Hydrological Impact of Heinrich Events in the Subtropical Northeast Atlantic

See allHide authors and affiliations

Science  25 Aug 2000:
Vol. 289, Issue 5483, pp. 1321-1324
DOI: 10.1126/science.289.5483.1321


Reconstructing the impact of Heinrich events outside the main belt of ice rafting is crucial to understanding the underlying causes of these abrupt climatic events. A high-resolution study of a marine sediment core from the Iberian margin demonstrates that this midlatitude area was strongly affected both by cooling and advection of low-salinity arctic water masses during the last three Heinrich events. These paleoclimatic time series reveal the internal complexity of each of the last three Heinrich events and illustrate the value of parallel studies of the organic and inorganic fractions of the sediments.

High-resolution paleoclimatic records show that the glacial climate was far more variable than previously thought. Studies of North Atlantic sediments show that the circulation of the surface and deep waters was repeatedly perturbed by massive surges and melting of icebergs. Although most researchers would agree that these so-called Heinrich (H) events originated from the Laurentide ice sheets (1–6), there is growing controversy about the precise behavior of other ice sheets during these events [see recent works on lithological indices (7, 8) and studies based on the Sr-Nd isotopic signatures of ice-rafted detritus (IRD) (9, 10)].

The influence of Heinrich events has sometimes been invoked to explain unusual high-frequency climatic events at low latitudes [e.g., (11)]. Abrupt decreases of sea surface temperature (SST) in phase with Heinrich events were detected in sediments of the Bermuda Rise (12) and the Mediterranean Sea (13). Further south, the tropical western Atlantic apparently warmed during H1 and the Younger Dryas (YD) (14). The strong, direct effect of melting icebergs is generally thought to be restricted to a North Atlantic belt between 40°N and 55°N, where IRD forms distinct layers in the sediments (15). The occurence and climatic impact of H events are still poorly documented in the northeast Atlantic, especially at subtropical latitudes.

To reveal the impact of H events on the northeast Atlantic, we focused our study on the paleoceanographic variability in a midlatitude core from the Iberian margin (SU8118 located at 37°46′N, 10°11′W and 3135 m depth) (Fig. 1). One of the main advantages of using SU8118 is that it is one of the best-dated cores available for the last deglaciation (16). These sediments are characterized by a high sedimentation rate (>20 cm/1000 years during the glacial and deglacial sections) which has been constrained by numerous 14C accelerator mass spectrometry (AMS) ages and stable isotope records (17).

Figure 1

Map showing North Atlantic cores located outside the main IRD belt and for which alkenone data are available to study the last 30,000 years. Black arrows indicate likely sources of icebergs, whereas the gray arrows represent the inferred circulation compiled by several authors (32).

SST changes were reconstructed with the alkenone unsaturation ratios (Fig. 2B and Web table 1) (18, 19), and counts of IRD (Fig. 2B and Web table 1) (18, 20) and magnetic susceptibility (MS; Fig. 2B) (21) were used to identify transported sediments. The alkenone temperature estimates are on the order of 18°C during the last 5000 years, which is between the modern annual mean (17.5°C) and the SST during summer months (19°C) when most of the local primary productivity takes place. With an average sampling resolution of ∼250 years, the alkenone-SST record shows the classical climate history of the Late Glacial: the Holocene starts at 11,500 calendar years before present (cal yr B.P.), the YD event is centered at about 12,000 cal yr B.P., and the Allerød-Bølling interstadial occurs between 13,000 and 15,000 cal yr B.P. Interestingly, the LGM in the strictest sense (21,000 ± 2000 cal yr B.P.) is a rather mild period, with SST on the order of 13°C, i.e., about 5°C lower than modern SST. In addition, the glacial period is characterized by a rather large variability with abrupt cold spells centered at about 16,000, 23,500, 26,000, and 31,000 cal yr B.P.

Figure 2

New paleoclimatic records measured in core SU8118 represented versus calendar age based on the14C-dated tie points shown as triangles in (C) [see (16) for details]. Climatic events are abbreviated as follows: Hol., Holocene; YD, Younger Dryas; A-B, Allerød-Bølling; H1, Heinrich 1; and LGM, Last Glacial Maximum. (A) The three curves represent the C37 alkenone unsaturation data converted in terms of SST by using three different calibrations published by Prahl et al. (40) and Mülleret al. (41) for the thick solid line with open dots [#1 in Web table 1 (17)], by Weaver et al.(42) for the thin solid line (#2 in Web table 1) and by Rosell-Melé et al. (26) for the thin dotted line (#3 in Web table 1). Note that the first two SST calibrations are based on the U37 K′ index ([C37:2]/[C37:2 + C37:3]), whereas the third uses the original U37 K index ([C37:2 – C37:4]/[C37:2 + C37:3 + C37:4]). As discussed in (19), the differences observed in some sections illustrate the overall uncertainty on the alkenone calibration. However, none of the conclusions of our study would be altered by selecting one instead of the other equation (19). (B) Black dots show the IRD in number per gram for the size fraction greater than 150 μm (Web table 1). Details about the mineralogy of the three IRD peaks are described in (20). The occurrences of detrital carbonates and of hematite-coated grains are specifically used to identify H events (2, 3). Presence of these mineralogical markers are marked on (B) by C for detrital carbonates and H for hematite-coated grains. The IRD peak centered at 23,000 cal yr B.P. (H2a) is characterized by presence of both detrital carbonates (3%) and hematite-coated grains (2%). The small IRD peak centered at 17,500 cal yr B.P. (H1b) is mainly composed of carbonates (80%), whereas the prominent IRD peak around 16,000 cal yr B.P. (H1a) exhibits hematite-coated grains (3%) but almost no detrital carbonate. The MS record has been measured at CEREGE on discrete samples (21). The upper part of the core is characterized by an oxidized zone, and the MS signal is perturbed for most of the Holocene (21). (C) Percentage of C37:4 among C37 alkenones, i.e., C37:4% = 100x[C37:4]/[C37:2+ C37:3 + C37:4] with quantities based on the chromatographic peak areas.

Figure 2B shows the IRD counts and the MS record which can be used to identify detrital material usually attributed to H events. These are clearly expressed in both records between 18,000 and 15,500 cal yr B.P. for H1 and between 26,000 and 23,000 cal yr B.P. for H2. The MS and mineralogical records (20) also suggest that each of these two events were complex and characterized by two depositional phases centered at 16,000 (H1a) and 17,500 (H1b) cal yr B.P. for H1 and at 23,500 (H2a) and 25,000 (H2b) cal yr B.P. for H2. Our results strengthen earlier suggestions of twinned IRD peaks in some cores of the Portuguese margin (22). Interestingly, only the late depositional phases are well expressed in the IRD records: there is a lack of IRD at about 25,000 cal yr B.P. and only a slight increase around 17,500 cal yr B.P. An alternative explanation of the multiple peaks is that they belong to the pervasive millennial-scale cycle that consistently punctuates the intervals between H events and the Holocene period (2). Furthermore, the alkenone-SST record is also characterized by small shifts at about 21,500 and 29,000 cal yr B.P. (Fig. 2A) which may be explained by the same cyclicity (the latter cooling is even associated with an MS peak).

H3 may be only partly expressed in core SU8118, because the oldest sediment in this core is about 31,500 cal yr B.P. However, it is clear that the SST decrease and MS increase measured in the lowermost section do indeed correspond to H3 even though there is no IRD increase in the same sediment levels. Low IRD concentrations during H3 have already been described in sediments of the main IRD belt (3), north of the Azores (23), and off Portugal (24) [although a small IRD peak was detected farther offshore (22)]. The sole presence of an MS signal could be accounted for if the material transported by icebergs was all finer than 150 μm.

From Fig. 2, A and B, it is clear that strong and sharp cooling events coincide with all phases of H events H1a and H1b, H2a and H2b, and H3. It is worth noting that Zhao et al. (25) observed alkenone-SST drops off Mauritania roughly synchronous with H events, although the tentative chronology of their cores was not directly based on 14C data. Their claim of an impact of H events on the Canary current was rather unexpected at the time and has awaited further confirmation. Our well-dated alkenone-SST record now provides a further proof that all latitudes of the eastern North Atlantic were strongly affected by these cooling events. By contrast, the western part of the low-latitude Atlantic appears to have warmed during YD and H1, as reconstructed with alkenones (14). Collectively, the observations based on alkenones indicate a significant increase (+3°C) in the east-west gradient of SST of the low-latitude North Atlantic during these two recent climatic events.

A by-product of the alkenone analyses in SU8118 is the discovery of large and systematic variations of the abundance of tetra-unsaturated C37 alkenones (C37:4, molecular mass of 526). Percentages higher than 5% of the total C37 alkenones are restricted to very distinct periods centered around 31,000, 23,500, and 16,000 cal yr B.P. (Fig. 2C). These distinct C37:4 peaks are clearly synchronous with H3, H2a, H1a, and H1b. We also note an absence of such a peak during event H2b and a rather weak signal during the YD event.

Although the abundance of C37:4 alkenone was originally used for paleothermometry (26, 27), Sikeset al. (28) analyzed core-tops from the Southern ocean and suggested that this compound may not actually be linked to SST. Subsequently, Rosell-Melé (29) proposed that this particular alkenone is linked to low-salinity water masses and that a C37:4 increase of about 5 to 10% corresponds to a freshening of one practical salinity unit (PSU). Further research is still needed to decipher if indeed salinity has a direct effect on the C37:4 biosynthesis or if the observed pattern is due to a metabolic difference of coccolithophorids endemic of arctic or coastal water masses (30). Even if this latter hypothesis is confirmed, the C37:4 abundances may still be used to study the surface advection of iceberg-bearing water masses together with their living arctic alkenone producers. In any case, C37:4 levels in SU8118 are similar during the LGM and the Holocene, which indicates that this molecule is not simply correlated to SST because it behaves differently than the more abundant di- (C37:2) and tri-unsaturated (C37:3) methyl ketones used in the U37 K′ index.

Applying the Rosell-Melé (29) calibration to our C37:4 record leads to the conclusion that the salinity dropped by 1 to 2 PSU off Portugal during the last of the H events. This is in agreement with a salinity decrease of 2 to 3 PSU inferred for H1 which was based on δ18O measurements in planktonic foraminifera from the same sediments (31). This salinity drop is comparable in magnitude to other salinity estimates inferred from cores located farther north in the Atlantic and thus closer to the former continental ice sheets (29, 32). Even if these salinity estimates remain semiquantitative, there is a rather impressive correlation between the C37:4fluctuations and the lithological variations used to identify H events (Fig. 2, B and C). Furthermore, the precise timing of these C37:4 peaks is obviously not accidental and matches previous geochronological estimates for the abrupt H events (3, 7).

All available data from SU8118 suggest that advected high-latitude water transported icebergs that melted in the Iberian margin and released fine-grained detrital material. However, several differences are observed between the various proxies measured, particularly for events H2a (high MS but lack of IRD and C37:4) and H3 (lack of IRD but high C37:4 and MS). These peculiar signatures could be explained if icebergs did not transport the same amount of IRD during all phases of H events. This may indicate that there were multiple origins of ice during the surge events, such as from different parts of the Laurentide Ice Sheet or even from other sources such as the Fennoscandian Ice Sheet. The observed complexity may also be related to the size fraction of IRD: in distal layers, remote from all glacial sources, part of the mineralogical signal is carried by relatively fine detrital particles which would be expressed in the MS but not in the IRD counts (>150 μm).

Several authors recognized that H3 presents some unusual features compared to other H events (14, 33). Gwiazdaet al. (33) even proposed that, when it is present, the IRD concentration peak could be due to a lack of foraminifera caused either by dissolution or low primary productivity, and they concluded that the iceberg discharge melted mostly in the western basin of the North Atlantic. The absence in core SU8118 of an IRD peak during H3 would be compatible with this theory, but the C37:4, SST, and MS records strongly suggest that meltwater reached the eastern basin of the North Atlantic during this period. Independent evaluation of salinity changes may also be possible from the simultaneous analyses of Mg/Ca and δ18O in the same shells of planktonic foraminifera (34).

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


View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article