View all hidden authors and organizations

Edited by Michael Manga, University of California, Berkeley, approved on July 21, 2021 (reviewed on July 13, 2020)

The end-Permian mass extinction event (approximately 252 Mya) was the most serious loss of biodiversity in the history of the earth, and a globally recognized rapid negative carbon isotope shift. However, the trigger of this incident is still controversial. The new paired specific carbon isotope records of terrestrial and marine compounds may provide clues to this mystery. By comparing the observed data with the results of the isotope-enabled Earth system model, we found that a large amount of rapid, mainly volcanic CO2 emissions during the volcanic activity of the Siberian trap may be the trigger for carbon isotope drift and severe species extinction. Our findings provide a quantitative limit on how the rapid increase in CO2 might affect the marine ecosystem 252 Mya.

The end-Permian mass extinction event (~252 Mya) is related to one of the world's largest carbon cycle disturbances in the Phanerozoic, and is believed to be triggered by volcanic activity in the Siberian trap. Considerable carbon isotope shifts (CIE) have been found in many locations around the world, which indicates that a large amount of CO2, which consumes 13C, is imported into ocean and atmospheric systems. However, the exact size and cause of CIE, the rate of CO2 emissions, and the total amount of CO2 are still poorly understood. Here, we quantify the CO2 emissions in the Earth system model based on the new compound-specific carbon isotope records of the Finnmark platform and the astronomically tuned age model. By quantitatively comparing the simulated surface ocean pH and boron isotope pH proxies, a large amount (~36,000 Gt C) and rapid emission (~5 Gt C yr−1) of major volcanic CO2 sources (~−15%) are important for driving observations The CIE model, the sudden drop in surface ocean pH, and the extreme increase in global temperature. This indicates that a large number of greenhouse gases may have pushed the earth system to a critical point, beyond which extreme changes in ocean pH and temperature have led to irreversible mass extinctions. The relatively enlarged CIE observed in the leaf wax of higher plants indicates that the surface water of the Finnmark platform may be out of balance with the first hundred-year-scale carbon release generated by large-scale Siberian trap volcanic activity, supporting the speed of carbon injection. Our modelling work shows that carbon emission pulses are accompanied by organic carbon burial, and extensive ocean hypoxia facilitates this process.

The end-Permian mass extinction (EPME) that occurred at 251.941 ± 0.037 Mya is considered the most serious loss of biodiversity in the history of the earth (1, 2). EPME coincided with the eruption of the Siberian Trap, a huge igneous province (LIP) covering 6 million square kilometers (km2) in Siberia (3⇓ –5), Russia. The volcanic activity of this LIP is related to the SO2 and CO2 degassing produced by the threshold intrusion (6⇓ ⇓ ⇓ –10). The injection of large amounts of carbon dioxide into the atmosphere is believed to have caused severe global warming (11⇓ ⇓ –14), catastrophic ocean hypoxia (15, 16), and extreme acidification of the ocean and land (17⇓ ⇓ ⇓ –21). It is fatal to life on land and in the sea (22). So far, no agreement has been reached on the source of 13C depleted carbon that causes disturbances in the global carbon cycle, lower ocean pH, and global warming of the entire EPME. In addition, the level of atmospheric CO2 after the initial pulse of volcanic activity in the Siberian trap and through the EPME is still poorly understood (23, 24), limiting our understanding of the climate feedback that occurs during the release of greenhouse gases during this period.

To address this critical gap in our knowledge, we use a medium-complexity earth system model (ie, a comprehensive earth system model enabled by the carbon center grid [cGENIE]; SI appendix) to limit the source, speed, and total amount of CO2 emissions . A new astronomically tuned δ13C record from the well-preserved lipid biomarkers in the sediments of the Finnmark platform in Norway. The Finnmark platform is located off the coast of northern Norway on the East Barents Sea shelf. It has an expanded shallow part (an ancient water depth of about 50 to 100 m), where two drill cores (7128/12-U-01 and 7129/10- U-01) crosses the Permian-Triassic transition (Figure 1). The previous bulk organic carbon isotope record (δ13Corg) generated from the same core shows a two-step decline, with a total carbon isotope offset (CIE) of approximately 4‰ (25). Although sedimentary organic carbon is considered to be mainly from land, a small contribution from ocean organic carbon production cannot be ruled out. Here, we use the compound-specific carbon isotope analysis of long-chain and short-chain n-alkanes preserved in the marine sediments of the Finnmark platform to generate separate but directly comparable records of δ13C in the terrestrial and marine fields, respectively. EPME. Long-chain n-alkanes (n-C27 and n-C29) with strong odd-number advantages are produced by higher plant leaf waxes, and their isotopic composition (δ13Cwax) and their main carbon source (ie atmospheric CO2) (26). On the other hand, short-chain alkanes (n-C17 and n-C19) are derived from seaweeds, and their δ13C value (δ13Calgae) represents carbon in the ocean domain (27, 28). To date, only a few studies on specific carbon isotopes of EPME compounds have been reported, all of which are limited by unfavorable sedimentary facies or high thermal maturity of organic materials (29, 30). In this study, the yellow color of pollen and spores indicates that the thermal maturity of organic matter is abnormally low, indicating that the color index on the thermal scale of Batten (31) is 2 out of 7, which is equivalent to a vitrinite reflectance R0 of 0.3% . In addition, the high deposition rate of siliceous clastic sediments at the study site (discussed in Carbon Cycle Quantification Using Astrochronology and Earth System Models) allows the study of ocean and land CIEs across EPMEs in unprecedented detail. All in all, Finnmark sediment records can reconstruct individual but directly comparable carbon isotope records in terrestrial and oceanic domains. These records can be astronomically adjusted and used to quantitatively assess the source, rate and total amount of 13C depleted carbon released during the period leading to EPME The Siberian trap erupts. Using our new specific compound carbon isotope records instead of ocean carbonates has several advantages: 1) The new astrochronology allows our paired ocean and terrestrial carbon isotope records to have a time resolution of 104 years; 2) we There is no need to assume that the sedimentation rate between connection points is constant or use diachronic biotaper to compare the age used in the global compilation (24) (see Figure 4A); 3) The δ13Calgae data is not artificially smoothed as in the reference. 32 Avoid underestimating the CIE magnitude; 4) Our records are not affected by dissolution or truncation, which is a common phenomenon of shallow sea carbonates caused by hypothetical ocean acidification that occurred during EPME (18, 33). In addition, directly comparable records of atmospheric and ocean δ13C provide further insights into the true CIE size and the rate and duration of carbon emissions.

(A) Paleogeographic map of the Late Permian, including previous and current coastlines. Indicated are 1) the location of Finnmark core 7128/12-U-01 and 7129/10-U-01, and 2) the East Greenland site of Kap Stosch discussed in the references. 52, 3) GSSP site at the bottom of the Triassic in Meishan, China, and 4) Kuh-e-Ali Bashi site in Iran (66, 107). The map was revised after reference. 61. (B) Late Permian paleogeography and paleo-deep ocean surveys used in cGENIE.

The long-chain n-alkanes in the Finnmark core show a strong parity advantage, as evidenced by a carbon preference index much higher than 2 (Figure 2B and SI appendix), supporting the hypothetical low thermal maturity of the sediment (34). The average chain length (ACLC25-C33) varies between 27.5 and 29.5, and shows a trend of upward core increase (Figure 2C), where higher ACL values ​​correspond to the same core (for example, at 110-to 95 meters core depth ), and the low ACL value is related to the ecosystem dominated by ferns (seed ferns) that existed before EPME (Figure 2A). The interval with a medium ACL (116 to 108 m core depth) is related to the first negative migration in the spore peak and sedimentary organic matter δ13C record (δ13Corg) (Figure 2F), indicating that the local terrestrial ecosystem reorganization may have modified the site The original δ13Corg signal across EPME (25, 35). The abundance of short-chain n-alkanes showed more variation throughout the profile, sometimes dominating than long-chain n-alkanes, indicating that the ocean and land contributions to the total carbon pool have changed (Figure 2D).

(A) The relative abundance of the main terrestrial sporopollen taxa. (B) Carbon preference index. (C) The ACL of normal alkanes C25-C33. (D) The ratio between the abundance of C17 and C27 normal alkanes. (E) The ratio of original alkane/phytane (Pr/Ph). (F) Bulk organic δ13C value (δ13Corg) measured in this study. (G) The compound specific δ13C value of C27 and C29 normal alkanes (δ13Cwax) and the weighted average of C17 and C19 normal alkanes (δ13Calgae). Error bars represent SD based on multiple analyses of the same sample. The individual values ​​and SD are given in Table S6 of the SI appendix. The core depth refers to core 7128/12-U-01 (indicated by solid marks), and the correlation of samples from 7129/10-U-01 (indicated by hollow marks) is SI appendix, shown in Figure S2 and Table S5. Figure 3 details the astronomically tuned age model (left).

In addition to δ13Corg (Figure 2F), we report δ13Calgae as the weighted average of n-C17 and n-C19 alkanes δ13C values, and δ13Cwax as the weighted average of n-C27 and n-C29 alkanes δ13C values ​​(Figure 2G) . The δ13​​Calgae record shows a two-step negative CIE, where the first displayed magnitude is ~5‰, and the second displayed magnitude is ~1.5‰ (Figure 2G), similar to the record of δ13Corg (~5‰ and ~2 ‰ during two negative CIE periods) (Figure 2F). δ13​​Cwax showed a similar trend to the δ13Calgae record during the two negative CIEs (Figure 2G), although the amplitude was twice that of δ13Calgae. Several other Permian-Triassic Boundary (PTB) sites in China (24, 36), Australia, and Antarctica (37) can also see this terrestrial CIE enlargement. The amplified CIE amplitude in δ13​​Cwax may be partly attributable to the increase in atmospheric CO2 levels (38) and increase due to changes in terrestrial vegetation types and hydrological cycles (39). However, another possible explanation for the amplified terrestrial CIE signal is that the surface water of the Finnmark platform is out of balance with the rapid release of carbon from large-scale volcanic activity. The astronomical tuning age model (see Materials and Methods) allows detailed correlation of two negative CIEs in biomarkers: the first CIE is correlated with EPME between 251.941 ± 0.037 Mya and 251.880 ± 0.031 Mya (1), the second The CIE appears to be immediately earlier than the second extinction pulse at 251.71 ± 0.06 Mya (40) (see Figure 5 and the SI appendix). The onset of the first CIE in δ13​​Calgae occurred at the core depth of 119.1-m (Figure 2, 251.941 Mya or time 0 in our model simulation) (1), and δ13Calgae quickly dropped to 115.7 m (251.926 Mya) -34.1‰ then rebounded slightly to -32.8‰ at 112.2 m (251.910 Mya), and then reached a minimum of -34.3‰ at 110.0 m (251.900 Mya). If we define the CIE start as the fastest decline of δ13Calgae, then the CIE start duration is calculated as 15 Kyr (from 251.941 to 251.926 Mya). δ13​​Calgae is still lower than the pre-CIE value by >2‰, while δ13Corg and δ13Cwax are both lower than the pre-CIE value by >6‰, indicating that changes in terrestrial vegetation types and hydrological cycles continue to be regional in the terrestrial domain.

For many reasons, the δ13Calgae record is considered to be representative of the entire EPME's ocean dissolved inorganic carbon (DIC) system. Two factors that may affect the ocean δ13Calgae record, ocean hypoxia and changes in microbial communities (for example, reference 41) do not work on the Finnmark platform. First, in Finnmark core, the extensive hypoxic condition interval (25, 42) density bioturbation (49, 50) and non-existence developed in many locations during EPME (15, 43⇓ ⇓ ⇓ ⇓ –48) were not observed in Finnmark core. Amorphous organic substances (50, 51). This is consistent with the original alkane/phytane (Pr/Ph) ratio of most samples greater than 1 (Figure 2E). Only short intervals of hypoxic conditions were found (50), which correlated with Pr/Ph ratios of 0.6 at 116 m, 103 m, and 90 m. However, the δ13Corg and δ13Calgae records did not show a correlation consistent with temporary hypoxic conditions indicated by low Pr/Ph values ​​(SI appendix). Due to this fairly stable marine sedimentary environment, we speculate that the composition of the marine microbial community remains unchanged. This is a specious scenario, because in a similar sedimentary environment, the Permian-Triassic boundary interval of Kap Stosch (Figure 1, position 2) in East Greenland has a fairly stable hopan/sterane ratio of approximately 1.5 to 2 (52). Similar to terrestrial higher plants, the carbon isotope fractionation of seaweeds may also be sensitive to changes in atmospheric CO2 (53, 54). Increased carbon dioxide may amplify Finnmark algae biomarkers, but marine biogeochemistry may attenuate this effect, with little net change in its overall CIE signal (Figure 2F). Nevertheless, the CIE magnitude of δ13Calgae is still similar to the range of 4 to 5‰ reported from Paleo-Tethys marine carbonate (55) (Figure 3). In order to better understand the impact of large-scale carbon dioxide release on climate and environmental conditions, and thus determine the exact cause of the extinction, it is important to limit the magnitude of the offset to estimate the rate and magnitude of carbon dioxide release (23, 32, 56). Many of the published Late Permian δ13C marine field records are based on large carbonates (δ13Ccarb), most of which are derived from the carbonate platform of the Paleo-Tethys Ocean (55). A potential deviation in the marine carbonate record is the mixing of sediments that may have an impact on δ13Ccarb, as shown in the deep-sea profile during the Paleocene-Eocene thermal maximum (PETM) period (57). In addition, due to the extremely low accumulation rate, some of these carbonate rocks may be highly condensed (2, 58), which makes it difficult to distinguish EPME from the Permian-Triassic boundary (1, 59). For example, the global boundary layer profile and point (GSSP) profile of Meishan, South China (Figure 1, position 3) are affected by the extremely low deposition rate (<1 cm Kyr-1), but the CIE (~4 to 5‰) is consistent with this study. The δ13Calgae in is similar. On the other hand, one of the most expanded carbonate sections in the Kuh-e-Ali Bashi section in Iran (Fig. 1, position 4 and Fig. 3F) shows the Late Permian and Early Triassic (60). During this period, the CIE magnitude associated with EPME was about 4‰, which was similar to the CIE magnitude observed in the δ13Calgae record reported here (Figure 3), which supports that δ13Calgae represents ocean DIC. In contrast, compared with δ13Calgae and δ13Ccarb, the δ13Cwax data and δ13Cplant data of the epidermis and wood from the remains of C3 land plants in South China (24) show an increase in the CIE magnitude (Figure 4A and B). Therefore, we believe that the n-alkanes in the Finnmark profile are a faithful record of the disturbance of the global surface ocean carbon cycle due to their original preservation. Although our sampling resolution is lower than the existing record of bulk ocean carbonates, our new carbon isotope record for a specific compound is the highest of its kind in this time interval, allowing us to separate land and ocean signals (Figure 4 A and B). In addition, our records are more suitable for isotope inversion modeling, because astrochronology can well limit the age of each sample. In contrast, despite the high temporal resolution of ocean bulk carbonate records, the age of these samples is only interpolated from petrography or biostratigraphy, which makes high sampling resolution less relevant. In addition, the high deposition rate of siliceous clastic sediments on the Finnmark platform provides us with the high-quality records needed to establish high-fidelity geochronology based on cyclic stratigraphy (61).

Time series analysis. (A) Weighted average U-Pb (U-Pb) The Siberian trap LIP windowsill (red) and pyroclastic rocks (blue) reported on the date have a 2σ analysis uncertainty (5). (B) ∼100-Kyr eccentricity (green dotted line) and 20-Kyr precession (red) Gaussian bandpass filter period (passbands are 0.0085 ± 0.0025 and 0.045 ± 0.01 cycles/Kyr, respectively). (C) The time-calibrated gamma-ray series from core 7128/12-U-01 in the American Petroleum Institute (API) unit. (D) Time-calibrated δ13Corg (red) and δ13Calgae (black) across the Permian-Triassic transition. (E) The δ13Ccarb record of the Meishan section in China (2) shows the number of beds and the conodont (108). Conodont: C. m. = Clarkina meishanensis, H. c. = Hindeodus changxingensis, C. t. = C. taylorae, H. p. = H. parvus, and Is = Isarcicella staeschei. (F) The δ13C data of Kuh-e-Ali Bashi section in Iran (66, 107) show the conodont (109, 110): 1) C. changxingensis, 2) C. bachmanni, 3) C. nodosa, 4) C. yini, 5) C. abadehensis, 6) C. hauschkei, 7) H. praeparvus-H. Changxing, 8) M. ultima-S. ?mostleri, 9) H. parvus, 10) H. lobota, And 11) I. staeschei.

Synthetic proxy records of carbon isotopes from marine carbonates and fossil C3 land plant remains, sea surface temperature and pH. (A) δ13Calgae and global ocean carbonate carbon from Abad, Iran, Kuh Ali Bashi, Shahreza and Zar, Meishan in South China, Wenbu Dansang and Yanggou, northern Bhawani, Hungary and Nhi Isotope comparison of Tao in Vietnam (24). (B) Comparison of δ13Cleaf wax and δ13C of deposited leaf epidermis and wood of C3 terrestrial plants in South China (24). (C) Sea surface temperature data reconstructed using conodont fossils (circles) (24) and brachiopods (triangles) (14). The temperature data based on the conodont come from the Paleo-Tethys Mountains, including Chanakhchi, Kuh-e Ali Bashi, Meshan, Shangsi and Zal. (D) The relative change in sea surface pH based on the boron isotope substitute in the reference. 17 and reference. 20. The pink and red circles are the data of scene 1 and scene 2 in the reference. 17. The green and blue diamonds are the data of scene 1 and scene 2 in the reference. 20.

In order to quantify the carbon cycle disturbance of the entire EPME, we constructed a high-resolution astrochronological age model for the Finnmark core. The time series analysis of the 7128/12-U-01 core total gamma-ray intensity provides a floating astronomical time scale for carbon isotope records (SI appendix, materials and methods). The power spectrum of the gamma ray series shows that the main periods are at 23 m, 7.4 m, 5.5 m, 4 m and 2.7 m (SI appendix, Figure S6-S8). The evolutionary power spectrum shows that the 23-meter cycle represents the main cycle of the entire series, and the 5--7-meter cycle is the main cycle in the Permian-Triassic boundary interval (about 125-105 meters core depth); SI appendix, Figure S9). Optimization analysis of correlation coefficient and time scale shows that the best average subsidence rate is about 22.4 cm/Kyr (SI appendix, Figures S10 and S11), which can generate 432-Kyr long floating celestial chronology for Finnmark core (Figure 3)) . The initial CIE is related to the EPME start of the Meishan GSSP in South China (1) (SI Appendix, Figures S13 and S14) and set to 251.941 ± 0.037 Miaoling (Meishan) (1) to construct an absolute astronomical time scale. The Finnmark platform CIE duration (115.7 to 93.3 m core depth) contains approximately one eccentricity cycle (Figure 3A), which is equivalent to the Meishan GSSP in the Lower Triassic (SI Appendix, Figure S14). Cyclic stratigraphy of the Meishan Uranium series showed that the two extinction pulses lasted about 40 Kyr, and the largest CIE lasted about 6 Kyr (61). Other estimates of Meishan EPME interval are 60 ± 48 Kyr based on U-Pb dating (1) and 83 Kyr based on cyclic stratigraphy (62). However, the highly condensed nature of the Meishan section limits the credibility of its astrochronology. Here, the high deposition rate associated with continuous deposition (22.4 cm/Kyr vs. <1 cm/Kyr at Meishan and Daxiakou in South China) (61) provides a more reliable estimate of the duration of CIE 109 Kyr and the starting duration of CIE is 15 Kiel. Compared with previous estimates, the duration of CIE at the end of the Permian is slightly longer, which means that the carbon cycle in the ocean and atmosphere systems has a longer time to respond and recover from disturbances. The CIE at the end of the Permian began at the peak of the 100-kyr cycle on the Finnmark platform and the GSSP Meishan section (SI appendix, Figure S14), providing additional evidence for the global relevance of carbon isotope records.

The previous modeling work reversed the δ13C record of marine carbonate in the Meishan section of GSSP (32, 63), which was affected by the highly condensed nature of the site (2). When determining the carbon emission flux by assuming the isotope composition of unknown sources, inverting δ13C alone will also encounter a non-unique solution (32, 63). We use surface ocean pH records (17, 20) and ocean surface temperature records (14, 64⇓ –66) as additional constraints to determine the most reasonable carbon source and related carbon emission rate by minimizing roots, thereby improving The previous working root mean square (RMSE) (SI appendix). This method can determine the best matching δ13C value of the emitted CO2, thereby quantifying the paleoclimatic impact of this important greenhouse gas. Using the newly constructed astrochronology in the Finnmark platform, through the carbon isotope inversion based on δ13Calgae and the above-mentioned age model using the cGENIE model and further constrained by δ11B as a proxy for the pH value, the rate of CO2 change in the entire EPME is estimated (17) δ18O as Representative of temperature. The initial and boundary conditions at the end of the Permian were adapted from references. 32, except that the initial pCO2 is set to 440 ppmv according to the recent study by Li et al. (67) Use stomatal ratio and Wu et al. (24) Using the δ13C of fossil C3 plant remains. The simulations we ran considered a wide range of model parameters and assumptions related to agent interpretation (SI Appendix, Tables S1 and S2). Our model set includes seven simulations (SI Appendix, Table S3), each of which is related to the unique isotope characteristics of the carbon source mandated by the specific carbon isotope records of seaweed compounds and the astrochronology generated in this study (Figure 3 and 4). Based on the minimum RMSE between the pH value we simulated and the proxy record, we determined that the best matching scenario is related to δ13C = -15‰ (for RMSE results, please refer to the SI appendix). This most suitable scenario is related to the large pulse of CO2 from the main volcanic volcanic activity of the Siberian trap (if we assume that the δ13C of CO2 derived from peridotite is -6‰, as in References 68 and 77%, if the recovered δ13C is The remaining CO2 comes from the oxidation of sedimentary organic matter (9) or coal combustion (70), the CO2 of the earth's crust is -12‰, as in Reference 69). This situation is in contrast to the large amount of CO2 emissions (close to 36,000 Gt C) during the simulation duration of 168 Kyr and the maximum emission rate of 4.5 Gt C yr−1 (Figure 5), which is approximately half of the current carbon emission rate (71, 72). ). When δ13Calgae reached its first minimum, the atmospheric pCO2 increased from ~440 (67) to ~7,390 ppm, and the maximum pCO2 value fell within the upper limit of the recent CO2 reconstruction using independent estimates based on the carbon isotope of fossil C3 land plant remains. The entire EPME (twenty four). According to the temperature proxy data based on well-preserved brachiopod shells (13, 14), a 13-fold increase in atmospheric CO2 resulted in a global temperature rise from 25 °C to 40 °C, but slightly less than the temperature proxy data using conodonts (64⇓ -66), indicating that the δ13C source may become higher than -15‰ during the warmest interval, which is consistent with the δ13C source that changes with the change of the degassing method. Due to changes in ocean circulation patterns, the global ocean’s carbon dioxide-driven warming response may vary. Our model results show that the extreme warmth of Eastern Paleo-Tethys (SI appendix, Figure S25) compared to western Paleo-Tethys provides spatial variability, which can be further correlated with the temperature proxy data of Eastern and Western Paleo-Tethys For comparison (11), this is a feature that the one-dimensional biogeochemical box model used in (17) cannot reveal. Interestingly, our preferred δ13C source is -15‰, which is consistent with independent estimates based on the calcium isotope mass balance model (19), and lies within the range (-11 to -17‰) recommended by Gutjahr et al. (73) For PETM, it shows that the two events share a similar trigger mechanism. In fact, PETM is already connected to the North Atlantic Igneous Province, with a volume of 150,000 square kilometers (74). It is worth noting that, due to the assumed slow rate of carbon emissions, PETM is only related to the small extinction of benthic foraminifera (75, 76). In the following, we compare the carbon emission history of EPME based on our reverse modeling and literature results.

The key model results come from the best fit determined by RMSE. (A) The relationship between EPME and the timing of volcanic activity in the Siberian trap. According to Shen et al., the age of mass extinction is 251.939 ± 0.031 Ma. (106) The age of the second stage of volcanic activity in the Siberian trap is 251.907 ± 0.067 Ma, from Burgess et al. (4) The duration of pipeline degassing, contact halo, and lava degassing comes from Svensen et al. (9) Assume that the age of onset is 251.907 ± 0.067 Ma. It also shows that the extinction cessation age based on Meishan Bed 28 (5) is 251.88 ± 0.031 Ma, and the earliest second extinction age in the Triassic is 251.761 ± 0.06 Ma (40). Please note that the red curve is based on δ13Calgae, assuming that the fractionation constant between algae and DIC is 31‰, which is within the maximum fractionation range of seaweed (111⇓ –113). (B) Comparison of the δ13C forcing of surface DIC in this study and (32) from the GSSP Meishan section after loess curve fitting. (C) The simulated carbon emission rate in Gt C yr-1 in the best-fit scenario (red) and compared with the carbon emission rate in the organic matter scenario in (32) (blue) and Jurikova et al. (17) (Green). (D) Simulate the cumulative carbon emission of Gt C in the best-fit scenario (red) and compare it with the carbon emission rate of the organic matter scenario of Cui et al. (32) (blue) and Jurikova et al. (17) (Green). (E) Simulated changes in atmospheric pCO2 in ppmv from the best-fit scenario (red) and reconstructed continuous pCO2 from Wu et al. (24) Carbon isotope based on fossil C3 plant remains. F The Paleo-Tethys ocean temperature reconstructed by δ18O is compared with triangles) (14). (G) The simulated pH value of the surface ocean dropped from the best-fitting scheme (red) and compared with Jurikova et al.'s boron isotope replacement pH reconstruction. (17) And Clarkson et al. (20). The red dotted line in BG represents the steady state condition of 200 Kyr long spin.

The simulated CO2 emissions show at least two impulses that appear to be global (pulses 1 and 2) (SI appendix). Before the main emission pulse, a small amount of carbon was emitted at a peak rate of 2.7 Gt C yr-1 (average rate of 1.6 Gt C yr-1) for 1.9 Kyr (keeping the emission rate at> 1 Gt C yr -1), which It reflects the almost instantaneous drop of δ13Calgae during the beginning of CIE. This initial emission pulse was not seen in the previous modeling work (17, 32) (Figure 5C), so it may be an artifact of local effects. The two main CO2 emission pulses (labeled pulse 1 and pulse 2) and maximum carbon emission rates of 4.5 Gt C yr-1 and 1.9 Gt C yr-1 (average 2.6 Gt C yr-1 and 1.5 Gt C yr-) Related 1) and lasted 8.3 Kyr and 9.1 Kyr respectively (keeping the emission rate at> 1 Gt C yr−1). The largest emission pulse (pulse 1) is twice larger than the previous model inversion (32) estimated using the GSSP Meishan carbon isotope record (Figure 5A), and 7 times the estimated value in the reference. 17 (0.7 Gt C yr-1) Based on a forward biogeochemical model. This difference may be due to the small CIE magnitude recorded in the shallow-ocean carbonate profile, the lack of astrochronology, and the low δ13C source assumed in these previous studies (17, 32). This model is derived from the slower emission rate and lower pCO2 (~4,400 ppm) of Jurikova et al. (17) The full range of proxy estimates for pH drop and temperature rise are not captured (their Figure 2 C and H), so higher emission rates and pCO2 are needed to better match the proxy data. In the bedrock complex and magma intrusion (251.907 ± 0.067 Mya) (4), it supports the direct link between the volcanic activity of the Siberian trap and CO2 emissions (Figure 5). Increasing temperature (>10 °C) and acidification degree (∆pH ∼1), CIE magnitude (∼5‰) and CIE onset duration (15 Kyr) determine that the carbon emission rate during peak EPME is up to 7 times faster than PETM ( 0.6 Gt C yr−1 in Reference 73 and 1.7 Gt C yr−1 in Reference 75). The final carbon emission pulse lasted longer (10.3 Kyr), but the peak emission rate was much smaller at 1.0 Gt C yr-1 (average 0.8 Gt C yr-1), although it was not seen in previous work And it may be a local signal (Figure 5C). Except for the rate of carbon injection during the peak EPME, cumulative carbon emissions (36,200 Gt C) are equivalent to 3 times that of the entire PETM (10,200 to 12,200 Gt C) (Figure 5D), and may be the final driving factor for the end-Permian change. For serious ecological consequences. The ~36,000 Gt C accumulated carbon added in the first 15,000 years of our simulation is entirely derived from the volcanic activity of the Siberian trap (9, 74, 77) (SI appendix) and is similar to the estimate in the reference. 17. Although these authors allow mantle-dominated carbon emissions to continue after CIE. Nevertheless, we found that the total degassing budget of volcanic activity in the Siberian trap is very uncertain, and the source value of δ13C may deviate from -15‰, depending on the eruption method. In fact, the δ13C source value may become higher throughout the emission period, and there will be more mantle carbon during the Siberian trap is in place. More mantle carbon will result in a higher pCO2 simulated here and can explain the low pH observed after "Pulse 2" in Figure 5. Materials in the ocean (78) and catastrophic soil erosion on land (79) may increase carbon emission fluxes. Each carbon emission pulse is followed by the negative flux necessary to explain carbon isotope recovery (maximum carbon sequestration after pulse 2, at -2.8 Gt C yr-1), which indicates that organic carbon is buried in the ocean (Figure 5). Extensive ocean hypoxia (15, 16, 80) and high phosphate concentrations caused by increased continental weathering promote the increase of organic carbon burial and the primary productivity of the surface ocean (81, 82).

Similar to the trophic level, the alkalinity of the ocean is expected to increase due to the increase in the rate of silicate weathering after a large CO2 emission pulse (83). The newly erupted Siberian trap volcanic basalt may have accelerated the consumption of carbon dioxide. However, volcanic eruptions place volcanic rocks in areas with relatively low weathering rates on the earth's surface, and the expected increase in continental weathering will affect previously existing transportation-restricted areas on the earth's surface (84⇓⇓⇓⇓-89). This limited impact of CO2 decline may lead to failure of silicate weathering and the extension of warm wells to the Early Triassic (90). As an important caveat, the model does not consider clearly expressed lithology, although it includes silicate weathering feedback related to temperature. We noticed that due to the increase in CO2 and temperature (SI appendix), the weathering flux increased by 2.7 times (30 Tmol yr-1 to 79 Tmol yr-1), but the increase in weathering flux did not promote the complete recovery of pCO2 and temperature , Indicating that silicate weathering feedback that only depends on temperature is not sufficient to restore the climate system within the duration of the model simulation. The increased weathering flux resulted in a higher alkalinity in the surface ocean. When the maximum alkalinity was reached at 41 Kyr starting from CIE, the alkalinity increased from 4.5 to 5.3 mmol kg-1. The pH of the surface ocean dropped by 1.1 units, from 8.3 to 7.2, which is consistent with the pH drop of boron isotope substitutes based on well-preserved brachiopods in the Southern Alps in northern Italy (17) (0.9 to 1.1 units) Two scenarios related to whether life effects affect the δ11B-pH dependence) and selected micrites and early cemented granular rocks from the Musandam Peninsula of the United Arab Emirates (20) (assuming -34 and -36.8‰ δ11B seawater) Value, down 0.9 to 1.1 units). It should be noted that if the life impact of ancient calcified brachiopods is greater than assumed in these studies, then these pH changes may be estimated to be conservative. Higher pCO2 also resulted in lower benthic oxygen levels, starting from CIE and significantly reducing from 152 μmol kg-1 to 0 μmol kg-1 at ~9.9 Kyr. After pulse 1, the cumulative carbon emissions reached 30,800 Gt C, corresponding to the maximum pCO2 (7,844 ppmv) and consistent with the maximum simulated temperature (33.7 °C), similar to the proxy record shell based on the well-preserved brachiopod oxygen isotope (33 °C; Figure 5 BD). After the emission pulse 2 is the second peak of pCO2 (5,530 ppmv) and temperature (32.2 °C). The lowest pH is also consistent with launch pulse 1, demonstrating the importance of rapid CO2 emissions associated with volcanic activity in the Siberian trap. The model inversion strongly supports that rapid and large amounts of CO2 led to a sudden drop in pH, extreme increases in ocean surface temperature, and a common environmental impact that led to the mass extinction of 251.939 ± 0.031 Mya. During the volcanic activity of the Siberian trap, the transition from mainly flooded lava to mainly bedrock intrusions (4) coincides with the start of global CIE and EPME, and provides an effective triggering mechanism for the release of CO2 from volcanoes. Elevated mercury content in the entire EPME found in many places including China (91, 92), Canadian Arctic (93, 94) and the United States (95) supports the volatility during threshold intrusion and lava degassing Gas emissions. Recent mercury mass balance models provide additional quantitative limits on mercury loads associated with volcanic eruptions and window sill intrusion (96, 97). The main point presented here is that if the volcanic CO2 scenario is correct, it can provide basic meaning for understanding the response of the earth system to a large number of CO2 emission pulses and the relationship between rapid climate change and mass extinctions over time.

In addition to the palaeoclimate impact studied by the carbon emission model, we can also learn more about the control of CIE amplification in δ13Cwax. This amplification can be attributed to three factors: 1) the increase in atmospheric CO2 levels (38); 2) changes in hydrological conditions and vegetation (39); 3) the imbalance between the surface ocean and the atmosphere (98). The amount of amplification caused by the increased pCO2 is highly dependent on the initial pCO2 before EPME (ie, a lower initial pCO2 will produce a larger photosynthesis fraction, and vice versa). The ~440 ppm CO2 level reconstructed from the stomatal agent (67) may underestimate the pCO2 of the latest Permian (23), which may lead to an overestimation of the increase in photosynthetic fractionation. If the initial atmospheric pCO2 is higher than the value indicated by the stomatal agent (ie, ~2,800 ppm in Ref. 32), the increase in photosynthetic discrimination due to the increase in pCO2 is only ~1‰ (SI appendix), which allows an alternative to magnifying the ground Explanation of CIE. The terrestrial plant leaf wax directly samples the atmospheric reservoirs, indicating that the atmospheric changes are much larger than those observed in the equilibrium ocean response, which means that during the EPME, the changes may occur more than the equilibrium time between these reservoirs. quick. For the very rapid carbon release scenarios of Kirtland Turner and Ridgwell (99) in PETM in cGENIE, the observed atmospheric CIE is greater than the CIE in surface water. This alternative possibility indicates that due to large-scale Siberian trap volcanic activity, the surface water of the Finnmark platform is not in balance with the carbon release of the first hundred years. The significantly larger atmospheric CIE magnitude during PTB means that carbon emissions are rapid and substantial, which is consistent with our modeling results. The critical rate of the Earth’s carbon cycle disturbance may have been reached, which triggered amplified feedback and eventually pushed the Earth system beyond the threshold and led to mass extinction (100, 101).

The Late Permian shale and siltstone of the Finnmark platform are ideal for compound-specific isotope analysis and subsequent carbon cycle quantification, because the organic matter is well preserved and the marine environmental conditions remain stable. The resulting compound-specific carbon isotope records show that the first simultaneous negative shifts of δ13Calgae and δ13Cwax are 4 to 5‰ and 10 to 11‰, respectively. The first negative CIE in the Finnmark δ13C record is related to the globally recognized negative offset marking the EPME, and the second negative CIE immediately follows the second small extinction of the earliest Triassic (Figure 5). The high deposition rate associated with continuous deposition in the Finnmark core enables more reliable astrochronological estimates of the CIE onset duration of 15 Kyr and the total CIE duration of 109 Kyr. The high-resolution δ13Calgae records presented in this study are similar in shape to global CIEs seen elsewhere (Figure 4A and B) because they show similar CIE amplitudes and exhibit multiple negative CIE pulses. A medium-complexity earth system model with realistic continental structure and ocean bathymetry with the latest Permian conditions, used to simulate carbon emission history using carbon isotope inversion recorded by δ13Calgae. Our best-fit model δ13C source value is determined to be close to -15‰ by minimizing the root mean square error between the simulated pH value and the proxy estimate of the boron isotope pH value and between the simulated sea surface temperature and the proxy record of the oxygen isotope temperature. . This indicates that atmospheric pCO2 has increased by 13 times its pre-extinction level, with at least two independent pulses with a maximum rate of 4.5 Gt C yr-1. Each pulse corresponds to the in-position of volcanic activity in the Siberian trap, which is similar to most volcanic sources. Consistent (53 to 77%), the contribution of coal combustion or thermal methane during the threshold intrusion is small. Our work also highlights the sea-atmosphere imbalance as evidenced by the significant amplification of the atmospheric CIE, which supports our main conclusions about rapid and massive carbon dioxide emissions. The temperature-dependent silicate weathering feedback and organic carbon burial are not enough to promote the complete recovery of carbon isotopes to their pre-extinction levels. Atmospheric pCO2 is maintained at more than three times the pre-event level, leading to long-term Early Triassic warming. In summary According to the above, we believe that a large number of rapid and massive volcanic carbon dioxide emissions and related feedback caused the catastrophic mass extinction. It should be pointed out that despite the emergence of new astrochronology, carbon isotope records based on specific compounds, and advanced earth system modeling results in this study, there are still some unresolved issues that need to be resolved in future research. First, we want to mention here 1) the chronology of a single volcanic pulse and its exact correlation with the δ13C displacement in the sedimentary record, 2) the link between the evolution of the δ13C source and the history of pulsed CO2 degassing, which depends on the host -Rock characteristics, and 3) Due to the low data density, caution is needed when interpreting the structure of carbon isotope records from a single site. Therefore, future modeling work requires more intensive global representative ocean and atmospheric carbon isotope data and astronomically tuned age models to produce more reliable inversions and compare them with current results.

Detailed information about lithology and stratigraphy can be found in the SI appendix and references. 49. In short, the samples were collected from two parallel cores in three different units: an Upper Permian siltstone unit, and a Late Permian rubble containing pebble shale and some local apatite/ The top of the granular rock unit, and a Lower Triassic siltstone/shale unit. A total of 19 samples were selected for lipid extraction from core 7128/12-U-01, and three samples were added from core 7129/10-U-01. The sample was taken from a dark, organic-rich shale/siltstone interval, and one sample was taken from a condensed phosphate deposit (7129/10-U-01, 66.5 m). Bugge et al. correlated the two cores with each other based on lithology. (49)

The analysis of total organic carbon and total organic carbon isotope is carried out in ISO-analytical (http://www.iso-analytical.co.uk). The samples were treated with HCl to remove carbonate before analysis. Repeated analysis of the standard product shows that the SD is less than 0.5‰.

Use a Soxhlet apparatus to extract 5 to 30 g of the dried precipitate from each sample with a mixture of DCM/MeOH (7.5/1 vol/vol). The solvent containing n-alkanes is then evaporated using rotary evaporation. A small column with activated alumina was used to separate the total lipid extract. The aliphatic hydrocarbon fraction was eluted with hexane, and the polar fraction was eluted with a mixture of DCM:MeOH (1:1 vol/vol). The hydrocarbon fraction is further purified using urea addition, and then the adduct fraction (containing linear hydrocarbons) is dissolved in hexane for further analysis. The adduct fraction is injected into the gas chromatograph to test the yield and check for contamination. Utrecht University uses mass spectrometry to identify compounds, and isotope ratio mass spectrometry is used for carbon isotope analysis. For more detailed information on chromatographic methods and alkane indices, please refer to the SI appendix.

At the University of Oslo, hydrochloric acid and hydrofluoric acid are used to extract sporopollen from powdered rock samples according to standard sporopollenology methods and sieved on a 7-micron sieve. Only the main pollen groups are distinguished, and then they are combined into the main vegetation types.

The total gamma-ray intensity cycle stratigraphy of the 7128/12-U-01 (49) site provides a floating astrochronology for the carbon isotope record. The time series analysis to construct the astronomical time scale follows the typical procedure described in (102). The analysis was performed using the software Acycle version 2.2 (102), which is fully described in the SI appendix.

In order to gain insight into the most reasonable 13C depleted carbon sources during the destruction of the global carbon cycle and EPME, we use isotope inversion and the compound-specific carbon isotope data we collected in cGENIE. cGENIE is set as the boundary conditions of the Late Permian (such as paleogeography and paleo-deep ocean survey; SI appendix). The model has a one-dimensional atmosphere and a three-dimensional ocean with 16 vertical layers. cGENIE considers the following geochemical tracers in the atmosphere and ocean through ATCHEM and BIOGEM modules: O2, CO2, DIC, alkalinity, carbonate ions, stable carbon isotopes (12C and 13C) and nutrients (nitrates and phosphates) .

On a time scale of more than 100,000 years, carbon input from terrestrial silicate weathering and volcanic venting balances carbonate burial output (103). In a short time scale (that is, millennium to tens of thousands of years), exogenous carbon sources, such as methane hydrate and organic matter oxidation, will cause an imbalance in the carbon cycle, thereby driving rapid changes in atmospheric pCO2. The SI appendix tables S1 and S2 summarize the biogeochemical model parameters and rock weathering parameters. The cGENIE model ran for 20,000 years in a closed system and another 200,000 years in an open system to achieve a balance between the input of silicate weathering and degassing and the output of carbonate burial. Afterwards, seven experiments were established using a series of reasonable carbon sources (δ13C source = -9, -15, -18, -25 and -30‰), covering the δ13C of CO2 in Cullo melt inclusions in Reference 13. Value range. 69; We also include δ13Csource = -45‰ and -60‰ to cover scenarios related to thermal and biogenic methane (104, 105) Use δ13Calgae records to derive surface ocean δ13CDIC in the model inversion (for evolution, please refer to Figure 5F Surface ocean δ13CDIC record and SI appendix of related assumptions. Based on the use of the stomatal ratio proxy of the well-preserved stratum corneum in southwestern China, the initial pCO2 is assumed to be ~440 ppm, and the cGENIE model is adjusted to ice unconditionally by adjusting the albedo. The model ran for 168,600 years to cover 251.941 to 251.772 Mya for short-chain, n-alkane-based inversion (from CIE to 168.6 Kyr after CIE; age 251.939 ± 0.031 Mya is marked as re-based in Figure 5) EPME age F. 106).

All the data provided in this study are available in the supporting information. The code required to run the cGENIE model is available at https://github.com/derpycode/cgenie.muffin, and the instructions for all model output and copying model results are available at https://zenodo.org/record/4543684.

This work was funded by the Norwegian Research Council (Project 234005) funded by WMK, the US NSF (funded 2026877) funded YC, and the National Natural Science Foundation of China (funded 4207040) funded ML. We thank K. Nierop, D. Kasjaniuk and Utrecht University. A. van Leeuwen-Tolboom and K. Backer Owe of the University of Oslo provided technical support for laboratory work. We also thank A. Mørk for sampling assistance and L. Kump's discussion. YC thanks A. Ridgwell for providing access to the University of California Riverside Domino Cluster and Y. Wu to set up the launch. We thank the three anonymous reviewers whose suggestions greatly improved the quality of the manuscript. The core materials are stored and provided by SINTEF Petroleum Research in Trondheim.

Author contributions: WMK design research; YC, ML, and EEvS conducted research; YC, ML, EEvS, FP, and WMK analyzed data; YC, ML, EEvS, FP, and WMK wrote this paper.

The author declares no competing interests.

This article is directly contributed by PNAS.

This article contains online support information at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2014701118/-/DCSupplemental.

This open access article is distributed under Creative Commons Attribution 4.0 (CC BY).

Thank you for your interest in advertising on PNAS.

Note: We only ask you to provide your email address so that the people you recommend the page to know that you want them to see it, and that it is not spam. We do not capture any email addresses.

Feedback privacy/legal

Copyright © 2021 National Academy of Sciences. Online ISSN 1091-6490. PNAS is a partner of CHORUS, COPE, CrossRef, ORCID and Research4Life.