Research Article

Atmospheric Pco2 Perturbations Associated with the Central Atlantic Magmatic Province

See allHide authors and affiliations

Science  18 Mar 2011:
Vol. 331, Issue 6023, pp. 1404-1409
DOI: 10.1126/science.1199011


The effects of a large igneous province on the concentration of atmospheric carbon dioxide (Pco2) are mostly unknown. In this study, we estimate Pco2 from stable isotopic values of pedogenic carbonates interbedded with volcanics of the Central Atlantic Magmatic Province (CAMP) in the Newark Basin, eastern North America. We find pre-CAMP Pco2 values of ~2000 parts per million (ppm), increasing to ~4400 ppm immediately after the first volcanic unit, followed by a steady decrease toward pre-eruptive levels over the subsequent 300 thousand years, a pattern that is repeated after the second and third flow units. We interpret each Pco2 increase as a direct response to magmatic activity (primary outgassing or contact metamorphism). The systematic decreases in Pco2 after each magmatic episode probably reflect consumption of atmospheric CO2 by weathering of silicates, stimulated by fresh CAMP volcanics.

Large igneous provinces (LIPs) are geologically rapid episodes of extensive volcanism, often flooding vast oceanic or continental regions with several million cubic kilometers of lava (1). In particular, continental flood basalts have the potential to directly perturb Earth’s climate system through the emission of gasses to the atmosphere: most notably, SO2 and CO2, which together may result in an immediate (1- to 10-year) cooling (2, 3), followed by a longer-term (102- to 105-year) warming (4). Of these, only CO2 has the potential to influence climate on both short and long time scales because of its relatively long atmospheric residence time and effectiveness as a greenhouse gas, leading some to conclude that CO2 is the primary driver of Phanerozoic climate (5).

If the concentration of atmospheric CO2 exerts an influence on climate over such broad time scales, what are the effects of a LIP on this essential parameter of the carbon cycle? Although the potential radiative effects of LIP CO2 degassing on the million-year scale have been considered inconsequential (6, 7), shorter (104- to 105-year)–time scale reconstructions of atmospheric partial pressure of CO2 (Pco2) before and after LIP eruptions have not been systematically determined because of inadequate chronostratigraphic resolution in most settings (8, 9). Consequently, the direct Pco2 effect of a LIP remains untested empirically.

Intriguingly, LIP volcanism is often temporally associated with mass extinction events throughout Earth’s history (10). The three largest continental LIPs of the Phanerozoic are the Siberian Traps, the Central Atlantic Magmatic Province (CAMP), and the Deccan Traps, each of which is linked to one of the “Big 5” Phanerozoic mass extinctions [the end-Permian, end-Triassic, and the Cretaceous–Paleogene events, respectively] (11, 12). Though attempts have been made to estimate the gaseous emissions attributable to the Deccan (6, 13, 14) and Siberian (15, 16) traps, it is difficult to demonstrate causality because the uncertainties in correlating these Pco2 estimates from afar to the volcanic stratigraphy itself are usually no better than the turnover time of an atmospheric Pco2 perturbation (17). Of these, only the CAMP is sequenced in high-resolution, temporally continuous sediments that contain paleosols appropriate for estimating Pco2 and have a well-established chronology (18, 19) and extinction level.

Extrusives from the CAMP (20) are preserved in direct stratigraphic succession with cyclic continental sediments in the Newark Basin of eastern North America (Fig. 1). Milankovitch cyclostratigraphy of the primarily lacustrine sediments interbedded within the CAMP extrusives have yielded precise age control (to the level of orbital precession) and an estimated total volcanic duration of ~600 ± 20 thousand years (ky) (21, 22). In this same Newark Basin section, palynofloral evidence of the end-Triassic extinction (ETE) is found stratigraphically just below the first of the CAMP volcanics, preceding the magmatism by ~20 ky [(23), see (24) for review]. Also interspersed throughout these sediments, and often forming from CAMP lava flows themselves, are pedogenic carbonate-bearing paleosols (Fig. 2, A and B), which can be used to estimate ancient atmospheric Pco2 (25). Thus, the Newark stratigraphy is ideally situated to directly test the Pco2 effect of a LIP. Previous attempts at reconstructing the Pco2 effect associated with CAMP extrusives had very sparse sampling resolution (26) or had to rely on imprecise long-distance correlation (8, 9).

Fig. 1

(A) Stratigraphy and lithology (22, 34, 41) of the upper Newark Basin stratigraphic section, based on assembly of a series of short cores taken by the ACE covering the extrusive interval in high resolution (49), with substantial overlap both internally and with the Newark Basin Coring Project (NBCP) Martinsville core (19, 22) and outcrop. Note that the ETE event (red) is several meters below the equivalent of the Orange Mountain Basalt (OMB, the first flow unit) in the Jacksonwald section of the Newark Basin (24). Stratigraphic thickness is scaled arbitrarily from the base of the laterally extensive OMB. J, Jurassic; Tr, Triassic; UU through MM are stratigraphic members. (B) Profile-equilibrated mean δ13C values of pedogenic carbonate in the Newark stratigraphic section. Error is ±SD of mean (Fig. 2 and table S1) (29). Circles, samples from core; squares, outcrop. PDB, Pee Dee belemnite. (C) Measured δ13C values of preserved soil organic matter from clay linings or as close to the paleosol surface as possible. (D) Results of the pedogenic carbonate paleobarometer based on the input variables from (B) and (C) at 25°C. The concentration of respired CO2 in the soil [S(z)] was estimated to be 3000 ± 1000 ppm (error bars), corresponding to a plausible range for midproductivity tropical soils and probably encompassing the range of calculated atmospheric Pco2 values. Carbon-cycle perturbations are built into the model because the carbon isotopic ratio of the atmosphere (δa) is calculated from the measured δ13Corg by: δa = (δ13Corg + 18.67)/1.10 (32), which assumes consistent fractionation by photosynthesis [see (29) and table S1 for numerical values].

Fig. 2

Carbonate-bearing calcic vertisols formed directly on the tops of the lava flows; (A) on top of the Orange Mountain Basalt in ACE core PT-26, box 18; and (B) on top of the Preakness Basalt in ACE core C-124, box 3 (see Fig. 1 for stratigraphic column, fig. S1 for core locations). Scale bar, 30 cm. Depth in core decreases from left (deepest) to right (shallowest); up-core direction is indicated by the arrow; holes are soil-carbonate sampling locations. Such superposition removes any stratigraphic uncertainty from these Pco2 estimates with regard to the flow units themselves. (C) Down-soil δ13C profiles of pedogenic carbonates (δcc) from representative paleosols in the Passaic, Feltville, Towaco, and Boonton Formations used in this study (symbols encompass analytical error), compared with the δcc predicted by the diffusion model at atmospheric CO2 concentrations of 2000 and 4000 ppm (blue and pink lines, respectively) [after (31)]. For this exercise only, atmospheric δ13CO2 was set to –6.5‰, and soil δ13Corg was set to –26.5‰, with an exponential production function and characteristic depth of production at 15 cm [see (29) for description of other parameters]. For all Pco2 estimates made in this study, the δ13Corg was measured directly and used as a model input. Soil carbonate above 20 cm in the profile was rare. Note that stabilization of measured δcc is often well below 50-cm soil depth. Using the mean of these depth-stabilized measurements ensures the mixing between the atmospheric and soil-respired reservoirs at equilibrium with respect to the diffusion model. Photographs of the soils used in (C), from samples NFPT26 and NTC124, are shown in (A) and (B).

We use δ13C measurements of pedogenic carbonate nodules from paleosols stratigraphically distributed before and after each extrusive unit to generate a high-resolution Pco2 record through the Newark Basin CAMP sequence (Fig. 1). The extrusion of ~2 to 4 × 106 km3 of volcanics (27, 28) in less than 1 million years (My) implies a measurable effect on atmospheric Pco2, which our temporal resolution should allow us to detect. According to the model of Dessert et al. (17) scaled to the Deccan Traps, the transient increase in Pco2 is on the time scale of the eruptions, after which continental silicate weathering should lower Pco2 to pre-eruption levels in ~1 My.

Estimating Pco2 from pedogenic carbonates. Organic and inorganic carbon isotope measurements on paleosols from outcrop—and from multiple, stratigraphically overlapping cores taken by the Army Corps of Engineers (ACE) through the extrusive interval (fig. S1) (29)—are used as inputs into the diffusion model of Cerling (30)Ca=S(z)δs1.0044δϕ4.4δaδs(1)where Ca is the concentration of atmospheric CO2, S(z) is the concentration of CO2 due to respiration of soil organic matter, δs is the δ13C of soil CO2, δϕ is the δ13C of soil-respired CO2, and δa is the δ13C of atmospheric CO2. All δ values are relative to Vienna Pee Dee belemnite (VPDB). The temperature of calcite precipitation is set at 25°C, relating the carbon isotopic ratio of soil carbonate (δcc) to δs (29). As an independent objective metric of soil applicability, multiple (three or more) down-profile δcc measurements were made on each paleosol to reproduce the expected exponential decrease toward stabilization with depth (Fig. 2) (29, 31). Using the mean of these soil-equilibrated measurements ensures that the mixing between the atmospheric and soil-respired reservoirs is at equilibrium with respect to the diffusion model. The measured carbon isotopic ratio of soil organic matter (δ13Corg) is related directly to δϕ (29). Carbon-cycle perturbations are built into the model because the carbon isotopic ratio of the atmosphere (δa) is calculated from the measured δ13Corg by assuming consistent fractionation by photosynthesis: δa = (δ13Corg + 18.67)/1.10 (32). We use an S(z) value of 3000 ± 1000 parts per million (ppm), appropriate for soils developed in a semi-arid climate with moderate productivity (30, 33).

Atmospheric Pco2 estimates in superposition with CAMP basalts. Pedogenic carbonates from the upper Passaic Formation deposited before the first CAMP volcanic unit (Orange Mountain Basalt) have δ13C values generally below –5 per mil (‰) VPDB, with ±1.5‰ variability (Fig. 1B). The carbon isotopic composition of soil organic matter (δ13Corg) is remarkably stable around –26‰ (Fig. 1C), consistent with but less variable overall than results from independent analyses of wood and total organic carbon from elsewhere in the Newark and Hartford stratigraphy (34). Model results of the pedogenic carbonate paleobarometer yield Pco2 estimates with a mean of ~2000 ppm [S(z) = 3000 ppm] in the pre-CAMP stratigraphy (Fig. 1D). These estimates from over 500 m of the uppermost Passaic Formation, deposited over ~2.5 My of the latest Triassic (18, 19), show internal consistency between outcrop and between cores and, with the tight stratigraphic resolution, suggest that there are no major CO2-producing magmatic events before the observed Orange Mountain Basalt. This stable Pco2 background is probably a measure of equilibrium continental silicate weathering in the Late Triassic.

The ~2000-ppm pre-CAMP Pco2 baseline found in this study is not inconsistent with widely (geographically and stratigraphically) dispersed pedogenic carbonate estimates from elsewhere in the Newark Supergroup (26, 35) or reconstructions from the Late Triassic Petrified Forest section (Fig. 3) (8). However, the Petrified Forest section lacks CAMP volcanics and can be correlated to the Newark succession only at the stage level (36), in which case the Rhaetian samples most likely overlap with the 2.5-My (18) pre-CAMP samples of this study but show considerably greater variability (500 to 3500 ppm) (Fig. 3). The high variability in the Petrified Forest section is probably due to the lack of down-profile isotope measurements and the use of temperatures estimated from the δ18O of pedogenic calcite, which lacks a reliable calibration for use in the paleo record (37).

Fig. 3

(A) Atmospheric Pco2 estimates previously made on Late Triassic and Early Jurassic sections using low-resolution pedogenic carbonates from other basins in the Newark Group (26) and the Rhaetian data of Cleveland et al. from the Petrified Forest section (8), stomatal densities from sections in Greenland and Sweden (9), and geochemical modeling (50). Approximate placement of the CAMP is shown in blue. The McElwain data were placed using the magnetic stratigraphy of Kent and Clemmensen (51) and Whiteside et al. (34). (B) Pco2 estimates of this study with time [error bars are S(z) = 3000 ± 1000 ppm]. The samples taken in this study are from the same series of ACE cores used to construct a time sequence through the magmatic interval in the Newark Basin (22). The heavy blue lines indicate the temporal placement (and duration) of each CAMP flow unit. Ma, million years ago.

The carbon isotope value of pedogenic carbonate formed directly on top of the Orange Mountain Basalt (first flow unit) is enriched in 13C by 1 to 2‰ above the pre-CAMP background (Fig. 1B). This increase and similar increases above the Preakness and Hook Mountain Basalts (the second and third flow units, respectively) reflect the increased influence of the atmospheric reservoir on soil carbonate formed at depth. Stratigraphically, above each flow unit, the carbon isotope composition of soil organic matter decreases by ~0.5 to 1.0‰, followed by a general return of δ13Corg values to around –26‰, but with increased variability (Fig. 1C). The δ13C values of pedogenic carbonate and soil organic matter directly above the Preakness Basalt have particularly good reproducibility between several laterally equivalent individual cored sections, allowing confidence in these measurements. Those soils formed on top of the Orange Mountain Basalt (Fig. 2B) show an increase in Pco2 to 4400 ppm (Fig. 1D), which amounts to a doubling of Pco2 above the pre-CAMP baseline. In every case, pedogenic carbonate samples of soils formed on the tops of the basaltic units yield Pco2 estimates that are distinctly higher than the immediately pre-eruptive background levels (Fig. 1): 4200 ppm on top of the Preakness Basalt compared with 3000 ppm in the uppermost portion of the underlying Feltville Formation, and 5000 ppm directly on the Hook Mountain Basalt compared with 2500 ppm in the uppermost portion of the underlying Towaco Formation. This pattern suggests that the volcanism associated with each lava-flow unit had a direct effect on atmospheric Pco2 by ~2000 ppm and was virtually immediate [to within the resolution of orbital precession (~20 ky)].

Chronostratigraphic control of Pco2 estimates. A previous low-resolution reconstruction from the Newark Supergroup (two localities over 10 My) seemed to suggest relatively stable Pco2 levels across the CAMP interval (26) and, when compared to our study, underscores the difficulty in attempting to capture a transient perturbation of the carbon system without adequate chronostratigraphic control (Fig. 3) (38, 39). The stratigraphic level of the McCoy Brook sample of Tanner et al. (26) from the Fundy Basin is not known precisely (40), and it is quite possible that the soil formed long enough after emplacement of the North Mountain Basalt (correlative to the Orange Mountain Basalt) (41) that the full Pco2 increase was not recorded. As demonstrated by this study, capturing such a transient signal requires <100-ky sampling resolution with respect to the volcanics. Other attempts to estimate Pco2 over this interval, most notably the approximations using stomatal densities of leaf fossils from end-Triassic event boundary sections in Greenland and Sweden (9), indicate a doubling to tripling of Pco2 from 800 to 2100 ppm (Fig. 3). These stomata-based estimates are substantially lower than those found here, but the stomata proxy is thought to underestimate Pco2 (42), in which case a reported doubling to tripling broadly agrees with our findings and is corroborated by other cuticular estimates (43).

If the increase in Pco2 after each major basaltic unit (Orange Mountain, Preakness, and Hook Mountain Basalts) can be ascribed to episodes of magmatic activity, which are likely to be very short (22), then the relatively high-resolution record of Pco2 taken as a whole seems to require just the three distinct episodes of volcanism. This does not necessarily imply that the local thicknesses of the CAMP volcanics in the Newark are representative of the global volume of basalt produced by each flow unit. For example, although the Newark and Hartford Basins show three distinct volcanic episodes in nearly identical chronostratigraphic sequence (41, 44), the thicknesses of the correlative flow units vary by a factor of 2. Nonetheless, the observed magnitude and singularity of the Pco2 response to each flow unit in the Newark Basin implies that the magmatic events that produced them were regionally extensive and voluminous, because atmospheric CO2 is globally homogeneous on the circulation time of the atmosphere. Although the second episode of magmatism produced the thickest lava flow unit (Preakness Basalt) in the Newark Basin, there is only a relatively small corollary increase in Pco2. Interestingly, a middle flow unit equivalent to the Preakness Basalt is not present in the South Atlas region of Morocco, whereas the uppermost flow unit there (Recurrent Basalt) is stratigraphically and geochemically equivalent to the Hook Mountain Basalt in the Newark Basin (41). Although the Hook Mountain Basalt is locally thin in the Newark, the associated Pco2 increase is one of the largest recorded, implying that greater unrecorded volumes were erupted elsewhere. This body of evidence demonstrates the global applicability of the Pco2 findings reported here; together, they make a compelling case that what has been recorded in the Newark Basin is a reasonable representation of a global sequence of events.

To gauge the plausibility of a volcanic CO2 source, we can compare the CO2 efflux potential of the CAMP basalt volume to the effect observed in the Newark. Assuming a total CAMP volume of 2.4 × 106 km3 (27) and a volcanic efflux of 1.4 × 1010 kg of CO2 per km3 (7), we estimate a total CO2 degassing potential of 3.36 × 1016 kg CO2. We focus on the lower volcanic unit (represented by the Orange Mountain Basalt in the Newark Basin), for which the pre-eruption background Pco2 is well established. That initial pulse of activity represents roughly one-third of the total CAMP volume and, therefore, could have produced 1.12 × 1016 kg of CO2, amounting to a ~1400-ppm increase in Pco2 (at 7.82 × 1012 kg CO2 per ppm). This instantaneous approximation is of the same order and within the error of our observed ~2000-ppm Pco2 increase after the first major episode of CAMP volcanism, implying that the volcanic release of CO2 was extremely rapid, which is consistent with the sporadic presence of only very thin and discontinuous sedimentary strata between flows (21, 41). However, this does not preclude that a major component of the observed Pco2 increase is contact metamorphic in origin (45), which is hinted at by the small decrease in δ13COM above each flow unit.

The influence of basalt weathering. An intriguing phenomenon recorded by the Newark Basin paleosol sequence is the gradual apparent decrease in Pco2 over time scales of 105 years after each successive episode of volcanism (Fig. 4). For example, elevated Pco2 values just after the Preakness Basalt in the ~300-ky-long Towaco Formation have nearly returned to pre-eruptive levels by the emplacement of the succeeding Hook Mountain Basalt. We attribute the systematic decrease in Pco2 as the enhanced response of continental silicate weathering consuming each volcanic input of CO2 (46). Given the vast aerial extent of CAMP extrusive activity (27), it is plausible to attribute the rapidity of the decrease in atmospheric CO2 to consumption by hydrolysis of the CAMP volcanics themselves, especially in the tropical humid belt where the engines of continental weathering are most effective (47). Geochemical modeling of the period after emplacement of the Deccan LIP and its corresponding CO2 input show a similar exponential decrease in Pco2 due to consumption by weathering (Fig. 4) (17). Marine osmium isotope evidence also indicates that an increase in continental weathering followed the CAMP interval (48), lending credibility to a weathering hypothesis. Though the CAMP data correspond well to the initial stages of this modeled decrease, the uppermost portion of the Newark Basin section is truncated, so the full extent of this relation cannot be evaluated here.

Fig. 4

(A) Calculated Pco2 of this study versus time relative to the top of the Orange Mountain Basalt in the Newark Basin (the first CAMP flow unit). The Pco2 estimates after each successive flow unit in the Newark have been normalized to the initial extrusive event for comparison to a silicate weathering model [colored symbols; error bars are S(z) = 3000 ± 1000 ppm]. (B) Modeled CO2 consumption due to weathering (17) after emplacement of the Deccan Traps. In the model, a pulse of CO2 was added to the atmosphere (over 100 ky) accompanying the extrusion of the Deccan LIP. The change in Pco2 represents the remaining fraction of the total Pco2 increase in this particular modeling scenario. The abrupt increase of Pco2 and subsequent postextrusive drawdown found in response to the CAMP volcanism in this study (A) is remarkably similar to the modeling results of Dessert et al. (17), bearing in mind the much longer interval of igneous activity assumed in the model.

Implications for the end-Triassic extinction. Neither the Feltville nor Towaco Formations show evidence of Pco2 changes that can be associated with magmatic events other than those directly related to the observed volcanics. Similarly, the stability of Pco2 estimates in the Pre-CAMP Passaic Formation leaves the ETE without an obvious Pco2 precursor. However, the youngest sample in the pre-extrusive section (Exeter Member, Fig. 1) formed in a soil pre-dating the first flow unit by only ~20 ky. This soil occurs two meters below the clay layer containing palynofloral evidence for the ETE in the same exposure, which itself predates the observed onset of volcanism in the Newark by ~19 ky (24). At these sedimentation rates, the uppermost paleosol sample that we studied in the Newark probably pre-dates the ETE by perhaps as little as ~1 ky. Therefore, it is possible that a pulse of CAMP volcanism and an attendant rapid rise in atmospheric Pco2 with associated climatic implications occurred within the ~20-ky paleosol sampling gap before the age-equivalent of the Orange Mountain Basalt but remain undocumented. Nonetheless, the tight stratigraphic constraint implies that whatever phenomenon caused the ETE must have been very abrupt (occurring within a narrow thousand-year window) or have had minimal effect on atmospheric Pco2 if it occurred earlier.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1


References and Notes

  1. Detailed methodology and all original analytical data are available as supporting material on Science Online.
  2. We are grateful to J. Quade and P. E. Olsen for their numerous helpful discussions and L. Godfrey for her technical lab assistance. This research was supported by 2009 Geological Society of America and the Society of Economic Mineralogists and Petrologists Graduate Research Awards and NSF grant EAR 0958867. This work is Lamont-Doherty Earth Observatory Contribution #7432.
View Abstract

Stay Connected to Science

Navigate This Article