The Permian–Triassic interval has attracted much attention due to significant biological and geochemical events, including the end-Permian mass extinction (EPME)—the most catastrophic extinction event of the Phanerozoic1. The EPME is associated with a reduction in marine species biodiversity on the order of 80–90%2, extinction amongst tetrapods, and a possible dieback of terrestrial vegetation3. Driven by volcanism from the Siberian Traps4, the EPME is intimately linked with increased CO2, CH4 and SO2 fluxes5,6,7, heightened global atmospheric and sea surface temperatures (SST)8, intensified chemical weathering9, ozone depletion10, a reduction in marine pH11 and an expansion of anoxic, and possibly euxinic, oceanic water masses12,13. It has been proposed that the Early Triassic represents a period of climatic, geochemical, and biological instability, delaying the recovery from the EPME14,15,16,17. Multiple SST changes8,18 likely coincided with major fluctuations in ocean chemistry expressed as excursions in the carbon and sulphur isotope geochemistry of marine carbonates and evaporites14,16,17,19, followed by conditions of relative stability in the Middle Triassic15.

Despite the biogeochemical significance of the Triassic, robust sulphur isotope data are sparse, with most studies focusing on specific, short periods of time, such as the EPME20 and the Smithian/Spathian boundary14,17. These records lack temporal coverage and fail to capture long-term biogeochemical conditions for the Triassic at high resolutions. One exception is by Song et al.15, who compiled a δ34S record of carbonate-associated sulphate (CAS) from the late Permian to Middle Triassic from sections in south China. However, CAS is prone to diagenetic alteration21,22, with much of the isotopic heterogeneity of δ34SCAS records across the EPME attributed to post-depositional alteration19,23. Bernasconi et al.19 compiled a δ34Sevap record of sedimentary evaporites from the late Permian to Middle Triassic, including multiple sections across several countries in Europe. Although evaporites are less prone to diagenetic alteration19, their coverage in the sedimentary record is often sparse and not continuous, thus resulting in a lack of high-resolution δ34Sevap curves.

Constructing a high-resolution δ34S record

To address the lack of a single geographic and stratigraphic record, we have generated a high-resolution δ34Sevap curve from the Staithes S-20 borehole (NZ71NE/14; grid reference, NZ 476034E 518000N), Yorkshire, England (Fig. 1). The Staithes S-20 borehole was chosen due to its stratigraphic coverage (~ 668 m) of evaporite-bearing strata that are lithostratigraphically dated between the late Permian to Late Triassic. The Hardegsen unconformity has removed much of the Early Triassic in the Staithes S-20 borehole, although a palynological age constraint acquired from immediately above the unconformity is determined to be earliest Anisian in age (Warrington, pers. comm., 2019; see Supplementary for more information).

Figure 1
figure 1

The location of the Staithes S-20 borehole displayed by the star, and the distribution of Permian–Triassic sedimentary basins of NW Europe marked by the blue tones. The darker tones represent thicker sedimentary sequences that accumulated along the main rift axes (adapted from Ref.24).

Full size image

A total of 364 individual evaporite samples (e.g., gypsum, anhydrite, and halite) were collected at regular intervals. For gypsum and anhydrite, a drill was used to produce a fine powder for isotopic analysis, whilst for halite the sulphate was obtained through barium sulphate precipitation (see Supplementary for Methodology).

We compiled and recalibrated the global δ34Sevap curve for the Permian and Triassic, consisting of ~ 1000 δ34Sevap results (see Supplementary); our new, continuous record from a single site adds 38% more data to the global curve. All results were double-checked for their age assignment against a standardised geological timescale25. Based upon trends and inflection points in the global δ34Sevap record, we correlated the Staithes S-20 curve to generate a more robust global δ34Sevap record of the late Permian–Late Triassic; especially the Middle and Late Triassic (Fig. 2; see Supplementary).

Figure 2
figure 2

Sulphur isotope records derived from sedimentary evaporites from the Staithes S-20 borehole, northeast Yorkshire (top). The lithostratigraphy of the borehole is also displayed. The correlation between the global composite curve and the Staithes record (middle) is based primarily upon the trends and inflection points in the isotope records. The Staithes record was combined with the global record to produce a single composite curve (bottom; see Supplementary).

Full size image

Late Permian: Early Triassic sulphur isotope instability

The composite late Permian—Early Triassic δ34Sevap record exhibits substantial variability (Fig. 2), interpreted as a product of environmental changes possibly induced by Siberian Traps volcanism4. The late Permian Zechstein evaporites have an average δ34Sevap of ~ 10.9‰, before lowering to ~ 8.2‰ at the PTB. Immediately following this, δ34Sevap values exhibit a sharp increase, reaching a maximum of ~ 32‰ at ~ 250 Ma in the Early Triassic (Fig. 2). Possibly facilitated by low sulphate concentrations17,19 due to deposition of the late Permian Zechstein evaporites19, this positive excursion reflects a major perturbation in the Early Triassic sulphur cycle. In addition, with the assistance of a palynological age constraint for the Hardegsen unconformity (see Supplementary), and stratigraphic correlation with the composite δ34Sevap curve, a new rapid negative δ34Sevap excursion (on the order of 15‰) is recorded at the Olenekian/Anisian boundary (OAB) (~ 247 Ma). Following this, the δ34Sevap record exhibits an abrupt recovery to pre-excursion values of 29‰ at ~ 246 Ma.

Middle–Late Triassic sulphur isotope stability

The extreme environmental conditions that persisted during the late Permian and Early Triassic were more subdued in the Middle Triassic15,17,19. Accordingly, our δ34Sevap record exhibits a gradual and persistent decline from ~ 246 Ma in the early Anisian, before stabilising at ~ 236 Ma in the early Carnian (Fig. 2). Relative stability is maintained throughout the Carnian and the majority of the Norian.

Interestingly, we see no evidence for a substantial change in δ34Sevap during the Carnian Pluvial Event (CPE), potentially suggesting the environmental changes during the CPE had little impact on the global sulphur cycle. This is of interest, as the CPE is associated with major carbon cycle perturbations, the emplacement of the Wrangellian LIP (large igneous province) and a mass extinction event (followed by biotic radiation)26,27. It is thus intriguing that our δ34Sevap record maintains relative stability across this time interval. Higher resolution δ34Sevap records spanning the CPE, accompanied by further biogeochemical modelling, are required to confirm the apparent disconnect between the carbon and sulphur cycles during the CPE.

Our new δ34Sevap record also highlights the presence of a small positive δ34Sevap excursion (~ 4‰) prior to the Norian/Rhaetian boundary (Fig. 2), which potentially coincides with the emplacement of the Angayucham Complex (see below). Additional data are required to confirm the precise age and magnitude of this δ34Sevap excursion.

Sulphur isotope box model and pyrite burial

To explore the mechanisms responsible for the observed trends in the δ34Sevap curve, we incorporated our δ34Sevap data (compiled global dataset and the Staithes S-20 borehole data) into a sulphur isotope box model28 (see Supplementary). The model outputs predict three pyrite burial events (PBEs) during the time interval of this study, at ~ 251 Ma, ~ 246 Ma, and ~ 213 Ma (Fig. 3).

Figure 3
figure 3

Sensitivity of the modelled pyrite burial flux to changes in the fractionation factor for the chemical reduction of sulphate to sulphide (Δ34S), and subsequent pyrite formation. The bar above each model output displays the value set for Δ34S at different intervals of time. The only parameter changed between each model run is Δ34S, enabling us to test the influence of this specific paramer on the inferred pyrite burial flux.

Full size image

It should be noted however, that the fractionation factor (δ34S) associated with microbial reduction of sulphate to sulphide (and subsequent pyrite formation/burial) has been shown to vary according to a range of biological and environmental factors29,30,31. Recent biogeochemical modelling approaches suggest that variability in the δ34S of seawater sulphate during the Cenozoic can be accounted for by a shift in δ34S, reflecting a change in the locus of pyrite burial to deeper more oxygen-sparse water masses, rather than a simple change in pyrite burial rates32. Unfortunately, previous work19 did not consider a possible change in δ34S when interpreting variability in the δ34S of seawater sulphate observed for the Early Triassic.

We completed a range of sensitivity tests to determine how shifts in the δ34S affected predicted pyrite burial rates (see Supplementary for details). We explored a range of values for δ34S between −35 and −50‰ for the Early Triassic (Fig. 3). Our results suggest that changing δ34S to more negative values supress the magnitude of the pyrite burial flux inferred for the PTB and earliest Triassic, but does not eliminate it entirely from the model outputs (Fig. 3). Thus, an increase in the magnitude of sulphur isotopic fractionation associated with pyrite formation is certainly possible, which is in line with evidence for an expansion of ocean anoxia during the PTB and Early Triassic time interval13,33,34,35. This may have contributed to the positive isotope excursion reported for the Early Triassic. However, our model outputs also predict that a change in δ34S within the range tested here would have been insufficient by itself to account for the positive shift in δ34Sevap during the Early Triassic. Thus, the Early Triassic δ34Sevap excursion must require an accompanying and substantial increase in the pyrite burial flux; a prediction in line with previous work19,36.It has been suggested that elevated CO2 and CH4 emissions associated with the Siberian Traps5,7 increased Earth’s surface temperature8,18. Along with the possible dieback of terrestrial vegetation3 and environmental acidity37, this likely increased continental weathering in the latest Permian and Early Triassic9,37,38,39. Weathering liberates bio-essential nutrients and may have heightened the supply of nitrogen and phosphorus to the surface oceans33, stimulating primary productivity34,40, and hence the flux of organic matter to the seafloor34. Oceanic oxygen solubility would have been low in a warm ocean, and combined with increased organic marine snow, this would have fuelled the expansion of anoxia/euxinia in the late Permian and Early Triassic13,14,33. Microbial sulphate reduction, encouraged by heightened nutrient fluxes and low oxygen concentrations would have driven the conversion of sulphate to sulphide and promote pyrite formation41 (and a “pyrite burial event”, PBE) in the presence of reduced iron. With the expansion of anoxia, pyrite formation may have occurred more readily within the water column34,35,41, heightening the magnitude of isotopic fractionation31,42. As suggested by our model results, this process would have sequestered isotopically light sulphur (32S) from the ocean reservoir, which would have contributed to the major positive δ34Sevap excursion in the Early Triassic (Fig. 4).

Figure 4
figure 4

Sulphur and oxygen8,18 isotope records, modelled pyrite burial flux (see Fig. 3a) and environmental and biological changes for the latest Permian and Triassic. The sulphur and oxygen isotope data were fitted with a smoothing spline to produce the curves. The pyrite burial flux was calculated with the use of a sulphur isotope box model28 (see Supplementary). The predicted pyrite burial events are indicated with the abbreviation ‘PBE’. The blue line represents the calculated values, and the red dashed line illustrates the results assuming steady state. Environmental and biological events43,44 of significance are included to display how our isotope records and model outputs relate to the broader environmental context of the latest Permian and Triassic.

Full size image

Our modelling outputs predict the subsequent negative δ34Sevap excursion at the OAB was preceded by a reduction in pyrite burial to a minimum of ~ –0.02 Tmol/year at 248 Ma (Fig. 4) (assuming a δ34S value of –40‰). As before, it was necessary to test for the sensitivity of inferred pyrite burial rates to changes in δ34S, and we thus completed several sensitivity tests with a range of values between –25 and –40‰ for the time interval 249 to 247 Ma (Fig. 3). For the above range of δ34S values, estimates for the pyrite burial flux minima at ~ 248 Ma varies between −0.03 and −0.02 Tmol/year, respectively. Thus, our modelling procedure suggests that the fractionation factor for sulphate reduction and pyrite formation had little control over the reduction in the pyrite burial flux across the OAB. The isotopic composition of pyrite (δ34Spyr) has been demonstrated to correlate with sea level fluctuations30, and is of interest considering the OAB coincides with a general fall in eustatic sea level45 (Fig. 4). It is intriguing that our modelling output suggests that changes in δ34S provide a relatively minor contribution to the decline in δ34Sevap values we report for the OAB. Therefore, this time interval may reflect the expansion of anoxia and shallowing of the chemocline46 inferred for much of the Early Triassic.

The available geochemical and sedimentological data fail to highlight any single mechanism for driving the observed negative δ34Sevap excursion, and therefore we propose several mechanisms.

Oxygen isotope data suggest a reduction in SSTs during the latest Spathian and early Anisian (Fig. 4)8. Cooling of marine waters would have likely been associated with invigoration of ocean circulation and lessened water column stratification14. Under such conditions, and in broad agreement with cerium-anomaly data for the latest Spathian47, the volume of anoxic water masses would have reduced, causing a decrease in pyrite burial (Fig. 4).

Coincident with the temperature decrease is a general fall in eustatic sea level45 that would have exposed either/or previously deposited (1) pyrite-rich shales from Early Triassic continental shelves) to weathering, (2) or extensive late Permian evaporite deposits (Zechstein). The sulphate released from pyrite oxidation and/or weathering of Permian Zechstein evaporites would be isotopically depleted (in comparison to Early Triassic δ34S values of + 32‰), thus contributing to the negative δ34Sevap excursion at the OAB. This is in line with our model outputs, which suggest a reduction in pyrite burial to ~ −0.02 Tmol/year (e.g., negative pyrite burial is equivalent to pyrite weathering because the model otherwise specifies constant pyrite weathering). Using either atmospheric oxygen and/or ferric iron as oxidants, the weathering of pyrite would yield sulphuric acid48, hence exacerbating weathering rates and contributing to the high 87Sr/86Sr values at the OAB49.

The recovery of δ34Sevap values to earliest Triassic levels of 29‰ immediately after the OAB is concomitant with an increase in the pyrite burial flux to ~ 1.54 Tmol/year at ~ 246 Ma (Fig. 4) (assuming a δ34S value of –40‰). We propose this reflects a recovery from the pyrite oxidation/evaporite weathering event responsible for causing negative δ34Sevap excursion at the OAB. In line with decreasing 87Sr/86Sr values in the early Anisian39,49, a relative decline in terrestrial weathering of sedimentary sulphides and evaporites would have reduced the flux of isotopically light sulphur into the ocean reservoir. In turn, this would have ensured rates of pyrite burial outpaced those of pyrite weathering, sequestering isotopically light sulphur from the seawater sulphate reservoir, facilitating a return to previous long-term δ34Sevap values (Fig. 4).

Although the predicted pyrite burial rates after the OAB return to positive values, they are lower than the Early Triassic peak (Fig. 4). This is to be expected, since predicted rates of pyrite burial began to decline prior to the weathering event at the OAB. This may indicate a gradual increase in sulphate concentrations and water column ventilation, in line with uranium isotope data that suggest a return to more oxygenated conditions in the early Anisian50. Although organic-rich claystones in the pelagic Panthalassic Ocean suggest deposition under anoxic conditions34,51, considering the uranium isotope record50, it is likely that anoxia was restricted to oxygen minimum zones and not the entire ocean as indicated for the earliest Triassic. In addition, our model outputs are based on long-term records and changes in the global δ34Sevap curve. Although it is likely that short-term events may coincide with minor changes in δ34S, our long-term δ34Sevap curve and box model outputs are insensitive to them.

Our Middle–Late Triassic δ34Sevap record from the Staithes S-20 core shows minimal variability around a consistent value of ~ 15‰ (Fig. 2); excluding δ34Sevap data that are grouped together from literature sources. In accordance with this, our pyrite burial model output also exhibits relative stability, with minor fluctuations around steady state (Fig. 4). The stabilisation observed in δ34Sevap, and inferred for pyrite burial, is likely related to growth in the seawater sulphate reservoir17,19. Hence, more significant environmental perturbations would be required to disturb the global δ34Sevap record. The global, and long-term impact of the Siberian Traps would have ended, enabling the Earth’s climate system to re-establish more equable conditions8. Coincident with this, strontium isotope data show a general decline in the continental weathering flux38,39, thus reducing nutrient fluxes into the ocean, and stabilising the sulphur cycle52.

Global δ34Sevap data for the Late Triassic are sparse; therefore the δ34Sevap curve and model output rely heavily on the Staithes S-20 record. Towards the Norian/Rhaetian boundary there is a positive δ34Sevap excursion, which indicates an increase in pyrite burial from ~ 0.17 Tmol/year at ~ 217 Ma to ~ 1.2 Tmol/year at ~ 213 Ma (Fig. 4) (assuming a δ34S value of −35‰). Again, sensitivity tests were performed with a range of δ34S values between −25 and −50‰, yielding estimates for pyrite burial between 1.67 and 1.15 Tmol/year, respectively (see Supplementary). As before, shifting δ34S to more negative values reduced the magnitude of the predicted increase in pyrite burial; nonetheless, we still consider it a noteworthy PBE.

The precise mechanism behind this δ34Sevap excursion is currently unclear. A likely candidate is the emplacement of the Angayucham complex (Alaska, USA) at 214 ± 7 Ma53, which coincides with an oceanic warming event18, high CO2 concentrations54, and increasing humidity in Eastern Europe55 and the Alps56. Such environmental responses would have invigorated the hydrological cycle, thus increasing weathering and nutrient fluxes57, driving oceanic productivity in surface waters and oxygen consumption at depth in the water column. These environmental changes would have stimulated pyrite burial, and hence a positive δ34Sevap excursion. Tighter age constraint of the Angayucham Complex and additional δ34Sevap records over this time interval are necessary to ascertain their linkage. Why a δ34Sevap excursion is not present during a similar environmental event (CPE) is unclear and requires further investigation.

A direct comparison between LIP-induced environmental change in the geologic record and anthropogenic climate forcing is complex and ambitious. However, the fact that modern CO2 emissions are potentially 14 times greater than peak emission rates during the EPME5,11 is a matter of grave concern. The environmental changes recorded in our δ34Sevap record and the EPME lasted on the order of 10 million years before the sulphur and carbon biogeochemical cycles became stabilised. Current anthropogenic emissions have already shown a measurable impact on marine ecosystems globally58, a reduction in the pH of surface waters59, a decline in oxygen concentration60, and an increase in ocean stratification61. Understanding the long-term record of global Earth system perturbations caused by an elevation in greenhouse gases will improve our understanding of marine anoxia, weathering and pyrite burial events in the geologic record.