Thermogenic methane release as a cause for the long duration of the PETM

Significance

The Paleocene–Eocene Thermal Maximum (PETM) was a period of global warming associated with rapid massive ¹³C-depleted carbon input, often mentioned as a paleoanalog for future climate change and associated feedbacks. One hypothesized carbon source is intrusive volcanism in the North Atlantic region, but rigid dating lacks. We date thermogenic methane release from a hydrothermal vent and find that it postdates the onset of the PETM but correlates to a period of additional carbon injection within the PETM. This study provides evidence of carbon release during the PETM from a reservoir (i.e., organic matter in sedimentary rocks) and implies that carbon release from the vent systems should be included in all future considerations regarding PETM carbon cycling.

Abstract

The Paleocene–Eocene Thermal Maximum (PETM) (∼56 Ma) was a ∼170,000-y (∼170-kyr) period of global warming associated with rapid and massive injections of ¹³C-depleted carbon into the ocean–atmosphere system, reflected in sedimentary components as a negative carbon isotope excursion (CIE). Carbon cycle modeling has indicated that the shape and magnitude of this CIE are generally explained by a large and rapid initial pulse, followed by ∼50 kyr of ¹³C-depleted carbon injection. Suggested sources include submarine methane hydrates, terrigenous organic matter, and thermogenic methane and CO₂ from hydrothermal vent complexes. Here, we test for the contribution of carbon release associated with volcanic intrusions in the North Atlantic Igneous Province. We use dinoflagellate cyst and stable carbon isotope stratigraphy to date the active phase of a hydrothermal vent system and find it to postdate massive carbon release at the onset of the PETM. Crucially, however, it correlates to the period within the PETM of longer-term ¹³C-depleted carbon release. This finding represents actual proof of PETM carbon release from a particular reservoir. Based on carbon cycle box model [i.e., Long-Term Ocean–Atmosphere–Sediment Carbon Cycle Reservoir (LOSCAR) model] experiments, we show that 4–12 pulses of carbon input from vent systems over 60 kyr with a total mass of 1,500 Pg of C, consistent with the vent literature, match the shape of the CIE and pattern of deep ocean carbonate dissolution as recorded in sediment records. We therefore conclude that CH₄ from the Norwegian Sea vent complexes was likely the main source of carbon during the PETM, following its dramatic onset.

The Paleocene–Eocene Thermal Maximum (PETM) (56 Ma) was a period of rapid global warming (1) associated with massive injections of ¹³C-depleted carbon into the global exogenic carbon pool and extensive environmental upheaval, including ocean acidification (2), global expansion of oxygen minimum zones, local photic zone euxinia, sea level rise (3), species migrations (4), and an accelerated hydrological cycle (5). The carbon injection is recognized as a negative carbon isotope excursion (CIE) averaging 3–4‰ in marine sedimentary components (6). The CIE associated with the PETM as recorded in sedimentary records typically has a rapid onset, likely in the order of millennia, followed by 70,000–100,000 y (70–100 kyr) of stable values, referred to as the “body” of the CIE, and a recovery phase (50–100 kyr) (7–10). This shape, in particular the body, distinguishes it from other early Eocene transient carbon cycle perturbations (11). It is generally explained by rapid and massive additions of ¹³C-depleted carbon at the onset (12, 13), slow continuous release across the body and subsequent sequestration (14, 15). Several mechanisms have been proposed to explain the CIE, either in combination or alone, including enhanced volcanism in the North Atlantic Igneous Province (NAIP) (16, 17), the dissociation of gas hydrates (18, 19) and organic matter oxidation (20), possibly from permafrost thawing (21). However, no field data show that carbon was released from any of these proposed source reservoirs during the PETM that might explain its onset and long duration.

A link between Paleocene–Eocene climate change and NAIP was first proposed in the early 90’s (16) and discussion has subsequently focused on volcanic degassing impacting long-term climate evolution as well as triggering the PETM (Fig. S1). Storey et al. (17) provide estimates of magma production rates across the late Paleocene and Eocene and show that sufficient masses of CO₂ were generated to affect the global carbon cycle. Because magmatic CO₂ is relatively ¹³C-enriched (about −5‰), it represents an improbable cause for a CIE in the global exogenic carbon cycle (22).

Fig. S1.

North Atlantic Volcanism and long-term Paleocene–Eocene climate. (A) Oxygen and carbon isotope compilation from Cramer et al. (75) showing long-term warming from 60 to 51 Ma. (B) Significant climatic and biological events. ELPE, Early Late Paleocene Event; ETM2, Eocene Thermal Maximum 2. (C) Volcanic events and approximate magma production from Storey et al. (17).

However, two pathways have been proposed linking this period of intense volcanism in the NAIP directly to release of substantial masses of ¹³C-depleted carbon during the PETM. First, Svensen et al. (23) proposed the release of ∼300–3,000 Pg of carbon - the latest conservative estimate is 1,100 Pg (24) - in the shape of thermogenic methane (CH₄) from the Norwegian Sea as a possible trigger and carbon source for the PETM. The validity of this hypothesis strongly depends on the timing of sill-intrusions in the Vøring and Møre basins. These sills are extensive (2 × 10⁵ km²) and partly emplaced in organic rich rocks (25) and roughly coincide with the PETM based on radiometric dating (26, 27). Numerous (>700) hydrothermal vents (Fig. 1) are directly linked to the sill intrusions and provide evidence for large-scale degassing (Figs. S2 and S3). Furthermore, 95% of the vents terminate at a seismic reflector regionally interpreted to represent the top of the Paleocene (25). Apectodinium augustum, a dinoflagellate cyst marker species that is diagnostic for the PETM CIE in the North Atlantic (28–30), was found in the only drilled hydrothermal vent complex (Fig. 2).

Fig. 1.

Map showing the location of wildcat 6607/12-1 (red rectangle) and other vent complexes (black dots and white triangles) and volcanics in the Vøring and Møre basins, redrawn from Svensen et al. (23).

Fig. S2.

Seismic data around the studied vent system. (A) Seismic profile across the hydrothermal vent system penetrated by 6607/12-1; green lines indicate sill intrusions, and the yellow line represents the top Paleocene reflector. Two-way travel time (twt) is indicated. (B) Three-dimensional visualization of the hydrothermal vent system, with pronounced crater and overlying eye structure.

Fig. S3.

Conceptual image of carbon release through hydrothermal venting.

Fig. 2.

Schematic view of vent structures. Ages of important regional sedimentary formations are indicated on the left and maturity of the organic matter on the right. Note that the base of the vent complex is at 1,730 mbss, ∼30 m lower than the average depth of the Top Paleocene horizon outside the vent complex. TD, total depth.

Second, Rampino (31) suggested that the melting of organic rich sediments may have generated even more CH₄ and CO₂ than was calculated by Svensen et al. (23) and inferred that 3,000–6,000 Pg of carbon was released based on volumetric calculations of igneous deposits and intrusions across the North Atlantic. Rampino (31) correlated between deposits by radiometric dating and identified the PETM using dinoflagellate cyst biostratigraphy, most notably the presence of A. augustum in sediments interbedded in a basalt sequence at Ocean Drilling Program Hole 642E (32). However, recent chemostratigraphic analyses excluded the presence of the CIE at Hole 642E, implying that the specimens of A. augustum are reworked into early Eocene sediments (33).

To test for a causal link between sill emplacement, generation of thermogenic methane and the PETM, we analyzed samples recovered in the wildcat 6607/12-1 borehole, drilled in 1986 at 390-m water depth reaching 3,521 m below sea surface (mbss) in the central part of a hydrothermal vent complex in the Vøring Basin (34) (Fig. 2). Based on seismic data, the vent complex is characterized by a 2-km wide-eye-shaped upper part, representing the crater and mound, at the top of the Paleocene series, overlying a zone of disrupted sediments, interpreted to reflect a chimney structure (23) (Figs. S2 and S3). The upper part of the chimney has low to intermediate organic maturity and the lower part high maturity, based on vitrinite reflectance (Fig. 2). The chimney connects the upper part of the vent complex to the termination of a high-amplitude seismic event at 5.0 s two way travel time, regionally interpreted as a sill intrusion (Fig. S2). The strata above the vent complex are domed as a result of differential compaction postdating the vent formation.

We performed detailed analysis of palynology, stable carbon isotope ratios of palynological residue (δ¹³C_paly), Rock-Eval, and vitrinite reflectance on 22 cutting samples from the chimney and eye structure (1,640–1,745 mbss) at 6607/12-1.

Dinoflagellate cyst and pollen assemblages in borehole 6607/12-1 are typical for the early Eocene of the Nordic Seas (SI Text, Fig. S4). We correlate our dinocyst biostratigraphy to a regional dinoflagellate cyst zonation (35), which confirms an earliest Eocene age for all studied samples (Fig. 3C and SI Text). We also record the presence and abundance of A. augustum. Crucially, δ¹³C_paly values are extremely low at −31‰ in between 1,710 and 1,745 mbss, during the Apectodinium acme, and subsequently rise to −27‰ at 1,660 mbss, which is consistent with the presence of the PETM CIE. However, changes in organic matter sources, here represented by the fractions of marine and terrestrial palynomorphs, could result in variations in δ¹³C_paly values (36, 37). Indeed, palynological assemblages are dominated by pollen grains, with variable marine contributions (5–30%; Fig. 3B). We correct for this potential bias using end-member modeling of terrestrial and marine organic matter δ¹³C and the relative abundance of terrestrial and marine contributions to the palynological assemblages (38) (SI Text). The correction leads to a synthetic δ¹³C record of terrestrial palynomorphs (δ¹³C_pollen, Fig. 3 A and B). The trends in this record follow those of the δ¹³C_paly record, although the thickness of the body of the PETM CIE is reduced. Collectively, biostratigraphy and carbon isotope stratigraphy imply the presence of the body of the PETM CIE, down to at least 1,745 mbss, within the disrupted sedimentary material in the chimney, continuing within the undisturbed sediments above the base of the vent, up to ∼1,715 mbss (Fig. 3A).

Fig. S4.

(A) δ¹³C records, color coded as in Fig. 3A: chimney (gray), body of the CIE (red), recovery (orange), and Early Eocene (blue). (B) Fraction of marine palynomorphs calculated as marine/marine + terrestrial [M/(T+M)]. (C) Dinocyst zonation based on Bujak and Mudge [B&M (35)] and stratigraphically important dinocyst events. (D) Dinoflagellate cyst assemblages (relative abundances in percentages). (E) Pollen assemblages (relative abundances in percentages).

Fig. 3.

Stratigraphy of the 6607/12-1 borehole. (A) Isotope records of isolated palynological residue (δ¹³C _paly) and pollen (δ¹³C _pollen). (B) Fraction of marine palynomorphs calculated as marine/marine plus terrestrial [M/(T+M)]. (C) Dinocyst zonation is based on Bujak and Mudge (B&M) (35) and stratigraphically important dinocyst events.

The new stratigraphic constraints indicate that both A. augustum and the CIE are prevalent from below the base of the vent system at 1,730 mbss. Moreover, the previously recorded thermally mature specimens of A. augustum (23) suggest that the vent system blew through PETM sediments. In addition, a simple sedimentation rate model supports the inference that vent activity at this site was limited to a short period during the body of the CIE (SI Text). Collectively, all available evidence indicates that this vent blew well within the body of the CIE rather than at its onset.

To test whether pulsed release of thermogenic CH₄ could have produced the prolonged period of stable δ¹³C values in the sedimentary record (the body of the CIE), we conducted simulations with the Long-Term Ocean–Atmosphere–Sediment Carbon Cycle Reservoir (LOSCAR) model (39). This model was used by Zeebe et al. (14) to explore carbon injection scenarios for the onset and body of the PETM.

To constrain the model simulations, assumptions regarding the δ¹³C of methane and emission scenarios must be made. The sills were emplaced in a few phases (4−12) (23), as high volume injections during a relatively short period (25). This scenario is consistent with seismic data and interpretations from other Large Igneous Provinces, showing that (i) vents are formed from the contact aureoles of sills (40) and (ii) single sills may represent injection of >3,000 km³ of melt, suggesting that even large volumes of melt in a sedimentary basin may have been derived from a few emplacement events or “pulses.” The gas generated from heating of marine organic matter in the contact aureoles is dominated by CH₄ at high temperatures (41). The range of published values for natural thermogenic CH₄ is −30 to −65‰ (42). In case of near-complete conversion to CH₄ and CO₂, it is likely that the δ¹³C of released carbon approaches that of sedimentary organic carbon (−25 to −35‰).

For our simulations, we force the onset of the 3–4‰ CIE identically to the scenario of Zeebe et al. (14), which includes an initial carbon injection of 3,000 Pg with a δ¹³C of −50‰ over 5,000 y and a circulation change to reproduce the recorded patterns in calcite compensation depth (CCD) change (SI Text). For the body of the CIE, after the initial carbon injection, we test scenarios with different δ¹³C of thermogenic CH₄ (−30 and −45‰), variable carbon input (300, 1,500, and 3,000 Pg), number of pulses (4, 8, 12), proportion of C released directly into the atmosphere, and additional CH₄ bleeding from hydrates (SI Text, Fig. 4, Figs. S5–S8, and Table S1).

Fig. 4.

LOSCAR model output of pulsed carbon release. (A) Emission scenario showing the release of carbon with eight distinct pulses of 187.5 Pg over 1 kyr each following an initial release of 3,000 Pg over 5 kyr. Selected background conditions were changed during the PETM body and are highlighted in blue (SI Text). (B) Response of CCD in different ocean boxes. (C) Response of δ¹³C of dissolved inorganic carbon in different ocean boxes. (D) Response of atmospheric CO₂ concentrations in parts per million and pH of different ocean boxes.

Fig. S5.

LOSCAR model output for carbon release with variable δ¹³C (−30 and −45‰) during the body of the CIE, showing CCD, pCO₂ and pH, and δ¹³C change in different ocean boxes. (A) Emission scenario showing the release of 1,500 Pg of carbon with a δ¹³C of −30‰. (B) Release of 2,250 Pg of carbon and δ¹³C of −30‰. (C) Release of 1,500 Pg of carbon and δ¹³C of −45‰ (identical to Fig. 4).

Fig. S8.

LOSCAR model output of other alternative carbon release during the body of the CIE showing CCD, pCO₂ and pH, and δ¹³C change in different ocean boxes. (A) Emission scenarios showing the pulsed release of carbon consisting of eight distinct pulses of 187.5 Pg (60% into the atmosphere, compared with 90% in Fig. 4). (B) Emission scenario showing pulsed release of carbon (1,000 Pg/−45‰) but with continuous background carbon “leakage” (500 Pg/−50‰). (C) Emission scenario showing continuous release of 1,500 Pg/−50‰ across 60 kyr.

Table S1.

LOSCAR model scenarios for the body of the PETM CIE

Scenario no.	Onset CIE					Body CIE										Change relative to main scenario	Figure
						Pulses						Bleed
	C, Pg	kyr	δ13C, ‰	% atm	% Deep Atlantic	n pulses	C, Pg	kyr/pulse	δ13C, ‰	% atm	% Deep Atlantic	C, Pg	δ13C, ‰	% atm	% Deep Atlantic
Main	3,000	5	−50	60	40	8	1,500	1	−45	90	10	None				None	Fig. 4
1	3,000	5	−50	60	40	8	1,500	1	−30	90	10	None				δ13C −30‰	Fig. S5A
2	3,000	5	−50	60	40	8	2,250	1	−30	90	10	None				δ13C −30‰ and 2,250 Pg input	Fig. S5B
3	3,000	5	−50	60	40	4	1,500	1	−45	90	10	None				4 Pulses instead of 8	Fig. S6A
4	3,000	5	−50	60	40	12	1,500	1	−45	90	10	None				12 Pulses instead of 8	Fig. S6C
5	3,000	5	−50	60	40	8	300	1	−45	90	10	None				300 Pg of C input during body	Fig. S7A
6	3,000	5	−50	60	40	8	3,000	1	−45	90	10	None				3,000 Pg of C input during body	Fig. S7C
7	3,000	5	−50	60	40	8	1,500	1	−45	60	40	None				60% Instead of 90% of pulsed C into atmosphere	Fig. S8A
8	3,000	5	−50	60	40	8	1,000	1	−45	90	10	500	−50	60	40	8 Pulses (1,000 Pg) plus bleed (500 Pg)	Fig. S8B
9	3,000	5	−50	60	40	None						1,500	−50	60	40	No pulses	Fig. S8C

As expected, our results regarding changes in CCD and overall δ¹³C trends are similar to the scenario explored by Zeebe et al. (14), as we use the same background changes and initial massive carbon release (SI Text). Our scenarios, hence, do not improve the fit between modeled and proxy-based magnitude of initial pH and δ¹³C excursions, compared with other studies (43), but here we focus solely on the effect of pulsed carbon input during the body of the PETM.

The forcing with pulsed carbon input results in a stable plateau of δ¹³C values of marine dissolved inorganic carbon in all ocean boxes, with superimposed short-lived, distinct spikes (Fig. 4). Ocean-mixing time dampens these spikes in the deep basins, and, even in most shallow marine sites, bioturbation is expected to remove most millennial scale δ¹³C fluctuations from isotope records (13). Intriguingly, some high-resolution terrestrial (44) and laminated marine sections (45) record high-frequency variability during the body of the CIE, which may be consistent with our scenarios.

From our different scenarios for the body of the CIE, we find that the highest (3,000 Pg) and lowest (300 Pg) carbon releases from the vent systems cannot reproduce the body of the CIE, given the ¹³C value of the released carbon in these scenarios (−45‰). The 4, 8, and 12 pulse scenarios are qualitatively similar and results are insensitive to changes in the proportion of carbon injected to the atmosphere directly versus into the ocean (SI Text). We also find that carbon input in ≥12 pulses (Fig. S7) produces results that are practically indistinguishable from those obtained from continuous input scenarios. We find that 2,250 Pg of C is required to produce the same δ¹³C trend if we assume a δ¹³C_CH4 closer to that of sedimentary organic matter (−30‰), and this scenario also properly simulates CCD patterns (Fig. S5 and SI Text). Based on these explored scenarios, we conclude that the release of 1,500–2,250 Pg of thermogenic CH₄ with an isotopic signature of −30 to −45‰ in greater than four pulses is a plausible explanation for stable δ¹³C values during the body of the CIE.

Fig. S7.

LOSCAR model output of variable carbon release during the body of the CIE (300, 1,500, and 3,000 Pg) in eight pulses, showing CCD, pCO₂ and pH, and δ¹³C change in different ocean boxes. (A) Emission scenario showing the release of 300 Pg of carbon. (B) Same for 1,500 Pg (see also Fig. 4). (C) Same for 3,000 Pg.

Zeebe (15) suggested that carbon emissions forcing the body of the CIE came from slow injections of biogenic methane from submarine hydrates, representing a positive feedback to catastrophic carbon release at the onset of the event. Although not mutually exclusive from our scenario (SI Text and Fig. S8), we stress that this hypothesis is purely theoretical, whereas our scenario is supported by the data presented here. Moreover, the hydrate scenario requires much of the C input to take place near the beginning of hydrate dissociation and negligible masses of C are released after ∼40 kyr into the CIE (15), which only represents about half of the duration of the body. Although carbon cycle feedbacks to warming are expected during the PETM (15, 19), carbon input from hydrothermal vents presently provides a more complete explanation for the body of the CIE.

Could the activity of hydrothermal vent complexes have caused precursor events (46) and the onset of the CIE, in addition to the body? Activity in some of the hydrothermal vent complexes in the Vøring and Møre basin or elsewhere in the NAIP close to the onset of the PETM cannot be excluded. However, we note that the amount of carbon needed to force the entire onset of the CIE (>2,500 Pg) is close to the upper estimate for the hydrothermal vents in total (19) (3,000 Pg). This scenario also implies that all this carbon was released within millennia (12, 13), which seems improbable.

Collectively, we conclude that carbon release from this particular vent system cannot have triggered the PETM. Most importantly, we conclude that the pulsed release of thermogenic methane from these vent complexes is a plausible explanation for long duration of the body of the CIE, as well as environmental effects such as warming and ocean acidification during the PETM.

Sampling and Methods

Drilling chips from borehole 6607/12-1 were originally collected and grouped from 5-m intervals during drilling. For this study, we focus on 22 samples from the 1,640- to 1,770-m interval, from the Maastrichtian–Paleocene unconformity to the top of the eye structure. We carefully subsampled washed chips and hand-picked them for bulk rock total organic carbon (TOC), Rock-Eval, and vitrinite reflectivity measurements at the Norwegian Petroleum Directorate in Stavanger, Norway.

Rock-Eval and Vitrinite Reflectance.

TOC contents, Rock-Eval, and vitrinite reflectance (%Ro) analyses were carried out on powdered samples using a Rock-Eval 6 instrument at Applied Petroleum Technology (APT), Kjeller, Norway. Rock-Eval pyrolysis is used to identify the type and maturity of organic matter. The TOC and Rock-Eval analyses are performed at temperatures between 300 °C and 850 °C over 25 min. The vitrinite reflectance is a widely used parameter to define the thermal maturity of organic matter in shales and coals. The vitrinite reflectance was determined at APT from polished slabs analyzed on a Zeiss Standard Universal research microscope photometer (MPM01K) equipped with a tungsten–halogen lamp (12 V; 100 W) and an Epiplan-Neofluar 40/0.90 oil objective. Quality ratings reported in Dataset S1 are based on various important aspects which may affect the measurements, such as the abundance of vitrinite, uncertainties in the identification of indigenous vitrinite, type of vitrinite, particle size, and particle surface quality.

Palynology.

Of each sample, 5–15 g was oven-dried and processed for palynology, using standard procedures (47), including the addition of one Lycopodium spore tablet (n = 18,583 ± 764) (48) for absolute quantitative analyses and treatment with HCl and HF. Afterward, samples were sieved with water over a 250-μm and a 15-μm sieve to remove large and small particles, respectively. Residues were concentrated in glycerine water and mounted on microscope slides using glycerine jelly. Slides were analyzed at 400× magnification to a minimum of 200 dinocysts, where possible. We follow dinocyst taxonomy of Fensome et al. (49). All material is stored in the collection of the Laboratory of Paleobotany and Palynology, Utrecht University.

Stable Carbon Isotope Analysis of Palynological Residues.

Palynological residues from the 6607/12-1 bore hole were used for stable carbon isotope analyses. Splits of the residue were again washed with distilled water to remove glycerin. Samples were dried in a stove at 50 °C and subsequently TOC content was measured on ∼1 mg of homogenized residue using an elemental analyzer (Fisons). Stable carbon isotope ratios were determined on 15–30 μg of residue using an isotope ratio mass spectrometer (Finnigan Mat Delta Plus) coupled online to the elemental analyzer. We correct our δ¹³C_paly for variable marine influences to obtain δ¹³C_pollen using the equation of Sluijs and Dickens (38). Absolute reproducibility, based on international and in-house standards, for TOC and δ¹³C_paly is better than 0.1% and 0.05‰, respectively.

LOSCAR Modeling.

A detailed description of the LOSCAR model is provided by Zeebe (39). Essentially, this box model is modified from Walker and Kasting (50) and simulates the cycling of carbon through atmospheric and oceanic reservoirs of which the latter are coupled to a sediment module. Concentrations of several biogeochemical tracers are calculated for each box, including dissolved inorganic carbon, total alkalinity, δ¹³C, oxygen, and phosphate. The Paleogene model ocean (39) consists of four main ocean basins (Atlantic, Indian, Pacific, and Tethys), which are in turn separated into surface (0- to 100-m water depth), intermediate (100–1,000 m), and deep (>1,000 m) boxes. The surface ocean boxes are in contact with one atmospheric box. Thermohaline circulation and ocean mixing are prescribed. For our simulations, we use default parameter settings (39) and alter selected background conditions during the PETM identically to Zeebe et al. (14). Simultaneous with our initial carbon release the following background changes are applied. (i) Southern Ocean deep-water formation is decreased, complemented by increased formation of North Pacific deep water (51). (ii) The locus of CaCO₃ deposition is shifted from the deep ocean to the continental shelf, consistent with records of PETM sea level rise (52) and CaCO₃ accumulation (53). These first two assumptions greatly improve the fit between simulated and recorded changes in Atlantic and Pacific CCD (54, 55). (iii) In addition, a PETM whole ocean temperature change of +4 °C was prescribed, as currently accepted values for climate sensitivity (1.5–4.5 °C per doubling of pCO₂) would result in underestimated temperature change compared with the records (14).

SI Text

North Atlantic Paleogene Volcanism.

A short period (<1 My) of voluminous breakup volcanism at about 56 Ma has been linked to the PETM (17). Seismic interpretations show that the breakup volcanism began with a period of intrusive activity in the Vøring and Møre basins, forming at least 700 (Fig. 1), but most likely thousands of hydrothermal vent complexes terminating at the reflector that represents the top Paleocene horizon (23, 25). It is well documented that sill intrusions alter the light element geochemistry of the intruded sediments (56, 57), forming methane and carbon dioxide followed by overpressure generation and in some cases explosive venting (40, 58). The contact metamorphic volume of sedimentary rocks is usually twice the sill volume, which can be used to calculate the mass of carbon generated (59). The estimates of carbon release during the main period of intrusive volcanism in the Vøring and Møre basins is 300–3,000 Pg (23, 56), and the latest conservative estimate is 1,100 Pg (24). Large uncertainties remain concerning the amount of released carbon due to the extrapolation of data over a large (200,000 km²) area and the assumption of synchronicity of 95% of these vent systems based on seismic profiles (25).

Dinocyst Biostratigraphy of Borehole 6607/12-1.

Following the North Sea dinocyst zonation of Bujak and Mudge (35) (Fig. 3C), the last occurrence of A. augustum (1,695 m) defines the start of earliest Eocene biozone E1a. Our findings are consistent with this zonation; sensu scricto, this zone is characterized by last occurrences of Phelodinium magnificum (1,700 m), Cerodinium speciosum and Cerodinium dartmoorium (1,670 m), and the first occurrence of common to abundant Hystrichosphaeridium tubiferum (here 1,685–1,640 m), which we also record near the top of A. augustum. The last occurrence of C. speciosum is not recorded, suggesting this subzone extends further up section. A single specimen of Wetzeliella meckelfeldensis at 1,730 m indicates some caving has occurred. However, the dinocyst assemblages show continuous events characteristic of the early Eocene North Sea, and it is hence concluded that caving is a minor factor.

Stable Carbon Isotope Records and Corrections.

With the present dataset, we cannot correct for varying contributions of angiosperm and gymnosperm pollen (60), carbon dioxide effects on ¹³C fractionation of marine (36) and terrestrial organisms (61), or changes in trophic state of the ecosystem. However, to correct for the influence of varying amounts of terrestrial and marine organic matter on the stable carbon isotope ratios of palynological residue (δ¹³C_paly), we use the total pollen versus total marine counts (TP%) as a proxy for the proportion of terrestrial organic matter the composition of the palynological residue. The difference in carbon isotopic signature between terrestrial and marine organic matter is relatively well constrained for the Early Paleogene at ∼4.4‰ (38, 62, 63) (Fig. 3).

Both the δ¹³C_pollen and δ¹³C_paly records show the body of the CIE, subsequent recovery, and stable lower Eocene values. The body in the δ¹³C_pollen record is substantially shorter than in the δ¹³C_paly record. Furthermore, sediments below 1,730 m are interpreted to be part of the chimney structure (Fig. 2) and have been (re)deposited during vent activity; this interval is therefore not taken into account in calculations of sedimentation rates. The body of the PETM, hence, spans 1,730 to ∼1,715 mbss in the δ¹³C_pollen record.

Sedimentation Rates.

Based on δ¹³C_paly and dinoflagellate biostratigraphy, the PETM CIE is at least 80 m thick (Fig. 3), which, given a total duration of ∼170–230 kyr for the CIE (8, 9), amounts to average sedimentation rates of 35–47 cm kyr⁻¹. It should be noted, however, that sedimentation rates immediately following crater formation were possibly much higher because of crater fill.

We construct a simple sedimentation model, based on the thickness of the recovery (∼50 m in the δ¹³C_pollen record) to crudely assess when sedimentation in the crater resumed after the vent was active. The duration of the recovery is ∼85 kyr (8, 9), leading to sedimentation rates of ∼60 cm kyr⁻¹. If we assume the same sedimentation rates during the body of the CIE, which is 15 m thick, only ∼25 kyr of the body of the CIE is present in this record. Alternatively, if the entire body of the PETM CIE (∼100 kyr) were present in our section, it would imply sedimentation rates of 15 cm kyr⁻¹ increased by a factor 4 during the recovery phase. We consider this scenario unlikely and therefore conclude that the vent system blew well into the body of the CIE.

The presence of mixed immature and mature A. augustum specimens in the top of the chimney structure (23) supports this inference. Importantly, there are no other records of thermally altered A. augustum in other records from the Nordic seas, indicating that the signal is local. The most likely explanation for this observation is that the vent system penetrated sediments deposited during the early phase of the PETM, thereby thermally altering the organic matter, including A. augustum. Part of this material fell back into the chimney structure and was mixed with immature PETM material from just outside the crater and perhaps also with new sediments. All of these observations require PETM sediments to be present at the time of explosive venting. Collectively, the data imply that activity in this vent complex occurred during the body of the CIE.

LOSCAR Experimental Setup.

LOSCAR represents a relatively simple box model designed to resolve large-scale processes and perturbations in the carbon cycle (14, 39). We use the Zeebe et al. (14) scenario (Z09) as a starting point for our simulations and include our new constraints from detailed dinoflagellate biostratigraphy, δ¹³C records, sedimentation, and previous research on thermogenic methane generation around sills. In all of our experimental scenarios, the system is first perturbed by the input of 3,000 Pg of ¹³C-depleted (−50‰) carbon over 5 kyr. Following the Z09 scenario (14), we invoke changes in ocean circulation and a shift in the locus of carbonate deposition from the deep ocean to the shelves during the body of the CIE. Contrasting with Z09, we then impose a pulsed input of ¹³C-depleted carbon, representing discrete, short-lasting (1 kyr) episodes of methane venting in the Vøring and Møre basins. All scenarios are set up so that the first pulse always directly follows the initial 3,000 Pg perturbation and the first and last pulses are always 60 kyr apart.

We conduct a series of sensitivity experiments where we change the δ¹³C of the released C, locus of carbon input, the number of pulses, and the total mass of released carbon. We note that at present, the available data do not allow for detailed constraints on size, number, and timing of individual pulses. Here, we aim to test whether such pulses would stand out in sedimentary carbon isotope records and hence whether pulsed carbon input could represent a plausible carbon source for the body of the CIE.

Isotopic signature of released carbon.

During contact metamorphism of organic matter in sedimentary rocks, carbon is released dominantly as thermogenic methane, a process somewhat similar to pyrolysis. The evolution and average δ¹³C of thermogenic methane (δ¹³C_CH4) during progressive heating in contact aureoles is however poorly constrained. Pyrolysis experiments on marine kerogen have shown that the δ¹³C_CH4 becomes increasingly ¹³C-depleted during laboratory heating (−30 to −40‰), but this fractionation is balanced by less ¹³C-depleted cogenetic CO₂ (−25‰) (41, 64). However, in the inner zones closest to the sills, all organic matter is lost from the sedimentary rocks (56). The average generated C should then have an isotopic value that matches the δ¹³C of the parent sedimentary organic matter (−25 to −35‰). In the outer parts of contact aureoles, where the temperatures were lower, the δ¹³C_CH4 values are expected to qualitatively compare with those from the pyrolysis experiments.

Although a bulk value of −35‰ is commonly used as a value for thermogenic CH₄ in the literature (65), a very large natural range that particularly extends to much lower values (−30 to −65‰) is observed (42), indicating a possible mismatch between experimental results and field observations. For our experiments, we therefore test two values (−30 and −45‰) representing the experimental and a median natural range, respectively. Increasing carbon input mass by a factor 1.5 in the −30‰ scenario (2,250 Pg versus 1,500 Pg) synchronizes the ¹³C curves (Fig. S5 B and C). A δ¹³C value of −45‰ is used for thermogenic methane in all other model scenarios.

Locus of the carbon input.

In the original Z09 scenario, carbon from methane hydrates at the start of the PETM is released into the atmosphere (60%) and the deep Atlantic (40%). We keep this ratio constant in the initial 3,000 Pg perturbation, but 90% of the subsequent pulsed input of 1,500 Pg enters the atmosphere directly. We feel this approach is more realistic when considering violent releases of methane, as witnessed by the presence of numerous large craters in the Vøring and Møre basins. As a consequence, the scenario in which 40% of the carbon input enters the Atlantic directly shows more variability in the deep Atlantic CCD and δ¹³C. However, the observed differences are only noticeable on a millennial scale and hardly affect δ¹³C in the other ocean basins (compare Fig. 4 and Fig. S8) and sedimentary records.

Four, eight, or twelve pulses.

We conduct three experiments with 4, 8, and 12 individual pulses of 1 kyr each, as constrained by seismic profiles (23) and sill cooling time (56) (Fig. S6), while total carbon input is held constant at 1,500 Pg (cf. Z09). All of the pulses are represented by discrete, albeit small, δ¹³C excursions. The largest excursions are found in the Atlantic, the smallest in the Pacific. In the Atlantic, the amplitude of the δ¹³C changes ranges from 0.6‰ in the 4-pulse scenario to 0.2‰ in the 12-pulse scenario. However, these excursions are smoothed in the Pacific, resulting in output that is practically indistinguishable from the continuous input scenario (Z09) (compare Figs. S5 and S8C). Arguably, δ¹³C excursions of ∼0.6‰ would have been recorded in some of the high-resolution shallow marine PETM records. However, after deposition, mixing processes, such as bioturbation, usually smooth out the largest excursions (66). Indeed, high-resolution terrestrial (67) and marine PETM records without substantial bioturbation (45) typically show more variability in δ¹³C. In the 8- and 12-pulse scenarios, δ¹³C excursions are limited to <0.2 and <0.1‰ in the deep Atlantic and deep Pacific, respectively. Such small excursions would not be picked up in any sedimentary δ¹³C record.

Fig. S6.

LOSCAR model output for carbon release scenarios with variable number of pulses during the body of the CIE, showing CCD, pCO₂ and pH, and δ¹³C change in different ocean boxes. (A) Emission scenario showing the release of carbon in four distinct pulses. (B) Same for eight pulses (identical to Fig. 4). (C) Same for 12 pulses.

Uncertainties in the amount of released carbon: 300, 1,500, or 3,000 Pg.

The proposed 1,500 Pg of carbon as continuous input was tuned to the CCD data from different ocean basins (23, 56) and close to the average estimate of methane release from the Vøring and Møre basins (23, 56). However, large uncertainties remain in the latter due to extrapolation over a large area (∼200,000 km²), and the assumption that 95% of the craters is synchronous with the top-Paleocene reflector. Therefore, we explore what the effects are of releasing 300 Pg and 3,000 Pg, the minimum and maximum estimated mass of released thermogenic methane, respectively. A total release of 300 Pg is unable to produce the observed prolonged excursions in δ¹³C and pH (15, 43) (Fig. S7A). In contrast, 3,000 Pg would cause a continued decrease of δ¹³C (Fig. S7C), whereas most records show stable or slightly increasing δ¹³C values within the body of the PETM. Importantly, the most recent conservative estimate of the total carbon production from the vent systems is ∼1,100 Pg (24), which is close to our optimal scenario.

Dinocyst Assemblages and Paleoecology.

The dinoflagellate cyst assemblages indicate a rather uniform paleoecological signal in the 1,745- to 1,640-m interval. Notably, the assemblages represent a mixture of open ocean, oligotrophic, and more proximal eutrophic species (Fig. S8). Such mixtures of distal and proximal species are usually interpreted as transported signals (68), in line with high sediment supply at the site location. Senegalinium and related genera, usually associated with (seasonally) lower than normal marine salinities and eutrophic conditions (69–71), make up ∼25% of the assemblage on average, perhaps slightly more in the lower part (1,705–1,745 m). This could indicate the prevalence of lower salinities during the greenhouse warmth of the PETM, as was previously found in Spitsbergen (72) and the Arctic Ocean (30, 73).

We further find close to 30% Apectodinium spp. at 1,725–1,730 m, a (sub)tropical, euryhaline genus that is usually associated with warm and eutrophic waters (71, 74). Deflandrea and Cerodinium species substitute for Apectodinium from 1,700 m upward, arguably representing a similarly eutrophic, albeit somewhat colder, environment. More open marine Spiniferites account for ∼35% of the assemblage, with a maximum of 60% at 1,695 m. In combination with a maximum in marine over terrestrial palynomorphs at 1,705 m, the increased abundance of open marine species suggests a maximum flooding surface.

Pollen Assemblages.

We record the typical late Paleocene and early Eocene Taxodiaceaepollenites hiatus acme (35), derived from extensive coastal Taxodium swamps. Concomitant with abundant Apectodinium spp. are high percentages (∼40%) of Caryapollenites spp., possibly related to warmer conditions during this period. Other common pollen include Alnipollenites spp. and Momipites spp., which is closely related to Caryapollenites. Bisaccate pollen, indicative of cooler, drier conditions make up ∼10% of the pollen total higher up in the section (1,675–1,640 m). These findings are in general agreement with the observations by Eldrett et al. (28) in the North Sea.