Holocene evolution of the North Atlantic subsurface transport

. Previous studies suggested that short term freshening events in the subpolar gyre can be counterbalanced by interactions with the subtropical gyre and thus stabilize the Atlantic Meridional Overturning Circulation (AMOC). However, little is known about the intergyre transport pathways. Here, we reconstruct surface and subsurface transport between the subtropical and polar North Atlantic during the last 10000 years, by combining new temperature and salinity reconstructions 10 obtained from surface and subsurface dwelling foraminifera with published data from the tropical and subpolar North Atlantic and published foraminiferal abundance data from the subtropical North Atlantic. These observations imply an overall stable warm surface water transport. Subsurface warm water transport started at about 8 ka with subtropical heat storage, and reached its full strength at about 7 ka, probably associated with the onset of the modern AMOC mode. Comparison of different potential forcing mechanisms suggests a freshwater control on these ocean transport changes. Sea (Brambilla and Talley, 2006; Foukal and Lozier, 2016; Lozier, 2012). In contrast, hydrographic data from ocean cross sections indicate a transport of warm, saline water between 10 and 1000 m water depth into the Irminger Basin and Labrador Sea (Sarafanov et al., 2012; Våge et al., 2011) with a recirculation and 20 mixing zone of NAC waters into the SPG in the North west corner of the SPG (Mertens et al., 2014) Most paleo reconstructions focus on the surface water structure in the subpolar North Atlantic (Came et al., 2007; Farmer et al., 2011; Staines-Uréas et al., 2013) and neglect the probably significant subsurface component of the transport. The two studies that investigated the interaction between the STG and SPG (Solignac et al., 2004a; Thornalley et al., 2009) focus on surface water pathways mainly, while changes in the tropical origin of the water masses are not considered. Thus a 25 systematic investigation of the subsurface transport from the tropical into the subpolar North Atlantic and its potential impact on the stabilization of the late Holocene AMOC mode is critically missing. Here, the role of surface and subsurface warm water transport from the tropics into the subpolar North Atlantic on the evolution and stabilization of the Holocene climate mode is investigated, and different forcing mechanisms are entertained. This knowledge is pivotal to establish robust scenarios for future AMOC changes under ongoing anthropogenic climate change leading to warming and potential 30 freshening of the subpolar NA. is ideal to trace the varying position of the STG and associated 20 varying influence of subtropical and temperate waters (Repschläger et al., 2015). Results are compared to published data from the tropical (Bahr et al., 2013; Cléroux et al., 2012) and subpolar North Atlantic (Thornalley et al., 2009) along the warm water route. We combine published δ 18 O data of planktonic surface and subsurface dwelling foraminifera G. ruber w. and G. truncatulinoides (Repschläger et al., 2015) with new Mg/Ca records from the same species. Mg/Ca analyses followed the 25 same procedure as in (Repschläger et al., 2015), with a 2 σ STD of +/-0.20 mmol/mol Mg/Ca for G. ruber w. and +/-0.14 mmol/mol Mg/Ca for G. truncatulinoides . To convert Mg/Ca values into water temperature estimates, we used a species-specific calibration of G. ruber w ., and an equation for mixed subsurface dwellers for G. truncatulinoides, both published by Cléroux et al., (2008). The summed analytical and calibration 2 σ STD is +/- 1 °C for G. ruber w. and +/- 2 °C for G. truncatulinoides . The calibrated Mg/Ca temperatures for both species from our core top match well with modern surface and 30 subsurface (200 m depth) temperatures at the Azores coring site, and seasonal overprints on foraminiferal calcite are

identification and understanding of the importance of the different driving forces and stabilizing mechanisms for the Holocene climate remains challenging. It is assumed that the stabilization is related to a strong Atlantic Meridional Overturning Circulation (AMOC), driven by deep water formation in the Nordic Seas and the Labrador Sea, and fueled by the northward transport of warm saline waters. In this scenario, the main stabilization occurred via the formation of Labrador Sea Water (LSW), which is inferred to have been established at about 7 ka BP (Hoogakker et al., 2011;Kissel et al., 2013;5 Solignac et al., 2004b;Thornalley et al., 2013). A stable LSW formation is related to the interaction between warm water transport in the clockwise circulating subtropical gyre (STG) and the subpolar gyre (SPG) circulating counterclockwise in the Irminger Basin and the Labrador Sea (Thornalley et al., 2009). Antiphase heat and salt storage within the SPG during times when STG freshens (Cléroux et al., 2012;Thornalley et al., 2009) and the later advection of the warm saline water into the SPG leads to a strengthening of the deepwater formation by convection within the SPG (e.g. Born and Stocker, 2014). 10 Investigation of the different pathways of the warm water masses from the tropical into the subpolar North Atlantic is limited to a few studies thus far. The transport into the subpolar North Atlantic and the Nordic Seas via surface flow within the NAC, on the other hand, is relatively well known from oceanographic studies  and paleoceanographic reconstructions (e.g. Andersen et al., 2004;Came et al., 2007;Farmer et al., 2011;Giraudeau et al., 2010;Moros et al., 2004;Staines-Uréas et al., 2013). 15 The warm water transport into the Labrador Sea, however, is still a matter of investigation. Drifter and modeling studies of the modern ocean indicate a lack of warm surface water transport vis-a-vis a major warm water transport into the SPG via a subsurface (>500m) pathway into the Irminger Sea (Brambilla and Talley, 2006;Foukal and Lozier, 2016;Lozier, 2012). In contrast, hydrographic data from ocean cross sections indicate a transport of warm, saline water between 10 and 1000 m water depth into the Irminger Basin and Labrador Sea Våge et al., 2011) with a recirculation and 20 mixing zone of NAC waters into the SPG in the North west corner of the SPG (Mertens et al., 2014) Most paleo reconstructions focus on the surface water structure in the subpolar North Atlantic (Came et al., 2007;Farmer et al., 2011;Staines-Uréas et al., 2013) and neglect the probably significant subsurface component of the transport. The two studies that investigated the interaction between the STG and SPG (Solignac et al., 2004a;Thornalley et al., 2009) focus on surface water pathways mainly, while changes in the tropical origin of the water masses are not considered. Thus a 25 systematic investigation of the subsurface transport from the tropical into the subpolar North Atlantic and its potential impact on the stabilization of the late Holocene AMOC mode is critically missing. Here, the role of surface and subsurface warm water transport from the tropics into the subpolar North Atlantic on the evolution and stabilization of the Holocene climate mode is investigated, and different forcing mechanisms are entertained. This knowledge is pivotal to establish robust scenarios for future AMOC changes under ongoing anthropogenic climate change leading to warming and potential 30 freshening of the subpolar NA. Clim. Past Discuss., doi:10.5194/cp-2016-115, 2016 Manuscript under review for journal Clim. Past Published: 21 November 2016 c Author(s) 2016. CC-BY 3.0 License.

Regional Setting
The warm water circulation in the North Atlantic is governed by the North Atlantic Current (NAC) that transports warm saline water (e.g. Roessler et al., 2015) into the high northern latitudes. On its northward path, several warm water currents branch off the main pathway ( Figure 1). The most distinct ones are the Azores Current (AC) and the Irminger Current (IC).
The eastward flowing AC leaves the NAC between 30 and 40°N and forms the northern boundary of the STG. The latter 5 rotates clockwise and recirculates warm water between 10 and 40°N. The AC is associated with the Azores Front (AF) that is characterized by strong eddy activity and forms the boundary between cool transitional and warm oligotrophic subtropical waters. The AF is characterized by the depth of the 15° isotherm (Gould, 1985) and is therefore not traceable in SST (Alves et al., 2002) but by distinct shifts in foraminiferal assemblages from subtropical to transitional/subpolar (Ottens and Nederbragt, 1992;Schiebel et al., 2002b) and thus also by abundances of G. ruber w. (Repschläger et al., 2015 and citations 10 therein). The strength and position of the westerly wind belt control the position of the AF and AC (Volkov and Fu, 2011) with increasing/decreasing wind strength leading to a southward/northward displacement of the AF, respectively.
South of Iceland at about 50°N, the IC branches off the main northward path of the NAC. It flows in north-westward direction following the bathymetry of the Irminger Basin. South of the Denmark Strait, The IC returns southward following the coast of Greenland into the Labrador Sea (Spall and Pickart, 2003). It thereby feeds warm saline waters into the 15 counterclockwise rotating SPG, and fuels deepwater convection in the Labrador Sea (Born and Stocker, 2014;Våge et al., 2011).

Methods
For our study, we used core GEOFAR KF16 taken at 38°N, south of the Azores Islands at the Mid Atlantic Ridge, from 3060 m water depth. This position in the vicinity of the AF is ideal to trace the varying position of the STG and associated 20 varying influence of subtropical and temperate waters (Repschläger et al., 2015). Results are compared to published data from the tropical (Bahr et al., 2013;Cléroux et al., 2012) and subpolar North Atlantic (Thornalley et al., 2009) along the warm water route.
We combine published δ 18 O data of planktonic surface and subsurface dwelling foraminifera G. ruber w. and G.
truncatulinoides (Repschläger et al., 2015) with new Mg/Ca records from the same species. Mg/Ca analyses followed the 25 same procedure as in (Repschläger et al., 2015), with a 2 σ STD of +/-0.20 mmol/mol Mg/Ca for G. ruber w. and +/-0.14 mmol/mol Mg/Ca for G. truncatulinoides. To convert Mg/Ca values into water temperature estimates, we used a speciesspecific calibration of G. ruber w., and an equation for mixed subsurface dwellers for G. truncatulinoides, both published by Cléroux et al., (2008). The summed analytical and calibration 2 σ STD is +/-1 °C for G. ruber w. and +/-2 °C for G.
truncatulinoides. The calibrated Mg/Ca temperatures for both species from our core top match well with modern surface and 30 subsurface (200 m depth) temperatures at the Azores coring site, and seasonal overprints on foraminiferal calcite are Clim. Past Discuss., doi:10.5194/cp-2016-115, 2016 Manuscript under review for journal Clim. Past Published: 21 November 2016 c Author(s) 2016. CC-BY 3.0 License. assumed to play a subordinate role at our coring site (Repschläger et al., 2015). In addition, we assume that the subsurface temperature signal of G. truncatulinoides is predominantly determined by the conditions at the AF and not by the migration of G. truncatulinoides to shallower, warmer water depths or by thermocline shoaling (see supplementary information and (Repschläger et al., 2015)) Changes in salinity are reconstructed following the procedure described in Repschläger et al. (2015). The temperature effect, 5 estimated from the Mg/Ca was removed from the foraminiferal δ 18 O carbonate using the general equation of Shackleton (1974) for G. ruber w. and G. truncatulinoides in order to be consistent with the datasets used for comparison.. A correction for VPDB to SMOW was included. The δ 18 O w records were corrected for ice volume, using the relative sea level composite curve of Waelbroeck (2002). For δ 18 O w uncertainty evaluation we followed the argumentation of Cleroux 2011 and used and used eq. S8 of Cleroux et al., (2011), leading to an calculated 2 σ STD of +/-0.35 ‰ for G. ruber w. and +/-0.68 ‰ for G. 10 truncatulinoides, respectively. Given the relative large error in reconstruction, we only used 5-point average data for interpretation to investigate longer-term trends in the datasets.
Changes in the AF front were reconstructed using the relative abundance of G. ruber w. published by (Weinelt et al., 2015) These abundance counts have an 2 σ STD +/-2.5 %. As G. ruber w. is most abundant within the STG (Ottens, 1991;Schiebel et al., 2002a;Storz et al., 2009), low/high abundances indicate a southward/northward movement of the AF relative 15 to the coring site. Because the position of the AF is related to changes in the westerly wind belt, the abundances of G. ruber w. indicate the relative contribution of subtropical water and can be used as tracer for the position of the northern STG rim and thus of the position of westerlies (Repschläger et al., 2015). records start to decrease to values between 1.5 ‰ 1.7 ‰ at 6 ka BP. After 6 ka BP surface and subsurface δ 18 O w-ivc records separate, only slightly exceeding the error of calculation.

Surface and subsurface
Low abundances of G. ruber w. (7-15 %) (Figure 2d) between 11 and 9.5 ka BP indicate a southward displacement of the AF and correlate with low T sub and low subsurface δ 18 O w-ice . High abundances of G. ruber w. (~20 %) between 8 and 4 ka BP indicate a northward movement of the AF, coeval to a warming of the subsurface waters and high δ 18 O w-ivc in G. 5 truncatulinoides. Between 6 and 7 ka BP a short excursion with a decrease in G. ruber w. abundances to as low as 10% coincide with an increase in the δ 18 O records and a decrease in T sub . This event is associated is related to a discontinuity in the sediment core.
After 4 ka BP, the abundances of G. ruber w. decrease to15%, indicating a southern position of the AF. This decrease does not coincide with any changes surface and subsurface T and in δ 18 O w-ivc . 10

Discussion
Based on the records presented here, and following the suggestion of the INTIMATE group (Walker et al., 2012), the Holocene can be divided into four sections.
The Early Holocene (11-8.2 ka BP) is characterized by a thermal difference between surface and subsurface waters. Together with the similarity in the δ 18 O w-ivc records, this points toward a T-driven stratification during a period when the AF was south 15 of its modern position. In the early Mid Holocene (8-6 ka BP) the thermal difference between surface and subsurface waters decreased while the difference in δ 18 O w-ivc remained small. This phase can be interpreted as an intermediate phase with a weak stratification that is accompanied by a northward movement of the AF, reaching its northernmost position. In the late Mid Holocene (6-4 ka BP) the AF remained at the northernmost position, evolving differences between surface and subsurface T and S indicate a stabilization of the thermohaline stratification. The Late Holocene (4-0 ka BP) differs from the 20 previous phase only in the position of the STG. The latter is displaced to a more southern position that is also observed under modern conditions.

Transport between tropical and subpolar North Atlantic
In order to investigate how changes in the mixed layer at our subtropical coring site can be related to changes in surface and subsurface warm water transport between the tropical and subpolar North Atlantic, we compare our records with records 25 from the tropical (Bahr et al., 2013;Cléroux et al., 2012) and the subpolar NA (Thornalley et al., 2009).
In general, SST estimates from the tropical, subtropical and subpolar sites ( determinations. Assuming that the surface water properties at the three coring sites are mainly driven by changes in the surface water transport, we conclude that the mean northward surface water transport remained stable over the Holocene. Thus the variability described above, must originate from changes in the subsurface transport. The subsurface signal from the tropical site is rather constant during the entire Holocene, with average tropical T sub fluctuating by +/-3°C around a mean of 17.5°C and δ 18 O w-ivc varying by 0.5 ‰ around a mean of 1.2 ‰, and with no apparent long-term trends In contrast, 5 distinct temporal differences are observed in the subsurface records at the subtropical and subpolar site (Figure 3b).
From 11 to 8 ka BP, coincident with a southern position of the AF, subtropical subsurface records are very similar to subpolar values with low T sub (12/10°C) and low subsurface δ 18 O w-ivc of about 0.5 ‰. The surface water records suggest active northward water transport, similar to the state described at the transition between the Allerød and the Younger Dryas (Repschläger et al., 2015). The low subsurface T and δ 18 O w-ice during the Early Holocene can be explained with an intrusion 10 of transitional Eastern North Atlantic Water (ENAW) reaching the Azores coring site (Figure 4a). This intrusion is probably related to the influence of meltwater from the Laurentide ice sheet that is also inferred from a freshening in the subpolar North Atlantic (Came et al., 2007;Thornalley et al., 2009), although the δ 18 O w-ivc needs to be confirmed by independent proxies.
During the Mid Holocene transitional phase (8-6 ka BP), subtropical and subpolar subsurface records diverge. The 15 subtropical T sub and subsurface δ 18 O w-ivc values converge towards those from the tropical record after 8.2 ka BP, while the subpolar T sub record remains in the Early Holocene mode and the δ 18 O w-ivc imply further freshening. This state coincides with a northward migration of the AF. Warm subsurface waters reached the Azores coring site, but was not transported further northward yet (Figure 4b).
At 7 ka BP subtropical and subpolar subsurface records start to converge toward the tropical record, with the subpolar record 20 lagging behind changes in the subtropical record by at least 1,000 years. This indicates the onset of modern transport of warm subsurface waters from the tropics into the subpolar region (Figure 4c). At the same time, LSW formation started with a temperature decrease in Labrador Sea surface waters (Solignac et al., 2004a) and decreasing warm water transport within the NAC into the Nordic seas (Staines-Uréas et al., 2013). These changes are potentially accompanied by a divergence of the warm water transport into the Irminger Basin and Labrador Sea at subsurface. To investigate this transport route further 25 high-resolution SST and T sub reconstructions from the subpolar gyre would be needed.
In the Late Holocene after 6 ka BP, the tropical, subtropical and subpolar records are showing parallel patterns with millennial-scale fluctuations in S and T that anti-correlate with records from the Labrador Sea as already described by Thornalley et al. (2009) andCléroux et al. (2012). With this circulation mode, short-term variability including anti-phasing between the SPG and STG temperature and salinity records is observed. Our data thus confirm earlier suggestions of anti-30 phasing between gyre properties during the Late Holocene (Cléroux et al., 2012;Thornalley et al., 2009) implying that shortterm freshening events in the Labrador Sea can be balanced by increased warm and saline water transport between the SPG Clim. Past Discuss., doi:10.5194/cp-2016Discuss., doi:10.5194/cp- -115, 2016 Manuscript under review for journal Clim. Past Published: 21 November 2016 c Author(s) 2016. CC-BY 3.0 License. and the STG stabilizing the AMOC. However, our data indicate that the major pathway in this stabilization mechanism must be the subsurface warm water route in the North Atlantic.
Surprisingly, no changes in the long-term trend of the Late Holocene are observed after 4 ka BP when changes in the G. ruber w. abundance indicate a southward movement of the AF position (Repschläger et al., 2015), that can be associated with a global southward shift of the wind systems (Wanner, 2008). 5

Driving factors for subsurface transport changes
In light of observed and anticipated anthropogenic climate change it is important to understand the underlying mechanism of Holocene climate variability. In addition to the forcing by increases in atmospheric CO 2, three driving factors for Holocene climate variability are discussed 1) changes in the atmospheric circulation related to the NH summer insolation (Renssen, 2005) 2) changes in meltwater forcing over the North Atlantic region due to melting of continental ice sheets (Mayewski et 10 al., 2004) and 3) changes in solar activity (Bond et al., 2001).In the following we will discuss which of those factors can be identified as main cause of the variability in subsurface transport.
Changes in NH insolation are coupled with changes in the position of the westerly wind belt (Renssen, 2005). Major changes in the atmospheric circulation are probably related to the early Holocene thermal maximum, however the position of the westerlies within this time period remains unclear. During the Late Holocene (since 4.3 ka BP), ITCZ changes (Wanner, 15 2008) are probably related to the southward movement of the westerly wind belt and southward displacement of the AF. The G. ruber w. data indicate a northward movement of the AF between 8 and 9 ka BP and southward movement of the AF and associated strengthening of the westerly wind belt at about 4 ka BP, the latter matching the tropical changes within the limitations of the age models (Repschläger et al., 2015). Despite the coincidence of frontal movements with Holocene changes in the northern hemisphere wind fields, inferred changes in the AF position do not coincide with the onset of the 20 northward subsurface water transport. The onset of the subsurface transport and LSW formation lag behind the northward movement of the AF by about 2 ka and lead the southward movement of the AF at 4 ka BP by 2 ka. Thus, the onset of the subsurface transport seems to be decoupled from changes in the atmospheric circulation.
The onset of the northward subsurface warm water transport at about 7 ka BP agrees with the onset of LSW formation and stabilization of the global sea level at that time (Wanner, 2008), and the end of meltwater flow into the Labrador Sea 25 (Jennings et al., 2015). This coincidence between changes in surface water transport and the end of meltwater input to the North Atlantic region argues for the cessation of meltwater input to the Labrador Sea and North Atlantic as a major driver for the onset of the modern mode of subsurface warm water transport and LSW formation.
In addition to this long-term evolution, short-term variability in the LSW subsurface transport and in the inter-gyre dynamics is observed during the last 7 ka BP. These short-term fluctuations match roughly a 1500 year cycle, possibly implying 30 another driving force such as solar activity or volcanism that generates smaller scale variability, not addressed in this study. Clim. Past Discuss., doi:10.5194/cp-2016-115, 2016 Manuscript under review for journal Clim. Past Published: 21 November 2016 c Author(s) 2016. CC-BY 3.0 License.

Conclusion
In this study we show a stepwise evolution of the mixed layer temperature and salinity at the AF during the Holocene that is closely linked to a northward migration of the AF and related transport of warm water within the NAC system.
The mean surface warm water transport from the tropics into the subpolar North Atlantic remained relatively stable over the last 11 ka BP. In contrast, subsurface transport into the North Atlantic evolved in three phases: an early Holocene meltwater 5 phase (11-8 ka BP) with no subsurface transport, a mid Holocene transitional phase (8-6 ka BP) with subsurface transport that reached the AF but not the subpolar NA and a late Holocene to modern phase (6-0 ka BP) with subsurface transport into the subpolar NA that coincides with the onset of LSW formation.
The subsurface transport changes are mainly driven by the amount of meltwater in the subpolar North Atlantic that are probably more important for the AMOC evolution than changes in the atmospheric circulation. Short-term variability in the 10 subsurface transport over the last 6 ka BP is anti-phased with Labrador Sea records, corroborating previous studies postulating inter gyre dynamics as a potential stabilizing mechanism for the modern LSW formation.
Subsurface transport pathways appear to have an important role in the Mid-to Late Holocene climate stabilization. However, this aspect needs to be investigated further in high latitudes in order to obtain a full understanding of underlying forcing mechanisms. 15