Reply to Pearson and Nicholas, Stassen et al., and Zeebe et al.: Teasing out the missing piece of the PETM puzzle

Understanding the Paleocene-Eocene Thermal Maximum (PETM) critically depends on knowing the rate at which the perturbation carbon was released. In our report (1) we argue that the layered Marlboro Clay may provide this important constraint. Our strongest evidence in support of the rapid release of carbon at the onset of the PETM is the differential response of the %CaCO₃ and δ¹³C in the Millville core. The sharp %CaCO₃ decrease occurred over 4 mm, compared with the δ¹³C decrease over an interval of 25 cm (figure 3 in ref. 1). Temporal differences are predicted by a rapid (instantaneous) release of light carbon, which would lower the surface ocean in a matter of months, in contrast to the carbon isotopic equilibrium exchange, which occurs on the scale of a decade (2) and can only be recorded in a core with a high sedimentation rates (1). Based on the rhythmic bedding in the Marlboro Clay, we argue that the drop in %CaCO₃ occurred in less than a year and the δ¹³C equilibrium was on the order of a decade.

Much Higher Resolution Models Are Needed to Simulate an Instantaneous Release of CO₂

Zeebe et al. (3) use results from two carbon-cycle models showing that a large instantaneous release of “light” carbon should produce a −20 to −22‰ δ¹³C excursion in surface water, compared with −3.5‰ recorded at Millville, thereby attempting to invalidate the rapid release hypothesis for the PETM (1). Zeebe (4) states that LOSCAR (Long-term Ocean-atmosphere-Sediment CArbon cycle Reservoir model) “is not designed to address carbon cycle problems on time scales much shorter than centuries,” treating the surface ocean as one box. This aspect is critical because the temporal/spatial scales of surface ocean carbon invasion from an instantaneous release vary significantly, and much of the excursion dissipates over decades to a century. Here, we show that this discrepancy results largely from a false equivalency between the global open ocean mixed-layer and shelf localities, and that shelf results are more appropriately interpreted in the context of a rapid release.

Zeebe et al. (3) assert that the Millville δ¹⁸O record is inconsistent with a rapid release because the large thermal inertia of the ocean prevents a decadal temperature response. However, results from general circulation models show both fast and equilibrium surface temperature responses to instantaneous CO₂ perturbations (5). The fast temperature response occurs within a decade (5), consistent with the timescale of our report (1). Contrary to Zeebe et al.’s (3) assertion, shelf localities—such as the Salisbury embayment—respond rapidly to this instantaneous forcing because the heat capacity of the shelves is miniscule compared with the open ocean. The modeled temperature responses of Zeebe et al. are clearly the global long-term equilibrium, the time-scale and magnitude of which cannot be applied to shallow shelf localities.

An instantaneous, transient carbon perturbation produces nonuniform and spatially heterogeneous carbon isotope changes in the surface ocean (2, 6). Thus, any parameterization of the “surface mixed-layer” as a single box (e.g., LOSCAR) is inappropriate for simulating these conditions. This nonuniform response is expressed clearly in bomb radiocarbon records, which vary in magnitude and duration as a function of deep-water influence (2). Similar heterogeneity is preserved in the PETM, as evidenced by the differential response of shelf localities to an atmospheric perturbation (e.g., ∼−8‰ δ¹³C decrease at Medford vs. −3.5‰ decrease at Bass River). Furthermore, a δ¹³C excursion approaching −20‰ in the southern Salisbury Embayment was recently reported (7). LOSCAR has not yet replicated these results.

In addition to the differential response of %CaCO₃ and δ¹³C on the shelves (1), single specimen planktonic foraminiferal analyses at Maud Rise (8) record an abrupt δ¹³C decrease with no intermediate values, implying that the surface ocean δ¹³C shift occurred within the lifetime of a single foraminifera. Both the shelf and single specimen planktonic foraminiferal δ¹³C observations are easily explained by a rapid perturbation but are more difficult reconcile with a longer (5 kyr) release. The modeling efforts of Zeebe et al. (3) offer no explanation for these key observations.

Description of Rhythmic Bedding in the Marlboro Clay from Surface Exposures

Pearson and Nicholas (9) express concern that the observed layering reported in our paper (1) in the Millville core is because of drilling disturbance via the injection of drilling muds. This was an apprehension of ours as well, but as we noted (1), rhythmic bedding is preserved in outcrops of the Marlboro Clay. The presence of this sedimentary cyclicity in otherwise undisturbed outcrop (Fig. 1) precludes a drilling-induced origin for the layers, indicating that the rhythmic bedding is a primary depositional feature of the Marlboro. Here, we describe the nature of this bedding in natural exposures near Medford, NJ.

Fig. 1.

Photograph of the rhythmic bedding in the Marlboro Clay exposed in the Rancocas Creek, Medford, NJ. Pencil is ∼15 cm. The inset map shows the location of the exposures relative to the other locations used by Wright and Schaller (1). The blue/gray clay is interrupted at regular (∼2 cm) intervals by very thinly bedded silts and very fine sands. These areas also provide zones of weakness along which fractures will form when hand samples from the exposure are dried in the laboratory.

The rhythmic bedding observed in exposures of the Marlboro Clay (“Facies 2” of ref. 10) is characterized by a blue/gray kaolinitic clayey silt, interrupted at ∼2-cm intervals by thin <1-mm layers of sandy silts. These thin-bedded silts are occasionally accompanied by an increase in smectite (11), although it is unclear if this is primary or diagenetic. The regularity of the bedding is locally persistent at the outcrop scale, although the swelling nature of the clays obfuscates the delicate silt layers unless examined in detail (Fig. 1). No such bedding has been reported from the shallower Facies 1 of the Marlboro in the southern embayment (10).

We are aware of the potential for mud injection into the formation/core barrel while drilling clays and mudstones, and we readily acknowledge that this may have occurred in the Millville core because of overpressure. However, injection and “biscuiting” during coring generally follow preexisting zones of weakness, here provided by the rhythmic sandy-silt beds observed in outcrop. We also note that the Marlboro recovered in Wilson Lake B shows the same rhythmic bedding, but there was no overpressure of the drill string while coring this site.

Assembling the Marlboro Clay Puzzle

Stassen et al. (12) question our interpretation of the rhythmic bedding as seasonal, pointing to phenomena such as storms/hurricanes, freshwater discharge, and processes similar to the Atlantic Multidecadal Oscillation. We agree that these are possibilities, but note that they have durations ranging from weeks to a few decades. None of Stassen et al.’s suggested mechanisms provide the hundreds or thousands of years required for the duration of the section to be consistent with published age estimates for the PETM onset (1–20 kyr).

Stassen et al. (12) note that foraminifera are common in the Marlboro Clay, offering this as evidence of nonturbid waters in a deep-shelf setting. Foraminiferal counts (>100 µm) from Wilson Lake B core depth ∼110 m (Fig. 2), show ∼1,000 × 10³ individuals per square meter per layer, which is within the range of modern annual planktonic foraminiferal fluxes along the mid-Atlantic shelves: 750–1,500 × 10³ individuals per square meter per year (13). Moreover, foraminiferal fluxes near large rivers vary from 200 × 10³ (Amazon) to 2,000 × 10³ individuals per square meter per year (South China Sea) (13), contra Stassen et al.’s (12) claim. We observe systematic variability in foraminiferal abundance within the layers, an expectation if they are indeed seasonally forced (Fig. 2).

Fig. 2.

Total number of foraminifera (>100 µm) per gram of sediment across two couplets from the Wilson Lake core. Core was sampled at 2-mm intervals starting at 109.965 m (360.78–360.88 ft). Note that within the layers there are peaks in foraminifera (Forams) and quartz (Qtz) grains that are independent of each other, ruling out winnowing as a cause for these peaks.

The argument of Stassen et al. (12) that Millville-PETM water depths were closer to 120–150 m is based on an idealized ecological model of normal marine conditions on the shelf at steady state. An enhanced hydrologic cycle brought on by extreme warming also results in much higher sediment (10) and nutrients fluxes to the shelves, and alters the carbonate chemistry. The combination of these effects nullifies the use of such paleo-slope ecologic models to interpret water depths during the PETM. The expectation that ecological zonation be maintained under such extreme environmental perturbation is unreasonable. Furthermore, depth estimates just before the PETM are around 50 m (14); therefore, the PETM estimates of 120–150 m favored by Stassen et al. (12) require a depth increase of >70 m. The source of this volume of water (equivalent to the expected rise if all of the modern ice were melted) in a relatively ice-free world is indeed a far greater puzzle.

Stassen et al. (12) claim that the trend toward larger δ¹³C excursions with decreasing water depth is observed only in bulk carbonate records, comparing them to a benthic foraminiferal compilation. Coccolithophorids typically comprise most of the bulk carbonate and are more relevant than the δ¹³C excursion recorded by benthic foraminifera (see figure 1B of ref 12) because they sample the mixed layer, while benthics are bathed by thermocline waters (Fig. 3) (15). In our report (1), we note that it is the relative contribution of thermocline water (free of perturbation carbon) that buffers each site from the light carbon, and as such we would expect little to no cross-shelf δ¹³C gradient in the benthic foraminifera from Stassen’s deeper-water sites.

Fig. 3.

(A) Temperature profile across the mid Atlantic shelf from ref. 15 for conditions in November. Positions for the New Jersey Core locations based on estimated water depths are placed at the top of the panel. Note that bottom waters deeper than 40-m water-depth record similar temperatures and originate as open ocean thermocline waters. (B) Temperature profiles from April and August. Bottom water temperatures show little interannual variability and are isolated from the shelf surface waters. Benthic organisms on the shelf at depths of 40 m and deeper are not be expected to record surface water conditions and potentially have similar geochemical signatures.

The argument of our paper (1) for a rapid carbon release does not critically rest on the interpretation of the sedimentary layering as having approximately annual periodicity. The much more fundamental geochemical evidence for rapid acidification of shelf waters, compared with the much more protracted carbon isotope equilibration, is exactly what is predicted in response to an instantaneous atmospheric perturbation. None of the comments in the letters of these authors (3, 9, 12) have attempted to explain or even acknowledge these very basic observations.