Introduction

The Siberian Traps Large Igneous Province (STLIP) is regarded as the ultimate trigger for the latest Permian mass extinction (LPME, ca. 252 Ma) and associated global-scale environmental perturbations. High-resolution zircon U–Pb dating records make the Permian–Triassic transition one of the most well-studied intervals for cause-and-effect-relationships between LIP volcanism and ecosystem perturbations1,2,3. However, U–Pb dating relies on the use of zircons in volcanic ash beds. The widespread occurrence of such zircons is mainly in the Paleo-Tethys region4,5, limiting the linkage of environmental and biological signals to well-dated fossiliferous marine successions. Although geochemical proxies such as the concentrations and isotopes of Zn6 and Ni7,8,9 have been employed to track volcanic inputs to Permian–Triassic transitional successions more widely, their utility in this regard is limited by facies dependency and susceptibility to surficial recycling processes.

For the past decade, mercury (Hg) enrichments and isotopes have been widely used to track volcanic inputs to both marine and terrestrial environments during the Permian–Triassic transition (Fig. 1 and Supplementary Fig. 1) as well as other intervals10,11,12. Massive volcanogenic Hg inputs can overwhelm normal buffering mechanisms, leading to spikes in both raw and normalized Hg concentrations (i.e., ratios of Hg to total organic carbon, Hg/TOC) in sediments10,11,12,13. Sediment Hg enrichments associated with the LPME at  far-flung sites (e.g., n = 38, Fig. 1, Supplementary Fig. 1) document a synchronous widespread increase in Hg fluxes13,14. Furthermore, Hg isotopes, especially Δ199Hg, which is stable (exhibiting no or limited fractionation) under diverse physical, chemical, and biological conditions, can provide information regarding Hg provenance and processing15. Earlier studies have demonstrated that volcanogenic Hg sources tend to dominate in offshore settings14,16 and terrestrial Hg sources in nearshore and terrestrial facies16,17,18,19. However, terrestrial sections have received far less attention (n = 9) compared to marine sections (n = 29; Fig. 1, Supplementary Fig. 1). In addition, most published records are from the paleo-Northern Hemisphere and Equatorial regions (especially South China, which accounts for 8 out of 9 terrestrial records) with no studies from the paleo-Southern Hemisphere (Fig. 1 and Supplementary Fig. 1). This geographically uneven distribution of Hg records limits our ability to assess the global footprint and impacts of the STLIP eruptions.

Fig. 1: Global paleogeography of the Early Triassic (~250 Ma).
figure 1

Adapted from Ron Blakey, http://jan.ucc.nau.edu/~rcb7/, © 2016 Colorado Plateau Geosystems Inc. Red triangles represent the study sites. The Bethulie and Ripplemead sections are located in the mid-latitude Karoo Basin, South Africa (paleo ~30–60°S), and the Eveleigh and Bunnerong sections are located in the high-latitude Sydney Basin, Australia (paleo ~60°–90°S). Circles and diamonds represent other marine and continental sections, respectively, for which mercury data have been generated. The name and locations of 9 marine and 8 terrestrial Permian–Triassic boundary sites in South China are shown in Supplementary Fig. 1, and the sources for each site are given in Supplementary Table 1.

In this study, we present Hg concentration data from four terrestrial sites in the Southern Hemisphere, including two cores (Bunnerong, 33.97°S, 151.23°E and Eveleigh, 33.90°S, 151.19°E) from the Sydney Basin, Australia, and two outcrop sections (Ripplemead, 31.49°S, 24.25°E and Bethel, 30.42°S, 26.26°E) from the Karoo Basin, South Africa (Fig. 1). To date, these are the most distant terrestrial sites from the STLIP to have been investigated for volcanogenic Hg, which was located at high northern paleolatitudes (50–80°N). We also analyzed Hg isotopes for a subset of samples from the Bunnerong and Ripplemead sections in order to track Hg sources. Furthermore, our work provides constraints on the stratigraphic position of the LPME interval in the Karoo Basin, which has been a subject of uncertainty (n.b., in thick terrestrial successions, this biological event spans a discrete stratigraphic interval rather than being recorded at a single horizon).

Results

Mercury concentration records

In the Sydney Basin, both carbon isotopic compositions of organic matter (δ13Corg) and ratios of mercury to total organic carbon (Hg/TOC) exhibit significant variations through the Permian–Triassic transition (Fig. 2 and Supplementary Fig. 2). In the Bunnerong core (Fig. 2a), the Upper Permian exhibits relatively uniform δ13Corg values of ~‒24‰ and Hg/TOC of <30 ppb/%, followed by rapid shifts near the Permian–Triassic boundary to −29‰ and >100 ppb/%, respectively. Almost identical patterns are observed in the Eveleigh core (Fig. 2b), with a few Hg/TOC peaks (to ~100 ppb/%) in the Lower Triassic. In the Karoo Basin, δ13Corg at Ripplemead varies from −27 ‰ to −24 ‰ with more negative values spanning the Permian–Triassic transition, accompanying an increase in Hg/TOC (to ~50-100 ppb/%) from background values of <40 ppb/% (Fig. 2c, Supplementary Fig. 3). At Bethulie, background Hg/TOC is <100 ppb/% but peaks to ~1000 ppb/% are observed during the Permian–Triassic transition (Fig. 2d and Supplementary Fig. 3).

Fig. 2: Timeframe, carbon isotope compositions of organic matter (δ13Corg, ‰) and carbonate (δ13Ccarb, ‰), and mercury to total organic carbon (Hg/TOC, ppb/%) profiles of selected global sections.
figure 2

a Bunnerong (Fielding et al.35 and this study for δ13Corg and Hg/TOC, respectively); b Eveleigh (Fielding et al.57 and this study for δ13Corg and Hg/TOC, respectively); c Ripplemead (this study); d Bethulie (this study); e Dalongkou17; f Lubei17; g Meishan D14; h Buchanan Lake13; and i Akkamori 214. These sections represent various depositional environments (e.g., terrestrial, shelf, slope, and pelagic) from multiple locations in the paleo-Southern Hemisphere (Sydney and Karoo basins), paleo-Equator (South China) and Paleo-Northern Hemisphere (Northwest China, Northeast (NE) Pangea, and Panthalassa). The red and blue dashed lines represent the onset of Hg enrichment and negative carbon isotope excursion, respectively. The gray rectangle represents the extinction interval for each section. The Permian–Triassic transition (P–T transition) in columns c and d represent the transitional interval within which placement of the Permian–Triassic boundary in Karoo Basin is debated (refer to Fig. 4 for more information). The red arrows in panel A represent the distribution of volcanic ash beds yielding U–Pb ages in Bunnerong35,57,62. Abbreviations: Sy = system, St = stage, Z = palynozone, M = member; Fa = facies; Changh. = Changhsingian; G. = Griesbachian; P. = Permian; c. = Playfordiaspora crenulata, m. = Protohaploxypinus microcorpus, p. = Lunatisporites pellucidus.

Mercury isotope records

The mass independent fractionation (MIF) of odd Hg isotopes (Δ199Hg) exhibits a similar increasing trend through the LPME in both study areas (Fig. 3). At Bunnerong (Sydney Basin), Δ199Hg shows a positive shift from the pre-LPME Upper Permian (+0.1 to +0.2‰) to the Permian–Triassic transition interval (~‒0.1‰) (Fig. 3a), whereas at Ripplemead (Karoo Basin), Δ199Hg values are close to 0‰ or slightly negative both below and above the LPME, with a significant increase to +0.1 to +0.2‰ within the Permian–Triassic transition interval (Fig. 3b).

Fig. 3: Mass-independent fractionation (Δ199Hg, ‰) for.
figure 3

a Bunnerong (this study); b Ripplemead (this study); c Lubei17, d Guanbachong19, e Meishan D14, f Xiakou14, g Buchanan Lake16, h Shangsi5, and i Gujo Hachiman14. The red dashed line represents the latest Permian mass extinction (LPME); the vertical gray rectangle represents the Δ199Hg values of volcanic sources (+0.02 ± 0.06‰29). The horizontal blue rectangle represents the mercury enriched interval (MEI) of each section. The horizontal bars of the Δ199Hg profiles represent standard deviations (2σ), which are smaller than the symbol size for some samples. Abbreviations: Sy = system, St = stage; Fa = facies; Changh. = Changhsingian; G. = Griesbachian; P. = Permian.

Discussion

Hg concentration data must be normalized to some other sediment parameter to avoid influences related to lithologic variation, in order to meaningfully assess levels of Hg enrichment12. Recognition of the dominant host phase of Hg is essential for normalization of Hg concentrations. Because organic matter is the most common host phase of Hg, normalization to total organic carbon (Hg/TOC) has been widely used11, but other phases such as sulfides (proxied by total sulfur, TS) or clay minerals (proxied by aluminium, Al) can host Hg under some conditions (e.g., euxinia), making TOC normalization unsuitable12,20. In the Sydney Basin, Hg exhibits a stronger correlation to TOC (r = +0.60, p < 0.01, and r = +0.79, p < 0.01, for Bunnerong and Eveleigh, respectively; Supplementary Fig. 4a) than to TS (r = +0.29, p > 0.05, and r = +0.44, p < 0.01; Supplementary Fig. 4b) or Al (r = +0.05, p > 0.05, and r = +0.08, p > 0.05; Supplementary Fig. 4c), suggesting that organic matter is the dominant host for Hg. However, in the Karoo Basin, Hg exhibits no or weak correlations to TOC (r = +0.02, p > 0.05, and r = −0.03, p > 0.05, for Ripplemead and Bethulie, respectively; Supplementary Fig. 4d), TS (r = −0.12, p > 0.05, and r = −0.07, p > 0.05; Supplementary Fig. 4e), and Al (r = −0.16, p > 0.05, and r = +0.00, p > 0.05; Supplementary Fig. 4f; note: n variably 40–152 above), implying the lack of a dominant Hg host phase and making optimal normalization of Hg concentration data uncertain. We chose to normalize Hg to TOC rather than to TS or Al for the sake of uniformity of data presentation with the Sydney Basin sections (this study) as well as other, earlier-studied sites (Fig. 2). However, caution must be exercised in such cases to avoid small denominator values (e.g., TOC < 0.2%) yielding spurious peaks in normalized Hg concentrations (e.g., Hg/TOC21).

Substantial Hg enrichments near the extinction interval support excess Hg inputs into both the Sydney and Karoo basins. Increases of Hg/TOC near the Permian–Triassic boundary regardless of TOC content (which ranges from <1% to > 30%; Fig. 2, Supplementary Fig. 2, Supplementary Note 1) document excess Hg inputs in the Sydney Basin (Fig. 2 and Supplementary Fig. 2). Although the dominant host phase(s) of Hg in the Karoo Basin are uncertain, increases in both raw and normalized (i.e., Hg/TOC, Hg/TS, Hg/Al) values indicate Hg enrichment through the extinction interval in the Ripplemead and Bethulie sections (Fig. 2, Supplementary Fig. 3). The observations of significant increases in both raw (~2–10× and ~2–8× for the Sydney and Karoo basins, respectively) and TOC-normalized Hg values (~2–10× and ~4–20×, respectively) for two separate Southern Hemisphere regions that were thousands of kilometers from the STLIP, as well as of the similarity of their Hg records to those of coeval Northern Hemisphere and Equatorial terrestrial (e.g., Dalongkou, Lubei; Fig. 2e, f) and marine successions (e.g., Meishan D, Buchanan Lake, and Akkamori-2; Fig. 2g–i), lead to the inference that these Hg enrichments reflect a global-scale process.

Although Δ199Hg has been widely used to track Hg sources during the Permian–Triassic transition14,16,18, interpretation of Hg isotopes in sediments can be complicated. Several factors influence Δ199Hg: (1) the proportions of Hg sourced from the mantle (near-zero MIF22) versus organic-rich sediments by sills and/or combustion of vegetation and soil by wildfires (negative MIF23); (2) the magnitude of Hg fractionation during atmospheric transport, which ranges from 0.1 to 0.8‰ in the modern atmosphere24; and (3) mixing of multiple Hg sources, including terrestrial runoff, atmospheric removal, and seawater loading, in terrestrial and nearshore facies16,17,25,26. In Permian–Triassic successions, Δ199Hg shows large variations in different settings, with near-zero values in deep-shelf to slope settings but negative excursions in continental, shallow-shelf, deep-basinal, and pelagic settings (Fig. 3). More negative Δ199Hg values in terrestrial (e.g., Lubei, Guanbachong, Fig. 3c, d) and nearshore settings (e.g., Meishan D, Fig. 3e) have been attributed to a larger proportion of terrestrial Hg16,17,19. On the other hand, deep-basinal and pelagic facies (e.g., Shangsi and Gujo-Hachiman, Fig. 3h, i), in which terrestrial Hg inputs are limited, provide a promising reservoir for tracking atmospheric fluxes of Hg23. For the pelagic Gujo-Hachiman section, the negative or near-zero Δ199Hg values associated with Hg concentration peaks near the LPME, document elevated fluxes of isotopically light Hg because Δ199Hg becomes more positive during atmospheric transport24. Combustion of terrestrial organic-rich sediments (e.g., coal27), which can lead to elevated Hg concentrations28 with a negative MIF in marine deposits29, was the likely dominant source of Hg at a global scale14 (Fig. 3) during the second stage of the STLIP eruptions3.

At the Southern Hemisphere study sites examined here, we infer that the positive Δ199Hg values associated with Hg peaks document long-distance transfer of volcanogenic Hg of STLIP through the atmosphere from the Northern Hemisphere, although exact transport vectors cannot be determined. Mercury in terrestrial facies is mainly derived from clastic inputs and atmospheric precipitation30, which can be readily distinguished based on negative MIF for terrestrially sourced and positive MIF for atmospherically sourced Hg15,26,30. The slope of Δ199Hg versus Δ201Hg covariation is between 1.0 and 1.36 for Ripplemead (Karoo Basin) and Bunnerong (Sydney Basin) (Supplementary Fig. 5), consistent with photoreduction of aqueous Hg(II) driven by dissolved organic matter15. Differences in the pre-LPME Permian background Δ199Hg values between these basins (ca. −0.1‰ and 0‰ for Bunnerong and Ripplemead, respectively; Fig. 3) are consistent with the Δ199Hg ranges of modern29 and ancient terrestrial facies17,25,29,30 and, thus, probably due to the basin-specific Δ199Hg values of terrestrial materials, or to variable proportions of terrestrial and atmospheric Hg delivered to each study site. However, the significant Δ199Hg increases (~0.1 to 0.2 ‰) associated with Hg peaks close to the Permian–Triassic boundary in both successions support a dominant atmospheric source of Hg to these sites during the LPME interval. Increasing inputs of terrestrially sourced Hg17,29,31 played a limited role, at most, in generating these Hg peaks because: (1) Hg contents show no or weak correlations to Al (Supplementary Fig. 4); and (2) the positive Δ199Hg shifts associated with the peaks are inconsistent with terrestrial Hg inputs, which generally have negative Δ199Hg values29. Given that the two study basins were separated by ~8,000-10,000 km, the similar increases in Δ199Hg (~0.1 to 0.2‰) in each basin are likely to represent a transregional (i.e., global) flux of Hg from a volcanogenic source (Fig. 1). These positive excursions were likely due to photoreduction of Hg(II) during long-distance transport through the atmosphere24. Moreover, the sharp decrease of pyrite sulfur isotope (δ34S) values coinciding with the extinction interval in the Bunnerong core provides evidence for increasing atmospheric sulfate concentrations, which have been previously linked to aerosol production by the STLIP32. Similarly, the 1783–1784 eruptions of the Laki volcano in Iceland (~70°N) caused coeval climate perturbations in the Southern Hemisphere33.

Increased delivery of atmospheric Hg to these far-flung basins is unlikely to have been due to emissions from regional or local volcanic sources. Apart from the STLIP, volcanism linked to regional subduction is believed to have generally intensified during the Permian–Triassic transition interval (sensu lato) due to the coalescence of Pangea4. However, the most active volcanic arcs were distributed around the Paleo-Tethys Ocean (especially its eastern margin), as evidenced by the abundance of volcanic ash beds as well as Hg-enriched beds in the South China area4,5,19. In contrast, arc volcanism in the Sydney and Karoo basins was not particularly intense (compared to South China) as Permian–Triassic boundary ash beds there are relatively rare34. Furthermore, nine volcanic ash beds that are present in the Bunnerong core around the Permian–Triassic boundary are not associated with Hg enrichments35 (Fig. 2a), consistent with the hypothesis that they represent regional volcanic eruptions, and thus played a limited role in the Hg enrichments in these basins.

Integrated analysis of Hg concentration and isotope data provides a new tool for inferring the influence of volcanism in both marine and terrestrial sedimentary records. Although negative carbon isotope excursions (NCIE) have been used previously to infer volcanogenic carbon inputs to Permian–Triassic boundary successions36 (Fig. 2), many processes can emit carbon and alter sedimentary carbon isotopic signals, complicating their linkage to volcanic events37. Mercury, on the other hand, can be transferred long distances through the atmosphere at short timescales (months to years38). It is rapidly deposited, resulting in a high-fidelity, correlatable signal of volcanic activity in the stratigraphic record. The near-synchronicity among Hg peaks, the onset of the end-Permian NCIE, and the LPME at Meishan D, where all of these features are associated with Bed 25, as well as comparable relationships in many other marine Permian–Triassic sections, points to a causal link between Hg enrichment and changes in the global carbon cycle and marine ecosystems during the latest Permian14 (Fig. 2). Owing to expanded stratigraphic resolution, the apparent temporal correpsondence among these events varies somewhat in terrestrial facies, however, being nearly synchronous in some sections18,19 and modestly diachronous in others17,31,39, e.g., Lubei and Dalongkou, where excess Hg inputs began slightly below the LPME interval and its associated NCIE17. Similarly, the initial Hg peaks in the two study cores from the Sydney Basin are also slightly below the terrestrial extinction interval and the onset of the associated NCIE (Fig. 2a, b and Supplementary Fig. 6), representing a temporal lag of ~200 to 300 kyr based on U-Pb dating of the Bunnerong core35 (Supplementary Fig. 6 and Supplementary Note 2). Volcanogenic Hg inputs to all Permian–Triassic boundary sections appear to have been synchronous at a temporal resolution of ~105 years and, therefore, are likely to have shared a dominant common source (e.g., STLIP). This finding is analogous to that for the Triassic–Jurassic boundary, at which numerous marine and terrestrial sites have yielded Hg anomalies linked to eruptions of the Central Atlantic Magmatic Province at a temporal resolution of ~105 years23,30,40,41,42.

The Hg results of the present study provide insights into the stratigraphic positions of the LPME and Permian–Triassic boundary in the Karoo Basin, about which no consensus has been reached to date (Fig. 4). Smith and Botha-Brink43 proposed a three-phased extinction event in which the major second extinction wave coincided with deposition of a laminated mudstone interval (their Facies C) in the lower Palingkloof Member. Botha et al.44 generated a U-Pb date of 251.7 ± 0.3 Ma from detrital zircons in Facies B, slightly beneath the previously proposed Permian–Triassic boundary in the Karoo Basin, and consistent with radiometric dating of marine sections in China and elsewhere (e.g., Australia). However, Gastaldo et al.45 published a U–Pb date of 252.24 ± 0.11 Ma for a putative tuff bed at the base of the Katberg Formation (Facies E) that resulted in their placement of the LPME some 25 m higher in the succession, within the lower Katberg Formation. Recently, Viglietti et al.46 undertook a statistical analysis of binned biostratigraphic data that concluded that the LPME had an extended stratigraphic range in the Karoo Basin. Based on widespread evidence from around the world that the initial Hg enrichments in Permian–Triassic boundary sections were associated with the mass extinction event in both terrestrial and marine settings14,16 (Fig. 1), we propose that the LPME interval in the Karoo Basin is in the uppermost Palinkloof Formation, slightly higher (+10 m, in the lower part of Facies D) than inferred by Botha et al.44 but substantially lower (−30 to −50 m) relative to that of Gastaldo et al.45 (Fig. 4). This level represents the upper part of the extinction interval (i.e., third extinction wave) proposed by Smith and Botha-Brink43, and it coincided with the nadir of the carbon isotope excursion47 as well as with the rapid faunal turnover phase of the protracted extinction interval proposed by Viglietti et al.46. Evidence is mounting that the LPME was not a single bioevent but multiple extinction episodes over an interval of 104–105 years in both marine48 and terrestrial ecosystems43,49.

Fig. 4: Comparison of previously proposed placements of the tetrapod extinction interval in the Karoo Basin with that of the present study.
figure 4

The absolute ages of 251.7 ± 0.3 Ma44 and 252.24 ± 0.11 Ma45 are radiometric dates from different research groups. Four different levels have been proposed for the LPME horizon, based on either U-Pb data44,45 or biostratigraphic data43,46. The Hg concentration [Hg (ppb)] profile is from the Bethulie section (this study). G = group, F = formation, M = member, AZ = tetrapod assemblage zone, Fa = facies.

Some researchers have argued that the LPME was diachronous between marine and terrestrial successions35,45, although relatively limited radiometric dating of terrestrial sections, uncertainty regarding the significance of their C-isotope records50, and the taxonomic dissimilarity of marine and terrestrial fossil records render this inference highly uncertain. Regardless, the onset of Hg enrichments either preceding or synchronous with the mass extinction interval in both marine and terrestrial sections (Fig. 2) supports the STLIP having been a global trigger for this event. The gap between the onset of Hg peaks and extinction interval on land may be due to the stratigraphically expanded nature of terrestrial successions17,39. The paleontological record in the Karoo Basin shows the same general pattern: a rich and diverse assemblage of therapsids and reptiles in the Upper Permian based on numerous finds in fluvial deposits43,51,52. During the latest Permian, within the Daptocephalus Assemblage Zone, a shift from meandering to braided streams in many regions, including South Africa53, Antarctica54, and North China55, was accompanied by the last occurrence of the zonal index fossil Daptocephalus, a changeover from non-mammalian synapsids to an archosauromorph-dominated assemblage, and a restructuring of Karoo terrestrial ecosystems. Biodiversity continued to decline until the acme of the extinction near the Daptocephalus-Lystrosaurus declivis Assemblage Zone boundary. Regional origination rates increased abruptly above this boundary, co-occurring with high extinction rates to drive rapid turnover and an assemblage of short-lived species symptomatic of ecosystem instability46, coinciding with the Permian–Triassic transition based on carbon isotope and magnetostratigraphic evidence47, and concurrent with the peak of environmental disturbance in marine sections as represented by microbial facies in China and elsewhere56.

Understanding the nature and extent of the environmental footprint of the STLIP is an important issue with regard to better interpretation of the processes surrounding the LPME. In this study, we analyzed Hg concentrations in four terrestrial Permian–Triassic boundary successions in the paleo-Southern Hemisphere as well as Hg isotopes in two of them. Our data show elevated Hg concentrations associated with the biostratigraphically determined extinction interval in all four sections. Positive Δ199Hg values indicate that the Hg inputs were atmospherically sourced and likely traveled a long distance. Our findings are consistent with the global distribution of Hg derived from the STLIP eruptions. Furthermore, these Hg anomalies allow us to establish a higher-resolution geochronological framework for the extinction interval in the Karoo Basin than has been possible to date from biostratigraphic data alone. Our findings validate the use of Hg data as a volcanic proxy in ancient sediments, demonstrating that peaks in Hg concentrations can serve as a potential chemostratigraphic fingerprint for stratigraphic correlation and event-bed markers in a range of depositional settings.

Methods

Study sections

The Sydney and Karoo basins preserve continuous terrestrial successions of Late Permian to Early Triassic age51,57. Detailed paleontological work has been carried out in these basins (palynological and vertebrate research in the Sydney and Karoo basins, respectively), allowing detailed studies of the relationships between volcanism, environmental perturbations, and ecosystem effects. The Eveleigh and Bunnerong cores represent high-paleolatitude (~60–90°S) alluvial and coastal plain successions in the Sydney Basin, Australia, and the Bethulie and Ripplemead sections represent mid-paleolatitude (~30–60°S) fluvial successions in the Karoo Basin, South Africa.

The Sydney Basin, located in eastern Australia, contains a >5000-m-thick succession of Upper Carboniferous to Middle Triassic strata58. The Upper Permian succession is especially significant in hosting numerous bituminous coal seams that represent some of the world’s largest steaming- and coking-coal resources59. Coastal exposures of Permian–Triassic transition strata occur in the northern (near Catherine Hill Bay) and southern (near Wollongong) parts of the basin35,57,60,61. The Bunnerong (33.97°S, 151.23°E) and Eveleigh (33.90°S, 151.19°E) cores are among the most complete Permian–Triassic boundary successions in the Sydney Basin35,57,62. A series of seven formations has been recognized, including the uppermost Permian (Changhsingian Stage) Wongawilli Coal and Eckersley Formation, and the lowermost Triassic (Induan Stage) Coal Cliff Sandstone, Wombarra Shale, Scarborough Sandstone, Stanwell Park Claystone, and Bulgo Sandstone. The Upper Permian strata mainly consist of mudstone and coals (deltaic to coastal facies) yielding an abundance of plant fossils (e.g., Glossopteris, Vertebraria, Lepidopteris, and Dicroidium). The Lower Triassic strata dominantly consist of sandstones with a few mudstone interbeds (alluvial facies) yielding relatively few fossils. Several volcanic ash beds are preserved in the study sections.

The Karoo Basin is one of most intensely studied areas of Permian–Triassic transition facies. The Ripplemead section (31.49°S, 24.25°E) is located near Nieu Bethesda in the Sarah Baartman District, and the Bethulie section (30.42°S, 26.26°E) is located on the farm Bethel 763 in the Xhariep District of South Africa43. These sections, situated in the southwestern part of the main Karoo Basin, contain a ~130 m thickness of the Balfour Formation (including the Elandsberg and Palingkloof members) and the Katberg Formation, which belong to the Beaufort Group (Figs. 2, 4). The Balfour Formation consists mainly of green and gray massive mudstone, greenish-gray siltstone and rare laterally accreted sandstone bodies, whereas the Katberg Formation consists of dark reddish-brown/olive-gray mudstone, siltstone and vertically accreted tabular channel sandstone bodies. The Permian–Triassic boundary extinction interval is marked by a dramatic faunal turnover from the Daptocephalus Assemblage Zone46 to the Lystrosaurus declivis Assemblage Zone43,52. These sections have been the subject of vertebrate taphonomy, sedimentology, and geochemistry investigations, making them well-suited for the present study due to their stratigraphic continuity and high-resolution biostratigraphic framework43,47,63.

Geochemical analysis

Samples were trimmed to remove visible veins and weathered surfaces and pulverized to ~200 mesh in an agate mortar for geochemical analysis. Aliquots of each sample were prepared for various analytical procedures. Organic carbon isotopes (n = 128 for Ripplemead) were analyzed at the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences (Wuhan). Samples were reacted offline with 100% H3PO4 for 24 h at 250 °C, following which the carbon isotope composition of the generated CO2 was measured on a Finnigan MAT 253 mass spectrometer. All isotope data are reported as per mille (‰) variation relative to the Vienna Pee Dee belemnite (VPDB) standard. The analytical precision is better than ±0.1‰ for δ13C based on duplicate analyses.

Hg concentrations (n = 70, 62, 127, and 152 for Bunnerong, Eveleigh, Ripplemead, and Bethulie, respectively) were analyzed using a Direct Mercury Analyzer (DMA80) at the China University of Geosciences (Wuhan). About 150 mg for siltstone samples and 100 mg for mudstone samples were used in this analysis. Results of low-Hg and high-Hg samples were calibrated to the GBW07424 (33 ± 4 ppb Hg) and GBW07403 standards (590 ± 80 ppb Hg), respectively. One replicate sample and one standard were analyzed for every ten unknowns. Data quality was monitored via multiple analyses of GBW07424 and GBW07403, yielding an analytical precision (2σ) of ± 0.5% of the reported Hg concentrations.

Carbon and sulfur concentrations (n = 128 and 152 for Ripplemead and Bethulie, respectively) were measured using an Eltra 2000 C–S analyzer at the University of Cincinnati (USA). Data quality was monitored via multiple analyses of the USGS SDO-1 standard with an analytical precision (2σ) of ± 2.5% of reported values for carbon and ± 5% for sulfur. An aliquot of each sample was digested in 2 N HCl at 50 °C for 12 h to dissolve carbonate minerals, and the residue was analyzed for total organic carbon (TOC), with total inorganic carbon (TIC) obtained by difference.

Major element abundances (n = 60 and 99 for Ripplemead and Bethulie, respectively) were determined by X-ray fluorescence (XRF) analysis of pressed powder pellets using a wavelength-dispersive Zetium spectrometer at the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences (Wuhan). Results were calibrated using GSS-35. Analytical precision based on replicate analyses was better than ±2% for major elements and ±5% for trace elements.

A subset of 43 samples (17 and 26 from Bunnerong and Ripplemead, respectively) was analyzed for mercury isotopes using a Neptune Plus multi-collector inductively coupled plasma mass spectrometer (Thermo Electron Corp, Bremen, Germany). The Bunnerong samples were analyzed at Tianjin University5, and the Ripplemead samples at the State Key Laboratory of Environmental Geochemistry, Institute of Geochemistry, Chinese Academy of Sciences, Guiyang, per methods as reported in Shen et al.23,30. Hg isotopic results are expressed as delta (δ) values in units of per mille (‰) variation relative to the bracketed NIST 3133 Hg standard, as follows:

$${\delta }^{202}{{{{{\rm{Hg}}}}}}=[{({\,}^{202}{{{{{\rm{H}}}}}}{{{{{\rm{g}}}}}}/{\,}^{198}{{{{{\rm{H}}}}}}{{{{{\rm{g}}}}}})}_{{{{{{\rm{sample}}}}}}}/{({\,}^{202}{{{{{\rm{H}}}}}}{{{{{\rm{g}}}}}}/{\,}^{198}{{{{{\rm{H}}}}}}{{{{{\rm{g}}}}}})}_{{{{{{\rm{standard}}}}}}}-1]\times 1000$$
(1)

Any Hg-isotopic value that does not follow the theoretical mass-dependent fractionation (MDF) was considered an isotopic anomaly caused by MIF. MIF values were calculated for 199Hg and expressed as per mille deviations from the predicted values based on the MDF law:

$${\varDelta }^{199}{{{{{\rm{Hg}}}}}}={\delta }^{199}{{{{{\rm{Hg}}}}}}{-}0.252\times {\delta }^{202}{{{{{\rm{Hg}}}}}}$$
(2)

Analytical uncertainty was estimated based on replicate analyses of the UM-Almadén secondary standard solution and full procedural analyses of MESS-2.