Abrupt Climate Events 500,000 to 340,000 Years Ago: Evidence from Subpolar North Atlantic Sediments

See allHide authors and affiliations

Science  27 Feb 1998:
Vol. 279, Issue 5355, pp. 1335-1338
DOI: 10.1126/science.279.5355.1335


Subpolar North Atlantic proxy records document millennial-scale climate variations 500,000 to 340,000 years ago. The cycles have an approximately constant pacing that is similar to that documented for the last glacial cycle. These findings suggest that such climate variations are inherent to the late Pleistocene, regardless of glacial state. Sea surface temperature during the warm peak of Marine Isotope Stage 11 (MIS 11) varied by 0.5° to 1°C, less than the 4° to 4.5°C estimated during times of ice growth and the 3°C estimated for glacial maxima. Coherent deep ocean circulation changes were associated with glacial oscillations in sea surface temperature.

During the last glaciation (MIS 2 to 4) and deglaciation, sea surface temperatures (SSTs) oscillated in the subpolar North Atlantic at several time scales. Discrete ice-rafting events marked times of cool SSTs. A series of gradual cooling intervals 6000 to 10,000 years (6 to 10 ky) long were terminated by massive iceberg discharge into the North Atlantic (Heinrich events) (1-3). Shorter SST cycles of 2 to 3 ky [Dansgaard-Oeschger cycles (4)], each terminated by a cold ice-rafting event, occurred between Heinrich events (5). New evidence indicates that there may have been more frequent sea surface changes, spaced ∼1.5 ky apart (6). A similar hierarchy is emerging from Greenland ice core records: glaciochemical time series indicate that the strength of the polar atmospheric circulation varied over cycles of between 6 and 1.45 ky (7), comparable to the approximate spacing of events deduced from the marine record. Such millennial climate oscillations also occurred during the Holocene (6-8).

Deep-water circulation variability may play a critical role in driving and amplifying millennial climate oscillations (6, 9), as well as communicating the response outside of the North Atlantic (10). Cold glacial events were often associated with weakened production of North Atlantic Deep Water (NADW) (11-14).

The finding of similar cycles during glacial times and the Holocene (MIS 1) suggest that they characterize Earth's climate system independent of ice volume. If this is true, then older sediments should reveal millennial-scale climate oscillations at a similar pacing. Indeed, ice-rafting events every 2 to 3 ky have been identified in glacial intervals of the early Pleistocene (13), a time when climate varied at the 40-ky obliquity cycle (15) and ice sheets were about one-quarter to two-thirds the size of those of the last glacial maximum (13).

To explore the persistence of a link between SST and deep-water millennial-scale variability throughout the Pleistocene, we studied sediments from Ocean Drilling Project site 980 on the Feni Drift (55.5°N, 14.7°W, 2179 m below sea level) (16), dated ∼340 to 500 thousand years ago (ka) (17). By this time, late Pleistocene 100-ky climate cycles, characterized by rapid deglaciations, or terminations (18), were firmly established (15). This time interval includes MIS 12 and 11, which are among the most extreme glacial and interglacial intervals, respectively, of the past 500 ky (19). The associated deglaciation, Termination V, had a large amplitude even though insolation forcing was weaker than at the time of most other terminations (20). The interval includes two ice-growth transitions (13/12 and 11/10) that are comparable to MIS 3 through early MIS 2, the part of the last glacial cycle containing the highest amplitude millennial-scale variations (7, 21). We collected δ18O and δ13C records (22) of the benthic foraminifera Cibicidoides wuellerstorfi and δ18O records of the planktonic foraminifera right-coilingNeogloboquadrina pachyderma (N. pachy-R δ18O) to provide detailed estimates of ice volume and deep-water circulation and to study surface water hydrography. The samples were spaced about 300 years apart.

At the Bjorn Drift, to the northwest of site 980, SST estimates across the MIS 6/5 boundary (Termination II) based on N. pachy-R δ18O and the modern analog technique are similar (23), suggesting that N. pachy-R δ18O more accurately record SST than the δ18O of other high-latitude planktonic foraminifera (24). To further examine the extent to which N. pachy-R δ18O records SST variations, we measured the percentage of the polar planktonic foraminifera left-coiling N. pachyderma (%N. pachy-L), a more widely used but qualitative proxy for SST, on the same samples in an 87-ky time slice (350 to 437 ka) that included peak MIS 12 through the early part of MIS 10. Lithic fragments, or ice-rafted debris, were also counted in the same time slice.

The proxy records (Fig. 1) exhibit millennial-scale variability superimposed on variability forced by slowly changing insolation. During the ice-growth transition between MIS 11 and 10, N. pachy-R δ18O oscillated by 1 per mil, and %N. pachy-L varied over a 50% range (Fig.2A). The amplitude of both of these signals corresponds to SST oscillations of about 4° to 4.5°C (25), suggesting that N. pachy-R δ18O records the full range of SST oscillations during this transition. The agreement between the two proxies improves when the ice volume component of the N. pachy-R δ18O signal is removed by subtracting the benthic δ18O record (Fig. 2B). The %N. pachy-L record does not extend to the end of MIS 10, but N. pachy-R δ18O oscillations through 340 ka imply that SST oscillations also occurred during MIS 10. Five of the six coldest events during the MIS 11/10 transition (events c throughg) and one event within MIS 10 (event b) are associated with lithic evidence of iceberg discharge (Fig. 2C). In all, eight to nine strong coolings occurred within a 50-ky interval, for an average spacing of 6 ky, comparable to the spacing of Heinrich events during the last glacial cycle (1-3, 7, 26). Weaker cold events occurred more frequently. The spacing between severe cold events appears to have decreased with increasing ice volume. A series of gradual cooling intervals culminated in cold iceberg discharge eventsg, f, d, and b. Thus, the character of the climate signal is reminiscent of that of MIS 3 (2). Furthermore, as during the last glacial cycle (11, 12), the coldest events were generally associated with low benthic δ13C values (Fig. 2C), implying that the contribution of high- δ13C NADW to the site was reduced. Other δ13C minima are associated with weaker cool events, which may have been more pronounced to the north of site 980.

Figure 1

Summary of site 980 data. (A)C. wuellerstorfi δ13C values (‰, per mil), (B) C. wuellerstorfi δ18O values, (C) N. pachyderma (right-coiling) δ18O values, (D) lithic counts (ice-rafted debris) expressed as a percentage relative to total entities (%IRD, blue) and as a ratio relative to weight of dry bulk sediment (IRD/g, black), and (E) an enlarged version of the data in (D) for 415 to 420 ka. Deglacial warm events (DWE) are labeled. (A) and (C) also include the 1- to 7-ky bandpass filter (31) of the data (black). Shaded interval in (B) indicates the peak of MIS 11. Dashed and dotted lines denote times of cooler temperatures. Inset between (C) and (D) shows N. pachy-R δ18O (red) and %N. pachy-L (black) scaled to equivalent temperature (25) for MIS 11.

Figure 2

(A) N. pachy-R δ18O (red) and %N. pachy-L (black) scaled to equivalent temperature (25). Lithic count (%IRD as in Fig.1) (blue, left axis) is also shown. (B) N. pachy-R δ18O minus benthic δ18O (red) versus %N. pachy-L (black) scaled to equivalent temperature. (C) Benthic δ13C values. Severe cold and ice-rafting events are labeled.

During the MIS 13/12 ice-growth transition, N. pachy-R δ18O oscillated by >1 per mil (Fig. 1C), indicating that SST oscillated by 4° to 4.5°C on this transition as well. Severe cold events occurred more frequently as ice volume increased. During the peak of MIS 12, δ18O varied by ∼0.75 per mil (∼3°C), much less than during the two ice-growth transitions. Like cold events during the MIS 11/10 transition and during the last glacial cycle, cold events during the MIS 13/12 transition and within MIS 12 were associated with low benthic δ13C values (Figs. 1 and3), indicative of reduced NADW production (27). The lower amplitude of benthic δ13C (deep water) variations within MIS 12 compared with those during the 11/10 and 13/12 transitions is consistent with recent modeling experiments that suggest that convective overturn in the North Atlantic is more oscillatory when freshwater discharge is moderate, as might occur during intervals of ice growth, than when discharge is high (maximum glacial) or low (interglacial) (28).

Figure 3

Summary of spectral analysis. (A) Spectra of the %N. pachy-L (red) and N. pachy-R δ18O (black) records. The green line shows the coherency between the two shorter records. Solid and dashed lines are results using the entire (to 437 ka, the length of the %N. pachy-L record) records and only 350 to 410 ka, respectively. The dashed horizontal lines are the 80% and 95% test statistics for non-zero coherency. (B) As in (A), except the black lines are spectra of benthic δ13C, and the green lines are the coherency between %N. pachy-L and benthic δ13C. Bandwidth (BW) is shown.

The end of MIS 12 is marked by a drop in benthic δ18O (Fig. 1B) associated with sea-level rise at ∼420 ka (16). A decrease in %N. pachy-L indicates that the beginning of sea-level rise was immediately followed by a brief deglacial warm event (DWE-1) (Fig. 1E). After DWE-1, a cold and severe ice-rafting event occurred. The presence of detrital carbonate suggests that it involved massive discharge from the Laurentide ice sheet, as did Heinrich events of Mis 2 and 3 (2, 3). Despite extreme cold, a decrease of ∼0.4 per mil in benthic δ18O (Fig. 1B) suggests that sea-level rise accelerated during the cold event. A decrease in %N. pachy-L and ice-rafted debris indicates that a second deglacial warm event (DWE-2) abruptly ended the Heinrich-like cold event (Fig. 1E). After DWE-2, which was itself punctuated by a brief cool event, %N. pachy-L rose to about 30%, indicating that a less severe cooling (2° to 3°C) occurred before the final warming into peak MIS 11.

After this last deglacial cold event, a decrease in N. pachy-R δ18O and %N. pachy-L and an increase in benthic δ13C (Fig. 1) indicate developing interglacial conditions, warming, and enhanced NADW production. Small oscillations in %N. pachy-L,N. pachy-R δ18O, and benthic δ13C punctuated the gradual climatic amelioration into MIS 11. Variations of N. pachy-L abundance by ∼5% suggest that SST oscillated by ∼0.5°C. Although the N. pachy-R δ18O data show more scatter, δ18O variations of ∼0.2 per mil suggest that SST varied by ∼1°C. At the low N. pachy-L percentages measured (1 to 7%), the relation between SST and %N. pachy-L is weak (29); thus, the amplitude of SST variability within MIS 11 was likely closer to the 1°C estimated by the N. pachy-R δ18O data. During the 10- to 12-ky interval of minimum ice volume (benthic δ18O ≤ 2.7 per mil) and maximum SSTs of peak MIS 11, three SST cycles are evident, giving a repeat time of 3 to 4 ky, comparable to cycles documented in marine and ice core records for the last 100 ky (5, 7). Although the benthic δ13C data do not provide clear evidence that NADW was reduced during cool MIS 11 events, we cannot rule out this possibility because the site is close to the deep-water source region and may not be sensitive to subtle variations in NADW during peak interglacial intervals, when NADW production is generally strong.

Spectral analysis of %N. pachy-L, N. pachy-R δ18O, and benthic δ13C records using Blackman-Tukey (30) methods confirm the presence of cycles (Fig. 3) with frequencies close to those noted in the glaciochemical record from the Greenland ice core and in the marine records for the last glacial cycle and the Holocene. For their interval of overlap (350 to 437 ka), the N. pachy-R δ18O and %N. pachy-L records are coherent and in phase in broad bands centered near 6, 2.6, 1.8, and 1.4 ky (Fig.3A). These relations further suggest the utility of N. pachy-R δ18O as a SST proxy, in particular when changes in ice volume are minor. For their interval of overlap (350 to 437 ka), the %N. pachy-L and benthic δ13C records are coherent at periods near 6.3 and 3.8 ky. Power occurs in both records in broad bands centered near 2.3 and 1.8 ky; however, their coherence is weak (Fig. 3B). For the 350- to 410-ka interval (omitting the termination), variations near 1.5 ky are also coherent (Fig. 3B). Benthic δ13C and %N. pachy-L changes are approximately in phase near the 6.3- and 1.5-ky periods, indicating that surface- and deep-water changes occurred together at these cycles. The in-phase behavior at the 6.3-ky period reflects the association of the larger millennial-scale SST minima, which occurred ∼6 ky apart, with low benthic δ13C values (Figs. 1 and2). By contrast, SST changes began ∼800 years after changes in NADW production near the 3.8-ky period. This phase difference accounts for some of the differences in timing of changes evident in the benthic δ13C and %N. pachy-L records (Fig. 1).

We used a bandpass filter (31) to examine the amplitude of higher frequency (1 to 7 ky) variations in the N. pachy-R δ18O values. The filtered signal underscores the small-amplitude millennial-scale variations during interglaciations MIS 11 and 13 relative to variations that characterized times of large ice volume or ice growth. Climate oscillations during deglaciation were part of the regular sequence of millennial-scale oscillations.

Our study indicates that variability in the climate from 350 to 500 ka was similar to that of the last glacial cycle, suggesting that millennial-scale variability persisted during the past half a million years. The amplitude of the variability was much smaller during interglacial intervals and greatest during ice growth. Severe events became more frequent near glacial maxima. The pacing of climatic events during our 150-ky-long core interval is indistinguishable from that during the last glacial cycle. Sea surface temperature and deep water varied together at a cycle near 6 ky. At shorter (3.8 and 1.5 ky) cycles, surface and deep-water changes occurred, on average, several hundred years apart. Our data suggest that variability in benthic δ13C (deep water) values was also greater during ice-growth transitions, consistent with a role of deep water in amplifying millennial climate cycles (6, 9, 28).


View Abstract

Navigate This Article