The Toarcian crisis (ca. 183 Ma) is characterized by extinctions in both the marine and terrestrial realm, localized ocean anoxia (The Toarcian Ocean Anoxic Event; T-OAE), global warming and major carbon cycle perturbations1,2,3,4,5,6,7,8,9,10,11,12,13,14,15. The carbon cycle disturbances are evidenced by increases in the atmospheric concentration of CO2 (pCO2;1,2,3) and negative carbon isotope excursions (CIEs) recorded in both carbonate and organic matter in marine and terrestrial archives4,5,6,7,8,9,10,11,12,13,14. A prominent negative CIE marks the onset of the T-OAE, and varies in magnitude in both carbonate and organic matter (δ13Ccarb ≈ −0.6 to −6.0‰; δ13Corg ≈ −0.8 to −8.6‰; see review in ref. 15). High-resolution δ13C curves from multiple sections show a stepwise character of the excursion, with several abrupt negative shifts making up the full body of the CIE3,4,6,7,8,9,11,12,14,15,16,17.

High-precision U-Pb geochronology demonstrates that the Karoo LIP in South Africa and Lesotho (Fig. 1) was emplaced synchronously with the T-OAE negative CIE (Fig. 2; ref. 18,19,20,21). This has led to the hypothesis that Karoo-derived emissions were responsible for the carbon cycle perturbations (e.g.,1,7,18,19,20,21,); however, the negative shift in δ13C is of such magnitude that mantle-derived carbon alone is unlikely to have been the source22. Alternative suggestions include marine methane clathrate dissociation (e.g.,4,8,22,), permafrost melting23, terrestrial organic matter24,25, or thermogenic carbon release generated by sill emplacement into organic-rich sedimentary rocks of the Karoo Basin (e.g.,1,19,22,26,).

Fig. 1: Location and cross section of the Karoo Basin.
figure 1

Top: a plate reconstruction at 180 Ma (modified from ref. 60; Fig. 12.2) showing the known and inferred extent of the Karoo Large Igneous Province (LIP)21,61. Bottom: a detailed map of the Karoo Basin stratigraphy in South Africa and Lesotho, with a schematic cross section across the province (modified from ref. 47). Photo A shows deep laminar sills of the Ecca Group in western South Africa, while photo B shows the continental flood basalt lavas exposed in Lesotho (positions are labelled as A and B on the map and cross section).

Fig. 2: Correlation between Toarcian carbon cycle perturbations and the timing of Karoo intrusions.
figure 2

Timeline of Karoo activity, the Toarcian Ocean Anoxic Event (T-OAE; green box), carbon isotope curves from shallow marine stratigraphic sections and pCO2 data (including uncertainties) after ref. 18,19,35. The age of the Pliensbachian-Toarcian (Pl-To) boundary (orange line) is from ref. 19. U-Pb ages of Karoo intrusives are from ref. 18,19,20,21. U-Pb color codes refer to sill emplacement levels in the Karoo Basin, and corresponds to the colors used in the map and cross section in Fig. 1. Numbers in boxes represent the total negative shift (δ13Cmax − δ13CpreT-OAE) in δ13Corg and δ13Ccarb.

Volcanic basins, such as the Karoo, are characterized by a pulsed emplacement of sills (e.g.,27,), leading to pulsed contact metamorphism and fluid release. Notably, a stepwise decrease in δ13C is suggestive of isotopically light carbon supplied in pulses to the ocean-atmosphere system28. Field investigations, borehole studies, and numerical modeling suggest thermogenic carbon generation on the order of several thousand gigatons within the Karoo basin26,29,30,31. The presence of thousands of explosion pipes and hydrothermal vent complexes rooted in the contact metamorphic aureoles represent a plausible release mechanism, and attests to rapid emissions of the thermogenic gases26,32. Available high-precision U-Pb ages of the sills and thermogenic emission estimates from the Karoo Basin allows to identify detailed and realistic emission scenarios from the Karoo LIP; however, these have not yet been tested by carbon cycle modeling.

Here, we test such scenarios using the GEOCLIM model33,34, in order to explore whether pulsed carbon release from the Karoo LIP could explain the full extent of observed Toarcian pCO2 and δ13C changes. The emission scenarios are grounded by U-Pb geochronology, detailed evolution of the Karoo LIP, and estimates of both the mantle-derived and thermogenic carbon from volcanic and contact metamorphic degassing, respectively.


In order to investigate the consequences of carbon release from the Karoo LIP, it is essential to constrain the timing between the negative CIEs, pCO2 excursion, and Karoo activity. Following the correlations used here (see Supplementary Note 1;18,19,35), the T-OAE lasted for 500 kyr, and overlaps with the full extent of the carbon isotope perturbations (Fig. 2). This corresponds to the onset of the negative CIEs, as well as peak negative values, and the subsequent interval of δ13C increase eventually reaching background levels. Reconstructed pCO2 curves from the Danish Basin show an increase in pCO2 of up to ~2000 ppm during the T-OAE (including uncertainties; ref. 1). Following our correlations, the pCO2 peak coincides with the peak negative δ13C values (Fig. 2). The pCO2 records do not include absolute time constraints, so the correlation of pCO2 data with the Karoo activity and various carbon isotope records was based on the extent of the T-OAE (see Supplementary Note 1). Available U-Pb ages of Karoo intrusives demonstrate that the Karoo igneous activity overlaps with the full extent of the Toarcian CIEs and by inference the pCO2 peak (Fig. 2; ref. 18,19,20,21).

The Karoo intrusives postdate a 1–2‰ negative CIE that occurs around the Pliensbachian-Toarcian (Pl-T) boundary in some sections (e.g.;5,36, Fig. 2), excluding a temporal link between Karoo igneous activity and this earlier disruption of the carbon cycle37. As this CIE is not systematically recorded, and there is no evidence for pCO2 increase or Karoo activity at this time, it is not considered in this study. Furthermore, we stress that there are alternative correlations and interpretations for the Toarcian carbon cycle perturbations that differ from those used here, particularly with respect to the duration of the full extent of the T-OAE CIE, which varies from 120 to 2400 kyr (Supplementary Note 1;4,7,8,18,19,35,38,39,40,41,42,43).

The total model run is set to 500 kyr, which corresponds to the total duration of the T-OAE (following the correlations used here; Fig. 2; Supplementary Note 1). Both volcanic and thermogenic processes associated with the Karoo LIP activity are accounted for in the emission scenario. Seven thermogenic carbon pulses (δ13C of −36‰; Supplementary Note 1) are released between model time (t) = 30 and 262 kyr (pulses #2-8; Table 1; Supplementary Fig. 1), which is constrained by the timing of negative CIEs. The thermogenic carbon pulses vary in magnitude, between 300 and 3800 Gt. The total magnitude of released mantle-derived carbon (with δ13C of −5‰) is set to 8000 Gt, following average estimates for the Karoo LIP (e.g., ref. 22; Supplementary Note 1). 4500 Gt mantle-derived carbon is released at t = 160 (pulse #6; Table 1), while the remaining 3500 Gt C is released as a long-lasting and continuous pulse between t = 0 and 262 kyr (pulse #1; Table 1; Supplementary Fig. 1; Supplementary Note 1). Note that carbon pulse #6 represents a mix of both mantle-derived and thermogenic carbon, with a total magnitude of 7000 Gt and a δ13C value of −16‰ (Table 1; Supplementary Note 1). Except for the continuous mantle-derived carbon pulse (pulse #1), the carbon releases have durations of 3–22 kyr, which is constrained by the duration of observed carbon cycle perturbations targeted, and in agreement with modeling studies suggesting that the generation of thermogenic fluids occurs on timescales of 100’s to 1000’s years29,44.

Table 1 Overview and input values for Karoo emission scenario.

The pulsed release of a total of 20,500 Gt C leads to a stepwise increase in pCO2 from ~650 ppm up to a maximum of ~1700 ppm. The modeled pCO2 curve plots generally within the range of observed data, with the exception of the fourth peak (at t = 111 kyr), which overestimates the pCO2 (Fig. 3). The modeled δ13Ccarb curve decreases in seven steps (between ~0.3 and ~1.8‰), with a total negative shift (δ13Cmax – δ13CpreT-OAE) of ~−4‰. Following the δ13Ccarb data, the modeled δ13Corg curve also decreases in seven steps (between ~0.5 and ~3.7‰), but with larger magnitudes, as well as a larger a total negative shift of ~−6‰. Both the modeled δ13Ccarb and δ13Corg curves generally plot within the range of observed data considering both the magnitude and shape of the negative shift (Fig. 3). There are however some exceptions: 1) the modeled δ13Ccarb curve predicts slightly lower δ13C values at the end of the model run during the recovery interval of δ13C increase, and 2) the modeled δ13Corg curve does not replicate all of the short-lived negative CIEs making up the prolonged interval of negative δ13C values following peak negative values. Note that all observed δ13C curves are from shallow marine stratigraphic sections, and is accordingly compared to model δ13C data representing a shallow ocean epicontinental sea or shelf environment (i.e., epicontinental surface reservoir boxes; see Methods).

Fig. 3: Modeled pCO2 and δ13C curves compared to observed Toarcian data.
figure 3

Model response (pink dashed lines) of δ13Corg and δ13Ccarb of the shallow ocean (GEOCLIM epicontinental surface reservoir boxes) and atmospheric pCO2 to a Karoo emission scenario including eight pulses of carbon release (see Table 1; Supplementary Fig. 1). Gray and blue outlines represent the range of observed Toarcian δ13Ccarb4,5,9, δ13Corg8,16,36, and pCO2 data1, which corresponds to the curves presented in Fig. 2.


The emission scenario presented here replicates an overall global trend in δ13C, including several negative steps, and a total negative shift (δ13Cmax – δ13CpreT-OAE) of ~−4‰ for δ13Ccarb and ~−6‰ for δ13Corg. This is comparable to observed Toarcian carbonate δ13C records showing generally lower CIE magnitudes compared to those of organic matter, with mean total CIE values of ~−3 vs. ~−5‰, respectively15. Globally, the total magnitude of the T-OAE CIE ranges considerably (up to ~−6 for δ13Ccarb and ~−9‰ for δ13Corg; ref. 15), however, the largest excursions are likely exaggerated due to local effects15,43. Furthermore, all high-resolution Toarcian carbon isotope curves show a stepwise character of the T-OAE excursion, and a series of at least five prominent negative shifts is recognized in numerous carbonate and organic matter Toarcian sections (e.g., Kaszewy, Mochras, Yorkshire, Peniche and Luxembourg;12,45). δ13C curves showing one main long-lived T-OAE CIE, or a limited number of negative shifts, have likely failed to record the full extent of the stepped descent due to low sample or stratigraphic resolution. We chose not to take the observed prolonged interval of negative δ13Corg values following peak negative values into account, because it is uncertain whether or not this represents a global trend. It has been suggested that, due to improved preservation of marine organic matter under oxygen-depleted conditions, artificial Toarcian δ13Corg excursions have been produced40. This could potentially explain why there is an apparent prolonged interval of δ13Corg CIEs, while δ13Ccarb values instead return to background levels (which is also consistent with the shape of the pCO2 curve). Although Toarcian δ13C curves vary with respect to the magnitude, number, timing, and durations of the negative CIEs, we consider our modeled δ13Ccarb and δ13Corg curves to be representative for the overall global trend of Toarcian carbon cycle changes.

While the modeled δ13C data is compared to several observed δ13C curves, available reconstructed Toarcian pCO2 curves are scarce. As seen in Fig. 3, the modeled pCO2 curve generally plots within the range of observed Toarcian pCO2 data from ref. 1, but potential target pCO2 values represent quite a wide range due to data points with large error bars. Consequently, the magnitude of released carbon could vary greatly and still replicate the observed pCO2 data within the range of uncertainties. However, by respecting both observed pCO2 and δ13C curves simultaneously, the possible magnitude of carbon release is narrowed down considerably. Furthermore, based on Toarcian seawater pH (ref. 2) and Δ13C data (δ13Cterrigenic – δ13Cmarine; ref. 3), a pCO2 increase from ~800–850 ppm to ~1750–1800 ppm has been calculated (when assuming a release of thermogenic carbon with δ13C of −40 to −30‰ for the latter; ref. 3). As seen in Fig. 3, an increase of this magnitude (i.e., 800–850 to 1750–1800 ppm) is in very close agreement with the modeled pCO2 data. The overestimation of pCO2 at t = 111 kyr shown by the modeled curve (Fig. 3) is likely the result of the low sample resolution of the observed data, as it is very unlikely that a shift in δ13C of up to 4‰ is not reflected by a change in pCO2. Therefore, we consider the modeled pCO2 curve presented here to be realistic as it reflects Toarcian pCO2 estimates constrained by several methods.

According to our emission scenario, a total release of 20,500 Gt carbon replicates the observed Toarcian pCO2 and δ13C data. This number includes 12,500 Gt thermogenic carbon, which is within the range of thermogenic carbon generation estimations from the Karoo Basin (~7500–23,000 Gt C; ref. 26,29,30,31). The ratio of trapped versus released thermogenic gases from volcanic basins is poorly constrained, although 100% degassing from intrusive LIP components is considered unrealistic (e.g., 46,). Consequently, the magnitude of released carbon must be lower than the magnitude of carbon generated. The lowermost thermogenic carbon generation estimates are lower than the release value constrained by our modeling. However, these numbers are likely significantly underestimated as they have been upscaled based on a much smaller area of affected host rock compared to that of ref. 31, and do not include devolatilization of Ecca coals or Beaufort, Dwyka and Stormberg organic-bearing rocks in the Karoo strata (see Fig. 1). According to ref. 31, an estimated 72 % of the generated carbon gases in the Karoo Basin were released, which corresponds to ~17,000 Gt C. We therefore consider 12,500 Gt to represent a realistic value of cumulative thermogenic carbon release from the Karoo Basin.

Considering the TOC content and sill distribution in the Karoo Basin, sill emplacement into Ecca and Beaufort shales volatilized significantly more carbon-bearing gases compared to sill emplacement at other stratigraphic levels. Within the Ecca and Beaufort series, sills are abundant and can be traced almost continuously with depth, as well as between the basin margins (Fig. 1; ref. 47). In contrast, sills are rare in the Stormberg Group, while devolatilization of Dwyka sedimentary rocks was likely restricted to the upper parts of the series based on the distribution of the sills. The Ecca shales have the highest TOC contents, reaching locally as much as 18 wt.%, while maximum values within the Stormberg, Beaufort and Dwyka Group sedimentary rocks range from 3 to 6 wt.%48,49,50,51,52,53. Although the TOC contents are overall lower compared to the Ecca Group, the presence of closely spaced large sills (up to 270 m thick; ref. 47) in the Beaufort Group with TOC up to ~6 wt.%52 suggests significant gas generation potential at these stratigraphic levels as well. Sill thicknesses and vertical spacing are important factors influencing the gas generation potential; gas generation can increase by up to ~35 % for more closely spaced sills, compared to separate sills emplaced into the same host-rock27.

According to our emission scenario, the magnitudes of the individual thermogenic carbon pulses vary throughout the model run (Table 1), which is constrained by the magnitudes of the observed carbon cycle perturbations targeted. The available U-Pb data cannot be used to directly assess the magnitude of the individual degassing pulses as 1) there is no systematic correlation of sill age with stratigraphic emplacement level, and 2) they lack the sufficient resolution (large error bars) to match that of the stepped δ13C curves. Importantly, however, the U-Pb geochronology links sill emplacements within organic-rich sedimentary rocks (i.e., Ecca and Beaufort) to the entire interval of Toarcian carbon cycle perturbations (Fig. 2). Furthermore, the proposed carbon emission fluxes of 0.2 to 0.3 Gt C yr−1 (Table 1) corresponds to those estimated for other LIPs (e.g., NAIP and CAMP; 0.1–0.5 Gt C yr−1; ref. 54,55). We therefore consider the magnitudes of the individual pulsed thermogenic carbon releases to represent realistic values.

Realistic temporal and volumetric estimates of Karoo LIP volatile emissions can explain the carbon cycle perturbations observed in the Toarcian. Pulsed sill intrusions and contact metamorphism of organic matter-bearing sedimentary rocks in the Karoo Basin represents a significant source of 13C-depleted carbon, which can explain the observed negative CIEs. As the lowest δ13C value of the released thermogenic carbon in our model is −36‰, an extremely isotopically depleted carbon source (e.g., methane clathrates; −60‰; ref. 22) is not a required component (see also Supplementary Fig. 2). High-precision U-Pb geochronology links sill emplacements within the organic-rich strata of the Karoo Basin to the negative CIEs and the abrupt pCO2 peak. Although the exact values for the different degassing pulses represent estimates, we consider our emission scenario realistic as it incorporates the available geological knowledge of the Karoo LIP and a representative framework for Earth system processes during the Toarcian.


Model description

GEOCLIM couples the 10-box model COMBINE56, describing different geochemical cycles (e.g., C, P, O2, alkalinity, Sr) and their associated isotopic cycles (δ13C, 87Sr/86Sr), to the 3D FOAM general circulation climate model57. COMBINE consists of one atmospheric box and nine ocean boxes, including 1) an epicontinental surface reservoir divided in two boxes (surface box up to 100 m depth; deep box from 100 to 200 m depth), 2) three mid-latitude ocean boxes (surface, thermocline, and deep), and 3) two polar oceans both divided in two boxes (surface and deep). A full description of the GEOCLIM model can be found in refs. 34 and 56.

Boundary conditions for the simulations presented here include a Jurassic (180 Ma) paleogeography, which is derived from a synthesis of paleomagnetic data, hot spot tracks and geological constraints34,58. The background volcanic degassing rate was set to 4 × 1012 mol/yr so that the pCO2 equals ~ 650 ppm at t = 0 (prior to any Karoo degassing), which represents an average value of observed pCO2 data (including uncertainties; ref. 1).

GEOCLIM calculates the δ13C of the various carbon species (e.g., H2CO3, POC; particulate organic carbon) in each reservoir/box (e.g., epicontinental surface box). The isotopic fractionation between H2CO3* (H2CO3 + CO2; Supplementary Note 1) and POC is set to depend on the calculated H2CO3* concentration at each timestep for each oceanic reservoir. As shown in Supplementary Fig. 3a, the fractionation between H2CO3* and the POC increases during the interval of carbon injections, but then decreases towards “background values”. A decrease in fractionation reduces the difference between the δ13C of carbonate and POC, which allows the δ13Corg to increase faster compared to δ13Ccarb59. This could potentially explain the apparent decoupling between the δ13Corg and δ13Ccarb modeled curves (Fig. 3; Supplementary Fig. 3b); while the modeled δ13Corg curve recovers faster and reach observed values during the recovery interval at the end of the model run, the modeled δ13C carb curve predicts slightly lower δ13C values compared to observed values.