High-Resolution Greenland Ice Core Data Show Abrupt Climate Change Happens in Few Years

See allHide authors and affiliations

Science  01 Aug 2008:
Vol. 321, Issue 5889, pp. 680-684
DOI: 10.1126/science.1157707


The last two abrupt warmings at the onset of our present warm interglacial period, interrupted by the Younger Dryas cooling event, were investigated at high temporal resolution from the North Greenland Ice Core Project ice core. The deuterium excess, a proxy of Greenland precipitation moisture source, switched mode within 1 to 3 years over these transitions and initiated a more gradual change (over 50 years) of the Greenland air temperature, as recorded by stable water isotopes. The onsets of both abrupt Greenland warmings were slightly preceded by decreasing Greenland dust deposition, reflecting the wetting of Asian deserts. A northern shift of the Intertropical Convergence Zone could be the trigger of these abrupt shifts of Northern Hemisphere atmospheric circulation, resulting in changes of 2 to 4 kelvin in Greenland moisture source temperature from one year to the next.

Ice core records from Greenland have been instrumental in investigating past abrupt climate change. As compared with other sedimentary records, the ice core records have unparalleled temporal resolution and continuity (13). The newest Greenland ice core, from the North Greenland Ice Core Project (NGRIP), has been measured at very high resolution for water isotope ratios, dust, and impurity concentrations. This allows researchers for the first time to follow the ice core proxies of Greenland temperature, accumulation, moisture origin, and aerosol deposition at subannual resolution over the very abrupt climate changes in the period from 15.5 to 11.0 thousand years ago (ka) (measured from 2000 AD throughout this study).

In the Northern Hemisphere, the last glacial period ended in a climatic oscillation composed of two abrupt warmings interrupted by one cooling event (36). The temperature changed rapidly from glacial to mild conditions in the Bølling and Allerød periods and then returned to glacial values in the Younger Dryas period before the onset of the present warm interglacial, the Holocene (Fig. 1, and see table S1 for classification of climate periods). The shape and duration of the abrupt climate change at the termination of the last glacial have previously been constrained by Greenland ice core records from DYE-3 (4, 7), Greenland Ice Core Project (GRIP) (8) and Greenland Ice Sheet Project 2 (GISP2) (3, 6, 9), but sampling of these cores did not typically achieve a resolution sufficient to resolve annual layers. Because of new continuous flow analysis (CFA) systems (1012), impurity and chemical records of the recent NGRIP ice core (1) have been obtained at subannual resolution, which allows for the multiple-proxy identification of annual-layer thickness and the construction of a new Greenland time scale, the Greenland Ice Core Chronology 2005 (GICC05) (2). Complementary highly detailed stable water isotope profiles (δ18O and δD) have been measured on the NGRIP ice core covering the period from 15.5 to 11.0 ka at 2.5-to-5.0-cm resolution, corresponding to one to three samples per year. They were compared with the concentrations of insoluble dust, soluble sodium (Na+), and calcium (Ca2+), each measured with CFA at subannual resolution (10, 12, 13) (Fig. 2) and, when available, with the highest-resolution data from GRIP, GISP2, and DYE-3 ice cores on the GICC05 time scale.

Fig. 1.

(Left) Location of drill sites on the Greenland Ice Sheet: DYE-3 (65.15°N, 43.82°W), GRIP (72.59°N, 37.64°W), GISP2 (72.58°N, 38.46°W), and NGRIP (75.10°N, 42.32°W). (Right) The NGRIP stable water isotope profile (δ18O) on the GICC05 time scale (1, 2). The zone studied (11.0 to 15.5 ka) is marked with gray shading.

Fig. 2.

Multiple-parameter records from the NGRIP ice core 11.0 to 15.5 ka. (A) d (red) and δ18O (dark blue) at 20-year resolution over the entire period and details of the transition zones: (B) from GS-1 into the Holocene at 11.7 ka, (C) from GI-1a into GS-1 at 12.9 ka, and (D) from GS-2 into GI-1e at 14.7 ka. [Left part of (B) to (D)] NGRIP records of d (red), and δ18O (dark blue) and logarithmic plots of dust content (yellow), calcium concentration ([Ca2+], light blue), sodium concentration ([Na+], purple), and annual layer thickness (λ, green) at annual resolution. Bold lines show the fitted ramp functions; gray vertical bars represent the 95% (2σ) confidence intervals of the ramp point locations. [Right part of (B) to (D)] Bars representing the locations of the fitted ramp functions for the NGRIP records shown to the left and for the corresponding results obtained using DYE-3, GRIP, and GISP2 data, where these are available at sufficient resolution (see list of records below the figure). See Table 1, SOM methods, and table S2 (15) for additional information on ramp fitting.

Across the warming transitions, the records exhibit clear shifts between two climate states. We characterize a shift to be significant if the mean values of the climate states on each side of the shift differ by more than the statistical standard error of the noise of a 150-year period of these climate states. A simple but objective approach to finding the best timing of the transition is to characterize the shift observed in each proxy as a “ramp”: a linear change from one stable state to another. We applied a ramp-fitting method (14) to determine the timing of the transitions. The method entails using weighted least-squares regression to determine the ramp location and a bootstrap simulation to estimate the uncertainty of the results [transition times are listed in Table 1; see supporting online material (SOM) methods and table S2 for more detailed information on the method and the ramp fit values and uncertainties (15)]. Data and fitted ramps are shown in Figs. 2 and 3. For annual layer thickness (λ), concentrations of dust, Ca2+, and Na+, logarithmic scales were used because these proxies are approximately log-normally distributed.

Fig. 3.

(A to C) d (red) at measured resolution and the fitted ramp curves across the same transitions as shown in Fig. 2, B to D. The mean values over the 150 years before and after the transitions are shown as bold red lines. (Left) To visualize the shifts, the areas within 1 SD from the mean values are shaded light gray. Where the data values are more than 1 SD away from the mean, the zone is colored light blue. (Right) Histograms (gray) of the distribution of the d values in the states before and after the rapid shifts. The mean values are shown as bold red lines. Because of the abruptness of the d shifts, the ramp-fitting method produces results with very small uncertainties. Therefore, we suggest that the d record be used for defining transition points between different climatic episodes, especially for the transitions studied here (15.5 to 11.0 ka) but possibly also for those found in the older part of the NGRIP record (123.0 to 15.5 ka).

Table 1.

Ramp-fitted transition times. Ramp-fitting results (14) for δ18O, d, dust content, Ca2+, and λ over the three transitions from Greenland Stadial 1 (GS-1) into the Holocene at 11.7 ka, from Greenland Interstadial 1a (GI-1a) into GS-1 at 12.9 ka, and from GS-2 into GI-1e at 14.7 ka. The GICC05 time scale (2), produced by multiple-proxy identification of annual layers using NGRIP impurity records, provides the ages. The timing and standard error of the ramp points are listed at the onset and termination of the transitions. Times are in years before 2000 AD.

View this table:

The δ18O record is a proxy for past air temperature at the ice core site (16, 17). Although the magnitude of Greenland δ18O changes can be influenced by changing site and source temperatures and by snowfall seasonality (16, 18, 19), the timing of δ18O changes is dominated by the changing site temperature (18). The δ18O warming transition at 14.7 ka was the most rapid and occurred within a remarkable 3 years, whereas the warming transition at 11.7 ka lasted 60 years; both correspond to a warming of more than 10 K (6, 20). δ18O records from the GRIP (9, 21), GISP2 (9), and DYE-3 (7, 17) ice cores across the 11.7 ka transition show a similar duration. The δ18O cooling transition at 12.9 ka lasted more than two centuries, much longer than the warming transitions, and does not meet the above criteria for being described as a ramp shift.

Annual layer thickness λ (estimated independently of δ18O by annual layer counting) increased by 40% during the two warmings over 3 and 40 years, respectively. During the cooling, λ decreased by 33% over a period of 152 years. When corrected for strain, λ represents the annual precipitation rate, which is linked to site temperature and to synoptic weather patterns. Both Greenland site temperature and accumulation are expected to be strongly related to the extent of the northern sea ice (22).

The most abrupt transitions are those of the deuterium excess d = δD – 8δ18O, a second-order isotopic parameter that contains information on fractionation effects caused by the evaporation of source water (16, 18, 23). The excess is considered to be mainly a proxy of past ocean surface temperatures at the moisture-source region (16, 18, 23). Our dust record, not showing a similar abrupt transition as the excess, rules out the alternative explanation, that a rapid shift in d was linked with changes in cloud condensation nuclei and kinetic fractionation taking place during cloud ice-crystal formation (24, 25). The excess record shows a 2 to 3 per mil (‰) decrease in d during the warmings, corresponding to a cooling by 2 to 4 K of the marine moisture-source region (16, 18) over 1 to 3 years, and a 2 to 3‰ increase during the cooling transition. Figure 3 presents the rapid changes of d on a more highly resolved time axis in order to clearly show the rapid change between the climate states. The moisture-source evaporation conditions can change either because of a shift in atmospheric circulation, resulting in relocation of the moisture source, or because of changing sea surface temperature, humidity, or wind conditions at a stationary moisture source. The extremely rapid shifts in both warmings and coolings rule out an explanation that is purely in terms of sea-ice extent, because the northern sea ice extended far south during the final phase of the cold stadials and is not expected to have broken down in just 1 to 3 years (26, 27). Finally, if d was reflecting only changes in weather pattern trajectories with respect to the NGRIP observation point, then it would be expected to detect simultaneous changes in other parameters, such as dust, Ca2+, and λ, which is not the case. Thus, we interpret the rapid shifts in d to be more likely a consequence of changed source regions of the water vapor reaching Greenland. This points to a reorganization of atmospheric circulation from one year to the next.

The concentrations of insoluble dust and Ca2+ reflect both source strength and transport conditions from terrestrial sources, which for Greenland are the low-latitude Asian deserts (10, 28). At the two warming events, the concentrations of dust and Ca2+decreased by a factor of 5 to 7 within four decades, slightly preceding the d shift by 10 ± 5 years. In contrast, during the cooling event, the dust and Ca2+concentrations increased by a factor of 5 over a period of more than a century, slightly lagging the d shift by 20 ± 10 years.

The concentration of Na+, which is mainly a marine sea-salt indicator, shows only moderate changes at the transitions as compared with its interannual variability. Changes in sea-ice extent are expected to influence sea-salt export to Greenland through changing the distance to open water and altering sea-salt aerosol contributions from sea-ice and open-water sources. However, meteorological conditions play an important role in modulating the sea-salt uplift, transport, and deposition on the Greenlandice sheet (29). Although insufficient quantitative understanding of the processes involved is available, the lack of a fast response of the marine sea-salt proxy may be seen as a gradual change in sea ice or as a combination of changes in sea ice and meteorology, compensating for each other to some extent.

This high-resolution study shows a previously unkown sequence of events and gives insights into both the onset and evolution of a rapid climate shift. Our records demonstrate that the last two major warming events followed the same general pattern. During Greenland cold phases, the thermohaline circulation (THC) was reduced, northern sea ice extended far south, and the Intertropical Convergence Zone (ITCZ) was shifted southward, resulting in dry conditions at the low-latitude dust-source regions (22, 30, 31). Meanwhile, southern high latitudes and tropical oceans accumulated heat and underwent gradual warming as reflected in the bipolar seesaw pattern (32, 33), because of a reduction in the North Atlantic overturning circulation. We suggest that this Southern Hemisphere/tropical warming induced first a northward shift of the ITCZ and, when a threshold was reached, an abrupt intensification of the Pacific monsoon. The wetter conditions at the Asian dust-source areas then caused decreased uplift and increased washout of atmospheric dust, leading to the first sign of change in Greenland: decreasing dust and Ca2+concentrations. This reorganization of the tropical atmospheric circulation was followed by a complete reorganization of the mid- to high-latitude atmospheric circulation almost from one year to the next, as identified by the 1-to-3-year transitions in d. Sea ice then started retreating in the North Atlantic, associated with increased advection of atmospheric heat and moisture, as indicated by Greenland changes in δ18O and λ (22, 34).

Both abrupt warming events of the last termination are characterized by this sequence of events, even though they occurred at different stages of deglaciation. The 14.7-ka event followed Heinrich event H1 at a time when the ice sheets in the north were still extensive, whereas the north was more deglaciated at 11.7 ka (35), reducing the amount of ice discharge available to change the density of North Atlantic ocean waters and thereby the THC before the warming onset (35). The NGRIP ice core has also revealed that the very first interstadial of the last glacial cycle occurred at the inception of the glacial period 110 ka, before the ice sheets were fully developed and the climate system had cooled to full glacial conditions (1). The detailed sequence of events obtained here for the most recent warming events suggests that the classical bipolar seesaw concept (32) involving the ocean THC reorganization must include the role of abrupt atmospheric circulation changes from the tropics to the high northern latitudes in the onset of abrupt warmings seen in the North Atlantic region.

The cooling at 12.9 ka is characterized by relatively longer transition times for all parameters except for d, and the sequence of events is notably different. In this case, changes in d and δ18O precede the dust and Ca2+reactions. The centennial scale change in δ18O follows anterior gradual cooling during the Allerød period, probably including gradual buildup of sea ice. Given the generally slow nature of the coolings, the persistent rapid switch of the atmospheric circulation as recorded by the excess is even more surprising and confirms the potential for extremely abrupt reorganizations of the Arctic atmospheric circulation, whether going from cold to warm or vice versa. The lag and longer duration of the dust and Ca2+responses may be due to the inertia of land surfaces drying out and vegetation dying off in the dust-source regions before large fluxes of dust could be reestablished.

The high-resolution records from the NGRIP ice core reveal that polar atmospheric circulation can shift in 1 to 3 years, resulting in decadal- to centennial-scale changes from cold stadials to warm interstadials/interglacials associated with large Greenland temperature changes of 10 K (6, 20). Neither the magnitude of such shifts nor their abruptness is currently captured by state-of-the-art climate models. We propose a series of events, beginning in the lower latitudes and leading to changes in the ocean and atmosphere, that reveals for the first time the anatomy of abrupt climate change. Although no large shifts in d can be identified over the course of the Holocene in the Greenland ice cores (36), past warming events now documented at subannual resolution offer important benchmarks with which to test climate models. If we are to be confident in the ability of those models to accurately predict the impacts of future abrupt change, their ability to match what we see in the past is crucial.

Supporting Online Material

Materials and Methods

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article