The end-Permian mass extinction, the largest biological crisis in Earth history, is currently understood in the context of Siberian Traps volcanism introducing large quantities of greenhouse gases to the atmosphere, culminating in the Early Triassic hothouse. In our study, the late Permian and Early Triassic atmospheric CO2 history was reconstructed by applying the paleosol pCO2 barometer. Atmospheric pCO2 shows an approximate 4× increase from mean concentrations of 412–919 ppmv in the late Permian (Changhsingian) to maximum levels between 2181 and 2610 ppmv in the Early Triassic (late Griesbachian). Mean CO2 estimates for the later Early Triassic are between 1261–1936 ppmv (Dienerian) and 1063–1757 ppmv (Spathian). Significantly lower concentrations ranging from 343 to 634 ppmv are reconstructed for the latest Early to Middle Triassic (Anisian). The 5 m.y. episode of elevated pCO2 suggests that negative feedback mechanisms such as silicate weathering were not effective enough to reduce atmospheric pCO2 to precrisis levels and that marine authigenic clay formation (i.e., reverse weathering) may have been an important component of the global carbon cycle keeping atmospheric pCO2 at elevated levels.

The end-Permian mass extinction (ca. 252 Ma) coincided with the onset of intrusive Siberian Traps volcanism, which was likely responsible for outgassing of large quantities of CO2, CH4, and halogens by thermogenic heating of volatile-rich sediments (Courtillot and Renne, 2003; Svensen et al., 2009; Burgess and Bowring, 2015). The inferred increase in greenhouse gas concentrations has been interpreted to have resulted in a dramatic 8–10 °C increase in low-latitude sea-surface temperature (SST), with high ocean temperatures persisting into the Early Triassic (e.g., Sun et al., 2012; Joachimski et al., 2020). However, a proxy record for atmospheric pCO2 has yet to be established for the late Permian to Early Triassic.

We reconstructed the late Permian to Middle Triassic atmospheric CO2 record by applying the carbonate paleosol pCO2 barometer to soil carbonates from sections in northwest China (Xinjiang Province), north China (Henan and Shanxi Provinces), Russia (South Ural foreland basin), South Africa (Karoo Basin), and the United Kingdom (Dorset) (Fig. 1; Table S1 in the Supplemental Material1). Stratigraphically, the samples cover the Changhsingian (late Permian) to earliest Griesbachian, late Griesbachian to Dienerian, and Spathian to Anisian (Early to Middle Triassic). Reconstructed atmospheric CO2 levels suggest an approximate 4-fold increase in pCO2 from the latest Permian to Early Triassic, high to intermediate CO2 levels in the Early Triassic, and a decline to precrisis levels in the latest Early Triassic.

We calculated atmospheric CO2 concentrations from the carbon isotopic composition of microsampled micritic soil carbonate precipitated in the well-drained soils by applying a two-component carbon isotope mixing model, given that soil CO2 is a mixture of two isotopically different CO2 sources (soil-respired CO2 and atmospheric CO2; Cerling, 1991):


Here, S(z) is the soil-derived (respired) component of total soil CO2, and δ13Cs, δ13Cresp, and δ13Catm represent the carbon isotopic compositions of total soil CO2, soil-respired CO2, and atmospheric CO2, respectively. The δ13Cs, δ13Cresp, and δ13Catm values were calculated from the measured δ13C value of pedogenic carbonate, soil organic matter (SOM) occluded in the carbonate nodules, and marine carbonate, respectively.

The soil-derived contribution of CO2, S(z), and the soil temperature must be assumed, presenting the main uncertainties for paleo-pCO2 reconstructions. While earlier studies used high S(z) values of 5000 ppmv (e.g., Ekart et al., 1999), representing soil CO2 concentrations of the mean growing season, later studies observed that pedogenic carbonate precipitation mainly occurs during warm and dry periods characterized by a relatively low soil pCO2 due to moisture-limited soil respiration (Breecker et al., 2009). We assumed S(z) values between 500 and 1500 ppmV following Montañez (2013), who used the δ13C difference between modern soil carbonate and SOM (Δcarb-org) to constrain S(z) for different soil orders (see the Supplemental Material for details).

Soil temperature has to be estimated in order to calculate δ13Cresp from the δ13C value of pedogenic carbonate, utilizing the temperature-dependent carbon isotope fractionation between CO2 and CO32– (Romanek et al., 1992). Soil temperatures for the calcisols formed at various paleolatitudes were estimated using low-latitude SSTs (Sun et al., 2012; Joachimski et al., 2020), a low-latitudinal temperature gradient for warm climatic conditions (0.2 °C per 1° latitude; Zhang et al., 2019), air temperatures 3 °C lower than SSTs (Zhang et al., 2019), and 2–4 °C warmer soil temperatures compared to atmospheric temperatures (Hu and Feng, 2003; Table S2).

Atmospheric pCO2 was calculated for each possible combination of S(z), soil temperature (Table S2), and δ13C of atmospheric CO2 (Table S3) by varying S(z) from 500 to 1500 ppmv in 100 ppm steps, increasing soil temperature in 1 °C steps (using a ±5 °C temperature range around estimated average soil temperature), and δ13C of atmospheric CO2 in 0.1‰ steps. Mean atmospheric CO2 concentrations for each sample were then calculated based on the distribution of all possible values. The 95% credible interval (CI) for each estimate and sample was calculated using the quantile intervals (95% CI [lower quantile, upper quantile]) of the distribution of all possible values of each sample (Table S4).

Studied Paleosols

All studied paleosols were classified as Calcisols because they are characterized by rare to common carbonate nodules (up to several centimeters in size; stage II nodules) or layers with stacked carbonate nodules and rhizocretions (stage II; Fig. S1). Of a total of 105 pedogenic carbonate samples, 46 carbonates showed characteristic pedogenic features (Fig. S2) as well as carbon isotope values indicative of carbonate precipitation under the influence of atmospheric CO2, and these were accepted as having formed in the nonsaturated zone and used for atmospheric pCO2 reconstruction (see the Supplemental Material).

Permian–Triassic Atmospheric pCO2

Mean atmospheric pCO2 showed a significant increase from the latest Permian (Changhsingian) to Early Triassic, with elevated pCO2 persisting until the latest Early Triassic (Fig. 1). Late Changhsingian mean atmospheric pCO2 estimates derived from Russian, North China, and Karoo samples are generally between 412 (95% CI [162, 688]) and 949 ppmv (95% CI [400, 1743]; n = 9). Atmospheric CO2 started to increase before the Permian-Triassic boundary (PTB) to levels between 1031 ppmv (95% CI [419, 1966]) and 1558 ppmv (95% CI [676, 2789]; n = 6). Sample PY2 (north China) yielded a lower atmospheric CO2 content of 483 ppmv (95% CI [214, 833]). In the earliest Griesbachian, the mean pCO2 was at 1606 ppmv (95% CI [689, 2919]). Atmospheric pCO2 in the late Griesbachian (Xinjiang) ranged from 2181 (95% CI [1025, 3516]) to 2610 ppmv (95% CI [1224, 4226]; n = 4), being on average 4× higher than Changhsingian background levels. Estimates for the Dienerian (Xinjiang) are between 1261 (95% CI [596, 2009]) and 1936 ppmv (95% CI [921, 3104]; n = 6), while estimates for the Spathian paleosols (Russia, North China, and Xinjiang) range from 1063 (95% CI [490, 1765]) to 1757 ppmv (95% CI [792, 2994]; n = 11), except sample DYLY, which yielded a CO2 concentration of 671 ppmv (95% CI [306, 1124]). Terminal hothouse (latest Spathian and Anisian) pCO2 estimates based on paleosols from Xinjiang, North China, and the United Kingdom are between 343 (95% CI [155, 575]) and 634 ppmv (95% CI [285, 1067]; n = 6), i.e., considerably lower than most Early Triassic pCO2 estimates but comparable to Changhsingian CO2 levels.

The atmospheric pCO2 estimates are in good agreement with published pCO2 reconstructions from paleosol nodules (Karoo Basin—Gastaldo et al., 2014; India—Roy et al., 2021), stomatal index data (Li et al., 2019; Retallack and Conde, 2020), and in part with estimates based on C3 plant carbon isotope fractionation (Wu et al., 2021; Fig. 2). However, the reliability of this latter method is controversial (e.g., Lomax et al., 2019). This could explain the significant drop in pCO2 above the PTB reconstructed by Wu et al. (2021), which is at odds with increasing low-latitude SSTs as well as high Dienerian and Spathian pCO2 values (this study; Fig. 2). Published Anisian paleosol pCO2 estimates were recalculated (Fig. 1), as they had originally been calculated with S(z) of up to 5000 ppmv (Ekart et al. 1999; Prochnow et al. 2006). Most importantly, the generally good agreement among pCO2 estimates reconstructed from time-equivalent paleosols from distant sites (this study; Fig. 1) and the comparable estimates derived from other CO2 proxies (Fig. 2) underline the validity of the atmospheric CO2 record presented here. However, the evolution of atmospheric pCO2 during most of the Griesbachian and the Smithian remains unresolved because no suitable pedogenic carbonates have been found for these periods.

Early Triassic Greenhouse

Siberian Traps volcanism is interpreted as a proximate cause of the 4× increase in atmospheric CO2 from the latest Permian (Changhsingian) to Early Triassic (late Griesbachian). The onset of Siberian effusive volcanism has been dated prior to 252.2 ± 0.1 Ma (Burgess and Bowring, 2015), ~300 k.y. before the end-Permian mass extinction. Subsequent intrusive magmatism starting at 251.9 ± 0.067 Ma probably produced >100.000 GT of CO2 and CH4 by thermogenic heating of sediments around large sill intrusions (Svensen et al., 2009; Augland et al., 2019). This massive emission of greenhouse gases (depleted in 13C) has been suggested as the main cause of dramatic global warming, as well as of the negative carbon isotope excursion recorded globally in the latest Changhsingian to early Griesbachian (Fig. 2).

Parallel to the reconstructed rise in atmospheric pCO2 from 412–949 ppmv in the latest Changhsingian to 2181–2610 ppmv in the late Griesbachian, low-latitude SSTs calculated from oxygen isotopes measured on conodont apatite (Fig. 2) increased by 7–10 °C, from 25–28 °C to >35 °C (Joachimski et al., 2020). With the decrease in pCO2 in the late Spathian to Anisian, SSTs decreased again (Sun et al., 2012; Fig. 2). Thus, pCO2 as well as SSTs persisted at high levels for almost 5 m.y. (Fig. 2), representing an unusually long time interval. High atmospheric pCO2 conditions could only be sustained either by continuous and massive CO2 outgassing from Siberian Traps or by reduced CO2 consumption by continental silicate weathering and biological uptake. The emplacement of large volumes of subvolcanic intrusions (sills and dikes) started in the latest Changhsingian but continued for only 0.5 m.y. into the Early Triassic (Augland et al., 2019; Burgess and Bowring, 2015). Although large igneous province volcanism has been reported to have been active until the end of the Middle Triassic (Ivanov, 2007), published geochronological data have large uncertainties of ~5 m.y. and cannot resolve whether contact metamorphism resulted in prolonged degassing after the initial violent pulse (Augland et al., 2019). Notably, Hg concentrations in marine carbonates argue for massive volcanic activity at the Permian-Triassic transition, in accord with the Siberian Traps record, but for reduced volcanic emissions in the later Early Triassic (Wang et al., 2018).

Early Triassic Atmospheric pCO2 Regulated by Weathering

Assuming that the outgassing of large volumes of volcanic CO2 faded after the initial 0.5-m.y.-long phase, atmospheric pCO2 is expected to have been drawn down relatively fast by continental silicate weathering—the most effective mechanism by which to extract CO2 from the atmosphere and to buffer Earth's climate. However, pCO2 stayed at elevated levels for ~4 m.y. after the Griesbachian CO2 maximum.

Warm temperatures, water availability, and continental plates located within the humid climatic belt are the main factors favoring silicate weathering. Weathering of freshly deposited continental flood basalts would have been particularly effective at consuming atmospheric CO2, with weathering rates of basalts being 10 times greater compared to weathering rates of granitic continental rocks (Dessert et al., 2003). This efficiency has been documented for Late Triassic Central Atlantic Magmatic Province (CAMP) volcanism, when individual volcanic pulses resulted in a doubling of atmospheric pCO2 followed by a striking decrease to pre-eruption levels within only ~300 k.y. (Schaller et al., 2012). While CAMP volcanism occurred in warm equatorial latitudes favorable for CO2 consumption by silicate weathering, the Siberian Traps erupted at ~60°N, with silicate weathering potentially less efficient due to colder conditions. However, the latest Permian increase of low-latitude SSTs likely resulted in an amplified warming of higher latitudes. In conjunction with intensified higher-latitude precipitation as suggested by climate modeling (Winguth et al., 2015), weathering of freshly deposited Siberian lavas should have had the potential to consume atmospheric CO2. In contrast, Kump (2018) argued that low continental fragmentation and high continentality in Pangea's interiors combined with minimum uplift rates depressed CO2 uptake by silicate weathering. However, this interpretation seems to be at odds with strontium as well as osmium isotope records (Song et al., 2015; Liu et al., 2020), which indicate an increase in continental weathering, especially in the Early Triassic (Fig. 2).

Silicate weathering can be modulated by reverse weathering, whereby non-kaolinite phyllosilicates form on the seafloor, leading to consumption of dissolved silica and alkalinity sourced from weathering on land and, most important, the addition of CO2 to the ocean-atmosphere system (Isson and Planavsky, 2018). Reverse weathering has been suggested as a mechanism to maintain high pCO2 in the Precambrian, when oceans were probably characterized by high dissolved silica concentrations (Maliva et al., 2005) before the advent of silica-secreting organisms. Interestingly, while the Permian is known for extensive chert deposition, cherts disappeared almost completely from the rock record in the Early Triassic (Beauchamp and Baud, 2002). A low abundance of silica-secreting organisms, warm ocean temperatures (increasing silica solubility), and supply of silica from land probably led to high dissolved silica concentrations in Early Triassic oceans, which should have promoted reverse weathering. Cherts re-occurred in the Spathian and Anisian (Sperling and Ingle, 2006) in conjunction with the diversification of radiolarians (O’Dogherty et al., 2010) and the late Spathian decrease in pCO2.

In summary, the 4× increase in atmospheric pCO2 across the Permian-Triassic boundary to high and intermediate CO2 levels in the Early Triassic is in agreement with low-latitude SSTs documenting greenhouse warming and hot Early Triassic oceans. Elevated atmospheric pCO2 persisted for ~5 m.y., suggesting that warm-climate–enhanced silicate weathering, although indicated by geochemical proxies, failed to draw down CO2 until the latest Early Triassic. This apparent contradiction may indicate that the exceptional conditions in Early Triassic oceans led to an intensification of marine authigenic clay formation and contribution of CO2 to the ocean-atmosphere system, counteracting CO2 consumption by silicate weathering.

Financial support by the National Natural Science Foundation of China (grant 41821001) is acknowledged. V. Silantiev and F. Mouraviev were supported by a subsidy allocated to Kazan Federal University (Russia) for the state assignment 671–2020–0049 in the sphere of scientific activities and cont. no. 14.Y26.31.0029, res. no. 220 of the Government of the Russian Federation. Helpful comments by anonymous reviewers on an earlier version of this manuscript are acknowledged. This is a contribution to Deutsche Forschungsgemeinschaft (DFG) Research Unit TERSANE (FOR 2332: Temperature-related stressors as a unifying principle in ancient extinctions; project Jo 219/15–1).

1Supplemental Material. Detailed description of locations, samples, and methods, Figures S1–S3, and Tables S1–S4. Please visit to access the supplemental material, and contact with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.