U and Th isotope constraints on the duration of Heinrich events H0‐H4 in the southeastern Labrador Sea

The duration and sequence of events recorded in Heinrich layers at sites near the Hudson Strait source area for ice-rat•ed material are still poorly constrained, otably because of the limit and uncertainties of the •4C chronology. Here we use high-resolution 23øTh-excess measurements, i a 6 rn sequence raised from Orphan Knoll (southern Labrador Sea), to constrain the duration of the deposition of the five most recent Heinrich (H) layers. On the basis of maximum/minimum estimates forthe mean glacial Z3øTh-excess flux at the studied site a minimum/maximum d ration of 1.0/0.6, 1.4/0.8, 1.3/0.8, 1.5/0.9, and 2.1/1.3 kyr is obtained for H0 (-Younger Dryas), H1, H2, H3, and H4, respectively. Thorium-230-excess inventories and other sedimentological features indicate a reduced but still significant lateral sedimentary supply by the Western Boundary Undercurrent during the glacial interval. U and Th series systematics also provide insights into source rocks of H layer sediments (i.e., into distal Irminger Basin/local Labrador Sea supplies).

Sedimentary supply at the site is also influenced by the Western Boundary Undercurrent (WBUC) whose highvelocity core is located slightly upslope [McCartney, 1992]. About 60% of the carbonate-free clay fraction in the Holocene section of the core is composed of smectites originating from the Irminger Sea and the Reykjanes Ridge areas (Figure 1) [Fagel et al., 1996]. Other clay minerals include illites and chlorites which reflect local terrigenous fluxes [Fagel et al., 1996]. Adding to this terrigenous component, authigenic carbonates linked to coccolithophorids production [Hillaire- Marcel et al., 1994b] represent almost 40% of the Holocene sediment. Their grain size averages 4 grn [Veiga-Pires, 1998]. It is thus likely that these carbonates may have also been transported from remote production areas, possibly as far as the Irminger Basin, to the study site.
On the basis ofsedimentological and clay mineral studies of the late Quaternary section of 91-045-094-P, Fagel et al.  Stoner et al., 1996]. The interlayered and overlying sediments were analyzed at 5 crn intervals when possible (i.e., when enough matehal was left from previous studies) and otherwise were analyzed at 10 cm intervals. Organic and inorganic carbon contents were measured using an elemental analyzer (Carlo-Erbam). The inorganic carbon content is expressed in equivalent CaCO3 (dry weight percent). The average overall analytical uncertainty of both inorganic and organic carbon (Corg) is 3% (ñ1 o).
The oxygen isotope stratigraphy was established on N. pachyderma (left coiling Npl) assemblages using an Isocarb TM preparation device on line with a triple-collector VG-Prism instrument. Results are expressed against Peedee belemnite (PDB) after applying the conventional corrections [Craig, 1957]. Overall analytical uncertainties determined from replicate measurements of standard carbonate were better than ñ0.05• (lo). Radiocarbon stratigraphy for 0-25 ka is based on accelerator mass spectrometry (AMS) 14C measurements on monospecific (Npl) assemblages made at the IsoTrace laboratory of the University of Toronto. Results were corrected by 400 years to account for the apparent age of the North Atlantic Ocean surface waters [Bard, 1988]  in activities (dpm g-1 of dry weight sediment) and in activity ratios. Thorium-230 excesses (henceforth 230Thxs) are considered to represent the fraction of 23øTh scavenged •m the water column by organomineral matter. In practice it is assimilated to the unsupported 23øTh in the sediment at the moment it settles (i.e., to the 23øTh activity above that of its 234U parent). In the present study, two distinct approaches ( where 23øThxs is the initial 23øTh excess, )• is the 23øTh decay constant equal to 9.1929x10 '6 yr '1, t is the sample age in calendar years, 232Th, 238U, 230Th, and 234U are the measured activities in dpm g-l, 1.14 is the modem seawater 234U/238U activity ratio [Chen et al., 1986], and Rd is the 238U/232Th activity ratio of the detrital supply (here 0.58 ñ 0.08; see discussion below).
Thorium-230 fluxes (in dpm cm '2 kyr '1) represent the decaycorrected 23øThxs (expressed in dpm g-l)multiplied by the sediment accumulation rate (gcm '2 kyr'l). Because of to weak constraints on many of the parameters which are involved in the calculations, we consider that 23øThxs and fluxes are known to no better than ñ10% level of uncertainty (see discussion).
These fluxes may provide a first-order estimate of 23øTh scavenging rates when advection and lateral transport are negligible. However, in the present context with a strong WBUC carrying the "young" water masses of the NW Atlantic   In P-094 all H layers display several common features (high carbonate content, low Corg, high coarse fraction, top light peak in •80, etc.), but also show discrete differences. In H0, H1, and to a lesser extent, H4 the coarse fraction content shows two peaks encompassing the high detrital carbonate pulse (Figure 2). This feature is not as clear in H2 and H3 either because it has never been recorded or because of artifacts (differences in biological mixing, variable penetration of IRD in the sediment, etc.). Other minor differences are observed, such as the maximum detrital carbonate contents, with values _> 40% in H0, H1, and H2 in contrast with lower maximums in H3 and H4 (25 and 3 5% respectively).  Figure   3). This is the ratio which was used to derive 23øThx, based on this approach (see (1) above). The large standard deviation of this ratio linked to the variability of (IRD/hemipelagic sediment) ratio represents the largest uncertainty in this method compared with the ñ 3% counting errors which, although almost negligible, were taken into account. Initial 230Thxs (vs,232Th) (dpm g-l)

Boundaries and Structure of Heinrich Layers
The precise definition of H layers is difficult when investigating H events in sequences where contrasting sedimentological regimes prevailed. In core P-094, H layers (H0-H4) all include a carbonate-rich subunit within an interval marked by a maximum in the grain size fraction > 125 pm (Figures 3 and 6). In the late Quaternary sediments of the Labrador Sea this large size fraction is mainly composed of IRD with a few foraminifer shells. Its relative abundance may  6) compared with that of H0. Therefore we will retain the limits of the H layers as defined above (i.e., CaCO3 color peak boundaries), which remain the best definition for the layer boundaries within the inherent uncertainty of such a deep-sea record.

23øTh• Duration of H Events
In order to estimate the duration of each H event fi'om H0 to H4 we calculated 2•oTh,• inventories in the corresponding H layers (as delimited above) and divided these inventories by the 2•oTh flux. We calculated confidence intervals for the inventories using standard errors. However, an empirical ñ10% estimate would be more reasonable taking into account all uncertainties. As a matter of fact, the largest uncertainties in the assessment of the H event duration lie in the difficulty to ascribe precise boundaries to the H layers and/or in the estimate of the 23oTh flux. Two distinct fluxes can be used, the vertical production rate of23øThx• (i.e., •-9 dpm cm '2 kyr '1) or the mean 23øThx, fluxofthe glacial sequence (i.e.,-•15 dpm cm '2 kyr'•), resulting in maximum/minimum estimates for the duration of the H events. In the first case one assumes an almost complete collapse of the WBUC sedimentary supply during H events and therefore no ventilation of the deep water mass. The second case implies a reduced but steady WBUC supply during the whole glacial interval (H events included). We consider that a "higher supply hypothesis" can be discarded as it would imply, for example, higher smectite contents than observed.
The following minimum-maximum duration was obtained (also see Table 1): H0, 0.6-1.0 kyr; H1, 0.8 -1.4 kyr; H2, 0.8 -1.3 kyr; H3, 0.9-1.5 kyr; and H4, 1.3 -2.1 kyr. The duration of H4 is questionable because the 23øThx, calculation for the corresponding layer is not well constrained. Indeed, the large variations in the 234U/2•SU activity ratios within this layer raise doubts about any precise assessment of the "supported" 23oTh fraction. The other age estimates are better constrained.
Although, the true maximum duration could be slightly lower than the above value, notably for H0, since the K, rm/K profiles suggest probable mixing by bioturbation and thus some possible addition of 2•øThx, on top of the layer from the overlaying hemipelagic clays. In Table 1, values from the present study are compared with others from the literature based either on •4C data and/or on 23øThx• calculations. We will limit our discussion to those based on radiometric methods, which should be directly comparable to this study. In most cases the estimates for any given event vary between authors by 100%, depending on the time series studied and the method used, but remain within a range of 0.5-2 kyr for all sets of events. Two main observations can be mme. First, the depositional times based on observations or calculations as above are generally greater than those yielded by theoretical models [i.e. Alley and MacAyeal, 1994; Matsumoto, 1996]. Second, the maximum values from our study are close to •4Cderived estimates, thus suggesting that the 23øTh scavenging rate of~9 9 dpm crrf 2 kyr '1 (i.e., the vertical production) is a better estimate of the unsupported 23øTh flux during these events than the mean glacial flux of ~15 9 dpm cm '2 kyr '•, which represents the entire glacial sequence. This would imply a reduced WBUC supply (and thus a reduced outflow) during the H events compared with the glacial period as a whole. This interpretation would be in agreement with a change in the thermohaline circulation during H events, notably a lesser production of North Atlantic Deep Water

Conclusion
The sequence which has been used to constrain depositional times for H0, HI, H2, H3, and H4 layers in the deep Labrador Sea has very specific features which strengthen some of the conclusions which can be made with respect to the duration of the corresponding depositional events using U series data. First, it is located directly under the trajectory of the icebergs released by one of the most active margins of the Laurentidian Ice Sheet, i.e., the Hudson Strait area [Andrews et al., 1994] and is in an area that receives detrital carbonate pulses which were triggered from the same source area and channeled by the NAMOC [Andrews et al., 1995]. Second, the site is located below the high-velocity core of the WBUC and sheltered by Orphan Knoll (Figure 1). It thus lies out of the direct erosional influence of this current, but it is likely influenced by its distal sedimentary supplies [Fagel et al., 1996]. Therefore the H layers of the studied core do represent the most exhaustive record for reconstructing the Laurentide ice dynamics along its NE margin. Furthermore, the 23øthxs maxim• duration for these events is well constrained since the unsupported 23øth flux used to calculate it corresponds to the vertical production rate of 23øTh, i.e., to the minimum 23øTh flux conceivable (as sediment winnowing can be discarded here). This maximin duration apparently matches estimates based on 14C chronologies elsewhere and thus constitutes the most probable value for the depositional time. As a consequence, the minimum duration, which was based on the assumption of stronger 230Th fluxes induced by enhanced lateral supplies through WBUC transportation, seems invalid.
If ascertained, this leads to the conclusion that the WBUC outflow was significantly reduced during H events and that 23øth export from the Labrador Sea by deep ventilation was similarly reduced.
Nevertheless, the maxim umdepositional times of~l.0 ñ 0.1, •-1.4 ñ 0.1, •-1.3 ñ 0.1, ~1.5 ñ 0.1, and •-2.1 ñ 0.1 kyr, which we obtained for H events H0-H4, respectively, are generally compatible with estimates based on other data when they exist. In this sequence the duration of H4 remains questionable because of the very peculiar U-Th systematics of this layer compared with the others. All other events yield a duration of~l-l.5 kyr. This suggests that the mechanisms involved in the deposition of these layers did not differ drastically from one to the other. This conclusion would be in favor of the glacier internal forcing mechanism for the H events