A glacial epibenthic δ13C record from 773 m water depth in the SE Nordic seas reveals moderate, but distinct changes in the ventilation ∼65,000−14,000 years ago. The Last Glacial Maximum, the warm interstadials, and the shorter stadials are characterized by high δ13C values indicating well-ventilated intermediate water masses in the Nordic seas. Decreasing δ13C values during the cold Heinrich events signify a reduction in intermediate water ventilation. We attribute the reduction to the development of a halocline causing a stop in convection and outflow from the Nordic seas. The well-ventilated outflow water is replaced by warmer Atlantic water, which due to the stratification is isolated from the atmosphere. Its initial high δ13C values are reduced due to 'ageing'. We ascribe the lack of response in the subsurface Atlantic Water during the stadials to the smaller geographical extent of these events as compared to the Heinrich events.
 The thermohaline circulation and the climate of the north Atlantic region are closely linked. Warm surface water derived from the Gulf Stream system is cooled and convected to cold deep and intermediate water in the Nordic seas (Figure 1). The cold water flows across the Greenland-Scotland Ridge into the North Atlantic, where it contributes to the North Atlantic Deep Water (NADW). The newly formed NADW is oxygen rich and nutrient poor and has high δ13C values [Kroopnick, 1985]. In the North Atlantic, the NADW constitutes the intermediate and deep water down to a depth of approximately 4 km. Antarctic Bottom Water (AABW) fills the deeper basins below ∼4 km. The AABW is generated around Antarctica during winter time. It incorporates low δ13C values from nutrient-rich upwelling deeper water that is not sufficiently aerated before it sinks.
 Carbon isotope ratios measured in epibenthic foraminifera are considered to be a consistent proxy for bottom water ventilation and nutrient distribution [e.g., Zahn et al., 1986; Boyle, 1998]. Variations in δ13C values in deep-sea records may therefore reveal past changes in ventilation and give clues to previous shifts in the thermohaline circulation. Results of core studies from the North Atlantic Ocean show that during the last glacial maximum AABW with low benthic δ13C values spread below 2 km water depth indicating a reduction in the formation of NADW [Duplessy et al., 1988; Oppo and Lehman, 1993; Sarnthein et al., 1994]. The intermediate water above the AABW (∼1–2 km water depth) was replaced by Glacial North Atlantic Intermediate Water (GNAIW) with very high benthic δ13C values. This water mass was apparently generated in the North Atlantic south of Iceland [Oppo and Lehman, 1993].
 From the last glacial period prior to the LGM, numerous oscillations between warm interstadials and cold stadials, termed Dansgaard-Oeschger (DO) events, have been recorded in the Greenland ice cores. Rapid changes in marine faunas and isotope values indicate that these oscillations coincide with pronounced shifts in the thermohaline circulation [e.g., Bond et al., 1993; Oppo and Lehman, 1995; Rasmussen et al., 1996a, 1996b]. However, our understanding of the role of the thermohaline circulation is hampered by the fact that, so far, no epibenthic δ13C records have been obtained from the Nordic seas. The main reason for this is the scarcity or total absence of epibenthic foraminifera in the cores hitherto retrieved from north of the Greenland-Scotland Ridge [e.g., Bauch et al., 2001].
 In this study, we present a high-resolution epibenthic δ13C record from the southeastern Nordic seas comprising marine isotope stages (MIS) 4−2. The core is from 773 m water depth and the site lies in the strongest outflow of cold water generated by the convection in the Nordic seas (Figure 1). The core is unusual by containing numerous specimens of the epibenthic foraminifera species Cibicides lobatulus, thus, allowing us to generate a continuous and high-resolution benthic δ13C record for the coldest parts of the last glacial period. We compare the data from LINK16 with records from the North Atlantic, in particular with core MD01-2461 from west of Ireland [Peck et al., 2007] (Figure 1). This is one of the few cores from mid depth in the North Atlantic with a benthic δ13C record of sufficient resolution to reveal millennial scale climatic events.
 The purpose of the investigation is to increase our knowledge of the origin and properties of the intermediate-depth water masses in the Nordic seas during the last glacial period, in particular with respect to changes in the thermohaline circulation during the Dansgaard-Oeschger events.
2. Material and Methods
 Piston core LINK16 was retrieved in August 2000 during a cruise of RV Dana to the Fugloyar Ridge, northeast of the Faeroe-Shetland Channel [Nielsen et al., 2007] (Figure 1). The core is approximately 10 m long. The upper 4 m, comprising MIS 4-2, was studied in great detail with regard to stable oxygen and carbon isotopes of both benthic and planktic foraminifera. The core was sampled at 1–2.5 cm intervals in 0.7 mm thick slices. The planktic δ13C and δ18O measurements were performed on Neogloboquadrina pachyderma s. The benthic measurements were performed on the epibenthic foraminifera Cibicides lobatulus. This species is considered to reliably record δ13C changes in the bottom water [see Millo et al., 2006]. For Heinrich events H1 and H6, Cibicides aff. C. floridanus was also analyzed. For both benthic species the measured δ18O values were corrected by +0.64‰ to adjust for isotopic disequilibrium [McCorkle et al., 1997]. For each sample 3–6 specimens of C. lobatulus, single individuals of C. aff. C. floridanus and 7–10 specimens of N. pachyderma s were analyzed at Woods Hole Oceanographic Institution [see Ostermann and Curry, 2000]. In addition, for a low-resolution survey of faunal development and productivity, 200–300 specimens of each benthic and planktic foraminifera were counted and identified for every second sample (Figure 2). IRD > 150μm was counted at the same high resolution as stable isotope analyses. The same methods as those of Rasmussen et al. [1996a] were followed. We compare and correlate the new record with stable isotope results from nearby piston core ENAM93-21 taken at 1020 m water depth and described by Rasmussen et al. [1996a, 1996b] (Figure 1).
 Four AMS-14C dates and the position of known tephra layers in combination with variations in oxygen isotopes, magnetic susceptibility and the distribution of foraminifera establish the stratigraphy and chronology of the core (Figures 2 and 3and Table S1 of the auxiliary material). These parameters have previously been shown with great precision to identify the Dansgaard-Oeschger events and Heinrich events in the Faeroe area [Rasmussen et al., 1996a]. The stadials and Heinrich events are marked by low magnetic susceptibility, low δ18O values, a strong dominance of the polar planktic foraminifera species N. pachyderma s, and a benthic foraminifera fauna dominated by C. teretis and the Atlantic Species group [see Rasmussen et al., 1996a]. The interstadials are characterized by high magnetic susceptibility, high δ18O values, lower relative abundance of N. pachyderma s and a benthic fauna dominated by Melonis barleeanum. Interstadials have high accumulation rates of both planktic and benthic foraminifera, whereas the stadials and Heinrich events generally have high planktic, but low benthic accumulation rates (Figure 2).
 The Dansgaard-Oeschger events of LINK16 can be correlated with ENAM93-21 and the Greenland ice cores with great confidence [Rasmussen et al., 1996a]. The age models of LINK16 and ENAM93-21 are therefore based on a one-to-one correlation with the Dansgaard-Oeschger events of the NGRIP ice core using the newly developed GICC05 time scale [Svensson et al., 2008] (Figure 3 and Table S2).
 The planktic δ13C values vary more in ENAM93-21 than in LINK16, but taken as a whole the two records are fairly similar. In LINK16, the benthic and planktic δ13C records follow parallel courses with the exception of MIS 2 (and, in part, MIS 4 and IS14), where the difference between the two records increases (Figure 3). Excluding Heinrich events, the benthic δ13C values of LINK16 increase from about 1.3‰ during early MIS 3 (c. 58−45 ka BP), to > 1.5‰ during late MIS 3 (c. 45−30 ka BP) and more than 1.7‰ during MIS 2 (c. 27−17 ka BP). The LGM interval (23−17 ka BP) produces values around or slightly higher than 1.6‰. The values of Heinrich events H6, H5, H4 and H3 are roughly 0.5–0.6‰ lower than the values of the surrounding interstadials with minimum values typically ranging between c. 1.0 and 1.2‰. In Heinrich events H6 and H1, C. aff. C. floridanus gives lower δ13C values than C. lobatulus, whereas there is little difference between their δ18O values (Figure 3). It is possible that the lenticular-shaped C. aff. C. floridanus is not truly epibenthic, but lives at the sediment-water interface. It may therefore be influenced by pore-water chemistry [cf. McCorkle et al., 1997].
 LINK16 displays a pattern of fairly small benthic δ13C variations and large benthic δ18O variations (Figure 3). The overall high benthic δ13C values of LINK16 indicate the presence of well-ventilated bottom waters in the SE Nordic seas throughout MIS 4-2. The increase during MIS 2 probably reflects an increase in the formation of intermediate water as also seen in the North Atlantic Ocean [Oppo and Lehman, 1993]. The very high δ13C values measured in the LGM interval equal the values measured by Millo et al.  in the Denmark Strait and support their conclusion that during the LGM, GNAIW was generated in the Nordic seas as well as in the North Atlantic.
 Examined in more detail, it appears that the benthic δ13C values of the interstadials and stadials are similar. This is in contrast to the clear decreases observed in the Heinrich events. However, it is also in contrast to the planktic δ13C record, in which both the stadials and the Heinrich events show a decline (Figure 3). Benthic δ13C values are not only influenced by ventilation, but also by local changes in the productivity and degradation of organic material [e.g., Zahn et al., 1986]. A higher supply of organic material to the sea floor may lower the benthic δ13C values and vice versa. However, the lower δ13C values observed in the Heinrich intervals of LINK16 cannot be explained by increased productivity, as the accumulation rate of benthic foraminifera decreased during the stadials and Heinrich events (Figure 2), indicating a lower influx of food particles. Thus, poorer ventilation seems to be the most likely explanation.
 In several cores from the central North Atlantic, e.g., V29-202 [Oppo and Lehman, 1995], V29-204 and OC205-103GGC [Curry et al., 1999], and SO82-05 [van Kreveld et al., 2000], the interstadial δ13C values are mostly higher than the values from the Heinrich events. The cores are situated in the pathway of the NADW and the high interstadial values are associated with high production rates of well-ventilated NADW [Oppo and Lehman, 1995; Curry et al., 1999]. The fluctuating values are interpreted to reflect variations in the strength of NADW production from high during the interstadials to low during the stadials and Heinrich events [Oppo and Lehman, 1995]. This interpretation is in accord with changes in the distribution of benthic foraminifera in the SE Norwegian Sea and NE Atlantic Ocean, which indicate strong convection in the Nordic seas and overflow across the Scotland-Greenland ridge during interstadials and a stop in the convection and little or no overflow during stadials and Heinrich events [Rasmussen et al., 1996a, 1996b; Rasmussen and Thomsen, 2004]. The decrease in the δ13C values during periods with reduced NADW production are then attributed to increased mixing with water with low δ13C values originating from the Southern Ocean [Curry et al., 1999; Peck et al., 2007].
 This interpretation has also been applied to core MD01-2461 from the western slope of Ireland (Figure 1). The MD01-2461 record is from 1153 m water depth and spans the time period 43−8 ka BP. The benthic δ13C records of LINK16 and MD01-2461 are remarkably alike with values in both cores fluctuating around 1.5–1.6‰ and decreasing up to 0.6‰ during Heinrich events (Figure 3). According to Peck et al. , the benthic δ13C record of MD01-2461 indicates that the generally well-ventilated bottom water during Heinrich events was infiltrated by poorly ventilated deep water of a southern origin. In agreement with the previous studies they suggest that the southern water most likely was Antarctic bottom water or intermediate water advancing northward in the eastern Atlantic Ocean.
 Core site MD01-2461 is situated about 1200 km southeast of LINK16 (Figure 1). Considering the relatively short distance between the two cores and the similarity between their δ13C records (Figure 3) it seems probable that the two sites were affected by the same oceanographic changes. On the other hand, it seems highly unlikely that southern water could cross the shallow Greenland-Scotland Ridge and reach the site of LINK16. Another factor, besides southern water, that could cause poorer ventilation in the upper intermediate water masses during Heinrich events is the development of a shallow halocline. In the Nordic seas and in the northern North Atlantic the stadials and Heinrich events are generally characterized by low planktic δ18O values. The low values are considered to be the result of an increase in the supply of meltwater mainly from icebergs [e.g., Bond et al., 1993]. The fresh water supply supposedly created a strong halocline and a density stratification of the Nordic seas, much as in the Arctic Ocean today. The stratification would isolate the subsurface waters from the atmosphere and, furthermore, decrease or stop deep convection [Rasmussen and Thomsen, 2004; Praetorius et al., 2008]. During Heinrich events, a halocline was developed in the Heinrich belt south of the Greenland-Scotland Ridge down to 40°N and the area covered by meltwater was much larger than during the smaller stadials [e.g., Bond et al., 1993]. We suggest that the lower benthic δ13C values observed at mid-depth in LINK16 and in the North Atlantic cores during Heinrich events are mainly the result of this stratification.
 In the northeast Atlantic and in the Nordic seas the Heinrich events and stadials at mid depth are characterized by significant decreases in the benthic δ18O values. These shifts have been explained as the result of a temperature increase [Rasmussen et al., 1996a, 1996b; Curry et al., 1999; Rasmussen and Thomsen, 2004] due to the replacement of the cold overflow water with warmer Atlantic water, which continued to flow below the halocline. This relatively warm water mass was generally well-ventilated as indicated by the high benthic δ13C values obtained from both the stadials and Heinrich events in core OC205-103GGC from 965 m water depth off the Little Bahama Bank in the western North Atlantic. In this core, the benthic δ13C values do not decrease below ca 1.3‰ [Curry et al., 1999]. We suggest that the benthic δ13C decreases during Heinrich events in LINK16 and in the northeast Atlantic are due to ‘aging’ of this generally well-ventilated water mass, which due to the stratification was isolated from the atmosphere. As a modern analogue we refer to the southward flow of the NADW in the Atlantic Ocean, where such ‘aging’occurs [e.g., Kroopnick, 1985]. The lack of a clear response in the benthic δ13C values during stadials in both LINK16 and MD01-2461 may reflect the smaller geographical area covered by a meltwater layer during the stadial events.
 The results from the SE Nordic seas show good ventilation of the upper intermediate water during MIS 4-2 with high benthic δ13C values. The benthic δ13C values decrease during Heinrich events in the intermediate-depth water of the northern North Atlantic and Nordic seas indicating a reduction in the rate of ventilation. The lower values are mainly due to the development of a stratified upper ocean, a less vigorous mid-depth circulation pattern, and ‘aging’ of the subsurface Atlantic Water inflow to the Nordic seas. In the North Atlantic, infiltration by low δ13C southern water, such as described by Peck et al. , may have contributed to lower values observed in the deeper part of the intermediate water.
 The study of LINK16 was financed by the University of Tromsø and the University of Aarhus as a part of the ESF-EuroClimate Program RESOLuTION (grant 04-ECLIM-FP33). We thank V. L. Peck for kindly sharing her MD01-2461 core data with us. Tove Nielsen, A. Kuijpers and J. Boserup (GEUS) are thanked for their great help in the coring procedures. Peter Konradi and I. S. Nielsen (GEUS) are thanked for help in the post cruise core handling and logging.