The Permian-Triassic mass extinction is widely attributed to the global environmental changes caused by the eruption of the Siberian Traps. However, the precise temporal link between marine and terrestrial crises and volcanism is unclear. Here, we report anomalously high mercury (Hg) concentrations in terrestrial strata from southwestern China, synchronous with Hg anomalies in the marine Permian-Triassic type section. The terrestrial sediments also record increased abundance of fossil charcoal coincident with the onset of a negative carbon isotope excursion and the loss of tropical rainforest vegetation, both of which occurred immediately before the peak of Hg concentrations. The organic carbon isotope data show an ∼5‰–6‰ negative excursion in terrestrial organic matter (bulk organic, cuticles, and charcoal), reflecting change in atmospheric CO2 carbon-isotope composition coincident with enhanced wildfire indicated by increased charcoal. Hg spikes provide a correlative tool between terrestrial and marine records along with carbon isotope trends. These data demonstrate that ecological deterioration occurred in tropical peatlands prior to the main marine mass extinction.

The Permian-Triassic mass extinction (PTME) was the most severe extinction event of the Phanerozoic, both in marine and terrestrial settings (Wignall, 2015), but the relative timing of these crises is debated (e.g., S.Z. Shen et al., 2011; Gastaldo et al., 2015; Fielding et al., 2019). A negative carbon isotope excursion (CIE) in both carbonate and organic matter (OM) is seen at the main extinction horizon and is usually attributed to release of volcanic carbon (Korte et al., 2010; Cui et al., 2013). Most proposed kill mechanisms are linked to the effects of Siberian Traps eruptions (Wignall, 2001; Grasby et al. 2011). A spike in mercury (Hg) concentrations observed at the onset of the PTME, thought to be derived from Siberian eruptions, provides a chemostratigraphic marker in marine records (Sanei et al., 2012; Grasby et al., 2013; Wang et al., 2018, 2019; J. Shen et al., 2019a). A similar Hg enrichment event has also been documented in contemporaneous terrestrial sediments (J. Shen et al., 2019b).

Marine records show widespread environmental instability prior to the PTME (e.g., S.Z. Shen et al., 2019), and a new study of the Sydney Basin (New South Wales, Australia), suggests that the collapse of southern high-latitude floras occurred significantly before the onset of marine extinctions (Fielding et al., 2019), roughly coincident with onset of northern high-latitude marine stress (Grasby et al., 2011, 2015).

To evaluate the timing and nature of the terrestrial crisis, we examined variations in fossil charcoal abundance from paleo–tropical peatland in southwestern China to explore changes in wildfire occurrence (e.g., Scott, 2010), and the carbon-isotope composition (δ13Corg) of land plant cuticles, charcoal, and bulk OM to track changes in the isotopic composition of atmospheric CO2 (Arens et al., 2000; Schneebeli-Hermann et al., 2013). In addition, we investigated sedimentary Hg concentrations, and integration of our Hg record with carbon isotope values permits chemostratigraphic correlation of terrestrial and marine records.

We examined the continental Permian-Triassic (P-Tr) transition in cored borehole ZK4703, drilled 15 km south of Fuyuan County, China (25.54151°N, 104.28994°E), and the Chinahe outcrop section, 30 km southeastern of Xuanwei City (26.13077°N, 104.35637°E), both from the border area between western Guizhou and eastern Yunnan in southwestern China (Fig. 1; Fig. DR1 in the GSA Data Repository1). Latest Permian to earliest Triassic terrestrial strata in this region include, in ascending order, the fluvial-coastal swamp facies of the Xuanwei and Kayitou Formations (Chu et al., 2016; Zhang et al., 2016). The former consists of sandstone, mudstone, and common coal beds. The associated plant fossils belong to the Gigantopteris flora and include pecopterids, gigantopterids, lycopsiales, and equisetales taxa, collectively regarded as tropical rainforest-type vegetation (e.g., S.Z. Shen et al., 2011; Chu et al., 2016). The Kayitou Formation (latest Permian to earliest Triassic age) is similar to the underlying Xuanwei Formation, but lacks coal and is shale dominated. Previous studies showed that the loss of the Gigantopteris flora occurred in the lowest Kayitou Formation (Chu et al., 2016; Zhang et al., 2016).

Organic carbon isotopes, charcoal abundance, fossil plant ranges, total organic carbon (TOC), total sulfur concentrations (TS), and aluminum (Al) and Hg contents were assessed through the P-Tr transition in the ZK4703 core and at the Chinahe outcrop. To avoid facies variation issues, only mudstone samples were processed for charcoal. Some charcoal was examined under scanning electron microscope (SEM) to confirm identification. To ensure that the charcoal concentrations were not affected by variations in the nature or abundance of organic material, its abundance was normalized to phytoclast abundance and TOC. The δ13Corg values of decarbonated samples were analyzed using an elemental analyzer (EA) coupled with an isotope ratio mass spectrometer (Thermo Finnegan DeltaV). A Multi EA 4000 analyzer was used for TOC and TS determination. Hg content was measured using a LECO AMA254 mercury analyzer (ZK4703) and a Lumex RA-915 portable mercury analyzer with PYRO-915 pyrolyzer (Chinahe). Major element abundances were measured using an Agilent 7700e inductively coupled plasma–mass spectrometer. More sampling and laboratory details are given in the Data Repository.

Organic Carbon Isotopes

The δ13Corg of bulk OM declines by –5‰ (from –25.2‰ to –30.2‰) in the lower part of the Kayitou Formation at Chinahe. Charcoal δ13Corg values also decline from –24.5‰ to a minimum value of –29.5‰ in the lower part of the Kayitou Formation (Fig. 2B). For the ZK4703 core section, palynomorph color indicates stage 2–3 on the thermal alteration scale (Fig. DR2; Batten, 1996), demonstrating the cuticles’ suitability for the carbon isotope investigation. The three ZK4703 δ13Corg curves are similar (to each other and to the Chinahe record) and show a synchronous onset of a negative CIE, although absolute values differ (Fig. 2F): bulk OM–based and charcoal-based values are 1‰–2‰ heavier than the cuticle-based values. Overall, all the δ13Corg records in the Xuanwei Formation initially show stable values around –25.5‰ to –24.5‰ before a decrease of ∼5‰ in the lowest part of the Kayitou Formation.

Paleobotanical Record

The abundant, peat-forming Gigantopteris flora is seen at six levels in the Xuanwei Formation at Chinahe, and is dominated by well-preserved, large leaves (Fig. DR3). Both diversity and abundance of this flora decline drastically at the very top of the formation at a level that corresponds to the onset of the negative CIE (Fig. 2A; Fig. DR4). Thereafter, the flora consists of a monotonous assemblage of small plants, mostly Annalepis and Peltaspermum.

Charcoal Abundances

At Chinahe, the charcoal abundance is <300 particles per 100 g rock prior to the negative CIE, but rises briefly above background levels during the onset of the negative CIE (1524 particles per 100 g at 25 m log height), and ranges from 400 to 1600 particles per 100 g in the ∼4 m interval of the uppermost part of the Xuanwei Formation to lower part of the Kayitou Formation (Fig. 2C). Similarly, in the ZK4703 record there is a sharp increase in charcoal abundance, from <400 particles per 100 g below 15 m, to >2400 particles per 100 g above 16.5 m height at the base of the Kayitou Formation (Fig. 2G). SEM observation shows that the charcoal preserves anatomical details (Fig. DR2) and has similar preservation and structures with variable size, indicating minimal transport sorting. The reported variations in charcoal abundance do not appear to be an artifact of preservation or changes in terrestrial organic delivery, because variations in preserved phytoclasts and TOC show little covariation with charcoal abundance (Fig. DR5).

Hg Concentrations and Hg/TOC Ratios

Mudstone TOC concentrations are relatively high (∼1.5 wt%) in the Xuanwei Formation (varying from 0.4 to 2.8 wt% in Chinahe, and 0.8 to 3.1 wt% in ZK4703) and modestly enriched concentrations persist into the lower part of Kayitou Formation before dropping to 0.2–0.75 wt% at 17.5 m height in ZK4703 and 0.1–0.6 wt% at 27 m log height at Chinahe (Figs. 2D and 2H). Hg and Hg/TOC rise above background levels immediately above the interval with the onset of the CIE and increased charcoal abundance (Figs. 2E and 2I). High Hg concentrations and Hg/TOC can also be observed at higher stratigraphic levels, with a peak value at 19.75 m (324 ppb) in the ZK4703 core, which is ∼50× background levels. Hg and Hg/TOC drop to the previous baseline values above 37 m at Chinahe and 25 m in ZK4703. The weak correlation between Hg and TOC concentrations (R2 <0.05, P >0.05; Fig. DR6) suggests that the Hg fluctuations are not affected by changes in TOC. Additionally, the ZK4703 core has low total sulfur contents (mostly <0.15%) which show no significant covariation with Hg values (R2 <0.01, P >0.05). Correlation between Al and Hg concentrations is also weak (R2 <0.15, P <0.05), indicating that Hg fluctuations are not controlled primarily by clay content, even if some Hg is probably adsorbed onto clay. Nonetheless, there is secular variability in the Hg/Al ratio, with very low background Hg/Al values below and above the Hg anomaly and enriched Hg/Al values within the interval (Fig. DR7).

Hg Anomaly in the Terrestrial Setting of Southwestern China

Volcanic emissions represent one of the largest natural inputs of Hg to the atmosphere (Pirrone et al., 2010), and the Hg enrichment seen in many marine P-Tr boundary sequences is thought to record large-scale Siberian Traps eruptions (Sanei et al., 2012; Grasby et al., 2015, 2017; Wang et al., 2018). Volcanic Hg emissions from this source may have been up to 10,000 Mg yr−1 (∼14× natural background; Grasby et al., 2015). Thermogenic release of Hg from baking of organic-rich sediments on contact with Siberian Traps intrusions is another potential source of Hg (Wang et al., 2019). Terrestrial plants constitute a large Hg reservoir, and so wildfires can also contribute significantly to Hg fluxes to the atmosphere and freshwater environments such as those studied here (e.g., Them et al., 2019). We propose that the Hg spikes observed in terrestrial (this work) and marine (Sanei et al., 2012; Grasby et al., 2015, 2017; Wang et al., 2018) successions provide a useful correlative tool between terrestrial and marine records along with δ13Corg trends (Fig. 3).

The onset of the main phase of marine extinctions in South China, at the top of the Clarkina yini Zone, correlates with a peak in Hg concentrations and Hg/TOC, while a second phase of extinctions at the top of the Isarcicella staeschi Zone corresponds to a rise in Hg concentrations and Hg/TOC that peaks in the following I. isarcica Zone (Wang et al., 2018, 2019). The relative magnitude of these peaks varies between sections: at Meishan, South China, the lower Hg/TOC peak is the largest (Wang et al., 2018; Fig. 3), whereas at Guryul Ravine, Kashmir, the second peak is larger (Wang et al., 2019). δ13C began to decline somewhat before the marine extinctions in the Clarkina changxingensis Zone (Fig. 3).

In the terrestrial sections of southwestern China, the floral mass extinction (and charcoal peak) starts with the onset of the CIE. Hg concentrations begin to slightly rise at the same time, while Hg/TOC shows a sharp spike at the minimum of the CIE (Fig. 2). The Hg and Hg/TOC peak 4–6 m above the terrestrial extinction level and likely correlate with the rise in Hg/TOC values seen at the end–I. staeschi Zone that saw diverse taxa disappear from Triassic oceans (Song et al., 2013). Thus, the terrestrial crisis seen in equatorial sections of China appears to predate the main marine extinction phase (which occurred near the low point of CIE), and likely dates to the late C. changxingensis Zone (Fig. 3).

Carbon Cycle Perturbation and Atmospheric Carbon Injection

Our results demonstrate that a synchronous onset of the negative CIE is present in the bulk OM, cuticle, and charcoal carbon isotope records from terrestrial settings (Fig. 2). Changes in δ13Corg values of land plant cuticles record changes in atmospheric CO2 (e.g., Schneebeli-Hermann et al., 2013). We suggest that the observed negative CIE in cuticles and fossil charcoal reflects an injection of 13C-depleted emissions associated with the Siberian Traps. Interestingly, in our study, the peak in Hg concentrations and Hg/TOC ratios also occurred after the onset of the negative CIE (Fig. 2), suggesting a decoupling between the C and the Hg records that could result from the source of these two elements, be it volcanic, thermogenic, continental runoff, wildfire, or a combination of different reservoirs. Such decoupling deserves further investigation because it suggests different mechanisms of C and Hg release and/or processing in end-Permian environments.

Recurrent Wildfire and Climate Instability

Studies have shown insignificant or constant fractionation of carbon isotopes during the burning process, and so charcoal δ13C values are a direct record of the original wood tissue δ13C composition (e.g., Hall et al., 2008). The gradual decrease in the δ13C of terrestrial plant-derived charcoal in the studied successions indicates, as discussed also for cuticle δ13C, a change in δ13C of the original peat and vegetation due to changes in the δ13C of the atmospheric CO2. The charcoal δ13C negative shift is coeval with an increase in charcoal abundance, i.e., with increased wildfire activity (Fig. 2), suggesting that the latest Permian forests experienced recurring wildfires and regrowth while the atmosphere became more 13C depleted. Additionally, burning of terrestrial plant biomass can also increase the emission of Hg into the atmosphere, and then this Hg can be scavenged and buried in sediments (e.g., Grasby et al., 2017; Them et al., 2019).

The intensification of wildfire activity at the time of terrestrial mass extinction provides evidence of the harmful climatic changes in the lead-up to terrestrial crisis. The Gigantopteris coastal swamp flora thrived in humid, warm equatorial locations (S.Z. Shen et al., 2011) and was unlikely to have been adapted to intense levels of wildfire, as evidenced by the low charcoal abundance prior to the extinction interval (Figs. 2C and 2G). Thus, the increased wildfires suggest a transition to more unstable conditions punctuated by dry periods that would have been detrimental to coastal swamp floras (Zhang et al., 2016).

Our study sheds new light on the temporal links between the deterioration in the terrestrial environment and floral extinction, and the geochemical changes that mark the PTME. Our terrestrial Hg record from the P-Tr transition shows a sharp peak contemporaneous with the disappearance of Permian flora that correlates with marine Hg records. δ13C data from cuticles and fossil charcoal, thought to reflect changes in the carbon isotope composition of the atmosphere, show an ∼5‰–6‰ negative CIE during the terrestrial flora mass extinction interval. However, this was prior to the increase in Hg concentrations. Charcoal abundance shows that the floral extinctions coincided with an increase of wildfire activity and the carbon-cycle disruption. This likely reflects a change from persistent humidity to an unstable climate with frequent drought episodes. The temporal relationships between the events show that terrestrial disruption occurred shortly (but measurably) before the marine crisis.

We are very grateful to David Bond, Thierry Adatte, and an anonymous reviewer for their constructive comments, and to editor Mark Quigley for handling our paper. This study was supported by the National Natural Science Foundation of China (grants 41530104, 41661134047, 41702015) and the UK Natural Environment Research Council’s Eco-PT project (grant NE/P01377224/1), which is a part of the Biosphere Evolution, Transitions and Resilience (BETR) Program.

1GSA Data Repository item 2020079, descriptions of sample processing and studied sections, host of Hg in sediments, and supplementary figures and data tables, is available online at, or on request from
Gold Open Access: This paper is published under the terms of the CC-BY license.