Holocene shifts of the Subtropical Shelf Front off southeastern South America controlled by high and low latitude atmospheric forcings



Over the Uruguayan shelf and uppermost slope, the coalescence of northward flowing Subantarctic Shelf Water and southward flowing Subtropical Shelf Water forms a distinct thermohaline front termed the Subtropical Shelf Front (STSF). Running in a SW direction diagonally across the shelf from the coastal waters at 32°S toward the shelf break at ca. 36°S, the STSF represents the shelfward extension of the Brazil-Malvinas Confluence zone. This study reconstructs latitudinal STSF shifts during the Holocene based on benthic foraminifera δ18O and δ13C, total organic carbon, carbonate contents, Ti/Ca, and grain size distribution from a high-accumulation sedimentary record located at an uppermost continental-slope terrace. Our data provide direct evidence for: (1) a southern STSF position (to the South of the core site) at the beginning of the early Holocene (>9.4 cal ka BP) linked to a more southerly position of the Southern Westerly Winds in combination with restricted shelf circulation intensity due to lower sea level; (2) a gradual STSF northward migration (bypassing the core site toward the North) primarily forced by the northward migration of the Southern Westerly Winds from 9.4 cal ka BP onward; (3) a relatively stable position of the front in the interval between 7.2 and 4.0 cal ka BP; (4) millennial-scale latitudinal oscillations close to 36°S of the STSF after 4.0 cal ka BP probably linked to the intensification in El Niño Southern Oscillation; and (5) a southward migration of the STSF during the last 200 years possibly linked to anthropogenic influences on the atmosphere.

1 Introduction

The Holocene evolution of the western South Atlantic is poorly understood due to the lack of temporally high-resolution records [Leduc et al., 2010; Wanner et al., 2011]. Nevertheless, under modern conditions, the western South Atlantic Ocean is one of the most energetic regions of the world's oceans and is dominated by the encounter of the poleward-flowing Brazil Current (BC) and the equatorward-flowing Malvinas Current (MC) (Figure 1). The two currents meet at approximately 38°S, forming the Brazil-Malvinas Confluence (BMC) over most of the continental slope [Peterson and Stramma, 1991; Stramma and England, 1999]. On the shelf and uppermost slope cold, fresh and nutrient-rich Subantarctic Shelf Water (SASW) and warm, salty, and nutrient-poor Subtropical Shelf Water (STSW) collide in a similar way forming the Subtropical Shelf Front (STSF; Figure 1) [Piola et al., 2008; Piola et al., 2000]. The density-compensated thermohaline nature of the STSF makes it a key component in the cross-shelf offshore transport of water masses in the western South Atlantic [Matano et al., 2010]. Though the STSF appears to be the northwestward shallow-water extension of the BMC, the processes controlling its latitudinal position are still under discussion [e.g., Matano et al., 2010]. It has been suggested that the cross-shelf scale of the pressure gradient enforced by the MC as well as the northernmost extension of the MC itself plays a major role in controlling the latitudinal position of the STSF [Matano et al., 2010; Palma et al., 2008]. Below the mixed surface layer, the STSF position in the modern system experiences no significant seasonality [Möller et al., 2008; Piola et al., 2008].

Figure 1.

(a) Location of sediment core GeoB13801-2 (yellow circle) at the SE South American continental margin. Oceanographic features are redrawn after Piola et al. [2008] summarizing the mean winter circulation patterns; BC: Brazil Current; MC: Malvinas Current; PPW: Plata Plume Water; SASW: Subantarctic Shelf Water; STSF: Subtropical Shelf Front; STSW: Subtropical Shelf Water. Latitudinal transects of annual mean (b) salinity [Antonov et al., 2010], (c) temperature [Locarnini et al., 2010], and (d) phosphate concentration [Garcia et al., 2010] along the outer shelf/uppermost slope in the study area from ca. 42 to 28°S. The yellow circles in Figures 1a–1d indicate the location of sediment core GeoB13801-2 below the Subtropical Shelf Front. This figure was partly produced with Ocean Data View [Schlitzer, 2012].

Due to the strong differences in temperature and nutrient content between the SASW and the STSW, the oxygen and carbon stable isotopic composition (δ18O, δ13C) of benthic foraminifera represent ideal tools to reconstruct latitudinal shifts of the front, in a similar way as described by Chiessi et al. [2007] for the BMC.

Here we present a high-resolution (i.e., 107 years on average between adjacent samples) paleoceanographic reconstruction from the western South Atlantic based on benthic foraminiferal δ18O and δ13C, grain size composition, total organic carbon (TOC), and calcium carbonate (CaCO3) contents as well as the calcium/titanium (Ca/Ti) ratio covering the past ~11 cal kyr BP. With this multiproxy approach, we are able to identify latitudinal shifts in the shallow-water currents and the frontal regime off southeastern South America and propose the respective forcing mechanisms.

2 Regional Setting

The study site is located at the uppermost slope off Uruguay that is largely dominated by the oceanographic regime on the shelf (Figure 1a). The inner shelf current circulation is strongly influenced by the La Plata River discharge (~23,000 m3/s) which, together with the Patos Lagoon outflow (~1750 m3/s), forms the distinctly buoyant, low-salinity (<26 psu) Plata Plume Water (PPW) [Piola et al., 2008; Piola et al., 2000] (Figure 1a). The PPW is commonly northeastward directed and can be traced as far north as 25°S [Piola et al., 2000]. The maximum northward penetration distance of PPW is primarily controlled by the pronounced seasonality of the along-shore wind stress. During austral winter under strong southwesterly winds, the PPW reaches its northernmost penetration, while during austral summer under northeasterly winds, the PPW retracts to the south and is advected offshore [Möller et al., 2008; Piola et al., 2005]. The PPW distribution seems also to respond to El Niño Southern Oscillation (ENSO) variability through changes in the interannual wind pattern [Piola et al., 2005]. During El Niño events, the regional atmospheric circulation resembles the austral summer situation, thus hindering the PPW northward extension, although enhanced precipitation over the drainage basin leads to an increased La Plata River outflow [Piola et al., 2005].

Subsurface (i.e., below 50 m water depth) circulation in the region is dominated by the southward-flowing, nutrient-poor, warm (>16°C) and saline (>34.8 psu) STSW and the northward-flowing, nutrient-rich, relatively cold (<11°C) and fresh (<34.5 psu) SASW [Brandini et al., 2000; Piola et al., 2008; Piola et al., 2000] (Figure 1). The encounter of STSW and SASW forms a steep thermohaline transition, the STSF [Piola et al., 2000]. The STSF runs southwestward diagonally from the 50 m isobath at 32°S toward the shelf break at around 36°S (Figure 1a) [Piola et al., 2000]. At the surface, the position of the STSF shows a ~5° northward displacement during austral winter. At the subsurface on the middle and outer shelves, the STSF can, in contrast, be considered stable throughout the year (Figure 1 in the supporting information) [Möller et al., 2008; Piola et al., 2008].

Due to its density-compensated thermohaline structure, the STSF favors offshore advection of the colliding shelf water masses, making the STSF a critical component in cross-shelf water movement [Matano et al., 2010; Piola et al., 2008]. Although the complex processes controlling the STSF position are still a matter of debate, numerical simulations indicate that under modern atmospheric circulation patterns, the barotropic pressure gradient enforced by the MC on the shelf circulation plays a major role in determining the latitudinal STSF location [Matano et al., 2010; Palma et al., 2008]. The strength of this barotropic pressure gradient forcing on shelf circulation is mainly determined by the width of the shelf. Under present-day shelf morphology (shelf width of ca. 160 km at 37°S), this effect leads to a northward expansion of the SASW of ~600 km to the northeast of the MC northernmost expansion (A. Piola, personal communication, 2011). Through influencing the MC northernmost expansion, the latitudinal position of the Southern Westerly Winds (SWW) also exerts some control on the STSF position. Numerical modeling shows that a northward migration of the SWW would lead to a northward MC expansion and consequently a northward shift of the STSF [Sijp and England, 2008]. While the influence of changes in regional wind field strength on the STSF behavior is largely unknown, it has been shown that out-of-phase changes in the BC and MC mass transport, associated with changes in local wind-stress patterns, lead to the BMC reaching its southernmost position during austral summer [e.g., Garzoli and Bianchi, 1987; Garzoli and Garraffo, 1989; Goni and Wainer, 2001; Olson et al., 1988].

3 Material and Methods

Gravity core GeoB13801-2 (36°08.49′S; 53°17.96′W) was retrieved at 241 m water depth during research cruise M78/3a with the German R/V Meteor in May/June 2009 (Figure 1a) [Krastel et al., 2012]. The core was collected from a small terrace at the uppermost continental slope. Total core recovery was 950 cm.

The age model for GeoB13801-2 is based on nine accelerator mass spectrometer radiocarbon (AMS 14C) ages from well-preserved (e.g., nonreworked) carbonate material, mainly benthic foraminifera and secondarily unbroken echinoderm spines and bivalves (Table 1). The AMS analyses were performed at the Poznan Radiocarbon Laboratory (Poland). The obtained 14C ages were converted into 1-sigma calibrated ages using the Calib 6.1.1 software and the Marine09 calibration data set, including the standard reservoir age of 405 years [Reimer et al., 2009]. All dates reported are given in calibrated thousands of years before present (cal  ka  BP). In order to provide a continuous depth-age model, the Bayesian accumulation model “Bacon” [Blaauw and Christen, 2011] was employed.

Table 1. GeoB13801-2 Radiocarbon Datings and Calibrated Ages
Lab No.Core Depth [cm]Material14C Age [14C a BP]1σ Calibrated Age Range [cal a BP]Calibrated Age (Intercept) [cal ka BP]
Poz-4242120Mixed, well-preserved benthic foraminifera; sized 250 – 1000 µm800 ± 30420 – 4790.45 ± 0.03
Poz-4242550Mixed, well-preserved benthic foraminifera; sized 250 – 1000 µm1155 ± 30665 – 7230.69 ± 0.03
Poz-42422209Uvigerina bifurcata3730 ± 503572 – 37203.65 ± 0.07
Poz-42423375Mixed, well-preserved benthic foraminifera and unbroken echinoderm spines; sized 250 – 1000 µm5890 ± 406265 – 63536.31 ± 0.04
Poz-42424404Mixed, well-preserved benthic foraminifera and unbroken echinoderm spines; sized 250 – 1000 µm6535 ± 356998 – 71187.06 ± 0.06
Poz-35195540Uvigerina bifurcata8180 ± 408589 – 87338.66 ± 0.07
Poz-42426686Uvigerina bifurcata8970 ± 509539 – 96889.61 ± 0.07
Poz-42427750Mixed, well-preserved benthic foraminifera; sized 125 – 1000 µm9130 ± 609771 – 10,0329.90 ± 0.13
Poz-35197945Fresh (periostracum preserved) bivalves of Yoldiella genus9560 ± 5010,366 – 10,50310.43 ± 0.07

Except for XRF scanning (1 cm steps), all analyses were performed at 10 cm intervals. Analyses of total (organic) carbon as well as stable oxygen and carbon isotopes were performed from the same set of samples, whereas grain size was analyzed on a set of samples taken with 1 cm offset.

Total (organic) carbon (TOC, TC) were quantified using the LECO CS-200 system. Calcium carbonate content was calculated employing the standard equation CaCO3 [wt.%] = (TC − TOC) * 8.333.

Calcium and titanium relative intensities from the upper 360 cm of GeoB13801-2 were obtained utilizing the AVAATECH X-ray fluorescence core scanner II at MARUM and are given in total counts per second (cps). Below 360 cm, the increasingly sandy and variable lithology of Core GeoB13801-2 impeded statistically reliable scanning [Jansen et al., 1998; Richter et al., 2006]. Please refer to the supporting information of this article for detailed information on this method. The Ca/Ti ratio is a well-known marine versus terrigenous sedimentation proxy [Govin et al., 2012] and was used to corroborate the low resolution carbonate content record.

Grain size distributions from 2000 to 0.4 µm were analyzed by means of a Coulter Laser Particle Sizer LS200. Each sample was split in two subsamples, for one being analyzed without any chemical treatment (bulk distribution) and the other being digested in successive steps with 35% H2O2, 10% HCl, and 6% NaOH to remove organic carbon, biogenic carbonate, and opal, respectively, prior to analysis (terrigenous distribution). Sand (2000 – 63 µm), silt (63 – 4 µm), and clay (<4 µm) contents were calculated from the cumulative grain size distributions (bulk and terrigenous).

Oxygen and carbon stable isotopes were analyzed on samples with about 40 specimens of benthic foraminifera Cibicides mckannai within the size range 125 – 150 µm using a Finnigan MAT 252 mass spectrometer coupled to an automatic carbonate preparation device. C. mckannai was used since this species was the only epibenthic foraminiferal species present throughout our core. The long-term laboratory analytical standard deviation is <0.07‰ and <0.05‰ for δ18O and δ13C, respectively. All values are given as ‰ deviations from the Vienna Pee Dee belemnite (VPDB) (calibrated by the NBS 19 standard). The δ18O data set was corrected for changes in continental ice volume (i.e., sea level) based on Lambeck and Chappell [2001] and Schrag et al. [2002], in the following denoted as ice-volume-corrected δ18O (δ18Oivc). The estimated maximum error imposed by uncertainties in our age model during alignment with the sea level δ18Osw is one order of magnitude lower than the analytical error. The combined maximum error is <0.073‰ for the late Holocene section and varies between 0.074 and 0.112‰ in the middle and early Holocene parts of the record. In addition, we calculated the predicted δ18O of calcite (δ18Opc) at four different stations along the western South Atlantic Ocean margin by solving the paleotemperature equation of Shackleton [1974]. δ18O of seawater (δ18Osw) profiles for these four stations were extracted from the global gridded data set of LeGrande and Schmidt [2006], and the respective annual temperature profiles were taken from the World Ocean Atlas [Locarnini et al., 2010] (s. supporting information for more details).

For ordination of the samples, a principal components analysis (PCA) was performed using the Multivariate Statistical Package (MVSP) Version 3.2 [Kovach, 2010] on the standardized data of calcium carbonate content, TOC content, bulk and terrigenous clay, silt and sand content, bulk and terrigenous mean grain size, δ18Oivc, and δ13C.

4 Results

4.1 Age Model

The core top is of modern age, since the 0.8‰ decrease in the uppermost samples of the δ13C record (see below) can be attributed to the oceanic Suess effect. This effect ascribes the observed historical decrease of δ13C to the oceanic uptake of isotopically light CO2 injected in the atmosphere by the industrial burning of fossil fuels [Suess, 1955].

The dates appear in consistent order and the basal sample of the core dated back to 10.4 cal ka BP (Table 1 and Figure 2). From the detailed lithologic core description, no hiati were recognized [Krastel et al., 2012]. Linear sedimentation rates (LSR), resulting from linear interpolation between the intercepts of the calibrate ages, show strong variations. They decrease from about 360 cm/kyr between 10.6 and 9.9 cal ka BP to 80 – 150 cm/kyr between 9.9 and 6.5 cal ka BP, and to about 50 cm/kyr from 6.5 cal ka BP until present (Figure 2).

Figure 2.

Depth/age model of sediment core GeoB13801-2 (grey), overlying the distributions of the individual dates (blue). Grey dotted lines: the model's 95% probability interval; red dotted line: mean. Corresponding linear sedimentation rates (LSR) and averages for lithologically similar units are shown on the right-hand side of the main panel. See section 3 for more information about the depth/age model.

4.2 Grain Size, TOC, CaCO3, and Ca/Ti

Core GeoB13801-2 is generally dominated by terrigenous fine sands at grain sizes centered at about 150 µm. The siliciclastic nature of these sediments is reflected by the bulk clay and silt contents running parallel to variations in the terrigenous clay and silt fractions (Figures 3a, 3c, and 3d).

Figure 3.

GeoB13801-2 temporal variation of (a) terrigenous grain size fractions (cumulative clay (c), silt (z) and sand (s)), (b) bulk total organic carbon (TOC), (c) bulk silt content, (d) bulk clay content, (e) bulk carbonate (CaCO3) content, and (f) calcium/titanium (Ca/Ti) ratio (a proxy for marine vs. terrigenous input). Blue triangles demark calibrated AMS 14C ages from GeoB13801-2, and vertical dashed lines mark important dates discussed in the text.

From 10.4 to 8.8 cal ka BP, grain size, TOC, and CaCO3 contents show high variability, what may in minor parts be owed to the higher sedimentation rate leading to better sample resolution. Nevertheless, terrigenous clay and silt reach extraordinary high contents (e.g., up to 67 vol.% clay at 10.05 cal ka BP, and up to 72 vol.% silt at 9.92 cal ka BP) during this period, indicating a significantly different oceanographic/sedimentary regime compared to the subsequent period (Figures 3a–3e). Bulk TOC content amounts to ca. 1 wt.% and bulk CaCO3 content to ca. 10 wt.% (Figures 3b and 3e).

Between 8.8 and 7.2 cal ka BP, sediment composition is relatively stable, with terrigenous clay and silt amounting at 20 vol.%, bulk carbonate content at ca. 8 wt.% and bulk silt content at ca. 35 vol.% (Figures 3a, 3c, and 3e).

After 7.2 cal ka BP and until 4.0 cal ka BP, the variability in grain size, TOC, CaCO3, and Ca/Ti increases again. Terrigenous clay and silt, as well as TOC and carbonate content, increase synchronously reaching a first maximum at ca. 6.4 cal ka BP of 65 vol.% (terrigenous clay and silt together), 3.2 wt.% (TOC) and 15 wt.% (CaCO3) (Figures 3a, 3b, and 3e). A second maximum at ca. 5.0 cal ka BP is most pronounced in the carbonate content (up to ca. 18 wt.%) and Ca/Ti ratio (Figures 3e and 3f).

After 4.0 cal ka BP, the variability in grain size, TOC, CaCO3, and Ca/Ti decreases. Terrigenous clay, silt, and bulk TOC content vary around 30 vol.% (terrigenous clay and silt together) and 1.25 wt.% (TOC), and only increase up to 60 vol.% and 3.8 wt.%, respectively, for the most recent 500 years (Figures 3a and 3b). In contrast, CaCO3 content (mirrored by Ca/Ti ratio) decreases from 10 to 2 wt.% from ca. 4.0 to 0.8 cal ka BP (Figure 3e), followed by a ca. 400 year long period of elevated Ca/Ti ratios and up to 13 wt.% carbonate at 0.6 cal ka BP, before decreasing down to ca. 6 wt.% carbonate at the top of the record (Figure 3e).

4.3 Oxygen and Carbon Isotopes

Between 10.6 and 10.0 cal ka BP, δ18Oivc values show the highest variability throughout the record ranging from 0.3‰ to 1.63‰ (Figure 4a). From ca. 10 until 9.4 cal ka BP, δ18Oivc values stabilize at ca. 1.8‰, followed by an abrupt decrease to 1.2‰ at 9.4 cal ka BP. After 9.4 cal ka BP, δ18Oivc values gradually increase again reaching 1.8‰ at about 7.2 cal ka BP. After 7.2 until ca. 0.2 cal ka BP, δ18Oivc values vary around 1.8‰. The most recent 200 years of the core are characterized by a remarkably abrupt decrease in δ18Oivc that reaches 0.95‰.

Figure 4.

GeoB13801-2 (a) ice-volume corrected Cibicides mckannai oxygen isotopic composition (δ18Oivc) and (b) Cibicides mckannai carbon isotopic composition (δ13C; exPP: export primary productivity). (c) Relative sea level as a composite of regional data from Angulo et al. [2006] (Santa Catarina data set South of 28°S only), Cavallotto et al. [2004], and Guilderson et al. [2000], as well as eustatic data from Hanebuth et al. [2011]; all sea level age points were recalibrated following the procedure described for the age model of GeoB13801-2; thin black lines represent a graphic envelope around all sea level data points, while thick black lines represent graphic interpolation of potential regional [Guilderson et al., 2000] and eustatic [Hanebuth et al., 2011] sea level history. (d) Temporal occurrence of the three groups identified from the PCA of GeoB13801-2 data (supporting information Figure 3). Blue triangles demark calibrated AMS 14C ages from GeoB13801-2, and vertical dashed lines mark important dates discussed in the text.

To identify whether temperature or salinity is the predominant influencing factor on the GeoB13801-2 C. mckannai δ18Oivc record, we calculated predicted calcite δ18O (δ18Opc) for four different stations along the western South Atlantic margin from 33.5°S up to 39.5°S (Figure 5). Based on those calculations, we interpret lighter (heavier) values in the GeoB13801-2 C. mckannai δ18Oivc record as corresponding to an increased influence of relatively warm (cold) STSW (SASW) over our core site (see supporting information for more details).

Figure 5.

(a) Predicted oxygen isotopic composition (δ18O) of calcite based on annual mean temperature profiles (δ18Opc annual) from the western South Atlantic compared to oxygen isotopic composition (δ18Oivc) values from GeoB13801-2 (gray dots). (b) Map showing the location of stations A – D with respect to the location of GeoB13801-2 and the regional oceanographic features (s. Figure 1 for details). δ18Opc annual was calculated based on the paleotemperature equation of Shackleton [1974], δ18O of sea water from LeGrande and Schmidt [2006], and mean annual temperature from Locarnini et al. [2010] (see text in the supporting information for details). Note that the δ18Opc annual of station A (3° north of GeoB13801-2) at the modern water depth of GeoB13801-2 is much closer to the value of station B (≈ GeoB13801-2), than station C, which is only 1° to the south of GeoB13801-2.

During the early Holocene, δ13C values decrease from 2.17‰ at 10.6 cal ka BP down to ca. 0.7‰ at 8 cal ka BP, with the highest variability found throughout the record between 10.6 and 10 cal ka BP (Figure 4b). Between ca. 8 and 0.2 cal ka BP, δ13C values slightly increase from ca. 0.7‰ to 0.9‰. Similar to the δ18Oivc record, δ13C values show a strong decrease during the most recent ca. 200 years reaching 0.4‰.

Since at our study site the bottom water mass is basically the same that bathes the photic zone, we interpret our benthic C. mckannai δ13C record in terms of relative export productivity, with low (high) isotopic ratios indicating increased (decreased) surface water export productivity under the influence of SASW (STSW) (see supporting information for more detail).

4.4 Principal Components Analysis

A PCA ordination of carbonate content, TOC content, bulk and terrigenous clay/silt/sand content, bulk and terrigenous mean grain sizes, together with δ18Oivc and δ13C values allowed for the identification of three sample clusters termed Groups I – III (cf. Figure 3 in the supporting information). Group I correlates positively with terrigenous silt, bulk clay, calcium carbonate content, and δ13C. The temporal occurrence of Group I is restricted to the time interval prior to 8.8 cal ka BP (Figure 4d). Group II is marked by a positive correlation of terrigenous sand content and mean grain sizes (bulk and terrigenous), having its main occurrence between 8.8 and 7.2 cal ka BP, as well as a single occurrence at 5.5 cal ka BP (Figure 4d). Group III positively correlates with TOC, bulk silt, and terrigenous clay content, and occurs dominantly after 7.2 cal ka BP (Figure 4d).

5 Discussion

5.1 Early Holocene

δ18Opc at the modern water depth of GeoB13801-2 shows that the low δ18Oivc values recorded prior to 10 cal ka BP (ca. 0.5‰; combined uncertainty of this time interval: < ±0.112‰, s. methods section) are typically found today associated with STSW located 3° to the north of GeoB13801-2 (Figure 5). The high δ18Oivc values between 7.2 and 0.2 cal ka BP (ca. 1.8 ± 0.7‰; combined uncertainty of this time interval: < ±0.073‰, s. methods section) are close to modern SASW δ18Opc values found 1° to the south of site GeoB13801-2 (Figure 5). Considering this δ18Opc-water mass relationship, our δ18Oivc record (Figure 6a) implies an early Holocene northward migration of the STSF that crossed the latitude of GeoB13801-2 (i.e., 36°S) at around 7.2 cal ka BP.

Figure 6.

(a) GeoB13801-2 ice-volume corrected Cibicides mckannai oxygen isotopic composition (δ18Oivc) together with selected paleoclimatic proxy records from the Southern Hemisphere. (b) Rainfall variability over central Chile recorded by iron intensity from GeoB3313-1 (41°S) [Lamy et al., 2001]. (c) Diatom-based sea surface temperature (SST) reconstruction from the South Atlantic [Bianchi and Gersonde, 2004]; (d) Alkenone-based SST from offshore Chile (41°S) [Lamy et al., 2010]; (e) Alkenone-based SST from offshore Chile (30°S) [Kaiser et al., 2008]; (c – e) reflecting the northward migration and expansion of the Southern westerlies discussed in detail by Lamy et al. [2010]. (f) Frequency of warm ENSO events in events per 100 years [Moy et al., 2002]. (g) Speleothem δ18O from SE Brazil (27°S) reflecting precipitation [Wang et al., 2007 ]. (h) Relative sea level (s. Figure 4 for details).

Two factors have probably acted simultaneously forcing a northward shift of the STSF: (i) deglacial sea level rise; and (ii) a northward migration of the northern boundary of the SWW, as discussed below.

Under modern conditions, the STSF latitudinal position is principally determined by the northernmost extent of the MC [Matano et al., 2010; Palma et al., 2008]. The effect of MC pressure gradient forcing on the shelf circulation is in turn mainly determined by the shelf width. At 10.4 cal ka BP, the lowered sea level reduced the shelf north of 37°S to approximately half of its modern width (cf. Figures 1 and 6h). Consequently, the MC cross-shelf barotropic pressure gradient forcing on shelf circulation was significantly reduced due to a drastically narrowed shelf water body, resulting in a reduced northward expansion of SASW, which presumably caused a more southerly location of the STSF. As sea level rose in early Holocene times, the width of the shelf increased gradually which enabled a northward expansion of SASW pushing the STSF northward. The northward migration of the STSF, as recorded at site GeoB13801-2 (i.e., 36°S), stopped at 7.2 cal ka BP, synchronous to sea level approaching the modern height (Figure 6). Higher δ13C values found in the early Holocene indicate that the core site was mainly bathed by oligotrophic STSW, supporting the interpretation of a more southerly STSF position during the early Holocene (Figures 4 and 6). The good correlation among sediment composition, surface water export productivity, and sea level dynamics suggests that the early and mid-Holocene sea level rise strongly controlled the sedimentary configuration of the shelf. Until 8.8 cal ka BP, the high lithologic variability of PCA Group I and the low export productivity are temporally coincident with a presumably strong transgressive reworking of the outer to middle shelf (Figure 4c). From 8.8 until 7.2 cal ka BP, the rather stable lithology of PCA Group II and increasing export productivity occurred during a time of decelerating sea level rise and the approach to close to modern sea level height (8.8 – 7.2 cal ka BP: −35 to −5 m; Figure 4c). After 7.2 cal ka BP, the dominance of PCA Group III and relatively high productivity correlate with a stabilizing sea level (between −5 m and +6.5 m; Figures 4b, 4c, and 4d). It is worth noting that these sea level changes are not explicitly paralleled by changes in δ18Oivc (Figures 4a, 4c, and 4d). We thus suggest that rising sea level during the early Holocene had an influence on the northward displacement of the STSF by increasing shelf width. However, neither a low sea level nor the successive shelf flooding exerted a major control on the more southerly STSF position during early Holocene and on the subsequent northward migration after 9.4 cal ka BP.

The second potential mechanism determining the latitudinal position of the STSF is the northernmost latitude of the MC along the middle slope. Numerical modeling has shown that a northward shift of the SWW would cause an enhanced inflow of Antarctic Intermediate Water into the western South Atlantic and consequently enhanced northward migration of the MC [Sijp and England, 2008]. A more poleward location of the SWW would in turn cause a retraction of the MC and consequently a southward displacement of the STSF [Biastoch et al., 2009; Sijp and England, 2008]. Multiproxy reconstructions of the SWW northern boundary and of SWW strength indicate a poleward confinement during the late-glacial to early-Holocene time interval, followed by a significant northward migration of the northern boundary between 8.5 and 5.5 cal ka BP [e.g., Hebbeln et al., 2002; Kaiser et al., 2008; Kilian and Lamy, 2012; Lamy et al., 2010; Mohtadi and Hebbeln, 2004]. The northward migration of the STSF shown in this study starts, thus, ca. 900 years earlier (at ca. 9.4 cal ka BP; Figure 6). We are aware that this lag in timing appears close to the combined 2-sigma uncertainties of the respective age models (9.4 ± 0.3 cal ka BP for GeoB13801-2; 8.5 ± 0.4 cal ka BP for PS2090; Bianchi and Gersonde, 2004). However, applying the more focused 1-sigma intervals (67% confidence), the observed minimum time lag is ca. 500 years, hence clearly outside of the combined uncertainty. We, thus, hypothesize that the earlier northward migration observed in the record of GeoB13801-2 is most probably related to shelf flooding due to the final interval of sea level rise.

The cessation of the STSF northward migration documented in the present study at around 7.2 cal ka BP (Figure 6) could indicate: (i) the point in time when the STSF crosses 36°S, at which a further northward advance would not be documented by our record; or (ii) the time of the northernmost position reached by the STSF during the Holocene. The fact that GeoB13801-2 δ18Oivc values do not further increase after 7.2 cal ka BP and stay closely δ18Opc values estimated for modern SASW 1° to the south of GeoB13801-2 suggests that the maximum Holocene northward STSF displacement was reached close to 36°S by this time (Figures 5 and 6).

Between 10.0 and 9.4 cal ka BP, δ18Oivc values seem to indicate an abrupt and significant northward STSF shift in the range of 2° to 3° in latitude (Figures 4a and 5). This rapid northward shift could be related to a contemporary northward migration of the SWW. The fact that no synchronous excursion in the SWW is apparent in the compilation of Lamy et al. [2010] might be related to the compilations' focus on orbital time scales (Figures 6c – 6d). Though only extending back to 7.7 cal ka BP, Lamy et al. [2001] show millennial-scale variability on the iron supply offshore Chile at 41°S linked to the variability of the northern boundary of the SWW (Figure 6b). Thus, it is plausible that the northern boundary of the SWW over the South Atlantic showed millennial-scale variability in the early Holocene, similar to the variability suggested for the mid- and late-Holocene by Lamy et al. [2001]. We consequently hypothesize that the abruptly terminated northward shift of the STSF between ca. 10 and 9.4 cal ka BP is associated to millennial scale variability of the northern boundary of the SWW. Since other appropriately located South Atlantic records are either scarce or lacking temporal resolution, we cannot further test this hypothesis yet.

5.2 Mid- and Late-Holocene

Sea level reached a position close to its modern height at around 7.2 cal ka BP. This coincided with a change in the sediment composition (i.e., predominance of PCA Group II) and favored an increased export primary productivity (Figures 4b and 4d) due to the establishment of the modern high-energy and high-nutrient shelf regime.

Following a relatively stable phase in the δ18Oivc record between 7.2 and 4.0 cal ka BP, the increased variability (± 0.3‰) of δ18Oivc suggests that the STSF position oscillated close to site GeoB13801 during the late Holocene (i.e., 36°S; Figure 4a). Also, the slight increase in the silt fraction and the moderate decrease in the clay fraction indicate a modification of the late Holocene shelf sediment export characteristics (Figure 4). Moreover, decreasing carbonate content as well as Ca/Ti ratios (Figure 3) along with unaltered LSR (Figure 2) indicates increased dilution with terrigenous material during the late Holocene, probably from La Plata outflow.

Offshore Chile at 41°S millennial-scale positive excursions of iron influx have been linked to more southerly located SWW during the mid-Holocene [Lamy et al., 2001] (Figure 6a). Moreover, a shift at ca. 4.0 cal ka BP from correlation to anticorrelation of this SWW latitudinal variability (from iron content) with Antarctic ice-core dust content has been related to the onset of the modern ENSO [Lamy et al., 2001]. Debret et al. [2009] suggested a change in the main global climatic driving mechanisms from external (i.e., solar insolation) to internal (i.e., oceanic-atmospheric coupling, e.g., ENSO) forcing at ca. 5 cal ka BP. Wavelet analysis of sedimentary proxy data from the Laguna Pallcacocha (Ecuador) indicates that the modern ENSO started around 7 cal ka BP, and that the frequency of El Niño events significantly increased at approximately 4 cal ka BP [Moy et al., 2002] (Figure 6f). Modern warm ENSO (i.e., El Niño) events lead to enhanced precipitation over southeastern South America [Garreaud et al., 2009], causing enhanced La Plata River outflow into the Uruguayan-SE Brazilian shelf. The prevailing Northeasterly winds that occur during El Niño conditions suppress the northeastward extension of PPW and favor the offshore advection of the freshwater plume [Möller et al., 2008; Piola et al., 2008]. Although the hydrographic influence of this plume on the STSF position is still a matter of debate, we suggest that increased influence of ENSO climatic forcing, which caused millennial-scale variability of the late Holocene northern SWW boundary, served as a reasonable driver for equivalent variability in the STSF latitudinal location (Figure 6).

5.3 Preindustrial and Modern (Past 200 Years)

We attribute the light δ13C values of the most recent 200 years in GeoB13801-2 to the Suess effect [Suess, 1955], implying an anthropogenic influence. Similar anthropogenic influences on marine proxy records have been observed elsewhere [e.g., McGregor et al., 2007; Mulitza et al., 2010]. In analogy, we propose that the trend toward light GeoB13801-2 δ18Oivc values of the most recent 200 years of our record indicates a recent southward shift of the STSF (Figure 6a). A similar postindustrial southward migration of the SWW linked to human activity has been recently suggested from variations in surface velocity fields in the western South Atlantic and climate modeling [Lumpkin and Garzoli, 2011; Toggweiler and Russell, 2008]. A poleward migration of the SWW leads to a more effective alignment of the SWW with the Antarctic Circumpolar Current, causing an increase in Agulhas leakage and a subsequent strengthening and more southward location of the South Atlantic subtropical gyre [Biastoch et al., 2009; Sijp and England, 2008].

6 Conclusions

Multiproxy data (i.e., benthic foraminiferal δ18O and δ13C; TOC, CaCO3 content, Ca/Ti, and grain size distribution) from a high-accumulation sediment record collected at the uppermost continental slope off Uruguay (36°S) add novel and direct evidence for latitudinal shifts of the STSF during the Holocene. Five stages can be differentiated:

  1. Prior to 10 cal ka BP, the STSF was positioned to the South of 36°S, most probably linked to a more southerly position of the northern boundary of the SWW as well as to the narrower continental shelf due to a ca. 50 m lowered sea level.

  2. An abruptly terminated northward shift of the STSF between ca. 10 and 9.4 cal ka BP possibly associated to millennial scale variability of the northern boundary of the SWW.

  3. Between ca. 9.4 and 7.2 cal ka BP, the STSF migrated northward reaching latitudes to the North of 36°S; this shift was primarily forced by the northward migration of the SWW.

  4. After 4.0 cal ka BP, the STSF oscillated at latitudes close to 36°S, but never reached the study site again; these oscillations may be linked to a coincident intensification in ENSO.

  5. During the past 200 years, the STSF experienced a southward migration, probably linked to a hemispheric southward shift of the SWW as a consequence of anthropogenic influence on global climate.

Our findings indicate that the Holocene paleoceanographic changes over the Uruguayan continental shelf and uppermost slope are strongly linked to Southern Hemisphere atmospheric circulation.


This study was funded through DFG-Graduate College “Proxies in Earth History (EUROPROX)” and through DFG-Research Center/Cluster of Excellence “The Ocean in the Earth System.” TH acknowledges a DFG Heisenberg fellowship (HA 4317/4-1). CMC acknowledges a Hanse-Wissenschaftskolleg fellowship and the financial support from FAPESP (10/09983-9, 11/50394-0 and 12/17517-3). Bastian Steinborn is thanked for his support in picking C. mckannai. The SESA meeting group and especially Renata H. Nagai is thanked for fruitful discussions. We appreciate the very constructive comments from three anonymous reviewers, who certainly helped to polish the final version of this manuscript. Maarten Blaauw is acknowledged for help with the age model. Sample material has been provided by the GeoB core repository at the MARUM-Center for Marine Environmental Sciences, University of Bremen, Germany. The data reported in this paper are archived in PANGAEA (www.pangaea.de).