The Permian Extinction: When Earth Almost Died

The boundary has a date, and it is one of the more precisely fixed moments in deep time. High-precision uranium-lead dating of zircon at Meishan in China, the global reference section for the boundary, brackets it between 251.941 million years ago, with a stated uncertainty of 0.037, and 251.880 million years ago, with a stated uncertainty of 0.031. That is usually rounded to 251.9 million years, and more loosely still to 252. It replaced an earlier figure from Bowring and colleagues in 1998 of 251.4 million years, uncertainty 0.3. Where this article says 252 million years ago, that is a convenience for reading and not a measurement.

At that boundary the marine fossil record empties. The figure carried by our own research file, and by the standard accounts, is that roughly 96 percent of marine species died. The trilobites ended here, after surviving three earlier mass extinctions. So did the tabulate and rugose corals, most brachiopods, the fusulinid foraminifera, and the blastoid echinoderms.

On land the toll was roughly 70 percent of terrestrial vertebrate species. Most therapsids, the lineage that would eventually produce mammals, were lost, along with most large amphibians and most early reptile lineages.

The insects took the only mass extinction their record has ever registered: roughly 57 percent of insect families gone. Plant losses were severe but uneven, with the Glossopteris flora collapsing across Gondwana and the lycopsid forests devastated. Counted at the higher ranks, about 57 percent of all biological families and about 83 percent of all genera disappeared, more families than any other extinction event has taken.

Recovery was the slowest on record. Reef ecosystems essentially ceased to exist and did not return for 10 million years. The boundary is not simply a bad interval in the strata. It is the division between the Paleozoic and the Mesozoic eras, the seam where one world of life ends and another begins.

The figure traces back to Raup's 1979 rarefaction analysis, which extrapolated a species-level extinction from genus-level extinction data. The genus-level numbers underneath it are themselves a range, roughly 55.7 to 82 percent of marine genera lost depending on the dataset. The 90 to 96 percent species figure is what the extrapolation produces from those genus counts. It has been the standard citation ever since, and both our own research file and our overview of the mass extinctions carry it.

In 2016 Stanley reassessed the whole family of estimates in PNAS, using methods that separate mass-extinction loss from ordinary background extinction and correct for clustering effects in the rarefaction procedure. The conclusion was that the true species-level loss was closer to 81 percent than to 90 or 96, and that the traditional figure overstates the event once background extinction is subtracted out. Both estimates remain in active use in the literature. So the honest statement of this article's headline number is a range and a caveat: somewhere between roughly 81 percent and 96 percent of marine species depending on which method is applied, derived from genus-level fossil data by extrapolation rather than counted.

The Siberian Traps, in central Siberia, Russia, are the primary cause of the extinction. Lava originally covered something like 7 million square kilometers, of which about 2 million square kilometers remain exposed today after erosion. The erupted volume is put at about 4 million cubic kilometers, some estimates higher. The eruptions ran for roughly 2 million years, with peak activity concentrated into about 300,000 years at the boundary itself.

What the eruptions put into the air matters more than the lava. Estimated volatile release runs to about 170,000 gigatonnes of carbon dioxide and about 7,800 gigatonnes of sulfur dioxide, plus fluorine, chlorine, and mercury. Uranium-lead dating places the onset of the eruptions in precise coincidence with the onset of the extinction, the result published by Burgess, Bowring and Shen in 2014.

The detail that turns a very large eruption into a planetary emergency is what the magma passed through on its way up. The Siberian magmas intruded coal, carbonate, and evaporite deposits and cooked them. Contact metamorphism of the coal alone may have released about 100,000 gigatonnes of carbon dioxide, dwarfing the lava's own direct emissions. The province did not simply erupt. It baked a continent's worth of buried carbon and vented it into the sky.

Two chemical signatures tie the extinction to that volcanism directly, without relying on timing alone. Mercury anomalies appear in boundary sediments worldwide, first identified by Sanei and colleagues in 2012, and nickel isotope anomalies in the boundary layers trace to Siberian magmatism. This is the volcano's own material, found in the rock that records the dying.

The numbers carry real uncertainty and it is worth saying where. Published estimates of the erupted volume span roughly 1 to 16 million cubic kilometers, depending on how much erosion and sedimentary burial each estimate corrects for, with 2 to 4 million cubic kilometers the common working figure. Estimates of the original area reach about 5 million square kilometers in the wider literature against the 7 million square kilometers our file gives. So much of the province is buried or eroded that the original extent is not fully recoverable. One caution on the arithmetic, since this article has to be honest about its own library: our overview file on the mass extinctions states the Siberian Traps volume as 7 million cubic kilometers, which appears to be the area figure wearing the wrong unit. Area and volume are different quantities, and this article keeps them apart.

Start with heat. Oxygen isotope data show tropical sea surface temperatures reaching about 40 degrees C, lethal for most marine life. For the global mean rise our own sources disagree, and this article will not pretend otherwise: the primary research file for this event gives 5 to 8 degrees C, while our overview of the mass extinctions gives 8 to 10 degrees C for the same event. Both ranges appear somewhere in the primary literature, depending on whether tropical sea surface or true global mean temperatures are meant and which proxy dataset is used. The disagreement is real and it has not been adjudicated here.

Then the oxygen goes. Uranium isotopes, framboidal pyrite, and black shales record widespread oxygen depletion in the oceans and euxinic, hydrogen-sulfide-bearing conditions in shallow water. The anoxia developed over roughly 100,000 years and peaked at the extinction. An anoxic ocean is not a stressed habitat. For most of what lived in it, it is not a habitat at all.

The boron isotope proxy records a fall in ocean pH of about 0.7 units, coinciding with the carbon dioxide pulse. That is ocean acidification, and it falls hardest on exactly the organisms that build carbonate skeletons, which is to say the corals and shelled invertebrates that dominate the list of the marine dead.

Above the water, sulfur isotope anomalies, vegetation collapse, and spikes in soil erosion point to acid rain driven by episodic pulses of volcanic sulfur dioxide. Separately, malformed spore tetrads in boundary sediments carry a signature of ultraviolet-B damage, consistent with an ozone layer thinned by the halogens the eruptions released.

Two entries close the list. A large negative carbon isotope excursion records a massive input of light carbon, from methane clathrates or from thermogenic gas cooked out of the coal, or both. And polycyclic aromatic hydrocarbons and charcoal layers at some localities record wildfire.

One hypothesis proposes that the worst of the carbon did not come from the volcano at all. Rothman and colleagues argued in 2014 that Siberian volcanism seeded the oceans with nickel, lifting a nutrient limit on methanogenic archaea, and that a proliferation of Methanosarcina then produced methane in quantities no eruption could match, amplifying the warming far beyond what volcanism alone could achieve. The carbon isotope excursion is consistent with biogenic methane, and a nickel concentration spike in South China sits where the argument needs one. It is an elegant hypothesis and a genuinely difficult one to test, which is exactly why it sits at Tier 2 and not higher.

Burgess, Bowring and Shen dated the extinction interval itself at Meishan in 2014, bracketing it between two volcanic ash beds. The interval lasted 60,000 years, with an uncertainty of 48,000. That is the duration of the documented extinction as a whole. In geological terms it is an instant. In any other terms it is a very long emergency.

Within the crisis, the extinction came in two distinct pulses, an initial phase and a more severe second one. That is a separate finding from a separate paper: Song and colleagues identified two pulses of marine extinction in the fossil record in China in 2013 and put roughly 180,000 years between them. Our own research file attributes the two-pulse separation to the 2014 dating paper and gives the gap as 60,000 years, which merges two different measurements from two different studies under one citation. The 60,000 year figure is the length of the whole documented interval. The gap between the pulses is a different number from a different source.

The two published figures are not straightforwardly compatible with each other, and this article does not reconcile them. They are not measuring the same thing: Burgess, Bowring and Shen timed a stratigraphic interval bracketed by two dated ash beds, while Song and colleagues measured the spacing between two extinction pulses read from fossil occurrences. Neither paper reports the other's number, and how the one sits inside the other is not something our sources settle. What can be said cleanly is that the version in our file, which fuses them into a single citation, is not what either paper reported.

So no, this was not one sudden catastrophe. The research file's own Tier 4 section names that framing as a popular misconception and corrects it: the extinction unfolded over 60,000 years or more, in at least two pulses. Geologically fast is not the same as instantaneous, and the difference is not a technicality. This was not a bad afternoon. It was a crisis with a beginning, a pause, and a second and worse act.

In boundary sections around the world, in Australia, Greenland, Israel and India, the fossil record shows a dramatic spike in spores, chiefly of the genus Reduviasporonites. The standard reading is that these are fungi, and that the spike records fungal decomposers dominating ecosystems after a mass die-off of plant and animal biomass on a scale that left the world covered in dead material. The spore-dominated interval lasted tens of thousands of years.

That reading is genuinely disputed. Multiple studies, gathered in a 2016 review in Paleogeography, Paleoclimatology, Paleoecology, argue on geochemical and morphological grounds that Reduviasporonites may be a fossil alga rather than a fungus at all. The spike is not found in every studied boundary section, and where it is found it does not always sit precisely at the boundary. None of that makes the spore layer unreal or unimportant. It does mean that a planet given over to fungus is an interpretation with a live rival, not a settled fact, and our file presents it as settled.

In 2001, and again in 2004, Becker and colleagues reported evidence for an asteroid impact at the Permian-Triassic boundary: fullerenes carrying trapped extraterrestrial noble gases, and shocked quartz. They proposed the Wilkes Land structure in Antarctica, a gravity anomaly roughly 500 kilometers across, as the impact site. The first of those papers ran in Science. This was a serious claim, seriously made.

It has not held. The fullerene results were not reproduced in independent analyses of boundary sections in China, Canada and Austria, and the extraterrestrial fullerene signature is equivocal in any case, since fullerenes also occur in volcanic ash. The reinterpretation of the shocked quartz is disputed and the Wilkes Land structure is poorly dated. The current consensus is that there is no convincing evidence for a major impact at this boundary and that Siberian Traps volcanism is sufficient to explain the extinction on its own. The impact hypothesis remains a minority position rather than an excluded one, and the distinction is worth keeping.

The status of Wilkes Land can be put more sharply than poorly dated. The structure lies buried under the Antarctic ice sheet with no bedrock samples available, so the diagnostic tests for an impact cannot be run at all. And at Permian-Triassic boundary outcrops in the Transantarctic Mountains there is no well-defined ejecta layer, which is what an impact of the proposed scale should have left behind. Some micrometeorite evidence in the boundary beds at Graphite Peak has been offered as a possible link. Possible is as far as it goes. This is unconfirmed, not refuted.

At the time of the extinction every continent was assembled into the single supercontinent Pangaea. The argument is that this geography amplified everything. A supercontinent has far less continental shelf than scattered continents, and shallow marine habitat is where most marine diversity lives. A single continuous ocean lets a toxin or an anoxic water mass reach everywhere. And a continental interior that far from any coast has a climate no fragmented landmass produces. Our research file names this a plausible but untestable counterfactual, which is the correct label for it.

Some models sharpen the picture. They give a Pangaean world of extreme seasonality, monsoon-dominated, with desert cores in the continental interiors, and marine life concentrated in the Panthalassic Ocean and the Tethys Sea, exposed to a single ocean-wide anoxic event with far fewer refugia than today's fragmented oceans offer. It is a coherent reconstruction and it follows from the geography. It is also a model result, and no model of Permian climate can be checked against a Permian measurement.

The strongest version of this material is the cascade reading: that the extinction was not one mechanism but a sequence of tipping points, each stage making the next more likely. Volcanism drives warming; warming releases methane; methane drives more warming; ocean circulation shuts down; anoxia spreads; hydrogen sulfide accumulates; the ozone layer is destroyed. Read that way, the Great Dying is not a larger version of an ordinary extinction but a different kind of event, a runaway. Modern climate science recognises cascade risks of the same shape at far lower magnitude, and the Permian-Triassic event may be the worst case the planet has on file for carbon-cycle disruption. It is a compelling account. It is also, so far, a story fitted to the evidence rather than one the evidence compels.

The same fundamental mechanism operates today: carbon dioxide driven warming, leading to ocean acidification and ocean anoxia. The scale is much smaller and the rate appears to be much faster. Ocean dead zones are already expanding. The Permian-Triassic extinction is increasingly cited in climate policy discussion as a deep-time analogue for unmitigated emissions, and as an analogue it does real work, because it is the one case where the whole cascade actually ran to completion.

The comparison needs one correction, and it is a citation problem rather than an argument problem. Our research file supports its rate claim with a paper it dates to 2021, by Cui and colleagues, on slow carbon release during the Paleocene-Eocene Thermal Maximum. That paper is real, but it was published in 2011, and it is about the PETM, a warming event roughly 56 million years ago and an entirely different episode from the one this article covers. It cannot support a Permian-Triassic rate claim, and this article does not use it. The paper that does address the comparison is Cui and Kump, 2015, in Earth-Science Reviews, and its finding is sharper than the blanket version: the peak rate of carbon addition during the end-Permian extinction was likely smaller than the present rate of fossil fuel burning, while the total carbon budget released across the whole event exceeds the entire modern fossil fuel reservoir. Slower, and far larger. That is the comparison worth making, and it is not the one usually made.

No, Earth was not nearly sterilized. The claim that the planet came close to becoming lifeless overstates what happened. The extinction was devastating, and significant biodiversity survived in all major ecosystems; life was severely reduced but never close to total extinction. This article's own title says Earth almost died, and that is a figure of speech about how near the biosphere came to collapse, not a claim about sterilization. The honest superlative is that nothing else in the fossil record comes this close, not that life nearly ended.

No, there is no alien hand in this. Fringe claims connecting the Permian-Triassic extinction to alien intervention or to directed panspermia have no supporting evidence and are not taken seriously in any scientific context. There is nothing here to weigh, and saying so is not a dismissal, it is the state of the evidence.