The Mid-Pleistocene Transition in the Tropical Pacific

See allHide authors and affiliations

Science  11 Nov 2005:
Vol. 310, Issue 5750, pp. 1009-1012
DOI: 10.1126/science.1115933


A sea surface temperature (SST) record based on planktonic foraminiferal magnesium/calcium ratios from a site in the western equatorial Pacific warm pool reveals that glacial-interglacial oscillations in SST shifted from a period of 41,000 to 100,000 years at the mid-Pleistocene transition, 950,000 years before the present. SST changes at both periodicities were synchronous with eastern Pacific cold-tongue SSTs but preceded changes in continental ice volume. The timing and nature of tropical Pacific SST changes over the mid-Pleistocene transition implicate a shift in the periodicity of radiative forcing by atmospheric carbon dioxide as the cause of the switch in climate periodicities at this time.

In the mid-Pleistocene, ∼950 thousand years (ky) before the present (B.P.), the climate of Earth underwent profound changes in the length and intensity of its glacial cycles. This mid-Pleistocene transition (MPT), as indicated by benthic foraminiferal δ18O, was characterized by a change in the dominant periodicity of high-latitude climate oscillations from 41 ky to 100 ky; a positive shift in mean benthic d18O, generally ascribed to continental ice-sheet expansion; and an increase in the amplitude variability of δ18O, attributed to more severe glaciations after 950 ky B.P. (13). Most of the hypotheses offered to explain these changes involve high-latitude Northern Hemisphere processes such as ice-sheet or sea-ice dynamics (2, 4, 5). Recent paleoclimatic reconstructions, however, have shown that during the MPT, the tropics also experienced major changes that resemble some aspects of high-latitude climate variability but also have their own unique patterns (6, 7). Current hypotheses cannot fully explain these observations and the common characteristics revealed by paleoclimatic reconstructions from low and high latitudes.

To test hypothesized causes for the mid-Pleistocene transition, we reconstructed detailed thermal and δ18O-seawater histories spanning the MPT from a site in the heart of the western equatorial Pacific (WEP) warm pool (Fig. 1). This location is ideal for testing hypotheses that address proposed forcing mechanisms of tropical climate variability because (i) warm-pool thermal variability is linked throughout the tropics by convection (8); (ii) the warm pool is less subject to regional oceanographic influences such as thermocline depth changes, as demonstrated by the small response of warm-pool sea surface temperatures to El Niño/Southern Oscillation (ENSO) variations (Fig. 1) (9); and (iii) the warm pool is remote from the direct radiative influence of continental ice sheets and has the most direct response to radiative forcing as a result of changes in atmospheric greenhouse gases (10, 11).

Fig. 1.

Map showing the correlation of Kaplan interannual SST anomalies between the site of Hole 806B (0°19.1′N, 159°21.7′E) and other regions of the tropics (28). Correlations are based on SST data on a 5 by 5 grid of monthly anomalies from 1856 to 2003. The base period used for the anomalies is 1951 to 1980. Locations of ODP Hole 846 (6) (3°5′S, 90°49′W) and MD97-2140 (7) (2°02′N, 141°46′E) are also indicated. Warm-pool SST anomalies near Hole 806B are positively correlated with temperature anomalies in a wide swath of the tropical oceans. Warm-pool anomalies are either uncorrelated or are anticorrelated with anomalies in the eastern equatorial Pacific cold tongue (i.e., Hole 846), a consequence of their opposite behavior during ENSO changes (20). The location of ODP Hole 677 [reference benthic foraminiferal δ18O record in (15)] is also shown (1°12′N, 83°44′W).

We determined sea surface temperatures (SSTs), δ18O, and δ18O-seawater (δ18Ow) from Ocean Drilling Program (ODP) Hole 806B (0°19.1′N, 159°21.7′E, 2520-m water depth) (12) on the Ontong Java Plateau, using the surface-dwelling planktonic foraminifer Globigerinoides ruber (Fig. 2). Our records reach back to 1.3 million years (My) B.P., with an average resolution of 2.3 ky, extending a previous study (13). We used the Mg-paleothermometry technique, which is based on the temperature dependence of Mg substitution in calcite, and calculated δ18Ow following previous protocols (13, 14). We constructed the Hole 806B age model by visual alignment of the G. ruber δ18O sequence to the ODP Hole 677 benthic δ18O record (14, 15). Hole 806B has remarkably constant sedimentation rates (2.0 ± 0.3 cm/ky) from 0.45 to 1.3 My B.P. and, because of its location above the present-day lysocline depth, it also has good preservation of foraminifer shells. There are only two coring gaps of ∼0.9 m (∼50 ky) that include parts of marine isotope stage (MIS) 19 and MIS 37 (14).

Fig. 2.

Western equatorial Pacific ODP Hole 806B records based on the surface-dwelling foraminifera G. ruber. This record includes a previous (13) reconstruction for the late Pleistocene (4 to 470 ky B.P.). Gaps in the records are the result of coring gaps. The chronology is based on wiggle matching to the target core ODP Hole 677 benthic foraminiferal δ18O (14, 15). (A) G. ruber Mg/Ca-derived SST record. Mg/Ca data were converted using the relationship: SST (C°) = 0.089–1*ln[Mg/Ca (m) / 0.3] (13). Each point is the average of two to four replicates. The mid-Pleistocene transition (MPT) at ∼950 ky B.P. is indicated by a gray bar. Evolutionary spectral analysis plot for SST (top) from 250 to 1100 ky B.P. is based on the multitaper method (14). The scale represents Log10 units2 per cycle per ky. The evolutionary spectral analysis reveals that the SST record is dominated by the ∼100-ky period over the last 950 ky B.P. and by ∼41-ky and ∼60-ky–related periods from 1346 ky to 950 ky B.P. (B) G. ruber d18O record. The G-I range of δ18O increases over the MPT. (C) δ18O-seawater record, calculated by extracting the component in planktonic δ18O explained by the Mg/Ca-derived SST using the low-light paleotemperature equation determined for Orbulina universa (29). Neither the calcite nor the seawater δ18O records shows a positive shift over the MPT.

The G. ruber δ18O data indicate 12 glacial-interglacial (G-I) oscillations from MIS 13 to 41 between 450 and 1348 ky B.P., in agreement with reference foraminiferal δ18O records (12, 15) (Figs. 2 and 3). Over the past 900 ky, the G-I range of δ18O is larger by about one-third than the corresponding early Pleistocene (900 to 1348 ky) range, but mean d18O remains the same [–1.60 ‰ (per mil) and –1.56 ‰, respectively]. Spectral analysis of the Hole 806B δ18O data indicates that over the past 900 ky, the 100-ky period component explains more than 70% of the variance in d18O, whereas during the early Pleistocene (EP) similar power is shared by ∼90-ky and 41-ky–related periodicities, with a minor contribution from the 23-ky period. The presence of significant power at a ∼90-ky period might be the result of a strong salinity component at site 806B during the early Pleistocene (16).

Fig. 3.

(A) ODP Hole 677 benthic foraminiferal δ18O record (15) showing a mean positive shift of 0.25‰ at ∼900 ky B.P. and long-term stability from 1340 to 900 ky B.P. (dashed lines are linear regressions). The linear regression of δ18O over this time interval is not statistically significant (17). Top: Spectrogram of benthic δ18O from 250 to 1100 ky B.P. showing the shift in the dominant period from 100 ky to 41 ky at ∼950 ky B.P. (gray bar) (14). The scale represents Log10 units2 per cycle per ky. (B) Hole 806B G. ruber SST record showing no significant secular change from 1340 to 800 ky B.P. and long-term thermal stability over the MPT (red line before the transition represents a nonsignificant linear regression; slope = zero, 95% CI). (C) SST record from eastern equatorial Pacific ODP Hole 846 (3°5′S, 90°49′W) based on the alkenone unsaturation index (6). The statistically significant linear regression (slope ≠ zero, 95% CI) showing a secular cooling trend from 1340 to 900 ky B.P. is indicated (green line). As a consequence, from 1340 to 900 ky the equatorial Pacific temperature zonal gradient increased by ∼1.3°C.

The observed Mg/Ca-derived SST average from the early and mid-Pleistocene time intervals combined (from 500 to 1348 ky B.P.) is 27.8°C, similar to the late Pleistocene (0 to 500 ky B.P.) SST average of 27.4°C (13). The G-I SST range over the mid-to-early Pleistocene is smaller than that during the late Pleistocene, 3°C versus 4.3°C, respectively. This difference is largely a consequence of warmer (∼0.7°C) glacial intervals, relative to the late Pleistocene; average interglacial SSTs (∼29°C) are similar throughout the record. The warmest temperatures during the mid- and early Pleistocene occurred in MIS 25 (29.8°C), 952 ky B.P., and the coldest in MIS 30 (26°C), 1052 ky B.P. (Fig. 2). The mid-Pleistocene transition is well represented in the Hole 806B SST record (Fig. 2). Mean SST and the average G-I change in SST do not shift over the MPT, in contrast to changes observed in benthic δ18O (Figs. 3 and 4). Average SSTs during the early and mid-Pleistocene (500 to 900 ky B.P.) are virtually identical, 27.9 ± 0.7°Cand 27.7 ± 0.7°C (1 SD), respectively (Fig. 2).

Fig. 4.

Blowup of the mid-Pleistocene transition as seen in the Hole 806B SST record (A) and the Hole 677 benthic foraminiferal δ18O record (B). Right panels show a blowup over the MPT of the spectrograms shown in Figs. 2 and 3. The interglacial peak centered on 950 ky B.P. is MIS 25. Note the presence of the positive shift in benthic δ18O at ∼900 ky B.P. and the absence of a similar shift in SST at that time. The transition between the 41-ky and 100-ky variability occurs between 950 and 1000 ky B.P. in both records. Point-to-point comparisons between the two signals suggest that SSTs lead benthic δ18O by ∼3 ky over the MPT. The same pattern can be seen between SST and planktonic δ18O in Hole 806B (Fig. 2).

The G-I range in δ18Ow, calculated from measured δ18O and inferred temperatures, is ∼0.7‰ over the full length of the record (Fig. 2). A previous study of the late Pleistocene record from Hole 806B (13) demonstrates that d18Ow at this site is strongly influenced by hydrological changes on G-I time scales. In addition to orbital frequencies, the δ18Ow time series also shows quasiperiodic ∼200-ky cycles during the early and mid-Pleistocene time interval. These cycles might be related to long-term hydrological evolution in this region, suggesting that d18Ow is not a simple proxy of ice volume at this site. The G-I range of Hole 806B δ18Ow increases by ∼0.16‰ during the MPT, from an early Pleistocene value of 0.72‰ to a mid-Pleistocene value of 0.88‰, which likely reflects increasing variability in continental ice as suggested by benthic foraminiferal records (13).

The Hole 806B SST record is spectrally similar to the ODP Hole 677 reference benthic foraminiferal δ18O record (15), with a characteristic dominance of ∼100-ky and 41-ky periods and a much weaker contribution at 23 ky (Figs. 2 and 3). As suggested by evolutionary spectral analysis of Hole 806B SST and Hole 677 benthic foraminiferal δ18O, the transition between the 41-ky and 100-ky–dominant modes of variability occurred at ∼950 ky B.P. (Fig. 4, right panels). Point-to-point comparison between these two records over the MPT reveals that G. ruber SST leads benthic δ18O by ∼3ky (Fig. 4). Furthermore, cross-spectral analysis between G. ruber SST and benthic foraminiferal δ18O reveals that SST leads benthic d18O by 3 ± 1.2 ky [95% confidence interval (CI)] at the 41-ky–dominant period during the early Pleistocene.

SST records from two other sites in the tropical Pacific, one in the eastern equatorial cold tongue (6) and a second in the area of strong intertropical convergence zone influence north-west of our site (7), provide basinwide context for our records (Fig. 1). Comparison of these SST records from 1348 to 900 ky B.P. suggests a strengthening of the zonal equatorial Pacific SST gradient by ∼1.3°C, due almost entirely to the cooling in the eastern Pacific. The development of this SST gradient occurred during a time interval in which there was no secular change in WEP SSTs, as revealed by the two western Pacific SST records (7) (Fig. 3). High-latitude climate, as indicated by the Hole 677 δ18O record, was also relatively stable at this time (Fig. 3). Statistical analysis of benthic foraminiferal records and the Hole 806B δ18Ow series reveal that high-latitude climate was relatively stable for more than 400 ky before the MPT (Fig. 3) (17). This observation suggests that the intensification of tropical Pacific zonal temperature gradients and the inferred enhancement of the Walker circulation at this time was not accompanied by regional long-term hydrological changes in the WEP and Northern Hemisphere high-latitude climate reorganizations, in contrast to previous suggestions (7, 18).

The long-term surface cooling of the eastern-boundary upwelling regions from 1350 to 900 ky B.P. (6), while the warm pool and Northern Hemisphere climate remained relatively stable, may be related to secular changes in the density of the deep ocean that influenced the depth of the thermocline, as previously predicted (9). This cooling has been invoked by a number of hypotheses addressing the MPT (7, 18). The thermal stability of the WEP, where SSTs are expected to respond thermodynamically to atmospheric radiative forcing, and the coinciding high-latitude climate stability from 900 to 1350 ky B.P. reflected by benthic δ18O records, do not support changes in radiative forcing as the cause of the inferred eastern equatorial Pacific (EEP) secular cooling trend (7). Benthic foraminiferal carbon isotopic records, interpreted as a proxy of the thermohaline circulation, show a decrease in the δ13C contrast between the North Atlantic and Pacific from 1.3 My B.P. to ∼800 ky B.P. (19). The cooling in the EEP might reflect shallowing of the thermocline resulting from an increase in stratification produced by the deep-ocean circulation rearrangements suggested by δ13C records. Model calculations suggest that a modest change in the temperature difference across the thermocline of only a few tenths of a degree can produce changes in the EEP SSTs of over 1°C (9).

The spectral properties of the Hole 806B SST record provide a powerful test of current hypotheses addressing Pleistocene tropical and high-latitude climate variability and the mid-Pleistocene transition. The spectral resemblance between the WEP (Hole 806B) and EEP (Hole 846) (6) SST records is striking (Fig. 3). Sea surface temperature variations in both end members of the equatorial Pacific are statistically coherent and in phase within the 2-ky resolution of the sites, and both records switch from 41-ky to 100-ky–dominant periods during the MPT (fig. S3). Furthermore, the early Pleistocene G-I SST range from both Hole 806B and Hole 846 is similar, 3°Cand4°C, respectively. Today, SSTs in the EEP are strongly influenced by wind-driven thermocline depth changes (20). In the WEP, where the thermocline is very deep (>100 m), SSTs are much less likely to be affected by thermocline depth changes (9, 20). Because of this difference, interannual SST anomalies in the EEP cold tongue associated with the ENSO phenomenon are not correlated with anomalies near site 806B (Fig. 1). Further support for differences in the thermal evolution of the two equatorial Pacific end members lies in the observed long-term thermal stability of the WEP during the interval in which the EEP became progressively colder, intensifying Pacific zonal gradients after 1350 ky B.P. (Fig. 3). These observations suggest that a mechanism that invokes changes in thermocline depth (6) is unlikely to explain the observed warm-pool SST variability, because such a mechanism would not produce strong 41-ky cycles in SST in the WEP. On the other hand, as pointed out by Liu and Herbert (6), the sense of annual insolation changes in the tropics as driven by obliquity variations is in the opposite direction of that required to drive the observed tropical SST changes. We suggest instead that both end members of the equatorial Pacific responded to a common factor: atmospheric CO2 forcing.

Consideration of the radiative forcing by different components potentially implicated in the Last Glacial Maximum suggests that atmospheric CO2 changes are the dominant source of radiative forcing in the tropical ocean regions (10). A crucial role of atmospheric CO2 in forcing tropical and Southern Hemisphere climate variability is strongly suggested by the observation that Antarctic air temperatures (21, 22), tropical SSTs (11), and bottom-water temperatures (23) are in phase with atmospheric CO2 and lead benthic foraminiferal δ18O by several thousand years during the late Pleistocene. In the same manner, spectral comparisons between tropical SST records from the three sites in the tropical Pacific and foraminiferal δ18O over the early Pleistocene reveal that all three SST records lead foraminiferal δ18O by 3 to 7 ky at the dominant 41-ky period (table S3). The inferred lead of SST over continental ice volume rules out the hypothesis that tropical SST variability is controlled by the direct radiative influence of Northern Hemisphere continental ice sheets. The observed pattern of early and mid-Pleistocene tropical climate variability, marked by synchronous and similar magnitude SST cycles in both the warm and cold end members of the tropical Pacific, and with a clear lead of both over continental ice volume changes, is remarkably similar to late Pleistocene climate observations (11). The character of Pleistocene climate evolution suggests that the shift in tropical climate variability from a 41-ky to a 100-ky–dominated system (Figs. 3 and 4) is the result of changes in greenhouse forcing as mediated by the radiative effect caused by variability in atmospheric CO2. We speculate that the global carbon system, acting as an internal self-sustained oscillator sensu (4), was paced by obliquity changes during the early Pleistocene (24); this response shifted to the eccentricity envelope of precession after the mid-Pleistocene transition. Future reconstructions of atmospheric CO2 extending back to the MPT, projected as part of the European Project for Ice Coring in Antarctica (EPICA) (22, 25), would be a direct test of this hypothesis.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

Tables S1 to S3


References and Notes

View Abstract

Navigate This Article