During the last ice age, the density gradient across the Florida Current was reduced, implying a reduction in the flow of the Gulf Stream through the Florida Straits. Here we investigate the possibility that a significant portion of this wind-driven western boundary flow bypassed the Florida Straits during glacial times due to either changes in bathymetry induced by the sea level drop or changes in wind patterns. Down core records of the oxygen isotope ratios of the planktonic foraminifer Globorotalia truncatulinoides are used to locate the density gradients and thus the locations of upper ocean currents in the western North Atlantic. We find that western boundary flow was largely confined within the Florida Straits during the Last Glacial Maximum as it is today. This finding supports the idea that the reduced density gradient across the Florida Current represents a reduction in the surface branch of the surface to deep meridional overturning circulation in the Atlantic rather than a reduction in the proportion of the wind-driven flow carried by the Florida Current.
 The surface currents in the Caribbean and western North Atlantic incorporate portions of both the wind-driven subtropical gyre and the meridional overturning circulation of the Atlantic. The northward flowing surface waters which compensate the outflow of North Atlantic Deep Water cross the equator and enter the Caribbean through its southernmost passages, eventually joining the southernmost extent of the Gulf Stream as it passes through the Florida Straits. The westward moving surface waters of the southern portion of the North Atlantic Subtropical gyre enter the Caribbean through various passages. The transport through the Florida Straits is about 30 Sv (1Sv = 106m3/s [Baringer and Larsen, 2001]), comprising both the subtropical gyre circulation in the western North Atlantic (17 Sv) as well as the northward flow of ∼13 Sv of surface waters that compensate for the southward flow of deep waters formed in the North Atlantic [Schmitz and McCartney, 1993] (see Figure 1). About 10% of the flow through the Florida Straits at 27°N arrives through the Santaren Channel, between Cay Sal and the Great Bahama Bank, and the northernmost passage, the Northwest Providence Channel, north of the Great Bahama Bank [Atkinson et al., 1995; Leaman et al., 1995]. North of the Florida Straits, the Gulf Stream flows along the coast until Cape Hatteras, where it turns seaward and separates from the coast. Wind-driven recirculation cells confined to the western side of the basin account for an enhancement in transport of the Gulf Stream from 30 Sv through the Florida Straits, to more than 100 Sv at its separation off of Cape Hatteras [Schmitz and McCartney, 1993].
 Down core ocean sediment records of δ18Ocalcite provide a means for examining patterns of upper-ocean flow in the past. Using benthic foraminifer δ18Ocalcite to reconstruct seawater density on either side of the Florida Straits, Lynch-Stieglitz et al.  assess the vertical shear of Florida Current and find it reduced during the last glacial maximum (LGM). The findings of Lynch-Stieglitz et al.  are consistent with a reduced transport of the Gulf Stream through the Florida Straits. However, glacial sea level was lower by ∼130 meters [Lambeck and Chappell, 2001; Siddall et al., 2003] which would have restricted the passages leading into the Caribbean and perhaps diverted some wind-driven flow seaward of the Florida Straits. The most constricted passage through the Florida Straits is only ∼760 meters deep, and during the LGM, its depth would have been reduced to ∼620 meters, which may have further impacted the flow in this region. Further north, Matsumoto and Lynch-Stieglitz  find that the modern separation latitude of the Gulf Stream further north off the coast of Cape Hatteras was unchanged during the LGM.
 In this paper we will attempt to determine the flow path of the surface currents of the western North Atlantic using the oxygen isotope composition from subsurface calcifying planktonic foraminifera.
2.1. Controls on the Oxygen Isotopic Composition of Gr. truncatulinoides
 To a first order approximation, large scale steady state upper-ocean flows are in geostrophic balance; thus the vertical shear of the flow is balanced by the horizontal density gradient across the flow. Locating horizontal density gradients provides a method to determine the path of upper-ocean flows since they occur in the same places as the shear of the upper-ocean flows. In the North Atlantic, given a specific depth, seawater density is lower seaward of the Gulf Stream and higher landward.
 Because the oxygen isotope ratio in the calcite shells of foraminifera (δ18Ocalcite), like seawater density, is strongly related to temperature and salinity, it can serve as a proxy for seawater density [Lynch-Stieglitz et al., 1999]. The δ18Ocalcite increases with decreasing temperature of calcification [Emiliani, 1955]. The isotopic composition of calcite is also determined by the isotopic composition of the water in which it calcifies. The isotopic composition of seawater is regionally linearly related to salinity since fluxes of fresh water, including evaporation and precipitation closely affect both [Craig and Gordon, 1965]. Higher δ18Ocalcite implies that the calcite formed in seawater with higher potential density.
 The oxygen isotope ratio in the shell of the planktonic foraminifera Globorotalia truncatulinoides (δ18Otrunc) provides a useful proxy for upper-ocean density gradients associated with the shear in upper-ocean flows because it approximates water properties from intermediate depths (300–500 meters). Examining water column properties from these depths is particularly useful because this range is below the surface layer where air-sea exchanges of heat and fresh water can complicate the structure of lateral density gradients. In addition, the correlation between upper-ocean horizontal density gradients and the location of upper-ocean flows is particularly strong because of a maximum in the vertical shear in velocity of upper-ocean flows at these intermediate depths.
 We have shown that the horizontal density gradients associated with upper-ocean flows can clearly be seen in the spatial pattern of core top δ18Otrunc [LeGrande et al., 2004]. While the average isotopic composition of the test of Gr. truncatulinoides corresponds to calcification at around 350 meters water depth, the life cycle of this organism is more complex. This species most likely begins calcification in surface winter waters and then drops to as deep as 800 meters adding a secondary δ18O enriched calcite crust [Lohmann, 1995]. Atlantic core top samples of Gr. truncatulinoidesδ18O most closely resemble that of an idealized calcite calculated from water column properties at 350 meters depth, equivalent to 30% calcite addition at the surface and 80% calcite addition at 800 meters [LeGrande et al., 2004], and this finding does not change when only core samples from the western North Atlantic are considered (using the same method). While Gr. truncatulinoides has its largest flux to the sediments during wintertime [Deuser and Ross, 1989], the isotopic composition approximates mean annual conditions deeper in the water column where there is less seasonality; e.g., the difference between winter and annual calcification for the western North Atlantic is <0.05‰ in the 350 meter calcification case and <0.1‰ in the 30% surface, 70% 800 meter calcification case.
 These uncertainties in the exact calcification depth(s) of Gr. truncatulinoides suggest that though it is a good qualitative measure of current location, and it is not an adequate tool to derive quantitative current strength. Matsumoto and Lynch-Stieglitz  used δ18Otrunc gradients to determine the separation latitude of the Gulf Stream at Cape Hatteras during the LGM. Here we examine down core records of δ18Otrunc from additional core sites in the western North Atlantic to infer the pattern of circulation during the LGM.
2.2. Sediment Core Data
 We have examined core sites in the Gulf of Mexico, on the landward side of the Florida Current near the Florida Keys, on the seaward side of the Florida Current near the Bahamas, and in the open western North Atlantic northeast of the Bahamas (Table 1; Figure 1). For the purposes of this LGM study, we required core sites to be greater than 450 meters depth during the LGM (580 meters depth for the present) to ensure that the δ18Otrunc would not be biased toward shallow (lighter δ18Otrunc) due to the inability of Gr. truncatulinoides to calcify over its full range. (The shallowest core site included in the study of LeGrande et al.  is 452 meters.) In the modern ocean there is a significant δ18Otrunc gradient between the Bahamas and the Florida margin, reflecting the presence of the Florida Current. If there was a significant flow east of the Bahamas during the LGM, we would anticipate a stronger δ18Otrunc gradient between the Bahamas and the open ocean site.
Table 1. Core Location and Depths Where the Oxygen Isotope Ratio in Gr. truncatulinoides, δ18Otrunc, Is Used to Infer the Location of the Florida Current With the Range of Values Indicateda
Mean δ18Otrunc, ‰
Min. δ18Otrunc, ‰
Max. δ18Otrunc, ‰
Mean δ18Otrunc, ‰
Min. δ18Otrunc, ‰
Max. δ18Otrunc, ‰
The stratigraphy column (STRAT., right) indicates either 14C dating or oxygen isotope stratigraphy from both Gr. truncatulinoides (425–500 μm) and the following: GS, G. sacculifer (355–425 μm); CW, C. wuellerstorfi (300–600 μm); GR, G. ruber white (300–355 μm); and CP, C. pachyderma (>300 μm).
 Samples from these cores were prepared according to the methods described by Matsumoto and Lynch-Stieglitz . Holocene and LGM ages were determined from down core oxygen isotope stratigraphy on surface dwelling planktonic foraminifera (Globigerinoides sacculifer and Globigerinoides ruber), benthic foraminifera (Cibicidoides wuellerstorfi, Cibicidoides pachyderma), or 14C dating on planktonic foraminifera (Figure 2, Table 1). We analyze three to five Gr. truncatulinoides shells from the sieve interval between 425–500 μm, a range in which most specimens contain significant δ18O enriched secondary calcite crust that is added at depth [Lohmann, 1995]. Mean δ18Otrunc values, and the range of values, are calculated at each core site by averaging δ18Otrunc values over Holocene and LGM intervals (Table 1).
3. Results and Discussion
 In the modern ocean, the average δ18Otrunc value landward of the Gulf Stream is 1.1‰, while the average seaward δ18Otrunc value ranges from 0.7 to 0.4‰. Intensified Gulf Stream transport coupled to North Atlantic gyre recirculation cells may contribute to the more depleted δ18Otrunc values of the more north and eastern sites that are more central to the North Atlantic gyre. The δ18Otrunc gradient across the Florida Current in the modern ocean is ∼0.4‰, with a further ∼0.3‰ seaward gradient most likely associated with the gyre recirculation cell (Figures 3a and 3b; Table 1). If there was a significant diversion of flow outside the Florida Straits, we would anticipate (1) a gradient between cores on the eastern edge of the Florida Straits (Sites F and G) and the core seaward of the Bahamas (Site H) and (2) no (or a significantly reduced) gradient across the Florida Straits.
 The average δ18Otrunc value for the Bahamas (Sites F and G, east of the Florida Straits) as well the core east of the Bahamas (Site H) during the LGM is ∼1.7‰; the δ18Otrunc value of sites on the west side of the Florida Straits is ∼2.0‰. The δ18Otrunc gradient across the Florida Straits during the LGM is ∼0.3‰, greater than the LGM temporal variability associated with each core site, and no further seaward gradient exists (Figure 3c). Since the location of the δ18Otrunc gradient across the Florida Straits did not change during the LGM, we infer that there was vertical shear associated with the LGM Florida Current inside the Florida Straits. The lack of a δ18Otrunc gradient in between the Bahamas (Sites F and G) and the site in the open Atlantic to the east (Site H) indicates that there was no significant deflection of the Florida Current outside of the Florida Straits during the LGM.
 The Holocene δ18Otrunc from the cores on the gyre side of the Gulf Stream south of 30°N (Sites F and G) and the open Atlantic cores north of 30°N (Sites I, J, K) [Matsumoto and Lynch-Stieglitz, 2003] do exhibit both LGM and modern gradients of ∼0.4‰. The LGM to modern change is that core K140-2 JPC22 (Site H) is apparently outside the recirculation cell during the LGM, but inside the recirculation cell in the present. This change seems to imply a shift in glacial wind patterns and perhaps gyre circulation; however, due the inherently qualitative nature of δ18Otrunc as a proxy for upper ocean flows, including uncertainty in the actual calcification depth and the season of calcification, this potential northward migration of the recirculation cell cannot be concluded with this data alone. More cores from either side of this recirculation cell boundary in the modern ocean and LGM and additional corroboration with other proxies would be required to describe such a potential change in circulation. On the basis of this δ18Otrunc proxy, it appears that the Florida Current flowed through the Florida Straits during the LGM much as it does today.
 Implicit in this interpretation is that the relationship between intermediate depth density and δ18Otrunc did not change from the modern to the LGM. Modeling studies have indicated that the relationship between the oxygen isotopic composition of seawater and salinity at millennial scales are different (steeper) than at decadal and shorter timescales [Schmidt et al., 2007], and the possibility exists that a modulation of the relationship between density and δ18O has also occurred. If the sensitivity of the δ18Otrunc proxy to these density changes was dampened or amplified, an altered upper ocean flow of the LGM might yield an identical δ18Otrunc gradient.
 The δ18Otrunc proxy faithfully indicates patterns of upper ocean flow in the modern ocean [LeGrande et al., 2004], and so we use it to indicate the basic pattern of past upper ocean flow in the western North Atlantic for the LGM. We find that there is a glacial δ18Otrunc gradient confined within the Florida Straits, similar to the modern δ18Otrunc gradient associated with the vertical shear in the Florida Current. There was no additional intermediate depth density gradient east of the Bahamas which implies no significant Gulf Stream flow outside of the Bahamas. Lynch-Stieglitz et al.  identified a reduced density gradient within the Florida Straits and inferred a reduction in Florida Current transport during the LGM; this result paired with the apparent absence of a significant flow seaward of the Bahamas indicates that there was a reduction of the transport of the southernmost portion of the Gulf Stream during the LGM and not a diversion of its path.
 Additionally, the location of the intermediate depth density gradient and implied separation latitude of the Gulf Stream of Cape Hatteras during the last ice age was the same as today [Matsumoto and Lynch-Stieglitz, 2003]. The pattern of upper ocean flows along the western North Atlantic during the last ice age was apparently similar to the modern pattern. The implication of these findings is that any changes in the wind or changes in ocean bathymetry due to sea level drop were not sufficient to significantly change the overall pattern of ocean circulation of the western North Atlantic from the modern to the last ice age.
 This work was supported by NSF awards OCE-0096472, OCE-9984989, and OCE-0428803 to J.L.-S. Sample material used in this project was provided by the Lamont-Doherty Earth Observatory Deep-Sea Sample Repository. Support for the collection and curating facilities of the core collection is provided by the National Science Foundation through grant OCE03-50504. A.N.L. was supported by a National Defense Science and Engineering Graduate Fellowship and received institutional support from the NASA Goddard Institute for Space Studies. We also thank Trond Dokken and Stefan Mulitza for their helpful reviews.