stress before the boundary, to the sheer scale of Deccan degassing, and to the observation that every other great extinction in the record has a LIP and no impact, so invoking a unique mechanism for this one is odd.

There is also an intriguing hybrid proposal: that the seismic energy of the impact itself accelerated Deccan eruption rates, in which case the two are not competing explanations but a single coupled event.

High-precision dating has been deployed by both sides, and — this is the detail I find most telling about the state of the question — two major dating studies published in the same journal in the same week in 2019 came to opposite conclusions about whether most of the Deccan lava erupted before or after the impact.

My own reading, offered as a reading: the impact is the proximate trigger, because the timing is too exact to dismiss, and Deccan volcanism very likely stressed ecosystems before it and prolonged the disruption afterward. But I want to be clear that I hold this loosely, and that people who have spent their careers on it disagree in both directions.

The uncomfortable mirror

I have deferred the obvious comparison and I am not going to avoid it, but I want to be exact about what it does and does not license, because this is territory where both overstatement and dismissal are common.

The end-Permian involved an injection of carbon into the ocean–atmosphere system that was, in total, very large — on any estimate, considerably larger than everything humanity has released to date.

It was also very much slower. The injection occurred over tens of thousands of years. The current one has occurred over a few centuries.

Rate is not a detail here; it is the operative variable, and the reason is mechanical rather than rhetorical. There is a hierarchy of buffers, each with its own response time. Ocean carbonate chemistry buffers added carbon dioxide over thousands of years. Silicate weathering — the thermostat of Chapter 7 — buffers it over hundreds of thousands. If carbon is added more slowly than a buffer responds, that buffer absorbs it and little happens. If it is added faster, the buffer is bypassed and the perturbation is felt in full.

The end-Permian was catastrophic in large part because its injection outran the ocean's buffering capacity. The measured consequence — acidification, recorded in boron isotopes — is precisely the signature of carbon arriving faster than carbonate chemistry could neutralize it.

What the comparison illuminates is that mechanism: that the damage from a carbon perturbation depends on its rate relative to the buffers, and that the geological record contains a worked example of what happens when the rate wins.

What the comparison does not license is a prediction of outcome. The end-Permian world had a different continental configuration, a different ocean circulation, different biota with different tolerances, and a total carbon release larger than ours. The mechanisms that operate are the same mechanisms; the magnitudes, the starting conditions and the ending conditions are not the same. Anyone using the Permian to forecast a specific future is over-reading it in one direction, and anyone dismissing it as irrelevant because the total was larger is over-reading it in the other.

The honest statement is narrower and, I think, sufficient: the geological record demonstrates that this planet's carbon buffers can be outrun, that the consequences when they are outrun are severe and long-lasting, and that the relevant comparison is rate rather than total.

Rate, not kind

Let me close Part IV's first argument, because it completes something Part III began.

Chapter 7 made the case that volcanic outgassing is one half of the mechanism that has kept this planet habitable — the source term in a thermostat whose sink is silicate weathering, running on a response time of hundreds of thousands of years.

This chapter has described the same process, with the same chemistry, doing the opposite.

Nothing different happened during the Siberian Traps. No new mechanism switched on. Volcanoes released carbon dioxide and sulfur, exactly as they always have and exactly as they were doing yesterday. Weathering consumed carbon dioxide, exactly as it always has.

What changed was the rate. The source term increased by a large factor and stayed elevated for tens of thousands of years, and the sink — which cannot respond faster than it responds — fell behind. The thermostat was not broken. It was outrun.

And it did eventually catch up. The carbon isotope excursions recover. The climate returns. The geochemical perturbation resolves on a timescale that is, satisfyingly and not coincidentally, the timescale of the weathering feedback. The thermostat fixed it — slowly, indifferently, and much too late for anything that was alive at the time. Biological recovery took far longer than chemical recovery, because ecosystems do not come back on the same schedule as pH.

So the engine that made the air and holds the temperature is the same engine that has, several times, come close to sterilizing the surface. There is no version of this book in which those are two different mechanisms, and I would be misleading you if I let Part III stand without this chapter.

The difference between a life-support system and a kill mechanism is not kind. It is rate.

That is the last chapter in this book standing entirely on deep time. The next one brings the same mechanism into human history — into years that people wrote down, harvests that failed, and a summer that did not arrive.

Draft notes — verification status

Standing convention. This chapter revises a claim made in Chapter 7 and makes a rate comparison in politically sensitive territory. Both need careful checking.

Correction to reconcile:

• RESOLVED. The "four of the five" formulation appeared in the accompanying thesis paper, not in Chapter 7. The thesis paper has been amended to the narrower ledger given here. No chapter-to-chapter contradiction remains.

Pending verification — high priority:

• Siberian Traps contact metamorphism: the Tunguska Basin coal, evaporite and organic shale deposits; sill intrusion as the dominant emplacement mode for the lethal phase; halocarbon generation from evaporites; the estimate that sediment-derived carbon may have exceeded magmatic carbon. Attribution to Svensen, Planke and colleagues. This is the chapter's central mechanistic claim.
• End-Permian sequence: magmatism beginning ~300 kyr before the extinction; extrusive phase preceding, sill phase coinciding; main extinction phase duration (<100 kyr, possibly a few tens of kyr) from high-precision U-Pb zircon dating (Burgess, Bowring and colleagues).
• Extinction magnitude: marine species loss (commonly cited 81–96%, with recent work revising downward) and terrestrial vertebrate genus loss. The draft says "eighty per cent or more" — check current best estimates, which have moved.
• End-Cretaceous: the two 2019 Science dating papers reaching opposite conclusions on Deccan timing (Sprain et al. and Schoene et al.). Confirm the characterization and that they were indeed same-journal, same-issue.
• The rate comparison: total end-Permian carbon release estimates; cumulative anthropogenic carbon release to date; and the claim that the current rate of CO₂ increase exceeds anything in the Phanerozoic record including the PETM. Verify each figure and the "faster than anything in the record" claim specifically — it is strong and frequently asserted loosely.

Pending verification — standard:

• LIP volume ranges (>10⁵ km³ threshold; Siberian Traps total estimates, which vary widely) and emplacement durations.
• Ontong Java Plateau as larger than continental LIPs.
• Plume-head arrival as the favoured LIP mechanism, and the strength of LIPs as evidence for deep plumes.
• Ages: Siberian ~252 Ma, CAMP ~201 Ma, Deccan ~66 Ma, Emeishan ~260 Ma, Karoo–Ferrar ~183 Ma, NAIP/PETM ~56 Ma.
• End-Ordovician and Late Devonian LIP associations and why they are weaker.
• UV-damaged pollen and spores at the end-Permian as evidence of ozone depletion.
• Boron isotope evidence for acidification at end-Permian and end-Triassic.
• Green sulfur bacteria biomarkers (isorenieratane) as evidence of photic zone euxinia.
• Mercury/TOC anomalies as a LIP proxy at extinction horizons.
• The end-Permian negative carbon isotope excursion magnitude.
• The coal gap and reef gap durations.
• Insect extinction at the end-Permian as their only major loss.
• Chicxulub crater diameter (~180 km) and impactor size (~10 km); the Alvarez iridium anomaly.
• Richards et al. (2015) on impact-triggered acceleration of Deccan eruption rates.
• Ocean carbonate buffering timescale (~10³–10⁴ yr) versus silicate weathering (~10⁵–10⁶ yr).