Synchronization of North Pacific and Greenland climates preceded abrupt deglacial warming

See allHide authors and affiliations

Science  25 Jul 2014:
Vol. 345, Issue 6195, pp. 444-448
DOI: 10.1126/science.1252000

Climates conspire together to make big changes

The regional climates of the North Pacific and North Atlantic fluttered between synchrony and asynchrony during the last deglaciation, with correspondingly more and less intense effects on the rest of the world, researchers have found. The climate system can be highly nonlinear, meaning that small changes in one part can lead to much larger changes elsewhere. This type of behavior is especially evident during transitions from glacial to interglacial conditions, when climate is affected by a wide variety of time-varying influences and is relatively unstable. Praetorius and Mix present a record of North Pacific climate over the past 18,000 years. When the climates of the more local high-latitude Pacific and Atlantic sectors varied in parallel, large, abrupt climate fluctuations occurred on a more global scale.

Science, this issue p. 444


Some proposed mechanisms for transmission of major climate change events between the North Pacific and North Atlantic predict opposing patterns of variations; others suggest synchronization. Resolving this conflict has implications for regulation of poleward heat transport and global climate change. New multidecadal-resolution foraminiferal oxygen isotope records from the Gulf of Alaska (GOA) reveal sudden shifts between intervals of synchroneity and asynchroneity with the North Greenland Ice Core Project (NGRIP) δ18O record over the past 18,000 years. Synchronization of these regions occurred 15,500 to 11,000 years ago, just prior to and throughout the most abrupt climate transitions of the last 20,000 years, suggesting that dynamic coupling of North Pacific and North Atlantic climates may lead to critical transitions in Earth’s climate system.

Abrupt climate transitions observed during the last deglaciation (1, 2) and within the last glacial interval (3) demonstrate that internal climate feedbacks can amplify the effects of relatively weak external climate forcing. Understanding the mechanisms involved in generating past abrupt transitions, which have led to regional warming events of ~10°C within 3 to 60 years (2), will help to assess the dynamic nature of climate tipping points.

Fluctuations in the Atlantic Meridional Overturning Circulation (AMOC) are often invoked to explain millennial-scale climate changes in the North Atlantic region (46), as well as the so-called bipolar seesaw, which reflects changes in net oceanic heat transport between the southern and northern hemispheres (7, 8). An interocean seesaw also has been proposed to operate between the North Atlantic and North Pacific, such that poleward heat transport and/or deep-water formation increases in the North Pacific during times of weakened AMOC strength (911). This remains uncertain, however, because models show conflicting responses for the North Pacific (1013). Paleoclimate reconstructions are similarly in conflict; some support an interocean seesaw (9, 11, 14), whereas others suggest in-phase behavior between the North Atlantic and North Pacific (15, 16), and still other studies suggest a blend of northern (atmospheric) and southern (oceanic) influences (17, 18). If an Atlantic-Pacific seesaw exists, low northward heat transport in one ocean might be partly compensated by high northward heat transport in the other. Conversely, synchronous variations in the two oceans would tend to amplify climate changes in the high northern latitudes by either enhancing or diminishing meridional heat transport.

Changes in the AMOC and Arctic sea ice have been identified as “tipping elements” in the climate system (19); both are influenced by poleward heat transport and have the potential for rapid transitions (10 to 100 years) to a new climate state, accompanied by climate and ecosystem effects that are to some degree irreversible (19). Several diagnostic signs of approach to a tipping point have been proposed, including enhanced spatial correlation (i.e., interconnection or dynamic coupling), increase in short-term variability (i.e., flickering), and critical slowing down (i.e., increased autocorrelation) (2022). Some paleoclimate records document flickering and enhanced variance preceding the abrupt onset of Holocene and Bølling warmth (23, 24), but evidence for increased autocorrelation before these transitions has been mixed (20, 24, 25), leading to debate as to whether these transitions are true climate bifurcations (24). No paleoclimate records have yet shown symptoms of dynamic coupling (see supplementary materials for illustrative model). Here, we document the onset of enhanced correlation between North Pacific and North Atlantic climate variability that shortly precedes the most abrupt warming events of the last deglaciation. This new finding supports the idea that these abrupt transitions were unstable tipping points and points to a possible mechanism that could have pushed the system toward a climatic bifurcation.

We developed a high-resolution planktonic foraminiferal δ18O record as a composite from three sediment cores in the Gulf of Alaska (GOA) spanning the last 18,000 years (Fig. 1 and fig. S4). Exceptionally high sedimentation rates provide sample resolutions similar to that of polar ice cores; average sample spacing is ~80 years in the Holocene [0 to 11 thousand years ago (ka)], ~10 years in the deglacial interval (11 to 14.6 ka), and ~35 years during late glacial time (14.7 to 18 ka). Age controls include 39 planktonic foraminiferal radiocarbon dates, and a marine reservoir age constrained by a tephra deposit dated on land (see supplementary materials).

Fig. 1 Map of Gulf of Alaska and NGRIP core locations.

Base map shows average sea surface temperature and sea-ice extent for December, 1985 to 2005 ( Pink arrows depict surface currents, with overturning indicated in the Norwegian seas. Dashed white line across North America depicts the approximate extent of the Laurentide ice sheet (LIS) during the Last Glacial Maximum (LGM) (45). Dashed white line across the North Atlantic depicts the extension of winter sea ice during the LGM (46). Changes in the height and extent of the LIS and sea-ice margins could have affected the atmospheric pathways connecting the GOA and Greenland. Blue arrows across North America represent the schematic differences in the position of the jetstream between the modern and LGM climate states.

The GOA δ18O record integrates local temperature, salinity, and global ice volume signals, whereas the North Greenland Ice Core Project (NGRIP) δ18O record (26) reflects local temperature, moisture transport, and global ice volume. The temperature component of these δ18O records is inversely related (in marine carbonates higher δ18O is colder, whereas in ice cores lower δ18O is colder); therefore, a high negative correlation implies synchronization. The cross correlation was evaluated in moving time windows for the raw records, as well as for ice-volume–corrected, and low-pass– and high-pass–filtered deconvolutions of the data. Although we cannot exclude local salinity effects in the GOA record, alkenone paleotemperature reconstructions from the North Pacific and Bering Sea track planktonic foraminiferal δ18O, indicating that a large component of the North Pacific δ18O changes reflect upper-ocean temperature (16, 27, 28) (fig. S7). Similarly, NGRIP δ18O covaries with marine paleotemperature records from the subpolar North Atlantic (29) (fig. S7).

The GOA δ18O and NGRIP δ18O are positively correlated (temperatures inversely related) from 17.5 to 15.5 ka (equivalent in time to Atlantic Heinrich Stadial 1, HS1) (Fig. 2). The NGRIP δ18O record documents a warming trend from 18 to 16 ka, followed by a cooling from 16 to 15 ka. The GOA record shows high δ18O values (colder and/or saltier) from 18 to 16 ka, and the onset of warming and/or freshening starting around 16 ka. Abundant ice-rafted debris (IRD) in the GOA sediments during this interval documents a sustained input of icebergs, which would tend to decrease the local surface salinity and δ18O through the addition of meltwater. Therefore, the relatively high δ18O during an interval of enhanced IRD delivery indicates that North Pacific temperatures remained cold until 16 ka. IRD deposition peaked in the GOA from 17.5 to 16.5 ka, consistent with other observations in the North Pacific (14, 18) and coincident with the timing of Heinrich Event 1 (H1) in the North Atlantic (30).

Fig. 2 Comparison of GOA and NGRIP climate variability.

(A) The 25-year linearly interpolated GOA δ18O record (rose line, inverted axis) is plotted with the NGRIP δ18O record (dark blue) (26). Radiocarbon age controls plotted as blue diamonds. The purple Xs near the top of the plot indicate times when the subtropical planktonic species O. universa was present in the faunal assemblages, whereas the black dots indicate samples in which they were not present. (B) A 2000-year moving windowed cross correlation between the records is shown for the original GOA age model (black line, inverted axis) and for an alternative age model that is tuned to Greenland (gray line). The pink shading represents the time period of synchronization between records (see Fig. 3 for expanded view), whereas the gray shading reflects times with either no significant correlation or a seesaw-like mode between records. The coarse fraction percent [green shading in (B)] is dominated by IRD from 18.5 to 14.7 ka and virtually disappears near the onset of the Bølling. The peaks in coarse fraction from 14.5 to 13.5 ka are all dominated by volcanic ash in the sediments (purple shading).

Initial GOA warming at 16 ka coincides with the shift from the “Big Dry” to “Big Wet” events in western North America (31), supporting a role for enhanced North Pacific heat and moisture transport to the Great Basin Lakes. This GOA warming also coincides with a pause of Antarctic warming from 16 to 15.5 ka (8) and a slowing of atmospheric CO2 rise from 16 to 14.5 ka (32, 33). North Pacific diatom species in the Bering Sea imply sea-ice retreat at ~16 ka (27), coincident with a pullback of Arctic sea ice (34).

Starting at ~15.5 ka, several hundred years before abrupt warming into the Bølling Interstade at 14.6 ka, NGRIP and GOA δ18O become synchronized (highly negatively correlated; Fig. 2). The negative correlation persists until ~11 ka, with a near one-to-one correlation for the intervals that encompass the abrupt warming transitions into the Bølling and Holocene interstadials. Disappearance of IRD at the onset of the Bølling Interstade in the GOA records indicates rapid retreat of Cordilleran outlet glaciers onto land at 14.8 ka (18). The presence of the subtropical planktonic foraminiferal species Orbulina universa in the faunal assemblages of the GOA record during the Bølling and early Holocene suggests that these intervals were warmer than today and involved the incursion of water masses from the south (35). Such warm events in the GOA coincide with intervals of high meridional overturning in the North Atlantic (6), implying that poleward heat transport increased in both oceans during these times.

After the abrupt warming into the Bølling Interstade (14.7 to 14.1 ka), the GOA and NGRIP δ18O records remain synchronized but with a weaker negative correlation (from –0.5 to –0.8) during stepwise cooling events within the Allerød (14.1 to 12.9 ka) and Younger Dryas (YD; 12.9 to 11.7 ka) intervals, reflecting slight offsets in the timing of these events between records (see supplementary materials). An alternative age model based on tuning the GOA record to the NGRIP record (without violating the ash date) results in only minor changes to the original age model (Fig. 3) and improves the cross correlation between the GOA and NGRIP records throughout this interval (from –0.65 to –0.90; Fig. 2). Variations in marine reservoir of up to 200 years are plausible and within the range of modern spatial variability, but the combination of the ash date on land and the close correlation between GOA and Greenland preclude larger surface-ocean reservoir ages in this interval (Fig. 3).

Fig. 3 Age model tuning.

(A and B) Comparison of the raw GOA data (pink, inverted axis) with the 20-year NGRIP record (dark blue) (26) on the original age model (A) and on an age model that is tuned to the NGRIP record (B). The blue diamonds in (A) represent radiocarbon age controls with ±2σ uncertainty, and the black diamond indicates the MEd tephra layer (see supplementary materials). The age controls for the tuned record are shown in (B), with the tie points to the NGRIP record shown as black stars for core EW0408-66JC and a green star for core EW0408-26JC. The blue diamonds are the retained radiocarbon age controls from the original age model, and the black diamond is the MEd tephra layer. (C) Changes in marine reservoir age that result from the difference in the original and tuned age models for core EW0408-66JC (black) and EW0408-26JC (green). These differences are within the combined uncertainty of the two age models. The gray bars encompass the major climate reversals in the NGRIP record: YD: Younger Dryas, IACP: Inter-Allerød Cold Period, OD: Older Dryas.

GOA and NGRIP again become highly synchronized (r = –0.90) just before abrupt warming into the Holocene. After this warming, the windowed correlation between GOA and NGRIP changes sign at ~10 ka, suggesting a return to Pacific-Atlantic seesaw or decoupled behavior in an interglacial state similar to that of late-glacial time. Within the Holocene, asynchronous behavior includes cooling of GOA from 10 to 8 ka, followed by gradual warming; in opposition to this pattern, cooling in NGRIP is initiated after 8 ka and persists throughout the Holocene. Ice-volume–corrected, low-pass–filtered δ18O records show a more pronounced seesaw pattern (high positive correlation) in the Holocene and HS1 than the raw or high-pass records (Fig. 4). This suggests that the long-term (>2000 years) trends are inversely related during these times, but the short-term variability is either not sufficiently resolved or decoupled between basins during the Holocene and HS1.

Fig. 4 Low-pass– and high-pass–filtered data.

(A and B) Evaluation of the cross correlation between the ice-volume–corrected, low-pass–filtered [2000-year moving average (A)] and high-pass–filtered records [residuals from 2000-year moving average (B)] of NGRIP (dark blue) (26) and GOA on the tuned age model (rose, inverted axis). The 2000-year moving window cross correlation between the GOA and NGRIP records are shown in black for the low-pass– (A) and high-pass–filtered (B) records. Dashed gray lines in (A) reflect one-to-one positive and negative correlations. (C) Panel shows the variance for the high-pass and low-pass records. A 200-year windowed variance is shown for NGRIP (dark green) and GOA (dark blue) for the high-pass records, and a 2000-year windowed variance is shown for NGRIP (light green) and GOA (light blue). The transitions into the Bølling and Holocene are clearly identifiable in the peaks in the high-pass variance and the abrupt shifts in the δ18O records at 14.7 and 11.7 ka.

Variance increases within the GOA and NGRIP records before the abrupt warming transitions into the Bølling and Holocene (Fig. 4). High-frequency variance in the GOA peaks from 15 to 11.5 ka, coincident with the interval of synchronization, whereas in NGRIP high-frequency variance peaks slightly earlier. Both records show greater short-term variance during abrupt warming into the Bølling relative to the later warming into the Holocene, suggesting that earlier warming, beginning from relatively extreme glacial conditions, was the more abrupt climatic event in these two locations.

Thus, we find evidence for intervals of both synchronous and asynchronous climate variability between the GOA and NGRIP regions. Synchronization and an increase in variance preceded and extended through abrupt warming events, consistent with theory that highly connected systems are more susceptible to unstable tipping points (21) (see supplementary materials). Possible mechanisms to regulate the modes of connectivity between regions could be related to (i) ice elevation in North America, (ii) the opening of Bering Strait, or (iii) routing of fresh water in the North Atlantic or Southern Ocean. During the Last Glacial Maximum (LGM), topographic steering of the jetstream around an expanded North American ice sheet and sea-ice margin would have reduced the direct atmospheric connection between the GOA and Greenland (36, 37). A gradual fall in ice height following H1 could have caused an abrupt reorganization of atmospheric circulation that set up a stronger meridional flow in the North Atlantic (38), thereby strengthening the interocean atmospheric coupling (Fig. 1). Linkages between the North Pacific and North Atlantic may also depend on whether the Bering Strait is open or closed (10, 13, 39), which could signal a role for the incursion of low-salinity waters from the Arctic in the reversion to interocean seesaw during the early Holocene (40). Alternatively, models suggest that freshwater forcing near West Antarctica triggers synchronous warming in the North Pacific and North Atlantic (41), whereas meltwater inputs near Greenland create antiphased responses in the North Atlantic and North Pacific (10, 11).

The last glacial termination demonstrates several symptoms of threshold behavior, including greater spatial organization through synchronization of the North Pacific and Greenland/North Atlantic variability (this study), elevated variance [this study and (23, 24)], and enhanced autocorrelation of some climatic indicators just before the abrupt Bølling and Holocene transitions (20, 25). The new highly resolved evidence we present here for interocean dynamic coupling between the NGRIP record in the Atlantic sector and GOA records in the Pacific Sector suggests that synchronization of poleward heat transport in both oceans is an important catalyst for abrupt warming transitions within the Northern Hemisphere. Although a tipping point may be crossed in an instant, large-scale climate systems that include ice sheets or deep ocean circulation may have substantial inertia, such that the full response may play out dynamically over an extended period of time, constituting a “tipping interval.”

In the modern climate system, models suggest a tight coupling of short-term (<200 years) ocean-atmosphere interactions between the North Pacific and North Atlantic and indicate the potential for amplification of decadal-scale variability through interbasin resonance (42, 43). Before the 1970s, variability in poleward heat fluxes and storm tracks in the North Pacific and North Atlantic regions were uncorrelated; more recently, highly correlated behavior has emerged (44). Our study documents that the development of such teleconnected variability between these regions is a fundamentally important phenomenon associated with rapid warming, suggesting that such properties may be high-priority targets for detailed monitoring in the future.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S11

Tables S1 and S2

References (4766)

References and Notes

  1. Acknowledgments: We thank B. Jensen and D. Froese for the tephra analyses; J. Southon for assistance with radiocarbon samples; A. Ross, J. Padman, and J. McKay of the College of Earth, Ocean and Atmospheric Sciences Stable Isotope Lab; and five anonymous reviewers. This work was supported by NSF grants AGS-0602395 (Project PALEOVAR) and OCE-1204204 to A.C.M., and an NSF graduate research fellowship to S.K.P. The data can be found in the supplementary online materials and at the National Oceanic and Atmospheric Administration Paleoclimate Database.
View Abstract

Stay Connected to Science

Navigate This Article