3.1. Records From MD01-2461
 Elevated benthic δ13C values of 1.65‰ VPDB during the last glacial reflect the influence of a well-ventilated mid-depth water mass, GNAIW [Oppo and Lehman, 1993] or Upper North Atlantic Deep Water (UNADW) [Boyle and Keigwin, 1987]. H events recorded in MD01-2461, identified through the occurrence of IRD deposition with contribution of distinctive Hudson Strait-sourced dolomitic carbonate (Figure 2f) [Peck et al., 2006], coincide with benthic δ13C depletions of 0.6‰ to 1.25‰. A smaller negative δ13C anomaly, of ∼0.3‰, is associated with an enhanced flux of IRD at ∼30 kyr B.P. likely recording the occurrence of H3. This suggestion is supported by the apparent lack of LIS-sourced IRD in this interval, which has been considered a type-characteristic of H3 along with a strong volcanic component to the IRD assemblage [e.g., Grousset et al., 1993; Gwiazda et al., 1996], likely of East Greenland/Icelandic origin [Peck et al., 2007]. Decreased benthic δ13C can potentially be ascribed to high vertical fluxes of organic carbon [Mackensen et al., 1993]. However, the negative benthic δ13C anomalies associated with H events at MD01-2461 are considered to reflect, principally, bottom water mass changes [cf. Willamowski and Zahn, 2000], most feasibly advances of SCW, likely a glacial version of AAIW, to our core site in response to decreased or halted GNAIW formation, confirming restriction of deep-intermediate convection in the northern North Atlantic in the course of the events [e.g., Cortijo et al., 1997; Vidal et al., 1997; Zahn et al., 1997; Rickaby and Elderfield, 2005].
 An additional control on the ventilation of deep-intermediate waters in the North Atlantic during meltwater events might involve a change in the mode of North Atlantic deep water formation, from open-ocean convection under glacial conditions to buoyancy loss due to brine formation in the course of sea-ice formation in the Nordic Seas following meltwater surges [Dokken and Jansen, 1999]. Recent studies have suggested that brine formation proximal to ice margins persisted throughout the last glacial in the northern North Atlantic concurrent with open-ocean convection in the North Atlantic [Labeyrie et al., 2005]. The most likely preformed poorly ventilated signal [Vidal et al., 1997; Elliot et al., 2002] and low δ18O signature of these intermediate brine waters being masked during ambient glacial conditions by the well-ventilated GNAIW and only becoming apparent during episodes of reduced MOC, when simultaneous negative anomalies in both planktonic and benthic δ18O and benthic δ13C are observed in North Atlantic records [Dokken and Jansen, 1999; Labeyrie et al., 2005; Waelbroeck et al., 2006]. Such an interpretation challenges the interpretation of North Atlantic benthic δ13C anomalies as documenting the northward penetration of southern sourced waters in response to reduced MOC as the rapid transmission of the North Atlantic-derived brine signal to both the Indian and Pacific Ocean necessitates active intermediate water formation [Labeyrie et al., 2005; Waelbroeck et al., 2006].
 At site MD01-2461 we also see some evidence for the injection of low δ18O intermediate waters, characteristic of formation via brine rejection, notably at meltwater event “b” (18–17.7 kyr B.P.) with a well defined decrease of benthic δ18O by 0.7‰ associated with a transient negative benthic δ13C anomaly. Further benthic δ18O (brine) anomalies are hinted at during H4, H3 and meltwater events at ∼26 kyr B.P. and “a” (21.8–20.9 kyr) although these anomalies are much less distinct and appear absent at H2. This lack of a clear brine signal together with the recently reported covariance of 231Paxs/230Thxs and benthic δ13C records from core DAPC2 [Hall et al., 2006] provides convincing evidence that changes in bottom water ventilation along the northern European margin (inferred from benthic δ13C) relate to changes in the rate of MOC and the interchange of poorly ventilated southern and well ventilated northern sourced water masses (at least up to ∼1700 m water depth) rather than the dominant injection of poorly ventilated brine waters from the Nordic Seas [Labeyrie et al., 2005]. This interpretation of benthic δ13C data provides the basis for the following discussion.
 While initial benthic δ13C decreases in MD01-2461 directly coincide with the onset of H layer deposition, peak minima in benthic δ13C coincide with maximum divergence in δ18O from surface- and subsurface-dwelling planktonic foraminifers (Figure 2b) that we interpret as indications of meltwater stratification of the upper water column [Peck et al., 2006]. Benthic δ13C values fall to their lowest values within the entire record, +0.4‰ VPDB, at 16.2 kyr B.P., some 0.6–1.1 kyr after the incursion of the detrital carbonate peak indicative of H1 deposition at this site, coincident with offset between δ18O G. bulloides and N. pachyderma sin. of >2.5‰ indicating prominent meltwater stratification. Deglacial meltwater forcing is evident from persistent upper ocean stratification through 16.4–14.0 kyr B.P., associated with significant sea level rise at this time [Bard et al., 1990, 1996]. Benthic δ13C values do not return to their elevated glacial levels but vary between 0.7–1.2‰ VPDB throughout the deglacial, plausibly reflecting a combination of changes involving the terrestrial biosphere and its influence on marine carbon reservoir δ13C and establishment of the modern mode of North Atlantic THC. A decrease in benthic δ13C of 0.5‰ at 13 kyr B.P. is captured by two data points only, but potentially reflects a convection slow-down in the North Atlantic during the Younger Dryas cold period [Boyle and Keigwin, 1987; Sarnthein et al., 1994; Rickaby and Elderfield, 2005]. In addition to the H events, short-lived negative δ13C anomalies are observed at 26–25.7 kyr, 21.8–20.9 kyr (“a”) and 18–17.7 kyr (“b”) B.P., and directly coincide with intermittent meltwater stratification of the upper water column, and enhanced deposition of NWEIS-sourced IRD. Our timings of the two most recent events, “a” and “b” support dates of British ice sheet (BIS) ice sheet retreat determined from cosmogenic nuclide (36Cl) surface-exposure dating of glacial erratic boulders and glacially smoothed bedrock sampled around the former ice margins in Ireland [Bowen et al., 2002]. IRD associated with meltwater event “b” and an IRD peak with a similar NWEIS-signature prior to H2 (at 25.0 kyr B.P.) likely correspond to the well-documented “European precursor” events [e.g., Grousset et al., 2000; Scourse et al., 2000]. Freshwater surging associated with these episodes of NWEIS instability conceivably reached the area of GNAIW formation and caused transient, 200–500 year, reduction in northern sourced intermediate water flux, allowing brief northward penetration of SCW to the MD01-2461 site [Peck et al., 2006].
3.2. Intermediate Water Ventilation Changes From NE Atlantic Records
 We compare the benthic δ13C record of MD01-2461 with similar records of other North Atlantic sites (Table 1) to determine regional variability of intermediate water masses. In particular, we use the high-resolution record from core SO75-26KL from the Portuguese Margin at a water depth of 1099 m, similar to that of MD01-2461 [Zahn et al., 1997]. Data records from SO75-26KL have been published on a radiocarbon timescale applying a constant 14C-marine reservoir age correction of 400 years [Zahn et al., 1997; Willamowski and Zahn, 2000]. Assessment of multiple 14C data [e.g., Voelker et al., 1998; Waelbroeck et al., 2001], in agreement with age modeling of the MD01-2461 records [Peck et al., 2006], suggest that NE Atlantic marine 14C reservoir ages were highly variable during the glacial period such that the radiocarbon timescale of SO75-26KL does not allow for a detailed comparison with MD01-2461. To attempt a synchronized timescale for SO75-26KL we graphically correlate the benthic δ18O records of both cores by “tuning” the record of benthic δ18O record from SO75-26KL to that of MD01-2461 across the deglaciation (24–8 kyr B.P.) (Figure 2b). This procedure receives independent support from the observation that an IRD peak at SO75-26KL, on the synchronized timescale, is concurrent with the lithologically and geochemically distinct H layer 1 at MD01-2461, consistent with synchronous deposition of H layers 4 and 2 at the two sites (Figure 2f). Additionally, the onset of deglacial warming/surface ocean freshening in the planktonic δ18O records is simultaneous at the two sites (Figure 2c). The synchronized timescale of SO75-26KL suggests marine 14C-reservoir ages between 0.4 kyr to 0.9–1.0 kyr in the period 17.3–16.4 kyr B.P. at this site (Figure 2a). A latitudinal gradient of reservoir ages during the deglaciation from ∼1 kyr at 37°N (SO75-26KL) up to ∼2 kyr at 52°N (MD01-2461) [Peck et al., 2006] is comparable with the findings of Waelbroeck et al. .
 The lower resolution benthic δ13C record from core NEAP 4K at Björn Drift (1627 m water depth; Figure 1) [Rickaby and Elderfield, 2005] is used as reference for comparison with mid-depth ventilation changes in the high-latitude North Atlantic. The age model of NEAP 4K (>13 kyr B.P.) is based on stratigraphical correlation (benthic and planktonic δ18O) with core BOFS 5K [Barker et al., 2004], which has a radiocarbon-based age model incorporating the elevated marine reservoir ages of Waelbroeck et al.  (1.9 kyr) for this time period. Temporal resolution at this site averages 500 years and does not allow for a detailed correlation of the benthic δ18O record with that of MD01-2461. A broad peak in weight% of the >1 mm size fraction, spanning ∼20–9 kyr B.P. is of little use for correlation to the IRD events at the European Margin sites. Therefore we have no firm control on the timing of isotope patterns along the benthic isotope records of NEAP 4K in relation to MD01-2461 and SO75-26KL and will use the record for a qualitative assessment of regional gradients only.
 During the last glacial, an eastern branch of GNAIW was advected along the European Margin toward the Portuguese Margin, while northward flowing SCW (likely AAIW) penetrated as far north as the Moroccan Margin as is indicated in paired benthic δ13C-Cd/Ca profiles (Figure 1a) [Willamowski and Zahn, 2000]. GNAIW therefore maintained the well-ventilated ambient bottom water conditions (δ13C > 1.4‰ VPDB) recorded at both MD01-2461 and SO75-26KL during mean glacial conditions. Conversely, the offset, of up to 0.5‰, between the MD02-2461 and SO75-26KL benthic δ18O records over the glacial interval suggests the additional influence of a warmer and/or low-δ18O glacial mid-depth water mass at the upper Portuguese margin [Zahn et al., 1997].
 Similar absolute values of benthic δ13C are recorded at both SO75-26KL and MD01-2461 associated with and following IRD deposition at H4, 2 and 1. However, decline into these benthic δ13C anomalies starts considerably earlier at the Portuguese margin and appears more gradual, notably in the periods prior to H4 and H1. That is, mid-depth ventilation appears to deteriorate at the Portuguese margin up to 5 kyr prior to these two H events, and importantly, before the collapse of mid-depth ventilation at the site of MD01-2461. This contrast in mid-depth ventilation plausibly reflects the proximity of the northerly MD01-2461 to the site of mid-depth convection that provided the site with well-ventilated mid-depth waters even though the production of these waters was in decline. If so, the early decrease in benthic δ13C at the Portuguese margin suggests that the production of mid-depth waters started to deteriorate well before H1 and H4 presumably because of a gradual built-up of meltwater surging before the LIS destabilized and large-scale iceberg calving occurred. The benthic δ13C proxy close to the centers of convection would not resolve such an early decline in mid-depth convection.
3.3. Precursory Meltwater Forcing From the NWEIS
 GNAIW convection in the open North Atlantic at the LGM [Sarnthein et al., 1994; Vidal et al., 1997], proximal to the fully advanced NWEIS, facilitated close coupling of NWEIS meltwater and overturning circulation. Within the H2 to H1 interval, benthic δ13C at MD01-2461 suggests that bottom waters at this site remained well ventilated by GNAIW until ∼16.6 kyr B.P., the exception being the apparent transient advance of the GNAIW/SCW hydrographic front north of ∼52°N at meltwater events “a” and “b.” At SO75-26KL, benthic δ13C fell steadily over a ∼5 kyr period approaching H1, following an abrupt shift by −0.5‰ at 21.4 kyr B.P. Initiation of this trend of decreasing ventilation at SO75-26KL is coincident with event “a” at site MD01-2461 when the apparent advance of SCW as displayed in a likewise brief episode of benthic δ13C depletion centered on 21.2 kyr B.P. On the basis of an IRD assemblage at MD01-2461 dominated by BIS-derived lithologies, this event has been suggested to constitute a reduction in GNAIW formation during an episode of NWEIS instability and associated meltwater surging into the NE Atlantic [Peck et al., 2006]. The end of this freshwater surge allowed intermediate water production to resume, reverting benthic δ13C values at MD01-2461 back to elevated/GNAIW values at 20.9 kyr B.P. However, continuing decrease of benthic δ13C values recorded at SO75-26KL suggest GNAIW production did not fully recover and that North Atlantic THC was progressively weakening prior to the incursion of H1 icebergs and meltwater to the NE Atlantic. As no further meltwater forcing is evident in the planktonic δ18O records of MD01-2461 until event “b,” at 17.8 kyr B.P., convection was plausibly reduced by freshwater forcing at higher latitudes [e.g., Elliot et al., 2002] or further to the west. 231Paxs/230Thxs ratios at DAPC2 (1709 m water depth) in the Rockall Trough, increase toward production values at ∼18.0 kyr B.P., suggest substantially reduced rates of overturning [Hall et al., 2006] concurrent with event “b” which may represent either a second transitory advance of SCW or an episode of brine injection to MD01-2461 preceding H1 by ∼0.9 kyr. However, Gherardi et al.  use 231Paxs/230Thxs records from the Iberian Margin to suggest that “shallow” overturning in the NE Atlantic basin was vigorous until 16.5 kyr B.P. Core SU81-18 used in their study is located at 3135 m water depth, some 2000 m deeper than both MD01-2461 and SO75-26KL. It may be possible that 231Pa export at depths below cores SO75-26KL, MD01-2461 and DAPC2 is recorded at SU81-18 accounting for these divergent signals, supporting a reduction in shallow overturning (≤1700 m), while deeper convection was perhaps maintained, accounting for the lower 231Paxs/230Thxs values recorded at SU81-18 at this time.
 The low-resolution record of NEAP-4K closely follows the structure of the benthic δ13C record of MD01-2461 and documents elevated δ13C values around 1.4‰ throughout the interval between H2 and H1, suggesting persistent bathing of this site with GNAIW. This pattern confirms our contention of continued production of GNAIW, albeit at lower rates in response to NWEIS instabilities, at “a” and “b.”
 Benthic δ13C values of <0.7‰ VPDB at all three intermediate water depth sites reflect the prominence SCW between 16.2–15.4 kyr B.P., implying large-scale collapse of GNAIW production following H1. Unlike the coupled H layer-benthic δ13C collapse records of SO75-26KL, NEAP-4K and DAPC2 [Knutz et al., 2002; Hall et al., 2006], significantly reduced ventilation at MD01-2461 is not recorded until several hundred years after H1. Geochemical (40Ar/39Ar dates of individual hornblende grains, Sr-Nd isotopic composition of carbonate-free IRD), magnetic susceptibility and lithological classification (dolomitic carbonate) each suggest that this is the only horizon within the deglacial interval (20–10 kyr B.P.) of MD01-2461 that contains a notable contribution of LIS-derived debris [Peck et al., 2006]. The H1 layer in MD01-2461 spans a few hundred years, whereas deposition at both DAPC2 and SO75-26KL is in the order of 1.0 kyr. While it appears plausible that MD01-12461 may have witnessed the earlier stages of H1 deposition in the NE Atlantic only, perhaps reflecting changing surface current patterns, it remains an issue to explain why benthic δ13C at MD01-2461 does not decrease until after the H1 layer in this core. One possible explanation is that the elevated δ13C value recorded immediately following H1 layer deposition and measured from a single specimen of C. wuellerstorfi (asterisk on Figure 2d) is not representative of ambient bottom water ventilation at this time and ventilation reduction at MD01-2461 was effectively simultaneous with H1. Alternatively, taking the benthic δ13C record at MD01-2461 at face value across H1, the elevated benthic δ13C levels may reflect a time-transgressive shoaling of reduced ventilation that reached the deeper sites in the north (NEAP-4K, DAPC2) first, before affecting MD01-2461.
 At H2, a “precursory” reduction in benthic δ13C is observed at SO75-26KL at 25 kyr B.P., ∼1 kyr prior to H layer deposition, accompanied by a brief anomaly at MD01-2461 at 24.6 kyr B.P. The abrupt decrease in benthic δ13C at SO75-26KL is synchronous with the equally abrupt increase in IRD from the NWEIS at MD01-2461, representing instability of the BIS immediately preceding H2 [Peck et al., 2007]. Surface and subsurface planktonic δ18O display an only minor divergence at this time (up to 0.5‰) indicating freshwater forcing at a smaller scale than during the “a” and “b” events, but its influence on GNAIW formation appears significant enough to produce the reduction in ventilation observed at SO75-26KL. Mid-depth ventilation recovers abruptly after this event before large-scale ventilation collapse occurs in the course of H2. Minimum ventilation of intermediate waters along the European Margin is seen in the benthic δ13C record of both SO75-26KL and MD01-2461 that immediately follows the H2 IRD deposition in MD01-2461. Significant destabilization of the BIS is suggested by a substantial increase in the flux of BIS-derived debris immediately following deposition of H2 at MD01-2461, perhaps triggered by sea level rise of up to 15 m associated with LIS-collapse [Yokoyama et al., 2001; Chappell, 2002]. Consistently the maximum (>2.5‰) offset between δ18O of the G. bulloides and N. pachyderma sin. which lags H layer deposition by ∼300 years is thought to be derived principally from NWEIS meltwater.
 Similarly, to H1, a progressive decrease in benthic δ13C preceding H4 by ∼2 kyr is observed in SO75-26KL, while benthic δ13C depletion at MD01-2461 occurred abruptly associated with H layer deposition. The short-lived increase in IRD flux associated with H4 plausibly reflects the juvenile state of the BIS at this time [Peck et al., 2007], a contention that appears to be confirmed by the lack of a coeval freshwater signal in planktonic δ18O (Figure 2c). The limited extent of the NWEIS, coupled with North Atlantic deep-intermediate water convection likely occurring in a similar location to the contemporary ocean prior to the stage 3-2 boundary [Sarnthein et al., 1994; Vidal et al., 1997] suggest the NWEIS was an unlikely source of meltwater for triggering the reduction in mid-depth ventilation prior to H4. Rather, initial reduction in GNAIW production likely reflects other meltwater sources [van Kreveld et al., 2000; Elliot et al., 2002].
3.4. Ice Sheet Instability in Response to North Atlantic THC Changes
 Several hypotheses attempting to explain the occurrence of H events, incorporating a range of internal and external forcing factors, have been proposed [e.g., MacAyeal, 1993; Johnson and Lauritzen, 1995; Marshall and Clarke, 1997; Hunt and Maslin, 1998; Arbic et al., 2004]. Recent concepts have used the disintegration of ice shelves fringing the Antarctic Peninsula as a modern analogue for the sudden iceberg releases during H events [Hulbe, 1997; Hulbe et al., 2004] and consider ocean subsurface temperatures, coupled with North Atlantic THC variability, as a factor that may have destabilized the LIS through their effect on ice shelves and fringing ice margins [e.g., Moros et al., 2002; Shaffer et al., 2004; Flückiger et al., 2006].
 The concept of recurrent meltwater release from the NWEIS and surface ocean stratification, promoting transient weakening of the North Atlantic THC and regional cooling particularly appears to apply for the period preceding H1. Such cool conditions may also have promoted the growth of a LIS-fringing ice-shelf, perhaps priming the LIS for H event collapse [Hulbe, 1997; Hemming, 2004; Hulbe et al., 2004]. Subsequent subsurface warming and sea level rise (0.3–0.5 m), associated with the THC reduction, may then have played a role in undermining the ice shelf thus removing the buttressing support exerted on the feeder ice streams, leading to large-scale surging of the ice sheet [Flückiger et al., 2006].