Centennial-Scale Holocene Climate Variability Revealed by a High-Resolution Speleothem δ18O Record from SW Ireland

See allHide authors and affiliations

Science  09 Nov 2001:
Vol. 294, Issue 5545, pp. 1328-1331
DOI: 10.1126/science.1063678

This article has a correction. Please see:


Evaluating the significance of Holocene submillennial δ18O variability in the Greenland ice cores is crucial for understanding how natural climate oscillations may modulate future anthropogenic warming. A high-resolution oxygen isotope record from a speleothem in southwestern Ireland provides evidence for centennial-scale δ18O variations that correlate with subtle δ18O changes in the Greenland ice cores, indicating regionally coherent variability in the early Holocene. Evidence for previously undetected early Holocene cooling events is presented, but mid- to late-Holocene ice rafting in the North Atlantic appears to have had little impact on δ18O at this ocean margin site.

It is widely accepted that climate variability on time scales of 103 to 105 years is driven primarily by orbital, or so-called Milankovitch, forcing. Less well understood is the cause of the centennial- to millennial-scale variability that characterizes the δ18O records of both the glacial and interglacial intervals of the GRIP and GISP2 ice cores (1, 2), yet this higher frequency variability may be important for predicting future climate change. Unlike the last interglacial (isotope stage 5), the Holocene [the last 11,700 years or 11.7 thousand years before the present (ky B.P.)] appeared to be anomalously stable because δ18O in the GRIP and GISP2 ice cores seemed to be relatively constant (1, 2). More recently, millennial-scale climate variability has been detected in several Holocene climate proxy records, but there is little consensus about the precise timing, amplitude, or cause of these fluctuations (3–10). An emerging paradigm is that sub-Milankovitch climate variability is driven by a weak internal quasi-periodic (1500 ± 500 year) forcing of unknown origin that operates irrespective of whether the system is in a glacial or an interglacial mode (5, 9, 10). In the context of concerns about the impact of anthropogenic greenhouse gases, it is important to establish whether these relatively low-frequency (∼1500 year) events or higher frequency oscillations might determine the natural trends in global mean temperatures over the next few centuries. Therefore, a key question is the extent to which subtle higher frequency (century-scale) δ18O variations in the Holocene sections of the Greenland ice cores reflect regional climatic signals rather than local effects or noise. Until now, that has been difficult to test, due to a paucity of sufficiently high resolution palaeoclimatic records. But here we present a newly found high-resolution O isotope time series for a well-dated stalagmite (CC3) from Crag cave in southwestern Ireland (Fig. 1), which shows that these subtle features are regional and not local signals.

Figure 1

Map of the North Atlantic region showing the locations of GRIP and GISP2 ice cores, cores VM28-14, VM29-191, 28-03 and Crag Cave in southwestern Ireland.

Approximately 1640 laser ablation δ18O measurements (11) were carried out along the growth axis of the 465-mm-long stalagmite, resulting in a high-resolution Holocene δ18O record (12–14). Chronological control is provided by 13 TIMS U-series dates [Web table 1 (13)]. Before 5 ky B.P., the resolution is 2 to 20 times better than that of the published (2-m segment) δ18O data for the GRIP and GISP2 ice cores and is about a factor of two worse than that of the ice cores since 5.3 ky B.P. The average resolution is approximately an order of magnitude better than in the North Atlantic cores that record evidence for quasi-periodic (1475 ± 500 year) ice rafting during the Holocene (5, 9,10). Such events should, therefore, be discernible in the new data set if they had a large impact on the δ18O of precipitation and/or mean annual air temperatures at this ocean margin site.

The new data are plotted against time (calendar years B.P.) in Fig. 2. δ18O varies from –11.65 to –0.82 per mil (‰) [Vienna Pee Dee Belemnite (VPDB)], but typically varies by ±1.75‰ around a mean value of –3.26‰ (VPDB). The approximate timing of historic climate variability (8) is also shown (e.g., the Little Ice Age, Medieval Warm Period, Dark Ages Cold Period, Roman Warm Period). Colder periods (e.g., Little Ice Age and Dark Ages Cold Period) appear to be associated with lower δ18O (Fig. 2A). Because δ18O may be modified by temporal changes in the oceanic moisture source and/or storm track trajectories, it is not possible to calculate temperature changes precisely (15). On the basis of present-day spatial δ18O-temperature relations, the magnitude of δ18O variability around the mean is probably too large to ascribe to changes in air temperature alone. Thus, temperature-driven changes in speleothem δ18O appear to have been accentuated by synchronous changes in the δ18O of water vapor supplied to the site, resulting in a strong overall apparent relation between temperature and speleothem δ18O.

Figure 2

(A) CC3 δ18O record for the period since 10 ky B.P. (upper curve) compared with the GISP2 record (lower curve). Individual U-Th dates and their ± 1σ error bars are shown in the upper part of the diagram. Open circles represent the published low-resolution conventionally sampled and analyzed δ18O data (15), demonstrating the accuracy of the new laser-ablation data. The 8200 year event was not detected in the conventional δ18O data because of coarse sampling resolution. The GISP2 curve is the 2-m data set with a resolution of approximately 24 years per δ18O analysis at 10 ky B.P., increasing to about 7 years per analysis at 0.2 ky B.P. Also shown is the timing of the Bond et al. (5) ice-rafting events (1 through 6). RWP, Roman Warm Period; DACP, Dark Ages Cold Period; MWP, Medieval Warm Period; LIA, Little Ice Age. Textural and mineralogical descriptions of stalagmite CC3 are given in given in (15). (B) Timing of events a through h and the 8200 year event in the GISP2 record (y axis) compared with that in CC3 (x axis). Event timing is defined as δ18O minima or maxima. Error bars shown at ±2.5% for the CC3 chronology and 2.0% for the GISP2 chronology are estimates of dating uncertainties at the 2σ level. The timing of events in both records is identical within the dating uncertainties. Age differences for events range from essentially zero (event a) to 2.6% (event h).

Also shown in Fig. 2A (arrows labeled 1 through 6) is the approximate timing of abrupt cooling associated with North Atlantic ice-rafting events (5, 9, 10). Only event 5 has a clear expression in either the CC3 record (upper curve,Fig. 2A) or in the GISP2 ice core data (lower curve, Fig. 2A). Event 5, the so-called “8200 year cooling event,” is defined in CC3 by eight data points centered on 8.32 ± 0.12 ky B.P., and it exhibits a large (∼8‰) decrease in δ18O. Event 6 (∼9.45 ky B.P.) (5) does not have a clear expression in the CC3 record, although it occurs just before several high-amplitude (>4‰) O isotope shifts in the CC3 speleothem record (Fig. 2A). Thus, several shifts to lower δ18O occur between approximately 9.4 and 8.8 ky B.P., some of which may be recorded in the lower resolution GISP2 record (arrows, Fig. 3). Detection of these events in sediments from the North Atlantic (5,10) is precluded by the relatively low resolution of the latter records.

Figure 3

Close up of period between 9500 and 8000 years showing the new laser ablation CC3 record (upper curve) and the GISP2 curve. The vertical shaded line at 8470 years B.P. labeled LLB denotes the approximate timing of the Laurentide Lakes Burst event (24). U-series ages and their 1σ error bars are shown in the upper part of the diagram. Arrows denote timing of δ18O fluctuations in the new high-resolution record, interpreted as meltwater release events.

The 8200 year event, recognized as the only major Holocene δ18O event in the GRIP and GISP2 ice cores (1, 2) has been interpreted to reflect a cooling of 7° ± 3°C (16). Cooling events dated between 8.4 to 8.0 ky B.P. have also been documented in Europe (17-21), North America (6, 22), Northern Canada (23, 24), the Asian and North African moonsonal domains (25,26), and the equatorial Cariaco Basin (27). The timing of the 8200 year event in speleothem CC3 is within the dating uncertainties of the GISP2 core, estimated at ±1 to 2% for this part of the Holocene (28). Thus, the maximum amplitude occurs at 8.32 ± 0.12 ky B.P. compared with 8.215 ky B.P. in GISP2 and is coeval with faunal evidence for cooling at 8.30 ± 0.06 ky B.P. in core 28-03 from the Norwegian Channel (17). Shifts to lower δ18O between 9.4 and 8.8 ky B.P., detected because of the very high resolution of this part of the record (1 to 3 years per analysis), are interpreted as precursor meltwater release events (29) to the catastrophic 8200 year event. The extent to which these centennial-scale changes in δ18O may be linked to ice-rafting events in the North Atlantic is impossible to assess at present because of the relatively coarse resolution of the latter records (5, 10).

The amplitude of the shift to lower δ18O at 8.32 ky B.P. (∼8‰) is too large to ascribe solely to a reduction in mean annual air temperature (29). Instead, we attribute it predominantly to freshening of the surface of the adjacent North Atlantic ocean by isotopically depleted meltwater. Evidence from planktonic foraminifera for a coeval shift in δ18O in North Atlantic surface waters is equivocal (5, 9,10, 30), possibly because enrichments in δ18O were offset by the lower δ18O values of waters derived from melting drift-ice or because of a change in depth habitat in response to surface cooling (5,10). The chronology of our new record is consistent with suggestions (19, 24) that the 8200 year event was triggered by a catastrophic release of meltwaters by sudden draining of large ice-dammed lakes on the margins of the Laurentide ice sheet in northeastern Canada (31), dated to 8.47 ± 0.30 ky B.P. (24). The absence of a clear shift in δ18O in the speleothem data during the other North Atlantic ice-rafting events at about 5.9, 4.3, 2.8, and 1.4 ky B.P. (5), despite the high resolution of the new data (7 to 18 years per analysis), is important. It suggests that unlike the 8200 year event, the later Holocene ice rafting events (5,10) failed to trigger large changes in δ18O and, by implication, may not have established a detectable meltwater cap on the mid-latitude North Atlantic (29).

Although precise comparisons between the GISP2 and CC3 records are hampered by dating uncertainties, the low amplitude of the variations, and differences in sampling resolution, the subtle pattern of δ18O variability between approximately 8.4 and 4.5 ky B.P. is strikingly similar in both records (Fig. 2A). Excluding the 8200 year event, which is clearly common to both records, approximately eight peaks and troughs may be correlated within the combined dating uncertainties in this time interval (labeled a through h, Fig. 2A).Figure 2B illustrates that the timing of each event coincides in both records within the dating uncertainties. Cooling events b, d, f, and h (Fig. 2A) may reflect weaker North Atlantic thermohaline overturning at 7.73, 7.01, 5.21, and 4.2 ky B.P., respectively (8). We interpret these covariations to mean that the subtle δ18O variability in the early and mid-Holocene sections of the Greenland ice cores reflects coherent climate-driven in-phase changes in the δ18O of precipitation over a large region around the eastern North Atlantic margin. Differences in the shape of individual peaks and troughs are attributed to second-order effects such as temporal changes in the vapor source, trajectory, and rainout efficiency of clouds supplying moisture to these sites. The GISP2 and CC3 data appear to be decoupled in the latter part of the Holocene, but the CC3 data exhibit variations that are broadly consistent with a Medieval Warm Period (MWP) at ∼1000 ± 200 years ago and a two-stage Little Ice Age (LIA), as reconstructed by inverse modeling of temperature profiles in the Greenland Ice Sheet (32).

The coherent δ18O variations in CC3 and GISP2 (events a through h, Fig. 2A) indicate that many of the subtle multicentury δ18O variations in the Greenland ice cores reflect regional North Atlantic margin climate signals rather than local effects. Spectral analysis of the data confirms the importance of multicentury variability with peaks centerd on 625, 169, and 78 years. Taken together with historically documented variability (e.g., LIA and MWP in Fig. 2A) we argue that high-frequency (centennial-scale) oscillations, perhaps reflecting North Atlantic thermohaline circulation changes (8, 33) may have a more detectable impact on the climates of North Atlantic ocean-margin sites than the lower frequency events exquisitely recorded by ice-rafting proxies.


View Abstract

Navigate This Article