δ18Oc values in core CR02-23 decrease in a stepwise manner from the base (∼3.0 ‰ at 45 cm) to the top of core, and reach a minimum value of ∼2.7 ‰ in the upper 3 cm (Figures 2 and S1 of the auxiliary material). In core MD99-2220, δ18Oc values below 28 cm depth average 3.09 ± 0.08 ‰ (Figure 3). These values are higher, and are constrained within a far narrower range than those recorded in the upper 28 cm (3.0 to 2.6 ‰) of cores MD99-2220 and CR02-23 (Figure 3). In addition, a set of heavier values averaging 3.15 ± 0.05 ‰. (n = 36) is recorded between 55 and 90 cm in core MD99-2220 (n = 36). This set of values is significantly heavier than those recorded within the 28 to 55 cm interval and below 90 cm (see Figure 3).
Figure 2. Available instrumental temperature (red), oxygen (green) and salinity (dark blue) records for the St. Lawrence bottom water at a depth >300m against age (from the CLIMATE database). Theoretical δ18Oc (dotted light blue) and smoothed theoretical δ18Oc (solid light blue) were calculated from the relationship between salinity, δ18Ow, and the instrumental temperature in the isotope paleotemperature equation (1) and is plotted beside δ18Oc measured in core CR02-23 (purple). δ18Oc measured in core CR02-23 was used to estimate paleotemperature with the equation (1) (black). Error bars report uncertainty for the estimated temperature calculation ( ±0.2°C), propagated from the analytical precision of the δ18Oc measurements ( ±0.05 ‰), and for the analytical precision itself.
Download figure to PowerPoint
 Core CR02-23 encompasses the last century, allowing a comparison with historical, instrumental temperature data. δ18Oc values were converted into temperatures using equation (1), assuming, as a first approximation, a constant isotopic composition of about −0.05 ± 0.04 ‰ for the bottom water mass. This value has been retained as a first estimate, based on seven δ18Ow measurements carried out in 2007 and 2009, near the study site , which yielded values of −0.04 ± 0.03 (n = 6) and −0.11‰ (n = 1), respectively. As illustrated in Figure 2, the calculated and measured temperature curves show very similar patterns but the magnitude of the temporal temperature variation estimated from isotopic data is +1.3 ± 0.2°C (Figure S1) vs +1.7 ± 0.3°C, based on instrumental data [Gilbert et al., 2005]. This suggests that part of the temperature-driven shift in δ18Oc−values may have been masked by a small positive shift in δ18Ow−values. In the observed temperature range, one would expect a nearly constant dδc/dt relationship of about −0.3‰/°C (from equation (1)). The actual 1.7°C increase in bottom water temperature should thus have induced a shift of about −0.5‰ in the δ18Oc−values of G. auriculata shells. The measured shift (−0.3‰) thus suggests a +0.2‰ increase in δ18Ow values to account for the difference. Gilbert et al.  proposed that the recent increase in SLBW temperature results from the decrease of the proportion of the cold and well oxygenated LCW relative to the warm and less oxygenated NACW in the water mass that feeds the Laurentian Channel. Using temperature and salinity as tracers, Gilbert et al.  estimated that the proportion of NACW in the water mass entering the Laurentian Channel trough Cabot Strait increased from 28 to 48% relative to the LCW between 1930 and 2003 AD. According to recent measurements, the two water masses have δ18Ow−values of about −0.5 (LCW) and +0.5 ‰ (NACW) [Khatiwala et al., 1999]. The δ18Ow values measured in the LSLE (∼−0.05‰) is consistent with Gilbert et al.'s estimate of the relative contribution (∼50:50) of the two water masses to the modern SLBW. The increasing proportion of NACW in this water mass, from ∼28% to 48%, since the 1930s, should have resulted in a ∼+0.2‰ shift of the SLBW δ18Ow values over the same period. This shift in SLBW δ18Ow value would account for the discrepancy between the temperatures inferred from the measured δ18Oc−record and the instrumental data since the 1930s. This conclusion rests on the assumption that the isotopic compositions of the NADW and LCW have been invariant since 1930. It should also be considered with caution given the small isotopic offsets involved. In order to evaluate the effect of a variation in δ18Ow on the estimated temperatures, we also calculated a theoretical δ18Oc from equation (1), using our estimates of δ18Ow based on salinity measurements and instrumental temperature data (Figures 2 and S2 of the auxiliary material). The resulting profile is similar to the measured δ18Oc in CR02-23, suggesting that the δ18Oc record can be used as a rough paleothermometer in the SLBW. Therefore, irrespective of the offset value, a significant warming (>1°C) during the course of the last century is indisputable. This is confirmed by a similar trend, with a comparable amplitude, observed at another site of the Laurentian Channel in the Gulf of St. Lawrence (Figure 3) [Genovesi et al., 2008]. The similarity of δ18Oc records from the three sites in the Estuary and Gulf of St. Lawrence demonstrates that the warming trend of the last century is a regional feature. In contrast, prior to the last 100 years, the isotopic record from core MD99-2220 is nearly invariant, with a mean δ18Oc value of 3.09 ± 0.08 ‰ below 28 cm (Figure 3), i.e., for the time interval spanning from about 1000 to 1900 AD. However, a possible feature of significance is the set of heavier values recorded in the 55 to 95 cm interval (i.e., between ∼1630 and 1800), peaking at about 70 cm (∼1740). It suggests a cooling of less than 0.5°C, possibly linked to the Little Ice Age (LIA) (Figure 4). Paradoxically, data from cores collected in the area of the Laurentian fan, at the outlet of the Laurentian Channel in the North Atlantic, indicate sea-surface warming during the LIA [Keigwin and Pickart, 1999]. This has been interpreted as a northward shift of the slope water current in response to a dominant negative North Atlantic oscillation mode. The hydrography in the area of the Laurentian Fan is complex, as it is located at a front marked by the mixing of three water masses (the Labrador Sea Current, the Labrador Sea Water, and the North Atlantic Drift). Any change in the strength and trajectory of any or several of these water masses may have resulted in changes of surface water characteristics. The hypothesis of Keigwin and Pickart  may be correct for surface waters south of 44° but it does not necessarily apply to a water mass collected below 400 m in the NW North Atlantic and carried into the SLBW.
 Following the interval which we associate with the LIA, the warming of the SLBW appears to have occurred in two main steps. The first one is tenuous (∼0.3°C) but, nonetheless significant. It started at the beginning of the 19th Century and is consistent with the Northern Hemisphere compilation of climate changes by Moberg et al.  (Figure 4). This early warming could reflect the recovery from the LIA, although one may argue that it results from anthropogenic forcing. The second warming phase started at the turn of the 20th Century and is more pronounced, >1°C over the last century. Such a warming is seen in a large array of paleotemperature records [e.g., Crowley, 2000]. In the SLBW, the warming likely results from the increased temperature of North Atlantic waters entering the Gulf of St. Lawrence through Cabot Strait. Hence, it would reflect either the change in proportion of parent water masses or the distal effect of heat accumulation in constituent waters masses or both. Whether or not the post-1900 warming is due to anthropogenic forcing is a matter of debate, which we do not address here. Nonetheless, irrespective of the precise mechanisms responsible for the temperature variations reconstructed from core MD99-2220, it is unquestionable that the last century has been marked there by a warming trend having no equivalent over the last millennium.