Structure of the 8200-Year Cold Event Revealed by a Speleothem Trace Element Record

See allHide authors and affiliations

Science  21 Jun 2002:
Vol. 296, Issue 5576, pp. 2203-2206
DOI: 10.1126/science.1071776


Abrupt first-order shifts in strontium and phosphorus concentrations in stalagmite calcite deposited in western Ireland during the 8200-year event (the major cooling episode 8200 years before the present) are interpreted as responses to a drier climate lasting about 37 years. Both shifts are centered on 8330 ± 80 years before the present, coinciding with a large oxygen isotope anomaly and a change in the calcite petrography. In this very high resolution (monthly) record, antipathetic second-order oscillations in phosphorus and strontium reveal decreased growth rates and increased rainfall seasonality. Growth rate variations within the event reveal a two-pronged structure consistent with recent model simulations.

The 8200-year event is widely regarded as the strongest Holocene cooling episode, with clear expressions in Greenland (1, 2), the North Atlantic (3), Europe (4–11), North America (12–14), North Africa (15), and the Venezuelan Cariaco Basin (16). Decreased snow accumulation rates, lower levels of atmospheric methane, and increased atmospheric dust and sea-salt loadings indicate widespread dry conditions (17,18). Explanations usually involve a perturbation of the North Atlantic thermohaline circulation (THC) by increased freshwater inputs asso- ciated with the decay of the Laurentide ice sheet (6, 19). A high-resolution global circulation model (GCM) indicates that a freshwater pulse of a magnitude similar to that associated with the catastrophic drainage of the large proglacial lakes Agassiz and Ojibway could have produced the 8200-year event, including a very brief warming episode within the event (20).

Crucial unresolved questions are whether the cooling occurred as a single event or as a more complex multipulse episode involving partial recovery, and whether the cooling resulted in enhanced seasonality in mid-latitude temperate regions. The latter is important because it may affect the detection of the event by some proxies such as tree rings. The coarse resolution of the available climate records has hampered investigation of these issues. Here, we present a very high resolution trace element record for the 8200-year event in stalagmite CC3, from Crag Cave in southwestern Ireland (21), showing that the event was characterized by rapid deterioration, a brief intra-event amelioration, enhanced seasonality, and an abrupt termination.

Trace elements in carbonate speleothems (stalagmites and stalactites) can provide qualitative proxies for paleo-recharge in well-characterized karst systems (22–24). In temperate regions, solute acquisition by karst waters depends on water residence time in the soil and epikarst, and trace element ratios such as Mg/Ca and Sr/Ca tend to increase during drier periods when residence times are longer (25–27). Additionally, phosphorus abundances in speleothems frequently exhibit cyclical variations interpreted as annual flushing events from the soil, allowing detailed reconstruction of speleothem growth rates from the spacing between successive annual P peaks (24); such data provide an independent proxy for paleo-recharge (25).

Approximately 1018 points were analyzed for Mg, Sr, P, Ca, H, and Si at an average spatial resolution of 7.3 μm (28) along a 7762-μm track, chosen to traverse a previously documented O isotope shift interpreted to reflect the 8200-year event (10, 11). A layer of clear, inclusion-poor calcite 2.5 mm thick occurs within the traverse, coinciding with a sharp increase in Sr and a decrease in P (Fig. 1). Several other clear calcite layers exist within CC3, but systematic sampling reveals no elevated Sr concentrations associated with them, hence the trace element anomalies are not simply the result of variable calcite petrography. Over this “first-order” trace element anomaly, Sr concentrations increase by 89%, from a baseline average of 43 ppm to 81 ppm, accompanied by a 39% decrease in P from an average of 165 ppm to 101 ppm. For the period 10,000 to 5000 years before the present (B.P.), the mean Sr concentration determined by inductively coupled plasma mass spectrometry on drilled calcite is 45 ± 16 ppm, similar to baseline Sr concentrations of 43 ± 7 ppm directly adjacent to the trace element anomaly. On the basis of the published U-Th chronology (10, 11), the first-order trace element anomaly commences abruptly at about 8350 ± 80 years B.P., is centered on 8330 ± 80 years B.P., and ends at 8310 ± 80 years B.P. (Fig. 1).

Figure 1

(A) Stalagmite CC3 laser ablation δ18O record for the period 9000 to 7500 years (9.00 to 7.50 ky) B.P. (10, 11). The bracketed area is the interval chosen for ion microprobe analysis. PDB refers to the Pee Dee belemnite standard. (B) Photomicrograph of CC3 after analysis, showing principal track. (C) CC3 Sr concentration record from 8370 ± 80 years B.P to 8260 ± 80 years B.P. Average Sr values increase markedly from 43 ppm to 81 ppm. (D) CC3 P concentration record. Average P values decrease from 165 ppm to 101 ppm. Both first-order trace element excursions are coincident with a lens of clear calcite. These anomalies are synchronous with a previously documented O isotope excursion within CC3, interpreted as reflecting the 8200-year event (10,11). Dashed lines in (C) and (D) represent the mean trace element values for the baseline outside of the event. Letters a, b, and c correspond to intervals expanded in Fig. 3.

The first-order trace element anomaly (Fig. 1) is synchronous, within the 2σ dating uncertainties, with abrupt shifts in climatic proxies elsewhere indicating the prevalence of cold, dry conditions around 8200 years B.P. (1, 2, 17, 18). The anomaly also coincides exactly with the substantial 8 per mil decrease in δ18O in the same stalagmite, attributed to climatic cooling and a change in the source of vapor supplying rainfall to southwestern Ireland (10, 11). The abrupt increase in Sr concentrations centered on 8330 ± 80 years B.P. is interpreted as reflecting an increase in the residence time of water in the glacial till and carbonate bedrock overlying the cave as a result of a drier climate. Although the predominant control on Sr in speleothem calcite is the Sr content of the water (23), Sr incorporation also tends to increase with increasing calcite growth rate (29,30). However, the calcite deposited during the first-order shift appears to have resulted from slow, near-equilibrium deposition (31), therefore eliminating increased growth rate as an explanation for the elevated Sr concentrations. The synchronous decrease in P is attributed to an overall decline in vegetative cover and rock/till weathering rates due to lower temperatures and decreased precipitation. Nearly all available aqueous phosphorus in soil water is incorporated into plant tissue during the growing season, and previous work indicates that phosphorus is released into groundwater during the autumnal decay of plant matter (24). Reduced rock and till weathering rates may have reduced the phosphorus available for vegetative uptake and subsequent autumnal release into the groundwater. The Sr/P ratio is similar before and after the 8200-year event, which suggests that the local hydrologic conditions are similar before and after the event (Fig. 2). However, a change in the parameters controlling the drip water chemistry is indicated by the higher Sr/P ratio of the calcite deposited during the 8200-year event compared with that deposited before and after, suggesting sudden climatic change.

Figure 2

Sr concentrations plotted against P concentrations. The slope of the line describing the points before (slope = –0.05) and after (slope = –0.06) is nearly identical and is significantly different from that during the event (slope = –0.33). The different relationships between Sr and P during and outside of the event indicate the existence of a different set of parameters controlling the geochemistry of the drip waters. Similar slopes before and after the first-order anomaly indicate a return to baseline conditions, eliminating the possibility that the trace element anomaly was caused by a permanent shift in the local hydrology unrelated to climate.

Second-order antipathetic P and Sr oscillations, having wavelengths (λ) typically around 75 μm, are apparent during and preceding the first-order trace element anomaly (Fig. 3, A and B) but are discontinuous after the anomaly (Fig. 3C). Identical antipathetic relationships between P and Sr were identified previously in stalagmites from strongly seasonal climates (26) and were attributed to seasonal changes in the concentration of these elements in the infiltrating drip waters. P maxima and Sr minima probably occur during periods of increased infiltration in the autumn (24, 26), and Sr maxima may result from drier summer weather (23). The second-order trace element oscillations represent annual cycles of calcite deposition, as in other stalagmites (22, 24, 32, 33). The amplitude of the Sr oscillations during the first-order anomaly is four times that of the oscillations preceding the anomaly (Fig. 3); this finding suggests increased seasonality in effective precipitation, perhaps due to drier summers. However, no accompanying statistical change in the amplitude of the P cycles is evident (34), which suggests that the 8200-year event was either too brief or too mild to alter the seasonal phosphorus cycle.

Figure 3

(A) Very well defined second-order antipathetic cycles (mean λSr = 120.9 μm; mean λP = 117.0 μm, inset) occurring before the 8200-year event, representing annual deposition of calcite. (B) Some of the antipathetic cycles (mean λSr = 64.7 μm; mean λP = 65.8 μm, inset) within the 8200-year event. The λSr and λP are lower than those preceding the event, suggesting a decrease in annual growth rates caused by a drier climate and colder temperatures. The amplitude of the Sr cycles increases markedly (from 4.2 ppm before to 18.2 ppm during the event), suggesting enhanced seasonality in rainfall. (C) Second-order trace element variations after the event. No cyclicity is present, suggesting decreased seasonality in rainfall. Recently deposited calcite from the top of CC3 is also devoid of annual cyclicity, reflecting the modern maritime climate. The locations of these three panels on the complete trace element record are shown at a, b, and c in Fig. 1.

Reconstruction of stalagmite growth rate changes is possible because trace element cycle wavelength is equivalent to the amount of stalagmite calcite deposited in 1 year. The mean wavelengths of Sr and P cycles during the event (Fig. 3B) are approximately half those associated with the period before the event (Fig. 3A); this result suggests reduced growth rates within the event, consistent with reduced rainfall (25) and reduced mean annual temperatures (35).

Estimates of the duration of the 8200-year event are possible by counting the second-order annual trace element cycles within the first-order anomaly (Fig. 4). Thirty-eight Sr cycles and 37 P cycles, independently counted, exist within the portion of the record defined by the first-order trace element excursion. Hence, the hydrologic response to the 8200-year event lasted about 37.5 years in western Ireland. Annual growth rates preceding the 8200-year event, reconstructed with the use of four collinear U-Th–dated points (10, 11), are similar to growth rates before the event (as assessed from the measured wavelengths of the Sr and P cycles), indicating that the cycles are annual. A decrease in annual growth rates is apparent at the inception of the event but gives way to a sudden, short-lived growth increase (Fig. 4). Growth rates decrease again and eventually increase gradually toward the termination of the event. A two-pronged structure is also evident in coarser resolution records of the 8200-year event, including CC3 δ18O data (10, 11), Greenland Ice Sheet Project 2 accumulation rate (36) and δ18O (2, 37) data, and numerous Norwegian lacustrine sediment cores (9). The overall decrease in growth rate is consistent with cold, dry conditions, because growth rate is predominantly dependent on surface temperature and the amount of precipitation (35, 38, 39).

Figure 4

Annual axial calcite deposition rates for stalagmite CC3 reconstructed using the wavelengths of annual Sr and P cycles. The bracketed area indicates the timing of the 8200-year event, as assessed from the first-order trace element shifts. The horizontal dashed line represents the approximate growth rate of stalagmites in temperate regions. Growth rates decrease into the event, increase briefly 27 years into the event, then increase until the termination of the event. Thirty-eight Sr cycles and 37 P cycles exist within the 8200-year event, suggesting that the hydrologic response to the event lasted about 37.5 years. Trace element cyclicity was nonexistent after the event, precluding cycle counting.

The rapid commencement of the first-order trace element excursion suggests a single catastrophic triggering event, potentially the draining of glacial lakes Agassiz and Ojibway through the St. Lawrence and into the North Atlantic (19). It is noteworthy that a GCM modeled response to a freshwater pulse equivalent to the volume of water drained from Lakes Agassiz and Ojibway, 4.67 × 1014 m3, distributed into the Labrador Sea over 20 years, predicts a two-pronged event similar to the trace element record (20). This implies a mean meltwater flux of 0.75 Sv (1 Sv = 106 m3 s−1), easily exceeding that required to trigger a collapse of the THC (40). The short duration of the event compared with that of high-latitude ice cores may reflect a nonlinear threshold response in atmospheric circulation patterns or in the hydrologic system at this maritime mid-latitude site. The rapid return to baseline trace element values in stalagmite CC3 at 8310 ± 80 years B.P. is consistent with a rapid reestablishment of North Atlantic THC (20) ending the cold, dry episode in Ireland.

  • * To whom correspondence should be addressed. E-mail: james.baldini{at}


View Abstract

Stay Connected to Science

Navigate This Article