A Transient Rise in Tropical Sea Surface Temperature During the Paleocene-Eocene Thermal Maximum

See allHide authors and affiliations

Science  28 Nov 2003:
Vol. 302, Issue 5650, pp. 1551-1554
DOI: 10.1126/science.1090110


The Paleocene-Eocene Thermal Maximum (PETM) has been attributed to a rapid rise in greenhouse gas levels. If so, warming should have occurred at all latitudes, although amplified toward the poles. Existing records reveal an increase in high-latitude sea surface temperatures (SSTs) (8° to 10°C) and in bottom water temperatures (4° to 5°C). To date, however, the character of the tropical SST response during this event remains unconstrained. Here we address this deficiency by using paired oxygen isotope and minor element (magnesium/calcium) ratios of planktonic foraminifera from a tropical Pacific core to estimate changes in SST. Using mixed-layer foraminifera, we found that the combined proxies imply a 4° to 5°C rise in Pacific SST during the PETM. These results would necessitate a rise in atmospheric pCO2 to levels three to four times as high as those estimated for the late Paleocene.

The Paleocene-Eocene Thermal Maximum [55 million years ago (Ma)] was accompanied by a number of environmental perturbations, including mass extinction of benthic foraminifera (1), widespread appearance and proliferation of exotic plankton in the open and coastal oceans (25), shifts in precipitation patterns and intensity (6), and migration and dispersal of terrestrial mammals (7). These perturbations are attributed to a global warming of unusual magnitude (810). That this warming was greenhouse gas–induced is supported by two independent lines of evidence: a >2.5‰ negative carbon isotope excursion (CIE) of the entire global exogenic carbon pool and a pronounced deep-sea carbonate dissolution horizon. Both are consistent with injections of a large mass of 12C-enriched carbon into the ocean/atmosphere (11). The primary source of this carbon has been postulated to be the dissociation of methane hydrate (12).

In theory, a substantial increase in atmospheric CO2 and/or CH4 levels as posited for the PETM should have induced warming at all latitudes, with SST increases amplified toward the poles. This pattern of warming is a robust feature of greenhouse simulations that use coupled ocean-atmosphere models with a strong ice-albedo feedback (13). Simulations with a fixed ice-albedo feedback also exhibit polar amplification, but to a lesser extent (14). In this regard, the PETM provides an opportunity to test climate theory on greenhouse warming. A reliable record of tropical SST change during the PETM, however, has yet to be established. The few marine boundary sections recovered from the tropics are either unconformable (2), barren of planktonic foraminifera, highly lithified (15), or markedly disturbed by coring (16). Even where complete, interpretation of the most common temperature proxy, foraminiferal δ18O, has proved ambiguous in identifying relatively small changes in SST (<4°C) because of other variables that influence the oxygen isotope signature.

The lack of tropical sections was recently addressed by Ocean Drilling Program Leg 198, which recovered stratigraphically complete Paleocene-Eocene boundary sequences in three holes at Site 1209 (32°39.1081′N, 158°30.3564′E) (water depth of 2387 m) on Shatsky Rise (fig. S1) (17). The PETM is represented by a 25-cm thick, dark, calcareous ooze embedded within a uniform sequence of white calcareous nannofossil ooze (Fig. 1). The basal contact, which coincides with the benthic foraminiferal extinction event and a marked turnover in nannofossil assemblages, is sharp, whereas the upper contact is gradational. The percentage of CaCO3 declines just at the base of the contact and then gradually recovers. This lithologic pattern, which is a common feature of all marine pelagic PETM sections (18), is attributed to the dissolution of calcareous sediments on the sea floor from a CO2-induced drop in ocean pH and carbonate ion concentration. Because carbonate content remains relatively high in the dissolution layer at Site 1209, however, planktonic foraminifera are well enough preserved for paleoenvironmental reconstruction. At 55 Ma, Site 1209 would have been located between 15° and 20°N latitude.

Fig. 1.

Hole 1209B was drilled at a water depth of 2387 m (17). The sampled 80-cm interval was recovered at a subbottom depth of 195.95 to 196.6 m. The sediment is calcareous nannofossil ooze. The sharp color contact lies just above the Paleocene-Eocene boundary at 1.35 m in this section. (A) Percent CaCO3 of bulk sediment. (B) Bulk magnetic susceptibility. (C) Percent fragmentation of planktonic foraminifera. (D and E) The δ13C and δ18O of individual specimens of M. velascoensis (blue circles) and A. soldadoensis (red triangles). (F and G) Mg/Ca and Sr/Ca of M. velascoensis and A. soldadoensis. The lower age listed on the right represents the assigned age of the boundary (55.0 Ma) and is placed at the base of the excursion. The upper age (54.85 Ma) is based on the estimated duration of the event as determined from orbital cycles (23) and He isotopes (24).

Sediment samples were collected at a 1- to 3-cm spacing over an 80-cm interval (195.8 to 196.6 meters below the sea floor) encompassing the PETM in Hole 1209B. Two planktonic foraminifer taxa, Morozovella velascoensis and Acarinina soldadoensis, were collected for analysis. As with modern mixed-layer foraminifera (e.g., Globigerinoides sacculifer), both species harbored photosymbionts, indicating a photic zone habitat (19). A total of 25 to 30 whole specimens in the 300- to 425-μm size range were collected from each sample, of which four to five single specimens were each analyzed individually for stable isotopes (20). For analyses of minor elements (magnesium, strontium, and manganese), 15 to 25 specimens were cleaned with a reductive-oxidative process (21). Total inorganic carbon content (%CaCO3) and an index of carbonate dissolution, percent fragmentation (%F), were determined as well (22).

All records, with the exception of Sr/Ca, exhibit considerable anomalies across the PETM (Fig. 1 and table S1). The δ13C records for both species show abrupt –3.0‰ excursions, the hallmark CIE that provides the primary chronostratigraphic control. The δ18O records exhibit –0.5 to –0.8‰ excursions. The Mg/Ca ratios of both species increase from ∼3.6 to 5.5 mmol/mol. The Mg/Ca excursions appear to initiate just at the color contact, reaching a peak ∼20 cm above. All records recover or stabilize between 50 and 60 cm above the contact. On the basis of studies of other deep-sea sites (23, 24), the CIE from onset to recovery (1.35 to 0.85 m) has been constrained to less than ∼150,000 years.

With modern foraminifera, an exponential relation is observed between calcification temperature and shell Mg/Ca. Exponential functions have been fitted to empirical data from core-top, sediment-trap, and culturing studies (2528). Although absolute ratios for different species can vary by as much as 15% at a given temperature, the regression exponents are nearly identical (0.09 to 0.095). This phenomenon is observed in the δ18O/temperature relation of planktonic foraminifera and other marine taxa as well (29, 30). Consequently, we apply the Mg/Ca and δ18O temperature regressions for the tropical species G. sacculifer to the extinct photosymbiont-bearing foraminifera analyzed here (Fig. 2). Because the absolute Mg/Ca temperature calibration and mean chemical composition of early Cenozoic seawater are both unknown, and because foraminiferal specimens show a minor degree of overgrowth, we computed temperature change relative to a preexcursion baseline. Given these constraints, the changes in Mg/Ca for M. velascoensis and A. soldadoensis imply peak warming of 4.0°C and 5.0°C, respectively, whereas the changes in δ18O indicate warming of 3.0°C and 2.0°C, respectively.

Fig. 2.

In the Hole 1209B PETM section, the δ18O and Mg/Ca anomalies were both calculated with respect to the sample at 1.485 m (sample 1209B-22H-1, 148 to 149 cm), the lowest pre-event sample in the data set. (A) The Mg/Ca temperature anomalies were derived with the following equation: Embedded Image where ΔT is the relative change in temperature (°C), m is an assumed temperature sensitivity (i.e., exponential constant), and C is the percentage change in Mg/Ca (with respect to the baseline Mg/Ca value) (see note S1 for full derivation). The calculated Mg/Ca temperature anomaly field assumes a range of exponential constant values between 0.085 and 0.107, within a range characteristic of most modern planktonic taxa (26, 27). The δ18O temperature anomalies were calculated with a Δδ18Ocalcite/ΔT relationship of –0.213‰/°C. (B) The salinity anomaly calculations assume that the discrepancy between the δ18O and Mg/Ca temperature anomalies results from a local surface-water δ18O increase during the PETM due to a shift in the precipitation-evaporation balance. For these calculations, an intermediate exponential constant value of 0.09 was used to obtain the Mg/Ca temperature anomalies. These anomalies were then converted to expected δ18O anomalies (with the –0.213‰/°C relationship), and the observed δ18O anomalies were subtracted from δ18O anomalies. The resulting δ18O “residual” (‰) was then assumed to reflect changes corresponding to surface-water salinity changes. The salinity anomaly fields represent a range of Δδ18Oseawater/Δsalinity relationships between 0.50‰/salinity unit and 0.25‰/salinity unit.

In terms of quantifying temperature change, characterization of such a short duration event affords several advantages. First, the global mean seawater δ18O (in the absence of large ice sheets) and Mg/Ca ratios are constant on these time scales (31). Second, the effects of minor secondary calcification should be uniform on the cm scale (32), particularly in the unlithified calcareous ooze of Site 1209. This supposition is confirmed by scanning electron microscope observations of foraminifera, which show constant minor overgrowth throughout the PETM (fig. S2). Sea-floor dissolution, on the other hand, can alter the chemical composition of mixed-layer tropical planktonic foraminifera by selectively removing the more solution-prone low δ18O and high Mg/Ca portions of the shell (28, 33). In modern foraminifera, Mg/Ca decreases as a function of depositional depth. The primary index of shell dissolution for Hole 1209B, percent fragmentation, however, does not covary with Mg/Ca. Dissolution peaks in the uppermost Paleocene sediments (Fig. 1), consistent with the posited sea floor carbonate dissolution pulse (11). The invariant Sr/Ca also supports the lack of a preservation artifact on Mg/Ca.

The apparent ∼2.5°C discrepancy between δ18O and Mg/Ca can be attributed to one of several processes. First, the δ18O signal might be attenuated by secondary calcite added at the sea floor at lower temperatures than in surface waters where primary calcite formed (32, 34). This process, which preserves primary trends but shifts mean values, can be ruled out because it should affect both the isotope- and minor elemental–based temperature estimates equally (33). Second, changes in surface ocean pH may also affect isotope ratios with increasing δ18O at lower pH (∼0.15‰/0.1) (35, 36). For example, an increase in partial pressure of CO2 (pCO2) from 1000 to 2000 parts per million by volume (ppmv) should lower surface-water pH by ∼0.16 units, which would account for 0.22‰ increase in δ18O and at most ∼1°C of the difference between the calculated Mg/Ca and δ18O temperature anomalies. A third possibility is that the Mg/Ca sensitivity to temperature change for these extinct species is different than it is for modern taxa. Here, we must assume that the basic biochemical processes that govern the incorporation and partitioning of cations as well as isotopes in modern foraminifera were similar to those of their ancestors (37). So, although the absolute ratios may differ at any given temperature, the exponential constant used to estimate the temperature anomalies for extinct taxa should be similar to that derived for modern taxa. That the absolute Mg/Ca, as well as the δ18O, ratios of M. velascoensis and A. soldadoensis are consistently offset supports this assumption. The final and most likely source of the discrepancy involves changes in local sea surface salinity (SSS) and δ18OSW. In the modern ocean, SSS and δ18OSW covary because of fractionation associated with evaporation, vapor transport, and precipitation such that a local decline in net precipitation (higher evaporation) would increase δ18OSW and partially offset the temperature-related decline in shell δ18O. For example, using the Mg/Ca-derived SST increase, we estimate that a concomitant increase in δ18OSW of 0.5 to 0.8‰, and in salinity of 1.0 to 2.6 parts per thousand, is sufficient to account for the observed δ18O anomaly (Fig. 2).

The implications of a 5°C rise in tropical SST during the PETM are important. According to numerous modeling studies, an abrupt and extreme increase in greenhouse gas levels should give rise to warming at all latitudes, although amplified toward the poles. The magnitude of SST change recorded near Antarctica during the PETM, ∼8°C to 10°C, would necessitate a rise of 1500 ppmv in pCO2 if starting from a baseline of 560 ppmv (e.g., twice the pre-industrial level), less if starting from a lower baseline (13). A pCO2 increase of this magnitude should also warm the tropics by 3°C to 5°C, which is essentially identical to our Mg/Ca-derived estimates. Thus, these findings reinforce the hypothesized greenhouse gas forcing for the PETM. Whether the primary radiative forcing was supplied directly by CH4 and/or CO2 is still unknown. The carbon isotope excursion has been attributed to the expulsion of 1200 to 2500 gigatons (Gt) of CH4 from gas hydrates over 10,000 years (12). Using the lower limit and assuming that the CH4 was quantitatively converted to CO2, numerical models indicate a modest rise in atmospheric pCO2 (<100 ppmv) (11, 38), an amount far below that required to drive the observed warming, even if the initial baseline pCO2 was much lower (39). This suggests either that the CH4, which is a more potent greenhouse gas, escaped immediate oxidation in the ocean and accumulated in the atmosphere (4042), that the mass of CH4 released was substantially greater (> 4 × 103 Gt) than estimated, and/or that additional greenhouse gas (CO2) was supplied by another source. The first option may still be insufficient to create the warming observed in the tropics, and the second option is constrained to some extent by the magnitude of the CIE to less than 2.5 × 103 (38). The third option is compelling because additional greenhouse gases could be generated by positive feedback such as ocean degassing or a decline in terrestrial carbon storage. One last possibility, albeit remote, is that the climate system is more sensitive to changes in greenhouse gas levels than current theory would predict. Future work should focus on confirming the estimates of tropical SST change and placing tighter constraints on the total mass of carbon released during the PETM.

Supporting Online Material

SOM Text

Figs. S1 and S2

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article