Volume 16, Issue 3 p. 10-1-10-13
Free Access

Biogeochemical constraints on the Triassic-Jurassic boundary carbon cycle event

D. J. Beerling

D. J. Beerling

Department of Animal and Plant Sciences, University of Sheffield, Sheffield, England, UK

Search for more papers by this author
R. A. Berner

R. A. Berner

Department of Geology and Geophysics, Yale University, New Haven, Connecticut, USA

Search for more papers by this author
First published: 18 July 2002
Citations: 147


[1] The end-Triassic mass extinctions represent one of the five most severe biotic crises in Earth history, yet remain one of the most enigmatic. Ongoing debate concerns the environmental effects of the Central Atlantic Magmatic Province (CAMP) eruptions and their linkage with the mass extinction event across the Triassic-Jurassic boundary. There is conflicting paleo-evidence for changes in atmospheric pCO2 during the extrusion of the CAMP basalts. Studies on sediments from European and Pacific localities have, however, identified a substantial negative isotopic anomaly (up to −3.5‰) across the TR-J boundary, providing an important indicator of changes in the operation of the ancient global carbon cycle. We sought to explain the paleo-evidence by utilizing a carbon cycle model for the “hothouse” world of the end-Triassic that emphasizes the chemical weathering of silicate and carbonate rocks and the ocean carbonate chemistry. We find that volcanic CO2 outgassing fails to fully account for either a sufficient rise in atmospheric pCO2 (indicated by the stomata of fossil leaves) or the sedimentary isotopic fingerprint. Instead, the scenario that best fits all of the geologic evidence is a positive feedback loop in which warming, due to a buildup of volcanically derived CO2, triggers destabilization of seafloor methane hydrates and the catastrophic release of CH4 [Pálfy et al., 2001]. We calculate that this carbon cycle perturbation was huge, involving the release of ∼8000–9000 Gt C as CO2 during the CAMP basaltic eruptions and ∼5000 Gt C as CH4. In the model the initial isotopic excursion is assumed to take place over ∼70 kyr, while complete reequilibration of the ocean-atmosphere system with respect to CO2 is accomplished over 700–1000 kyr. Our results thus provide a preliminary theoretical explanation for the bioevents, estimated pCO2 changes, and isotopic excursions observed in marine and continental sediments at this time.

1. Introduction

[2] The boundary between the Triassic (TR) and Jurassic (J) epochs, ∼200 million years (Myr) ago [Pálfy et al., 2000a], is marked by one of the five largest mass extinction events in Earth history [Sepkosky, 1996]. Fossil evidence indicates that ∼80% of living species went extinct; this represents a degree of severity exceeding that seen across the Cretaceous-Tertiary boundary and included the loss of 53% of marine genera and 50% of tetrapod species [Sepkosky, 1996]. Terrestrial vegetation also suffered a sudden and severe crisis at the boundary. North American palaeobotanical studies in the Newark basin reveal the loss of 60% of palynospecies at the boundary, together with a proliferation of fern species [Fowell and Olsen, 1993]. In the North Atlantic region, even more severe ecological trauma occurred, with a >95% turnover of megafloral species [Harris, 1935, 1937; Lundblad, 1959] and marked microfloral turnovers [Visscher and Brugman, 1981].

[3] Despite the severity of the TR-J boundary mass extinctions, associated environmental changes have been poorly documented in comparison to other extinction episodes [Wignall, 2001], a situation arising in part from the relative scarcity of complete marine sections lacking diagenesis and reworking of sediments. In consequence, the causal mechanisms driving the end-Triassic extinctions continue to remain enigmatic [Hallam and Wignall, 1997]. Key issues of uncertainty relate to whether the biogeochemical cycling of carbon, and other elements, underwent a major perturbation across the boundary and, if this occurred, how it was linked with changes in global climate and the isotopic composition of the oceanic, terrestrial, and atmospheric carbon reservoirs [McElwain et al., 1999; Ward et al., 2001; Pálfy et al., 2001, 2002; Beerling, 2002; Hesselbo et al., 2002]. It is also likely that mass extinctions in the terrestrial and marine realms themselves influenced the role of biota in the exchange of carbon between reservoirs [McRoberts et al., 1997; Ward et al., 2001].

[4] The first evidence pointing to a significant carbon cycle event across the TR-J boundary derives from detailed studies of the stomatal characters and isotopic composition of fossil Ginkgoalean and Cycadalean leaves from Greenland and Sweden [McElwain et al., 1999]. Using the stomatal approach to estimating paleo-CO2 levels, McElwain et al. [1999] reconstructed an increase in atmospheric CO2 concentration (pCO2) of ∼1400 ppm across the boundary. Coincident with the pCO2 rise and terrestrial mass extinctions [Pálfy et al., 2000a] was the geologically rapid [Olsen et al., 1996; Hames et al., 2000; Cohen and Coe, 2002] emplacement of the Central Atlantic Magmatic Province (CAMP), a continental flood basalt province extending over an area >2.5 × 106 km2 in northern and central Brazil, western Africa, and the eastern United States [Marzoli et al., 1999]. Taken together, these observations suggest a linkage between the TR-J boundary mass extinction, the release of significant amounts of CO2 by CAMP volcanic activity, and catastrophic greenhouse warming. However, this coupling between mantle CO2 outgassing and extinctions has been questioned by a study on the carbon isotope composition of North American Late Triassic and Early Jurassic pedogenic carbonates [Tanner et al., 2001]. Calibrated with a diffusion-reaction model [Cerling, 1991, 1992], the minor change in paleosol isotopic composition (0.3‰) corresponds to a modest CO2 rise of ∼250 ppm, i.e., a relatively stable atmospheric carbon reservoir across the TR-J boundary.

[5] A second line of evidence indicating a disturbance of the carbon cycle across the TR-J boundary comes in the form of a negative isotopic excursion recorded in both organic and inorganic carbon pools [McElwain et al., 1999; Ward et al., 2001; Pálfy et al., 2001, 2002; Hesselbo et al., 2002]. Fossilized land plant leaves from a lacustrine section in Greenland show a negative 3‰ excursion between the Rhaetian and Hettangian stages of the Triassic and Jurassic, respectively [McElwain et al., 1999]. Further studies from sites elsewhere in the world indicate that his might be a rather general phenomenon characterizing continental and marine TR-J boundary sediments, in common with mass extinctions during the Permo-Triassic [Holser et al., 1989] and Cretaceous-Tertiary [Hsü et al., 1982]. Ward et al. [2001] reported a negative 2‰ excursion in bulk organic matter from a sedimentary sequence in the Queen Charlotte Islands, British Columbia, Canada, and Pálfy et al. [2001] identified a negative 3.5 and 2‰ excursion in marine carbonates and bulk organic matter, respectively, from Hungary. The excursions in both Hungary and British Columbia sections are abrupt, well above the background variation in the rest of the sections and coincident with the sudden extinctions of marine organisms.

[6] Reconciling the paleo-evidence for atmospheric pCO2 and the coincident sedimentary carbon isotopic fingerprint across the TR-J boundary now represents a central task if we are to decipher cause and effect during the end-Triassic mass extinctions. This aim represents the main objective of the present paper. Our approach has been to first revisit the stomatal and paleosol evidence for atmospheric pCO2 change across the TR-J boundary to consider uncertainties inherent in both proxies [Royer et al., 2001] to develop a revised set of pCO2 curves. We next adopt a model-based approach to quantify the effect of mantle CO2 outgassing during the emplacement of the CAMP basalts on atmospheric pCO2 levels and the carbon isotopic composition of the organic and inorganic carbon reservoirs. Our model, which accounts for the chemical weathering of terrestrial silicate and carbonate rocks and for ocean chemistry, is a modified version of the time-dependent Berner, Lasaga, and Garrels (BLAG) model [Berner et al., 1983], specifically designed for analyzing carbon cycle perturbations during the end-Triassic “greenhouse” world. Included in our modeling analyses is an assessment of possible positive feedback effects of a sudden release of isotopically light CH4 from the marine sedimentary reservoir, postulated by Pálfy et al. [2001, 2002] to be a primary driver of the isotopic excursion. The revised atmospheric pCO2 curves, together with the stable carbon isotope records from marine and lacustrine sections, are then used as a basis for evaluating the model results to determine which scenario best represents the carbon cycle event across the TR-J boundary.

2. Paleobiological Evidence for a pCO2 Increase Across TR-J Boundary

2.1. Stomatal Evidence

[7] Quantitative reconstructions of pCO2 from stomatal characters of fossil leaves and the isotopic composition of pedogenic carbonates both have sources of error attached to them [Royer et al., 2001; Ekart et al., 1999]. The stomatal technique utilizes the established inverse correlation between the stomatal index (percentage of leaf surface cells that are stomatal pores) of land plant leaves and atmospheric pCO2 [Woodward, 1987]. Direct measurements of stomatal index on fossil cuticles of Ginkgoalean and Cycadalean leaves from sections in Jameson Land, east Greenland (Cape Stewart formation), and Scania, southern Sweden, indicate marked reductions across the TR-J boundary [McElwain et al., 1999]. In fact, fossil leaves of some taxa from particular horizons have exceptionally low stomatal indices (∼3–4), values not observed in modern plants and only previously recorded on the fossil axes of early land plants growing in the inferred Devonian high-pCO2 environment [Edwards, 1998]. These observations are qualitatively consistent with vegetation responding to a major paleo-CO2 increase ∼200 Myr ago.

[8] Atmospheric pCO2 levels were quantitatively reconstructed [McElwain et al., 1999] with stomatal ratios, whereby the stomatal index of the modern equivalent divided by that of the fossil (i.e., the stomatal ratio, SR) is directly related to the ratio of atmospheric CO2 in the past relative to the present day (RCO2) [McElwain and Chaloner, 1996; Chaloner and McElwain, 1997]. Two calibrations have been suggested on the basis of a comparison of fossil and modern plant materials against the CO2 curve calculated from a geochemical long-term carbon cycle model [Berner, 1994], yielding 1 SR = 600 ppm [McElwain and Chaloner, 1996], and 1 SR = 450 ppm [McElwain, 1998]. The method assumes a priori a nonlinear relationship between the stomatal index of a fossil and CO2. However, for Ginkgo, the SR-derived calibration curve matches reasonably well observations on modern historical sequences of leaves, and leaves grown experimentally under a range of atmospheric CO2 concentrations, providing support for the approach [Beerling and Royer, 2002].

[9] We used the two SR calibrations to calculate likely upper and lower bound pCO2 estimates from the stomatal indices of leaves at a given depth in a particular section. Proceeding on this basis, we recalculated atmospheric pCO2 levels using the stomatal index data reported for fossil leaves from east Greenland and southern Sweden [McElwain et al., 1999] (Figure 1). Our reanalysis confirms the reported rise in atmospheric pCO2 across the TR-J boundary reported by McElwain et al. [1999]. Taking the mean of the upper and lower pCO2 estimates, we reconstruct end-Triassic values of 872 and 884 ppm for Greenland and Sweden, respectively, which are both within the range calculated from a long-term geochemical model of the carbon cycle [Berner and Kothavala, 2001]. By the earliest Jurassic, atmospheric pCO2 is reconstructed to have risen by between +810 ppm (Greenland) and +1120 ppm (Sweden) (Figure 1), increases smaller than originally estimated [McElwain et al., 1999], but nevertheless substantially greater than that implied by the paleosol isotope data (250 ppm) [Tanner et al., 2001].

Details are in the caption following the image
Reconstructed changes in atmospheric pCO2 levels over Triassic-Jurassic boundary based on stomatal indices of fossil leaves from Greenland (top left) and from Sweden (top right). Range was calculated by calibrating stomatal ratios (SR) of fossils as being equal to either 1.2 or 2 times the ratio of atmospheric CO2 in the past relative to the present (RCO2). Range was recalculated from original data of McElwain et al. [1999]. For comparison, changes in atmospheric pCO2 reconstructed using paleosol pCO2 barometer with pedogenic carbonate data of Tanner et al. [2001] and stable carbon isotope signature of terrestrial organic matter from Greenland are displayed (bottom left) [after Beerling, 2002]. Upper and lower lines are determined by setting S(z) to either 7000 or 3000 ppm CO2, respectively. See text for further details.

2.2. Paleosol Evidence

[10] Estimation of any transient atmospheric pCO2 changes across the TR-J boundary using the paleosol CO2 barometer [Cerling, 1991, 1992] requires isotopic analyses of pedogenic carbonates and coeval terrestrial organic matter (δ13COM) with a high stratigraphic resolution. An initial study of terrestrial pedogenic carbonates between the late Triassic to early Jurassic showed a distinct negative isotope excursion over this time [Suchecki et al., 1988]. Calibrated with the paleosol CO2 model, the shift corresponds to a rise in atmospheric pCO2 of >1000 ppm. However, although this increase is intriguing, the use of some of the isotopic data for this purpose may be compromised by inadequate controls on sample selection [Ekart et al., 1999]. In a more recent study, Tanner et al. [2001] obtained pedogenic calcite samples from sites dated to the Upper Triassic (Carnian and Norian) and the Lower Jurassic (Hettangian). These samples thus ranged over 20 Myr, a temporal resolution considerably lower than the study of fossil leaves [McElwain et al., 1999]. The upper Triassic sample is likely to predate the onset of any CO2 outgassing from the CAMP eruptions [Hames et al., 2000; Pálfy et al., 2002; Beerling, 2002]. Their atmospheric pCO2 estimate for the Hettangian paleosol however is, at 2480 ppm, in line with the stomatal evidence (Figure 1).

[11] A well-characterized requirement of the paleosol CO2 barometer is the need to constrain the isotopic composition of organic matter (δ13COM) [Cerling, 1991, 1992]. This sensitivity arises because δ13COM represents the isotopic composition of soil respired CO2, a significant flux relative to the back diffusion of CO2 into the soil from the atmosphere. Tanner et al. [2001] assumed a constant δ13COM value to calculate atmospheric pCO2 values for the Norian and the Hettangian. However, the δ13COM excursion shown by fossil land plant leaves in Greenland, and bulk organic matter from Canada [Ward et al., 2001] and Hungary [Pálfy et al., 2001] indicates this is unlikely to be correct. An alternative interpretation of the paleosol data is therefore possible in which the late Triassic and early Jurassic samples provide end-member constraints, and the diffusion reaction model of Cerling [1991, 1992] is forced with the δ13COM records of terrestrial plant leaves [Beerling, 2002]. Calculated in this way, atmospheric pCO2 increases by 1032 ppm, a rise consistent with the CO2 reconstructions based on fossil leaf cuticles (Figure 1) [Beerling, 2002].

[12] Our reanalysis of the stomatal and paleosol paleo-pCO2 proxies consistently indicates that within the uncertainties associated with the different approaches, atmospheric pCO2 levels rose by around +810 to +1120 ppm across the TR-J boundary. The revised pCO2 reconstructions therefore require an explanation that, additionally, must also be consistent with the isotopic signals in sediments from Tethyan [Pálfy et al., 2001; Hesselbo et al., 2002] and Pacific [Ward et al., 2001] localities.

3. Carbon Cycle Greenhouse Model

3.1. Carbon Reservoir Masses, Fluxes, and Exchanges

[13] We have addressed these issues through the development and application of a carbon cycle model for the end-Triassic greenhouse world. The carbon cycle model is a modification of the earlier BLAG model of Berner et al. [1983] and is illustrated in Figure 2. The BLAG model is modified by lumping Ca with Mg and by introducing a methane flux to the atmosphere where CH4 is rapidly oxidized to CO2. Carbon is added to the atmosphere by volcanism (Fv), the decomposition of methane hydrates (Fm), and the precipitation of carbonates in the ocean (Fbc) and is lost from the atmosphere by the weathering of Ca-Mg carbonates (Fwc) and silicates (Fwsi). There is also exchange of carbon between the terrestrial biosphere and the atmosphere (Fbio). Carbon is added to the oceans by carbonate and silicate weathering and is lost via carbonate precipitation.

Details are in the caption following the image
Schematic diagram of the end-Triassic greenhouse world carbon cycle. Carbon inputs to atmosphere are denoted by Fv = CO2 from volcanism, Fm = CH4 from seafloor methane hydrates, and Fbc = precipitation of carbonates in the oceans. Carbon is removed from atmosphere by weathering of Ca-Mg carbonates (Fwc) and silicates (Fwsi). Carbon exchange between atmosphere and terrestrial biosphere is also indicated (Fbio).
[14] The appropriate generalized silicate and carbonate reactions [e.g., see Berner, 1998], with Ca + Mg represented simply as Ca, are For weathering
equation image
equation image
For carbonate deposition in the oceans
equation image
The long-term organic subcycle [Berner, 1998], involving the weathering and sedimentary burial of organic matter, is assumed to be at steady state over the short time scale considered here. Also, methane released to the atmosphere is oxidized very rapidly to CO2 with a mean residence time of 10 years (as in the modern atmosphere) so that a steady state mass of methane (M) is present at all times (M = Fm/k, where k = 105 Myr−1). Rapid oxidation, for the alternative situation of CH4 release directly to the oceans, is also assumed [Dickens, 2000].
[15] In our model, mass balance expressions are constructed for carbon and for calcium. For atmospheric carbon (C) this gives
equation image
Note that Fbio can be <0 for loss of terrestrial biomass.For oceanic carbon (Cw)
equation image
For oceanic calcium
equation image
where C is mass of carbon (as CO2) in the atmosphere, Cw is mass of carbon (as HCO3) dissolved in the oceans, Mca is the mass of calcium dissolved in the oceans, flux terms F are defined above, and t = time. For simplicity of calculation, all masses were expressed as 1018 mol and fluxes (F) were expressed in 1018 mol Myr−1. However, to compare with the literature, the results have been converted in this paper to Gt C and Gt C kyr−1.
[16] Silicate and carbonate weathering rate was assumed to be a simple function of temperature, which is in turn a greenhouse function of the CO2 and CH4 contents of the atmosphere and is given by [Berner and Kothavala, 2001]
equation image
equation image
where T = temperature and subscript 0 refers to the initial state. A simple representation of greenhouse warming is given by (modified from Berner and Kothavala [2001] and Schrag et al. [2002])
equation image
where Γmc refers to the combined methane plus CO2 greenhouse response, λ is the ratio of the greenhouse response of methane per molecule to that of CO2, C0 and M0 are mass of carbon dioxide and methane, respectively, in the initial atmosphere. From global climate model (GCM) results applied to a warm climate [Berner and Kothavala, 2001], Γmc = 4°C, and λ = 25. -(6) can then be combined and rearranged to solve for Fwsi(T) and Fwc(T).
[17] For the carbonate precipitation and burial flux, Fbc, the approach of Berner et al. [1983] was used:
equation image
where kpp is a rate constant and K is the “equilibrium constant” for reaction (iii) above in terms of global molar masses of oceanic calcium, oceanic bicarbonate, and atmospheric carbon dioxide. Values of kpp and K are calculated so that an initial steady state value of Fbc(0) = Fwsi(0) + Fwc(0) is obtained and also so that the product kppK = (mean oceanic turnover time)−1 [Berner et al., 1983]. To satisfy these criteria, the values K = 2000 (1018 mol)2 and kpp = 0.759 (1018moles)−2 Myr−1 were used, giving a mean oceanic turnover time of 660 years.

[18] The expression for the uptake or release of carbon by the terrestrial biosphere, Fbio, is based on a function integrating the effects of changes in global mean temperature and atmospheric CO2 level on the carbon stocks of vegetation and soils [Beerling et al., 2002]. The function represents a fitted response surface to multiple simulations defining the sensitivity of terrestrial carbon storage described by a process-based model of the terrestrial carbon cycle and a global paleoclimate model simulation of the late Jurassic [Beerling, 2000]. The terrestrial carbon cycle model [Woodward et al., 1995] accounts for the effects of climate and atmospheric composition on photosynthetic primary production, vegetation biomass, autotrophic and heterotrophic respiration, and soil organic matter dynamics. At each small time step, global net changes in carbon storage by the terrestrial biosphere are calculated, and the flux Fbio is estimated as the change in biomass between t and t − 1.

3.2. Carbon Isotope Mass Balance

[19] The carbon isotopic composition for seawater was calculated from mass balance expressions for the ocean (equation (2)) and for the combined ocean and atmosphere:
equation image
equation image
where δ?‰) = [(13C/12C)/(13C/12C)stnd − 1]1000, δa refers to atmospheric CO2, and δw refers to oceanic dissolved inorganic carbon. Introducing the simplifications that the ocean and atmosphere are in isotopic equilibrium (δa = δw − 7‰), that δbc is the same as that for the oceans δw, and that fractionation via terrestrial photosynthesis is ∼19‰ (δbio = δa − 19‰), we get, by combining equations (2), (8), and (9), and solving for δw,
equation image

3.3. Computational Method

[20] All simulations used a simple Eulerian numerical computation with a time step of 500 years. Initial values were taken for a steady state silicate-carbonate cycle based on the results of Berner and Kothavala [2001] for 200 Myr ago. They are (in 1018 mol or 1018 mol Myr−1) C0 = 0.25, (1500 ppm), M0 = 1.67 × 10−4 (1 ppm), Fv(0) = 7, Fm(0) = 0, Fwsi(0) = 7, Fwc(0) = 30, and Fbc(0) = 37. Initial oceanic values for Mca and Cw were calculated for equilibrium via reaction (iii) to be Mca = 15.569, Cw = 5.937. Isotopic compositions (‰) assumed were δv = −6, δm = −60, δwc = 2, and initial δw = 0.6. A control simulation with the initial steady state conditions for 2 Myr, and with no extra methane or volcanic CO2, showed a drift of <1 ppm CO2 and <0.1% in δw due to arithmetic rounding.

4. Simulations of Environmental Change Due to CAMP Eruptions

4.1. CO2 Degassing During CAMP Eruptions

[21] To calculate rates of CO2 production during the eruption of continental flood basalts, it is necessary to first have some idea of the mass of CO2 emitted per unit volume of basalt erupted. There are two such estimates that can be used to set the upper and lower bounds (Table 1). The higher proportion of CO2 is based on the volatile budget data for Kilauea derived from observations over a 27-year period [Gerlach and Graeber, 1985]. The lower proportion is derived from Leavitt's [1982] empirical relationship relating the number of moles of CO2 emitted during an eruption to its magnitude.

Table 1. CO2 Degassing Rates During Volcanic Eruptions and Estimated Volume of Central Atlantic Magmatic Province (CAMP) Basalts
CO2 Degassing, g km−3 basalt Reference CAMP Volume, km3a Total CO2 Released, Gt CO2b Total C Released, Gt Cb
Lower bounds 3.5 × 1012 Leavitt [1982] 2 × 106 7000 1900
1.6 × 1013 Gerlach and Graeber [1985] 2 × 106 32,000 8727
Upper bounds 3.5 × 1012 Leavitt [1982] 4 × 106 14,000 3800
1.6 × 1013 Gerlach and Graeber [1985] 4 × 106 64,000 17,454

[22] The total mass of CO2 emitted during the CAMP eruptions is calculated as the product of a given CO2 emission per unit volume of basalt and the total volume of CAMP volcanics. Initially, the volume of basalts associated with the breakup of the Central Atlantic region in the Early Jurassic was estimated to be rather modest (∼50,000 km3) [McHone, 1996]. However, the realization that the basalts were in place before continental rifting took place [Deckart et al., 1997] led to the recognition of their widespread distribution [Marzoli et al., 1999]. CAMP is estimated to cover a total area in excess of 7 × 106 km2, conservatively assumed to correspond to a volume of 2 × 106 km3. Circumstantial evidence suggests that seaward dipping reflectors off the eastern United States may also be part of CAMP, although the dating is poorly constrained [Olsen, 1999]. Inclusion of the volume of these purported terrestrial flows would double the total CAMP volume to 4 × 106 km3, and even this can be regarded as a conservative estimate if the intrusive volume of CAMP is considered. Therefore we set the upper and lower bounds for the CAMP volume at 2 × 106 km3 and 4 × 106 km3, respectively.

4.2. Mantle CO2 Degassing Scenarios

[23] Four simulations of the effects of volcanic CO2 emissions from the CAMP basalts on the evolution of atmospheric CO2, global mean surface temperature (equation (6)), the carbon isotopic composition and storage of terrestrial organic carbon (vegetation biomass and soil organic matter) were undertaken. In each simulation the magnitude of Fv (volcanic degassing) was varied to give total emissions of (1) 1326, (2) 7957, (3) 13,262, and (4) 21,220 Gt C (Figure 3). The first and second scenarios encompass the estimates assuming a CAMP volume of 2 × 106 km3 (Figure 3 and Table 1), while all four scenarios provide a range of possible emissions for the higher CAMP volume of 4 × 106 km3 (Figure 3 and Table 1). The fourth scenario was included to set an extreme upper limit of CO2 degassing during the CAMP eruptions.

Details are in the caption following the image
Example of Gaussian distribution of volcanic CO2 degassing rates (Fv) (top left) and CH4 (bottom left) release (Fm) from methane hydrates over time used as inputs into various carbon cycle simulations. Fv plot is an example corresponding to cumulative excess release of 7957 Gt C; Fv plots for other cumulative amounts show same time distribution but different peak values. Fm plot is same for all scenarios. Top and bottom right depict the four cumulative excess (above background) CO2 release scenarios (Table 1) and that of CH4, respectively. Note that Fv and Fm are not maintained at above background rates for more than 500 and 50 kyr, respectively.

[24] In order to run the simulations in a reasonably realistic manner, Fv was assumed to follow a Gaussian-type time distribution with a degassing peak at 250 kyr after the start of a simulation and with a peak half-height time of ±100 kyr. This sets the duration for CO2 production at 500 kyr, in agreement with dating evidence suggesting short-lived peak intensity at ∼200 Myr ago [Marzoli et al., 1999], lasting for ∼580 ± 100 kyr [Olsen et al., 1996].

[25] In a separate series of simulations we investigated the possible effects of a suggested [Pálfy et al., 2001, 2002] CH4 release from seafloor sedimentary methane hydrate reservoirs. We assigned an arbitrary lag time of 50 kyr for delay between peak volcanic degassing and CH4 addition to the atmosphere. Identification and dating of other episodic CH4 events in the geologic record during the Latest Paleocene Thermal Maximum [Norris and Röhl, 1999] and the Late Jurassic [Hesselbo et al., 2000; Padden et al., 2001] indicate that the isotopic excursion typically occurs over 100 kyr. Therefore we set the degassing peak for Fm at 300 kyr after the start of a simulation with a peak half-height time of ±20 kyr (Figure 3). This means that essentially all CH4 degassing occurred between 300 and 400 kyr after the start of each simulation.

5. Results and Discussion

5.1. Environmental Effects of CO2 Outgassing From CAMP Basalt Eruptions

[26] Model results indicate that three of the four volcanic CO2 emission scenarios we investigated for the emplacement of the CAMP across the TR-J boundary induced transient changes in some aspects of the global environment (Figure 4 and Table 2). In particular, atmospheric pCO2 increased over the late Triassic value (1500 ppm) by between +500 and +1500 ppm, global mean temperatures increased by +1.2° to +2.8°C, and terrestrial carbon sequestration increased by +265 to +571 Gt C (Table 2). Atmospheric pCO2 levels remain elevated after volcanic CO2 degassing (Fv) returns to the background rate because of the slow consumption of CO2 by weathering and reequilibration of the ocean-atmosphere system with CaCO3. CO2-enhanced greenhouse warming increases the weathering of silicate rocks and leads to enhanced fluxes of Ca2+ to the ocean and (ultimately) leads to the deposition of excess CaCO3 onto the ocean floor (reactions -(iii)). In effect, CO2 released by CAMP volcanism became buried as marine carbonate. Since global temperatures are dependent on atmospheric pCO2, and terrestrial carbon storage responds to both, these variables all have similar trajectories. The curves indicate that if the volcanic degassing took place over 0.5 Myr, then the carbon cycle still required ∼2 Myr for a complete return of CO2 to a lower steady state concentration.

Details are in the caption following the image
Modeled changes in concentration of atmospheric CO2 (pCO2, top left), global mean temperatures (bottom left), stable carbon isotope composition of bulk terrestrial organic matter (δ13COM, top right), and carbon storage by the terrestrial biosphere (bottom right) in response to CO2 emissions during eruption of Central Atlantic Magmatic Province (CAMP) across Triassic-Jurassic boundary, 200 Myr ago. Lines numbered 1–4 indicate cumulative releases given in Figure 3 and Table 2.
Table 2. Effects of CO2 Outgassing on Global Environment During Emplacement of Central Atlantic Magmatic Province 200 Myr Agoa
Simulationb Atmospheric pCO2 Level, ppm Global Mean Temperature Change (ΔT), °C δ13Com, ‰ Terrestrial Carbon Storage, Gt C
CO2 CO2 + CH4 CO2 CO2 + CH4 CO2 CO2 + CH4 CO2 CO2 + CH4
1 (1326 Gt C) +78 +492 +0.2 +1.1 −0.12 −3.03 +47 +259
2 (7957 Gt C) +504 +951 +1.2 +1.9 −0.23 −2.67 +265 +434
3 (13,262 Gt C) +861 +1364 +1.8 +2.6 −0.25 −2.47 +405 +543
4 (21,220 Gt C) +1516 +2058 +2.8 +3.5 −0.40 −2.29 +571 +622
  • a Results from simulations using carbon cycle model with (CO2 + CH4) and without CH4 release. Values are expressed as change relative to Latest Triassic situation: pCO2 = 1500 ppm, global mean temperature change = 0.0 °C, δ13COM = −25.4‰, and total carbon storage in terrestrial biosphere = 2960 Gt C.
  • b Simulations correspond to cumulative Fv release given in brackets (see also Figure 3). Results with CO2 and CH4 all used same cumulative Fm of 5317 Gt C.

[27] We calculate that the CAMP eruptions, limited by the total masses of CO2 estimated by others and listed in Table 1, could have increased atmospheric pCO2 by as much as ∼850 ppm above an end-Triassic value of ∼1500 ppm (scenario 3 in Figure 4). This is greater than that calculated for the eruption of the Deccan Traps in India during the end-Cretaceous (200–250 ppm) [Caldeira and Rampino, 1990], because of differences in the mass of CO2 emitted (Table 1), itself a function of the original volume of the basalts and CO2 production rates. For the Deccan Traps, the widely accepted total volume of 2 × 106 km3 [Wignall, 2001] is better constrained than that for the CAMP basalts [Olsen, 1999]. On the basis of a degassing rate of 1.1 × 1013 g CO2 km−3 of basalt, Caldeira and Rampino [1990] obtained the 200-ppm rise in atmospheric CO2 under end-Cretaceous conditions by simulating a total Deccan Traps release of 6000 Gt C. This represents only an intermediate case in our set of scenarios for the CAMP eruptions (Table 1).

[28] The isotopic composition of terrestrial organic matter (δ13COM) shows a negative excursion driven by the input to the carbon cycle of isotopically light mantle-derived CO2 (Figure 4). It also shows a very small rise, following the initial drop, which dampens with time. The damped oscillation in δ13COM is a consequence of the redistribution of 13C among the reservoirs, especially the exchange of carbon between the terrestrial biomass and the atmosphere (Fbio). Most notable is that the magnitude of the initial δ13COM excursion, even with the largest input of carbon by volcanic outgassing, is only −0.4‰, which is incompatible with observations (−2.0 to −3.5‰) [McElwain et al., 1999; Pálfy et al., 2001; Ward et al., 2001; Hesselbo et al., 2002]. Changes in δ13C values for carbonate, organic matter in marine sediments, and terrestrial organic matter should vary to approximately the same degree due to essentially constant fractionation during photosynthesis and to carbon isotopic equilibrium between the atmosphere and the oceans (see discussion leading to equation (10)).

[29] Similar effects of volcanic emissions on the global carbon isotope system in general have been reported by Kump and Arthur [1999] using an alternative geochemical carbon cycle model. Those authors showed that a huge volcanic CO2 input (∼200,000 Gt C) produces a small negative δ13C excursion (−0.5‰) in marine carbonates. Kump and Arthur [1999], however, obtained a −2‰ excursion in marine δ13COM due to the use of an asymptotic expression describing the CO2 dependence of photosynthetic isotopic fractionation by marine algae during the synthesis of organic matter. Inclusion of this function in our model increases the magnitude of the δ13COM excursion by a maximum of only 0.3‰ because photosynthetic isotopic fractionation is largely saturated at the high background atmospheric pCO2 levels of the late Triassic.

[30] A collapse in marine primary production [Pálfy et al., 2001; Ward et al., 2001] would be anticipated to have an additional effect on atmospheric pCO2 [Sarmiento and Toggweiler, 1984] and δ13COM [Kump, 1991; Kump and Arthur, 1999], which is not considered in the present one-box ocean modeling. Under present-day conditions a collapse in the biological pump results in an increase in pCO2 of from 300 to ∼500 ppm [Sarmiento and Toggweiler, 1984]; for a 1500-ppm equilibrated ocean, the analogous increase in pCO2 would be from 1500 to ∼2300 ppm due to the greater initial acidity of such an ocean (K. Caldeira, unpublished calculations, 2002). However, because of the high acidity, the surface ocean would immediately begin to dissolve the abundant carbonates in shallow water sediments. Therefore, after equilibration of the surface ocean with CaCO3, the atmospheric pCO2 would drop to only ∼1700 ppm, giving a rise of 200 ppm (K. Caldeira, unpublished calculations, 2002). Further, the modeling shows that this change in atmospheric CO2 could last only for a few thousand years before other changes accounted for in the present modeling would begin to dominate, including carbonate reequilibration. The instantaneous loss of net production would also result in a sudden loss of the ocean vertical δ13C gradient, with surface carbon becoming isotopically lighter [Kump, 1991], but by a maximum of only ∼1–1.5‰ [D'Hondt et al., 1998]

[31] To test the effect of a collapse in marine productivity on atmospheric CO2 and δ13C over the longer term, the organic subcycle was added to the modeling and was unbalanced by letting the sedimentation and burial of organic matter go to zero for a long time, between 300 and 700 kyr, while the weathering of sedimentary organic carbon continued during this period at the previous rate of 2 × 1018 mol Myr−1 [Berner and Kothavala, 2001]. This maximum effect was a rise of the peak CO2 value by ∼200 ppm and a further lowering of the negative peak value of δ13C by about −0.5‰. Therefore we suggest that any effects of the collapse in marine primary production on CO2 and the negative isotopic anomaly would be relatively rather small, including the initial changes occurring right after the collapse.

5.2. Transient CO2 Fertilization of Chemical Weathering by Terrestrial Plants

[32] The biogeochemical weathering of minerals is brought about by plants and the mycorrhizal fungi attached to their roots. Experimental work has shown that in situ exposure of a loblolly pine (Pinus taeda) forest to an atmosphere enriched in CO2 by 200 ppm, over 2 years, increased the flux of dissolved inorganic carbon (bicarbonate) to groundwater by 33% [Andrews and Schlesinger, 2001]. Other studies have demonstrated that as atmospheric CO2 levels increase, more carbon is exported belowground by the plants, increasing the activity and growth of ectomycorrhizal fungi, themselves important weathering agents of forest soils [Landeweert et al., 2001]. It is conceivable therefore that a transient atmospheric pCO2 increase across the TR-J boundary stimulated a negative feedback on CO2 levels by enhancing plant-mediated chemical weathering of terrestrial silicate rocks.

[33] We assessed the strength of this feedback by including a Michaelis-Menton function describing the relationship between vascular plants, CO2, and chemical weathering of Ca-Mg carbonates (Fwc) and silicates (Fwsi) [Berner and Kothavala, 2001]. The Fwsi(T) and Fwc(T) terms (equations (4) and (5)) were each modified via multiplication by the weathering feedback term fb(CO2):
equation image
where RCO2 is the ratio of mass of CO2 at time t to that at the present (t = 0), and FERT is an exponent reflecting the proportion of plants globally that are fertilized by increasing CO2 and accelerated mineral weathering. We set FERT = 0.4, to mimic the results of Andrews and Schlesinger [2001], but applied it to only ∼35% of all land vegetation, to express the fact that fertilization of plant growth can be limited by factors other than CO2, such as light, nutrients, and water.

[34] To maximize any potential effects of equation (11) on the modeled changes in atmospheric pCO2, global mean temperature, δ13COM, and terrestrial carbon storage, we used the highest degassing scenario, giving a cumulative release of 21,220 Gt C (Figure 3). Under this scenario the plant-CO2-weathering feedback reduced atmospheric pCO2 levels by 200 ppm and reduced global temperatures by 0.3°C compared to the control (without CO2 fertilized weathering) case (Figure 5). Atmospheric pCO2 and temperature effects marginally lower global terrestrial carbon storage (by ∼40 Gt C), and there is virtually no impact on δ13COM (Figure 5). With the other three scenarios, involving the release of lower amounts of CO2 during the CAMP eruptions, the effects become even lower.

Details are in the caption following the image
Effects of including (dotted line) and excluding (solid line) CO2 fertilization of plant-weathering on the concentration of atmospheric CO2 (pCO2, upper left hand panel), global mean temperatures (lower left hand panel), the stable carbon isotope composition of bulk terrestrial organic matter (δ13COM, upper right hand panel) and carbon storage by the terrestrial biosphere (lower right hand panel) in response to CO2 emissions during the eruption of the CAMP across the Triassic-Jurassic boundary, 200 Myr ago. Results are shown for scenario 4 (Figure 3, Table 2) only.

[35] It appears quite plausible that a rise in atmospheric CO2 across the TR-J boundary exerted a negative feedback effect by accelerating plant-mediated chemical weathering of sedimentary rocks (Figure 5). The pCO2 drawdown is, however, not significant relative to the CO2 fluxes associated with the volcanic activity. However, this conclusion depends crucially on the proportion of plants at the global scale that behaved like the loblolly pine forest in the experiment of Andrews and Schlesinger [2001], a significant unknown at present.

5.3. Environmental Effects of CO2 Outgassing With a Positive Feedback on CH4 Release

[36] Calculations with our greenhouse carbon cycle model indicate that CO2 emissions by volcanism fail to account for the 2.5–3.5‰ negative isotopic excursion shown by terrestrial and marine continental sediments [McElwain et al., 1999; Ward et al., 2001; Hesselbo et al., 2002] and the parallel bulk organic matter and carbonates from a Tethyan locality [Pálfy et al., 2001]. Even if we sum the likely contributions of volcanism (−0.4‰) and CO2 dependence of marine algal carbon isotope fractionation (−0.3‰) and allow a negative 0.5‰ excursion for a collapse in marine primary production due to the end-Triassic mass extinctions, the final negative isotopic anomaly of 1.2‰ is far short of that observed. Climate change related impacts on carbon isotope fractionation by the terrestrial biosphere may also be possible. However, these have not been considered because they will vary at the regional scale depending on the global climate system response to a rise in CO2, and, moreover, the fluxes of carbon involving the terrestrial biosphere are relatively small, so the net effect in the global mass balance (equation (10)) are likely to be correspondingly small.

[37] These considerations indicate the requirement for introducing into the global carbon cycle an isotopically light source of carbon, in addition to mantle-derived CO2. CH4 is extremely isotopically light (−60‰) and thus might provide an appropriate carbon input [Pálfy et al., 2001, 2002]. Global warming due to a rise in volcanogenic CO2 could have triggered the release of methane stored in methane hydrates, which was rapidly oxidized to CO2, thereby acting as a positive feedback. The mass of CH4 released from the seafloor across the TR-J boundary under such circumstances is uncertain. Here we determined it by sensitivity analyses and found that to obtain a negative 3‰ excursion in δ13COM under Late Triassic conditions during CAMP CO2 outgassing required the total release of ∼5000 Gt C as CH4 spread over a period of 100 kyr (Figures 2 and 6). Such a massive CH4 release is more than double that implicated in the perturbation of the global carbon cycle during Latest Paleocene Thermal Maximum (LPTM) (1000–2100 Gt C) [Dickens et al., 1997], but is comparable to that calculated for the isotopic anomalies in early Jurassic sediments [Hesselbo et al., 2002; Beerling et al., 2002].

Details are in the caption following the image
Modeled changes in concentration of atmospheric CO2 (pCO2, top left), global mean temperature (bottom left), stable carbon isotope composition of bulk terrestrial organic matter (δ13COM, top right), and carbon storage by terrestrial biosphere (bottom right) in response to CO2 emissions with a positive feedback including release of CH4 from seafloor hydrates during eruption of CAMP across Triassic-Jurassic boundary, 200 Myr ago. Lines numbered 1–4 indicate the cumulative releases given in Figure 3 and Table 2.

[38] When this positive feedback is included in the modeling, it amplifies earlier changes in the global environment, compared to the results obtained by CO2 degassing alone (Figure 6 and Table 2). Atmospheric pCO2 levels increased by an additional 400–500 ppm, induced further greenhouse warming of up to 0.7°C, and led to extra carbon sequestration by the terrestrial biosphere of ∼200 Gt (Table 2). Injection into the ocean-atmosphere system of ∼5000 Gt C as CH4 accounts for the magnitude of the observed shift in δ13COM, but also adds structure to the curve (Figure 6). The curve now shows an abrupt initial decline to its most negative value over a period of ∼70 kyr, followed by a slower period of recovery of ∼640 kyr (Figure 6). The overall pattern is reminiscent of that shown by high-resolution carbon isotope records across the LPTM (reviewed by Dickens [2001]). That event, however, was apparently much shorter, with most of the 12C release taking place over 52 kyr and with the excursion taking 220 kyr to reach steady state [Röhl et al., 2000].

[39] Similar results (within 200 ppm for CO2 and 0.5‰ for δ13C) were obtained if CH4 was assumed to be injected directly into the oceans rather than the atmosphere. As in the atmosphere, methane was assumed to be oxidized rapidly, in this case by methanotrophic bacteria. The oxygen demand at the assumed peak of CH4 degassing (100 Gt kyr−1; see Figure 3) is ∼20 times lower than the standing crop of O2 dissolved in the (modern) oceans, and because the oceanic overturn and O2 resupply time is ∼1 kyr, it is unlikely that CH4 oxidation led to an appreciable depletion of oceanic O2. Thus we have ignored depletion of oceanic O2 that otherwise could lead to somewhat different results [Dickens, 2000].

[40] Further analysis of the two sets of model results (with and without CH4 release) indicates that the rise in atmospheric pCO2 and global mean temperature are linearly related to the total mass of CO2 released during the extrusion of the CAMP basalts (Figure 7). Inclusion of CH4 addition to the ocean-atmosphere system simply displaces the line upward for each variable. From Figure 7 it can be seen that an atmospheric pCO2 rise of ∼1000 ppm, based on our stomatal index and paleosol-derived estimations, requires the cumulative release of 8000–9000 Gt C as CO2 across the TR-J boundary, in addition to release of sedimentary CH4. Our synthesis of model simulations (Figure 7), and the paleo-evidence, indicates, therefore, that the actual amount of CO2 degassing during the emplacement of the CAMP volcanic basalts has been underestimated (Table 1) and/or that the original volume of the province lies, as suspected [Marzoli et al., 1999; Olsen, 1999], somewhere between 2 and 4 × 106 km3.

Details are in the caption following the image
Maximum change in atmospheric CO2 level (pCO2, top) and global mean temperatures (bottom) to total mass of CO2 released during the extrusion of CAMP under end-Triassic greenhouse conditions. In each case, results for CO2 only (solid line) and CO2 + CH4 release (dashed line) are depicted. Regression details: (top) CO2 only case: slope = 0.072, intercept = −49.78 (r = 0.99); CO2 + CH4 case: slope, = 0.079, intercept = 353.38 (r = 0.99); (bottom) CO2 only case: slope = 0.00013, intercept = 0.0858 (r = 0.99); CO2 + CH4 case: slope = 0.00012, intercept = 0.948 (r = 0.99).

5.4. Wider Issues

[41] If our proposed sequence of events for a massive carbon cycle perturbation across the TR-J boundary is correct, it places the duration for the atmospheric pCO2 and δ13COM excursions at between 700 and 1000 kyr (Figure 6). Dating of the duration of the Rhaetian and Hettangian stages differ, with Harland et al. [1990] giving 2 and 4.5 Myr, respectively, Gradstein et al. [1994] giving 3.9 and 3.8 Myr, and Pálfy et al. [2000b] reporting, on the basis of radiometric dates, a duration of 3.1 Myr for the Hettangian. All indications suggest that the TR-J boundary isotopic and pCO2 fingerprints occur within only part of the Rhaetian and Hettangian stages. Therefore our timescale for the carbon cycle dynamics of the event is certainly compatible with all of these considerations. Indeed, a preliminary timescale attached to the δ13COM signal at the Kennecott Point section in Canada suggests a total duration for the excursion of ∼500 kyr [Ward et al., 2001] and for the sudden extinction of radiolarians of 50 kyr, the latter being comparable to that modeled for initial δ13COM excursion with the inclusion of CH4 release (70 kyr) (Figure 6).

[42] A rise in atmospheric pCO2 of 1000 ppm, and a global greenhouse warming of 2°–2.5°C, occurring against a background pCO2 level of ∼1000–1500 ppm, represents a highly challenging environment for land plants, especially in the hothouse world of late Triassic Pangea. These changes in the global environment indicate that high temperature damage to large leaved plant taxa, hypothesized to underlie the ecological crisis [McElwain et al., 1999], continues to remain a real possibility. Under this scenario, large entire-leaved canopy-dominant plants suffered severe temperature damage to the photosynthetic system, especially during periods of low wind speeds, because of their innate low boundary layer conductances and limited convective cooling capacities. The leaf megafossil record from Greenland [Harris, 1935, 1937; Lundblad, 1959] clearly supports this notion by indicating a pattern of species extinction and replacement whereby large-leaved taxa were lost and replaced by those with more dissected or finely divided leaves. We can envisage no plausible mechanism for disturbing the carbon cycle, or the global climate system, consistent with the available geologic data, that supports the suggestion, based on multivariate analyses and climate correlations of sporomorphs [Hubbard and Boulter, 2000], that the global climate actually cooled across the TR-J boundary.

6. Conclusion

[43] We conclude from our series of carbon cycle simulations for an end-Triassic greenhouse world that CO2 release from the Earth's mantle during the extrusion of the CAMP basalts across the TR-J boundary, 200 Myr ago, was probably insufficient to fully account for the ∼1000 ppm rise in atmospheric pCO2 indicated by stomatal and paleosol evidence. Furthermore, the cumulative injection of a wide range of CO2 masses over 500 kyr failed to reproduce the negative δ13C anomaly seen in marine and terrestrial sedimentary carbon across the TR-J boundary. This conclusion holds even when the model is modified to account for the effects of an instantaneous cessation of organic carbon burial due to a drop in productivity and for the effects of atmospheric pCO2 on carbon isotope fractionation by marine algae. If CO2 degassing occurred over >500 kyr, the effects would be even less than simulated here.

[44] The scenario that best explains the geologic evidence is global warming due to a buildup of volcanically derived CO2 in the atmosphere, triggering a positive feedback whereby seafloor methane hydrate reservoirs become destabilized and release massive amounts of CH4 into the ocean-atmosphere system. Our analyses suggest that this sequence of events involved the release of ∼8000–9000 Gt C as gaseous CO2 during the eruption of the CAMP continental flood basalts and an injection of ∼5000 Gt C from the methane hydrate reservoir. Together, the net effects were a transient peak in atmospheric pCO2 in excess of 2500 ppm and a greenhouse warming of 2°–2.5°C above end-Triassic conditions. Reequilibrium of the ocean-atmosphere system was accomplished over 700–1000 kyr. Our theoretical results therefore provide a preliminary explanation for the estimated pCO2 excursion and for the duration for the isotopic signal observed in sediments.


[45] D. J. B. gratefully acknowledges funding through a Royal Society University Research Fellowship and the Leverhulme Trust, and R. A. B. gratefully acknowledges financial support by the NSF grant EAR-0104797 and DOE grant DE-FG02-01ER15173. We have benefited greatly from the comments of K. Caldeira, including unpublished calculations that he provided to us. This is a contribution to the International Geological Correlation Programme Project 458.