Rapid Acidification of the Ocean During the Paleocene-Eocene Thermal Maximum

See allHide authors and affiliations

Science  10 Jun 2005:
Vol. 308, Issue 5728, pp. 1611-1615
DOI: 10.1126/science.1109004


The Paleocene-Eocene thermal maximum (PETM) has been attributed to the rapid release of ∼2000 × 109 metric tons of carbon in the form of methane. In theory, oxidation and ocean absorption of this carbon should have lowered deep-sea pH, thereby triggering a rapid (<10,000-year) shoaling of the calcite compensation depth (CCD), followed by gradual recovery. Here we present geochemical data from five new South Atlantic deep-sea sections that constrain the timing and extent of massive sea-floor carbonate dissolution coincident with the PETM. The sections, from between 2.7 and 4.8 kilometers water depth, are marked by a prominent clay layer, the character of which indicates that the CCD shoaled rapidly (<10,000 years) by more than 2 kilometers and recovered gradually (>100,000 years). These findings indicate that a large mass of carbon (»2000 × 109 metric tons of carbon) dissolved in the ocean at the Paleocene-Eocene boundary and that permanent sequestration of this carbon occurred through silicate weathering feedback.

During the Paleocene-Eocene thermal maximum (PETM), sea surface temperature (SST) rose by 5°C in the tropics and as much as 9°C at high latitudes (13), whereas bottom-water temperatures increased by 4° to 5°C (4). The initial SST rise was rapid, on the order of ∼103 years, although the full extent of warming was not reached until some ∼30,000 years (30 ky) later (5). The most compelling evidence for greenhouse forcing is a coeval global carbon isotope excursion (CIE) of roughly –3.0 per mil (°) in deep-sea cores (4). The pattern of the CIE—an initial rapid decrease (∼20 ky) followed by a more gradual recovery (130 to 190 ky) (1, 68)—indicates the input of a large mass of isotopically depleted carbon into the ocean and atmosphere. Quantitatively, methane hydrates, with a mean »13C of <–60°, appear to be the most plausible source of this carbon (9). For example, only ∼1200 × 109 metric tons of carbon (GtC) of biogenic methane would be required to produce a CIE of 2.5° (10, 11). Thermogenic methane has been implicated as well (12), although the mass required to produce the CIE would be roughly double that of the biogenic methane.

Regardless of its source, the released methane was rapidly oxidized to CO2. Subsequent oceanic dissolution of this CO2 would alter ocean carbon chemistry, principally by lowering the pH and carbonate ion content [CO 2–3] of seawater. These changes would be partially neutralized by a transient rise in the level of the lysocline and calcite compensation depth (CCD) (13), resulting in the widespread dissolution of sea-floor carbonate. Eventually, the CO2 would be sequestered and ocean carbonate chemistry would be restored, primarily through chemical weathering of silicate rocks (10). The extent and duration of lysocline/CCD shoaling and subsequent recovery would depend largely on the source, mass, and rate of carbon input. For example, modeling of a 1200-GtC input over 10 ky produces a lysocline shoaling of 300 m (less in the Pacific) with a recovery time of ∼40 ky (10). Such changes in [CO 2–3] should produce distinct patterns in pelagic carbonate sedimentation and lithology, characterized by an abrupt transition from carbonate-rich sediment to clay, followed by a gradual recovery to carbonate. Moreover, the clay layer should increase in thickness with increasing water depth.

Clay or low-carbonate layers coincident with the PETM were previously identified in several deep-sea cores and land-based marine sections (1416). However, these sections, which are either geographically isolated or not completely recovered, or both, are inadequate for constraining CCD variations and for testing the methane hypothesis. Ocean Drilling Program (ODP) Leg 208 was designed to recover an array of pelagic cores spanning the Paleocene-Eocene (P-E) boundary over a broad depth range. The primary drilling target was the Walvis Ridge, in the southeastern Atlantic (fig. S1), where the Deep Sea Drilling Project (DSDP) Leg 74 rotary cored portions of the P-E boundary sequence near the base and summit of the ridge (sites 527 and 525) (17). By using advanced piston coring in multiple offset holes at five sites (1262, 1263, 1265, 1266, and 1267), Leg 208 successfully recovered stratigraphically complete and undisturbed upper Paleocene–to–lower Eocene successions at four of five sites between 2.7 and 4.8 km water depth (18). At each site, the P-E boundary sequence was characterized by an abrupt transition from carbonate-rich ooze to a dark red “clay layer,” which then graded back into ooze (Fig. 1 and table S1). Carbonate content was <1 weight percent (wt %) in the clay layers, and >80 and 90 wt % in the underlying and overlying oozes, respectively; the only exception was site 1265, where the basal portion of the clay layer was not recovered. The thickness of the clay layers increased with depth, from 5 cm at the shallowest site (1263) [2717 m; paleodepth ∼1500 m (19)] to 35 cm in the deepest site (1262) [4755 m; paleodepth ∼3600 m (19)] (Fig. 1). The benthic foraminiferal extinction horizon, which is characterized by the disappearance of long-lived Paleocene species and a rapid drop in diversity, occurred at the base of the clay layer in each site (18).

Fig. 1.

Digital core photos and weight % CaCO3 content plotted versus meters of composite depth (MCD) across the P-E boundary interval at ODP sites 1262 (hole A), 1263 (hole C/D), 1265 (hole A), 1266 (hole C), and 1267 (hole B) on Walvis Ridge (fig. S1) (18). Records are plotted from left to right in order of increasing water depth. The core photos for each site represent composites of the following sections: 1262A-13H-5 and -6; 1263C-14H-1 and core catcher (CC); 1263D-4H-1 and -2; 1265A-29H-6 and -7; 1266C-17H-2, -3, and -4; 1267B-23H-1, -2, and -3.

Bulk sediment carbon isotope (»13C) records were constructed at 1- to 5-cm resolution for each boundary section (table S2) (20). Each record is marked by a decrease in »13C at the base of the clay layer, followed by gradual recovery. Minimum carbon isotope values within the clay layer are not uniform, but increase from the shallowest to the deepest site (minimums of –0.9 and 0.0° at sites 1263 and 1262, respectively), a feature we attribute to truncation by dissolution and the presence of residual pre-excursion calcite (21). Also, the base of the CIE differs across sites, occurring in two steps at site 1263 and in a single step at the deeper sites. As a result, the excursion layer, from the onset of the CIE to the point of full recovery (i.e., stability), decreases in thickness from 2.1 m at site 1263 to 1.0 m at 1262.

In this spatially tight array of sites, the production and export of carbonate and the accumulation of clay should be similar at any given time, leaving dissolution as the major process that drives differences in carbonate accumulation between sites. We can therefore infer from the weight % carbonate and carbon isotope data that rapid shoaling of the lysocline/CCD occurred, followed first by a more gradual descent or recovery of the CCD and then by the recovery of the lysocline. The duration of the lysocline/CCD descent from the shallowest to the deepest sites was estimated by first correlating several key inflection points in the carbon isotope records (Fig. 2, tie points A to G), as well as in the Fe concentration and bulk magnetic susceptibility (MS) records (fig. S2). The tie points, particularly E and F, were then verified with biostratigraphic data (table S3) (20). We then correlated the site 1263 carbon isotope record to that of south Atlantic ODP site 690 (22), which has an orbitally derived age model (8), and ordinated the weight % carbonate and isotope data for each site within that age model (Fig. 3 and table S4). An alternate age model based on 3He exists for site 690 (23), but the two models are roughly similar for the initial 100 ky of the PETM; thus, the choice of model makes little difference in our interpretation of events up to that point. The greatest uncertainty in the site-to-site correlations and age estimates is in the basal portion of the clay layer, where the carbon isotope and other records are compromised by dissolution. The correlations (Fig. 1, tie points D to G) are most reliable in the recovery interval where the weight % carbonate is higher and the ocean »13C is rapidly shifting.

Fig. 2.

Bulk sediment carbon isotope records for holes 1262A, 1263C/D, 1265A, 1266C, and 1267B plotted versus MCD. Also plotted are nannofossil horizons (N1 to N4, arrows in red) for holes 1262B and 1263C/D (20). Data for ODP site 690 (22) are plotted to the far left versus meters below the sea floor (MBSF). Lines of correlation are based on inflections in the carbon isotope (A to G above the P-E boundary, –A below), Fe/Ca, and magnetic susceptibility (MS) records (20). vPDB, Vienna PeeDee Belemnite.

Fig. 3.

Bulk sediment »13C and weight % carbonate content (gCaCO3/gTotal × 100) plotted versus age for ODP sites 1262, 1263, 1265, 1266, and 1267. Age (ky) relative to the P-E boundary is plotted on the left axis and absolute age (Ma) along the right. Age models (table S4) are based on correlation to site 690 (8) using the carbon isotope stratigraphy as verified with the nannofossil events in Fig. 2 and with the Fe and MS cycles in fig. S2. Transferring the 1263 age model to deeper sites with carbon isotopes could only be achieved where sufficient carbonate was present. Ages within the clay layers for sites 1266, 1267, and 1262 were derived through linear interpolation from tie points E and A. Paleodepths (∼55 Ma) are provided for sites 1263 (1500 m), 1266 (2600 m), and 1262 (3600 m). Key events in the evolution of south Atlantic carbonate chemistry were (i) the rapid drop in content to <1% for all sites with the exception of site 1265, where the lowermost Eocene is absent (marked I); (ii) the return of the CCD to site 1263 roughly 5 ky after the excursion (marked II); (iii) the return of the CCD to site 1262 at 60 ky (marked III); and (iv) the lysocline descending to a point below the deepest site at 110 ky after the excursion (marked IV). PEB, Paleocene-Eocene boundary.

Given these age constraints, the CCD is inferred to have shoaled more than 2 km within a few thousand years (Fig. 3). Recovery was gradual, with the CCD descending to the shallowest site (1263) within ∼10 to 15 ky of the CIE onset and to the deepest site (1262) within ∼60 ky. By +110 ky, carbonate content had fully recovered. This pattern of change, particularly the recovery, has important implications. According to theory, the initial uptake of CO2 and buffering should occur mainly via deep-sea calcite dissolution, but eventually, chemical weathering of silicate rocks takes over accelerating the flux of dissolved ions (including HCO 3) to the ocean, thereby increasing [CO 2–3] and the rate of calcite accumulation (24). The distribution of carbonate between +60 and +100 ky indicates that the CCD had descended, but the lysocline was still shallow and the deep sea was largely undersaturated. The percentage of CaCO3 continued to increase, and by +110 kyr, it had reached 90% over the entire transect, a state that implies that the lysocline descended below the deepest site (>3.6 km) as well as its pre-excursion level. This phenomenon is consistent with theory (10) and likely represents a transitional period during which the excess ions supplied to the ocean by the weathering of silicate rocks greatly increased deep-sea CO 2–3 concentration and thus carbonate accumulation. The site 690 record is marked by a similar pronounced interval of high carbonate content (23, 25), demonstrating that CO 2–3 oversaturation was not a local phenomenon.

This scenario for acidification of the deep sea and initial neutralization by calcite dissolution is not unlike that simulated by models in response to the anthropogenic rise in CO2 (2628). Because dissolution layers are also present in P-E boundary sections in the Pacific and Tethys Oceans and at depths <1 km (2933), it appears that for a brief period of time, much of the ocean beneath the thermocline was highly undersaturated with respect to calcite. The mass of CO2 required to shoal the CCD to <1 km water depth would be substantial. In a series of simulations with an ocean/sediment carbon-cycle model designed to evaluate the ocean-buffering capacity in response to a range of anthropogenic CO2 fluxes, 4500 GtC was required to terminate carbonate accumulation over the entire ocean (26).

For the PETM, the release of >4500 GtC would be more consistent with the magnitude of global temperature rise (2, 3, 9). Such a large mass of carbon, however, would require a reevaluation of the source of carbon and its isotopic composition. With bacterially produced methane at –60°, the total input from hydrates is limited by the »13C excursion to <2000 GtC (10). To increase the mass of carbon added while adhering to the isotope constraints requires the input of isotopically heavier carbon, such as thermogenic CH4/CO2 (∼–30 to –20°) or oxidation of organic carbon (standing or stored) (–20°) (34). In this regard, recent documentation of an unusual concentration of upper Paleocene fluid/gas seep conduits associated with volcanic intrusions in the North Atlantic (12) merits additional attention. An alternative explanation, that the magnitude of the marine CIE has been greatly underestimated because of dissolution or damping by pH affects, seems unlikely given the constraints provided by continental isotope records (35). Finally, proximity to where carbon (CO2 or CH4) enters the deep sea via circulation will dictate where neutralization by carbonate dissolution is most intense (36). For example, severe dissolution in the Atlantic may indicate direct input of methane into bottom waters entering this basin.

Excessive carbonate undersaturation of the deep ocean would likely impede calcification by marine organisms and therefore is a potential contributing factor to the mass extinction of benthic foraminifera at the P-E boundary. Although most plankton species survived, carbonate ion changes in the surface ocean might have contributed to the brief appearance of weakly calcified planktonic foraminifera (6) and the dominance of heavily calcified forms of calcareous algae (37). What, if any, implications might this have for the future? If combustion of the entire fossil fuel reservoir (∼4500 GtC) is assumed, the impacts on deep-sea pH and biota will likely be similar to those in the PETM. However, because the anthropogenic carbon input will occur within just 300 years, which is less than the mixing time of the ocean (38), the impacts on surface ocean pH and biota will probably be more severe.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article