Eruption of the Siberian Traps large igneous province (LIP) is thought to have triggered the Permian-Triassic biological crisis, the largest of the Phanerozoic mass extinctions. Mercury concentration enrichments have been widely used as a proxy for volcanic inputs to sediments, especially for ancient LIP eruptions. However, detailed correlations of magmatic pulses with extinction events in the terrestrial and marine realms are not fully resolved. Here we use paired coronene (a six-ring polycyclic aromatic hydrocarbon, a high-temperature combustion proxy) and mercury spikes as a refined proxy for LIP emplacement. In records from stratigraphic sections in south China and Italy, we identify two sets of paired coronene-mercury spikes accompanied by land plant biomarker spikes, followed by a rapid decrease coinciding with terrestrial ecological disturbance and extinction of marine metazoans. Each short-term episode is likely caused by high-temperature combustion of sedimentary hydrocarbons during initial sill emplacement of the Siberian Traps LIP. These data indicate that discrete volcanic eruptions could have caused the terrestrial ecosystem crisis followed by the marine ecosystem crisis in ∼60 k.y., and that the terrestrial ecosystem was disrupted by smaller global environmental changes than the marine ecosystem.

The Permian-Triassic (P-Tr) mass extinction is composed of two separate global events: (1) the end-Permian terrestrial ecological disturbance (EPTD); and (2) the sudden global end-Permian extinction (EPE), accompanied by soil erosion and surface-water anoxia (Song et al., 2013; Kaiho et al., 2016a). The EPTD predated the EPE by tens of thousands to hundreds of thousands of years (Fielding et al., 2019). These events are hypothesized outcomes of Siberian Traps volcanism (Shen et al., 2011; Bond and Grasby, 2017; Burgess et al., 2017).

Sedimentary mercury enrichments, proxies for massive volcanic events, have been detected in dozens of P-Tr boundary sections across the globe (e.g., Sanei et al., 2012; Wang et al., 2018; Grasby et al., 2019; Shen et al., 2019a), but uncertainty remains in their interpretation. First, the EPE lagged the initial spike of mercury concentration in many marine and nonmarine settings (e.g., Wang et al., 2018; Shen et al., 2019b; Chu et al., 2020). Further, elevated mercury can be sourced from either direct atmospheric deposition from volcanic emissions or riverine inputs from terrestrial organic-matter oxidation (Grasby et al., 2013, 2017, 2019; Wang et al., 2018; Shen et al., 2019a; Dal Corso et al., 2020). Thus, we use the concentration of coronene (a six-ring polyaromatic hydrocarbon [PAH] formed by high-temperature combustion) in conjunction with mercury to identify P-Tr volcanic events.


We analyzed abundances of combustion-related and terrestrial plant–related biomarkers, mercury abundance, and total organic carbon (TOC) in sedimentary rock samples from three low-latitude shallow marine sections: Liangfengya (29°30′29.65″N, 106°52′58.26″E), 13 km west of the city of Chongqing, southwestern China; Meishan (the Global Stratotype Section and Point [GSSP] for the P-Tr boundary; 31°04′47.28′′N, 119°42′20.88′′E), Zhejiang Province, southern China; and Bulla (46°33′52.16″N, 11°38′07.02″E), between the towns of Castelrotto and Ortisei, northern Italy (Fig. 1). These sections are correlated by conodont zones and carbonate carbon isotopes (Chen et al., 2016; Kaiho et al., 2016a), and for Meishan, high-precision isotope dilution–thermal ionization mass spectrometry (ID-TIMS) U-Pb dating provides a highly-resolved age model (Burgess et al., 2014; Yin et al., 2014). Each studied sample covers 1–3 cm in height, with no gaps between samples for the critical horizons (each bed was cut into several samples; see Tables S1–S3 in the Supplemental Material1).

We used five- to six-ring PAHs as (benzo[e]pyrene + benzo[ghi]perylene + coronene) / phenanthrene (egc/phe), the coronene/phe ratio, and the coronene index (coronene/egc) to estimate combustion events, high-temperature combustion events, and combustion temperature, respectively (Kaiho et al., 2016b; Fig. S1 in the Supplemental Material); and the terrestrial plant index [n-alkane ratio (n-C27+29+31)/(n-C17+19+21+27+29+31)] to investigate soil erosion and vegetation collapse (long chain n-alkanes are widely used terrestrial plant biomarkers; Bush and McInerney, 2013). Although total PAHs are usually used as combustion proxy, we used egc/phe and coronene/phe because phenanthrene (a three-ring PAH, the most common PAH) is also formed by diagenesis, and five- to six-ring PAHs are enriched in combustion materials. More detailed methods are provided in the Supplemental Material.

We measured biomarker concentrations well above detection limits and blank sample values. The 22S/(22S + 22R) ratio of C31 homohopanes and computed vitrinite reflectance (Rc) (methylphenanthrene ratio, MPR) values in the three sections studied indicate that all the samples are “mature” at the stage of early to peak oil generation, indicating no contamination by modern hydrocarbon (Tables S1–S3). Ten TOC measurements of a limestone from Liangfengya with very low TOC yield a mean of 0.03577% with standard deviation of 0.00053%, with measured mass of carbon in all studied rock samples more than five times the lowest detection limit; the results indicate the reliability of the TOC data despite their low concentrations (Tables S4–S6).

This study identifies coincidental high values of coronene/phe and Hg/TOC as volcanic events because terrestrial biomass oxidation does not cause enrichment of the five- to six-ring PAHs. At Liangfengya, the first volcanic event (Event 1) occurred in the upper part of bed 14 to basal bed 17, from the uppermost part of the Clarkina changxingensis to the basal C. yini zone, associated with the start of gradual negative carbon isotope shift at level C (Kaiho et al., 2016a), which is correlative to a spike in coronene/phe and Hg/TOC in the lower part of bed 8 at Bulla and the middle of bed 23 in the C. changxingensis zone at Meishan, ∼60 k.y. before the EPE (Fig. 2). A major Hg/TOC spike occurred in the ash of bed 16 and in the basal part of the superjacent limestone (twice background values) at Liangfengya. At Meishan, Hg/TOC reaches three times background values, coinciding with high coronene/phe in middle of bed 23 in Event 1. A significant spike in coronene/phe occurred in all three sections. This event corresponds to the EPTD, identified by a concurrent spike in the terrestrial plant index followed by a drastic decrease. Volcanic Event 2 (beds 18–20 at Liangfengya) corresponds to the marine EPE identified by a drastic decrease in species richness and disappearance of fossils in thin sections, which is correlative to a spike in coronene/phe and Hg/TOC at the EPE the upper part of bed 8 at Bulla and at the top of bed 24, black marl (Kaiho et al., 2006), at Meishan within the same conodont zone (C. yini zone; Fig. 2). A third spike in coronene/phe, with no corresponding spike in Hg/TOC, occurred in the Hindeodus praeparvus zone or the equivalent zone of the three sections at ∼30 k.y. after the EPE.

High correlation between Hg and Hg/TOC beginning below the events and continuing to Event 2 (r = +0.83; Fig. 3D) indicates that there is little influence of the low TOC on the spikes in Hg/TOC during the events in Liangfengya. Here, low TOC (<0.2%) does not produce artificial spikes in Hg/TOC for the two events. The correlation coefficient between Hg/TOC and coronene/phe and between Hg and coronene/phe is high during the pre-events to Event 2 (r = +0.47 and +0.32, respectively) and low after the Event 2 (r = –0.16 and –0.32, respectively; Figs. 3B and 3C). These correlation variations imply that the sources of Hg subsequent to Event 2 differed from those for the lower beds. The relatively high TOC in samples predating Event 1 at Bulla causes low correlation between Hg and the other proxies (Figs. 3H–3J) because the main host of the Hg is TOC. We normalized Hg by TOC, resulting in high correlation between Hg/TOC and coronene/phe and between Hg/TOC and the coronene index (r = +0.59 and +0.72, respectively; Figs. 3F and 3G).

The coronene/phe and Hg/TOC values indicate that Event 1 was a smaller environmental change than Event 2 in the three sections. High coronene indices occurred globally only during Events 1–2, and in the third spike (0.3–0.8 relative to background average value of 0.04–0.15) (Fig. 2; Tables S1–S3). The correlation between Hg and the terrestrial plant index is high during the events (r = +0.61, +0.44) but low in the other samples (r = –0.08, –0.21) at Liangfengya and Bulla, respectively (Figs. 3E and 3J). These findings imply that the terrestrial input due to the plant crisis contributed to the Hg enrichment in shallow marine sediments. Events 1 and 2 coincided with high values of the terrestrial plant index followed by a significant decrease, showing plant organic-matter influx and terrestrial ecosystem devastation (Figs. 2A and 2B). The long-lasting high values of the coronene index compared to the short durations of Events 1 and 2 observed in the Liangfengya and Bulla sections suggest that low-magnitude volcanic eruptions persisted during the events.

Our coronene record demonstrates very high-temperature combustion of organic matter coincident with volcanism and extinction. Coronene is a highly condensed six-ring PAH and a stable molecule in Earth surface environments and can be preserved across geological time. It requires abnormally high energy to form coronene from hydrocarbons compared to smaller PAHs, as evidenced by heating experiments and thermodynamic simulations (Norinaga et al., 2009). Wildfire combustion occurs at 700–1000 °C (Pyne et al., 1996) and produces coronene indices ∼0.1, which correspond to the background values for mass extinctions (Kaiho et al., 2016b). The ∼0.1 value is also formed by 900–1000 °C in shorter-duration (a few seconds) heating, and the high coronene indices (>0.3) observed in Events 1 and 2, and the third spike of coronene, require combustion temperatures >∼1200 °C (estimated from data of Norinaga et al. [2009]). Such high-temperature combustion of organic matter to form coronene can occur with large-scale volcanic activity or asteroid or comet impacts causing high-temperature combustion of sedimentary hydrocarbon and terrestrial plants in the emplacement or target areas. Five- to six-ring PAHs, including coronene enrichment, have been found associated with three mass extinctions (Late Devonian, end-Permian, and Cretaceous-Paleogene boundary) but are absent from strata immediately below and above the mass-extinction horizons (Kaiho et al., 2013, 2016b; this study).

The coronene indices of the two events identified here are indicative of temperatures produced by sill intrusion (Aarnes et al., 2010; French and Romanowicz, 2015), and the age of the events overlaps with the dates of Siberian Traps sill emplacement (Burgess et al., 2017). Fly ash loading Events I and III of Grasby et al. (2011) and Hg/TOC spikes at nonmarine sites (Chu et al., 2020; Fig. 4) are correlated to events 1 and 2 recognized in this study based on carbon isotope (CI) correlations (CI levels C and D of Kaiho et al. [2016a]).

The synchronous occurrence of elevated coronene index and elevated Hg requires volcanic injections into the stratosphere, rather the weathering on land. Therefore, the two events reported here are associated with pulsed volcanic emissions, with the second event larger in magnitude. These data link the EPTD and EPE to volcanic eruptions. The Hg could have been sourced from gas formed in the contact aureole of sedimentary rocks, volcanic gas, and wildfire ash deposited by lava flow, ejected together by eruptions of high-pressure gas forming in the aureole, resulting in global distribution of Hg. Additional Hg was supplied from soil oxidation. The third spike of coronene/phe in the three sections studied is correlated to the largest spike of Hg/TOC in the Chinahe area (Fig. 1) between the CI levels E and F, which may be the third volcanic event coinciding with a ∼10 °C global warming detected by δ18Oapatite in several sections including Liangfengya and Meishan (Chu et al., 2020; Chen et al., 2016; Figs. 2 and 4).

The terrestrial ecological disturbance is supported by spikes in dibenzofuran/phe ratios in Bulla and Meishan during Event 1 and in many sections (Italy, south China, Japan) during Event 2 (Kaiho et al., 2016a). In high southern latitudes, the last occurrence of coal and Glossopteris and initiation of chemical weathering (recorded by the extent of feldspar alteration to clay) reported in the Sydney core, Australia, and a terrestrial vertebrate turnover in the Karoo Basin, South Africa, predates the northern low-latitude EPTD by ∼240 k.y. based on U-Pb zircon dating (Figs. 1 and 4; Fielding et al., 2019; Gastaldo et al., 2020).

Based on our correlations, the ages of the two events correspond to ages of the initial stage of laterally extensive sill emplacement and pyroclastic eruptions in the Siberian Traps LIP (Burgess et al., 2017; Fig. 4). The presence of several thousand volcanic pipes, remnants of eruptions, in the eastern Siberian Tunguska Basin imply that the laterally extensive basaltic magma intrusion (sills) regionally heated Tonian–Cambrian oil, Permian coal, and hydrocarbons in the ∼3-km-thick sedimentary rock succession to generate high-pressure volatile gases (e.g., CH4, CO2, SO2, halogen; Svensen et al., 2009) along with coronene and mercury. A series of short intense eruptions distributed their volatile load and combustion products to the stratosphere; these were globally distributed, resulting in climate changes (Svensen et al., 2009; Grasby, et al., 2017; Black et al., 2018). These products accumulated in both terrestrial and marine environments and produced the two short-term coronene-mercury events identified here. Terrestrial ecological disturbance in the high southern latitudes coincided with initial flood lavas of Siberian Traps at 252.27 Ma based on U-Pb zircon dating (Fig. 4; Burgess et al., 2017; Fielding et al., 2019; Mays et al., 2020). The huge flood-lava eruptions may have caused climate change mainly in high latitudes, causing the localized and earlier terrestrial ecological disturbance.

The temporal correlations between biotic changes and volcanism-driven environmental devastation likely unfolded as follows (Fig. 4): (1) Siberian Traps flood-lava eruptions caused high-latitude local terrestrial ecological disturbance; (2) after ∼240 k.y., Siberian Traps sill-complex growth, represented here by sedimentary coronene-mercury anomalies, led to global devegetation on land and a decrease in global marine δ13C (CI levels C to E); (3) after another ∼60 k.y., the largest volcanic eruptions produced huge amounts of coronene and mercury and caused the second terrestrial devastation, global marine mass extinction, and the carbon isotope drop (CI level E).

We detected two pulsed volcanic eruption events coinciding with the initiation of the end-Permian terrestrial ecological disturbance and marine extinction. The coronene and mercury enrichment in the first event was smaller, suggesting that the terrestrial ecosystem was disrupted by smaller environmental changes than the marine ecosystem. We interpret the tight correlation of coronene and mercury during these events to indicate high-temperature combustion of sedimentary organic matter associated with Siberian Traps sill intrusion, direct atmospheric deposition of coronene and mercury from volcanic emissions, and the accompanying volatile degassing causing the terrestrial and marine ecological disturbance. While mercury may potentially also come from soil oxidation, coronene does not and represents significantly high-temperature combustion. Therefore, our paired coronene-mercury data show that Siberian Traps volcanism caused the end-Permian mass extinction. Paired coronene-mercury spikes indicating volcanic emission causing climate changes may be useful in understanding environmental changes associated with LIP eruptions at other times in Earth history.

This study was supported by the Japan Society for the Promotion of Science, the Natural Science Foundation of China, and Donors of the American Chemical Society Petroleum Research Fund. We thank Satoshi Takahashi, Masahiro Oba, and Haijun Song for their help sampling in the field; Hideko Takayanagi for measuring carbonate carbon isotope ratios; Megumu Fujibayashi for measuring organic carbon content; and Ryosuke Saito for discussion. The manuscript was improved by comments from Mike Benton and anonymous reviewers.

1Supplemental Material. Methods, geochemical data, and supplemental figures. Please visit to access the supplemental material, and contact with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.