Biological productivity and carbon export in the equatorial Atlantic are thought to have been dramatically higher during the last glacial period than during the Holocene. Here we reconstruct the pH and CO2 content of surface waters from the eastern equatorial Atlantic Ocean over the past ∼30 k.y. using the boron isotope composition of Globigerinoides ruber (a mixed-layer–dwelling planktic foraminifera). Our new record, combined with previously published data, indicates that during the last glacial, in contrast to today, a strong west to east gradient existed in the extent of air:sea equilibrium with respect to pCO2pCO2), with the eastern equatorial Atlantic acting as a significant source of CO2 (+100 µatm) while the western Atlantic remained close to equilibrium (+25 µatm). This pattern suggests that a fivefold increase in the upwelling rate of deeper waters drove increased Atlantic productivity and large-scale regional cooling during the last glacial, but the higher than modern ΔpCO2 in the east indicates that export production did not keep up with enhanced upwelling of nutrients. However, the downstream decline of ΔpCO2 provides evidence that the unused nutrients from the east were eventually used for biologic carbon export, thereby effectively negating the impact of changes in upwelling on atmospheric CO2 levels. Our findings indicate that the equatorial Atlantic exerted a minimal role in contributing to lower glacial-age atmospheric CO2.

Over at least the past 800 k.y., the CO2 content of the atmosphere has shifted from ∼240–280 ppm during the warm interglacial periods to 180–200 ppm during the cold glacial periods (Petit et al., 1999; Lüthi et al., 2008); most attention has been focused on the amount of carbon stored in the deep ocean during glacials to explain these CO2 changes (e.g., Sigman et al., 2010). One important mechanism in this regard is the biological pump: biomass produced in the surface ocean sinks to depth and decomposes, thereby pumping both nutrients and organic carbon into the deep ocean, where the carbon is sequestered away from the atmosphere and the nutrients are temporarily unavailable to fuel new biological production. Given their importance today in terms of oceanic primary production (Fig. 1B), attention has long focused on the equatorial oceans to at least partially explain the lower glacial CO2 levels (e.g., Mix, 1989).

A small proportion of the equatorial regions is termed high-nitrate, low-chlorophyll (HNLC; e.g., the eastern equatorial Pacific Ocean), where the plentiful supply of macronutrients (N, P) from upwelling of cold (Fig. 1A), nutrient-rich (Fig. 1C) and high-CO2 (Fig. 1D) water is often underutilized because of the relative paucity of essential micronutrients such as Fe (Moore et al., 2013). The incomplete utilization of macronutrients in HNLC regions gives rise to outgassing of excess CO2 to the atmosphere (Fig. 1D). Changing the efficiency of nutrient utilization (e.g., through enhanced productivity via dust fertilization of Fe-limited areas; Martin, 1990) clearly has the potential to lower the pCO2sw (sw is seawater) in these regions. However, most areas of the equatorial and low-latitude oceans are non-HNLC regions, where productivity is not micronutrient limited (e.g., by Fe), and organisms are eventually able to fully utilize all the available macronutrients (N, P; Fig. 1C) with correspondingly lower quantities (5–30 μatm) of excess CO2 (defined herein as ΔpCO2 = pCO2swpCO2atm; e.g., the eastern equatorial Atlantic; Fig. 1D). As a result, across the low-latitude non-HNLC regions macronutrient (as opposed to micronutrient) limitation currently exists (Moore et al., 2013). In these regions, there is more limited scope to decrease pCO2sw by enhanced nutrient utilization, since most of the nutrients are already nearly fully utilized.

Despite the apparent potential for changing pCO2sw in either HNLC or non-HNLC regions via alleviation of macronutrient or micronutrient limitation, it has been hypothesized that their role in glacial-interglacial CO2 change should have been minor (Sigman et al., 2010; Hain et al., 2014), because of the following:

  • (1) Nutrients supplied to the low-latitude surface from below are all eventually consumed by productivity as the nutrient-rich water flows away from the site of upwelling before the water is able to sink into the ocean interior (Sigman and Haug, 2003). Even if an increase in on-axis (where axis refers to the longitudinally extensive zone of upwelling) consumption of nutrients occurred in an equatorial upwelling region (e.g., due to a relief from Fe limitation), leading to a local decrease in pCO2sw, this may not change atmospheric CO2 significantly because off-axis productivity may be correspondingly reduced, causing little net change in the residual unused nutrient concentration when the water ultimately descends back into the ocean interior. This almost complete utilization of available nutrients is in stark contrast to higher latitudes, such as the Southern Ocean, where the utilization of nutrients is currently inefficient (and consequently is associated with outgassing of CO2), but may have been more efficient during glacial times (e.g., Sigman et al., 2010).

  • (2) The high-latitude surface ocean is in direct communication with the deep ocean, whereas the low-latitude surface ocean is not, so changes in low-latitude pCO2sw are hypothesized to be somewhat buffered from driving atmospheric CO2 changes because of the much larger size of the high-latitude/deep-ocean reservoir (Broecker et al., 1999).

One region of the oceans that has well-documented changes in surface ocean productivity and export production across glacial-interglacial cycles is the equatorial Atlantic (Bradtmiller et al., 2007; Kohfeld et al., 2005, and references therein, and their figure 2c). While this has been interpreted as evidence of a strengthening of the biological pump and hence driving a portion of glacial CO2 drawdown (e.g., Mix, 1989), this remains to be quantitatively tested. Here we use the boron isotopic composition (expressed in terms of δ11B) of mixed-layer-dwelling planktic foraminifera Globigerinoides ruber to reconstruct surface water pH and therefore pCO2sw across a transect of sites spanning the equatorial Atlantic Ocean over the past ∼30 k.y., allowing us to directly test the role of this region in glacial-interglacial pCO2atm variations.

Samples from ca. 30 ka to 10 ka were selected from Site GeoB1105-4 (R/V Meteor cruise M9/4; herein GeoB1105) from the eastern equatorial Atlantic (Fig. 1), located near the center of modern upwelling. Age models for the studied period were discussed by Henehan et al. (2013) for Ocean Drilling Program (ODP) Site 999 and Site GeoB1523, and based on 14C dating for GeoB1105 (Bickert and Mackensen, 2003). Our new data, combined with published δ11B data for G. ruber from ODP Site 999A (herein ODP 999) and Site GeoB1523-1 (herein GeoB1523) from the Caribbean and western equatorial Atlantic, respectively (Foster, 2008; Henehan et al., 2013), allow us to examine zonal gradients in surface ocean carbonate chemistry (Fig. 1). All boron isotopic measurements presented here were performed by multicollector–inductively coupled plasma–mass spectrometry at the University of Bristol (UK), closely following the methodologies in Foster (2008) and Rae et al. (2011). The basis for the boron isotope-pH proxy was discussed extensively elsewhere (e.g., Foster, 2008; Rae et al., 2011), and we use the new δ11B-pH calibration for this species (Henehan et al., 2013). We also generated sea-surface temperature (SST) records for the same samples using Mg/Ca ratio of G. ruber and the calibration of Anand et al. (2003). See the GSA Data Repository1 for more details of the methodologies used to calculate pH and pCO2sw from δ11B; the relevant data are in Table DR1.

Figure 2 shows data from the three sites in the eastern Atlantic Ocean (blue), western Atlantic (black), and Caribbean (red). Our Mg/Ca-derived SST records reveal a persistent pattern throughout the past 30 k.y. of an eastward decrease in SST (i.e., SST999A > SSTGeoB1523 > SSTGeoB1105; Fig. 2), but with larger differences between the sites before ca. 18 ka and the greatest extent of glacial-aged cooling at our easternmost site (GeoB1105; Fig. 2A). δ11B values of G. ruber and the calculated surface-water pH exhibit similar values across the three sites during Marine Isotope Stage 1 (MIS 1; Figs. 2B and 2C), consistent with the absence of a major pH gradient in the Atlantic today. However, δ11B and calculated pH diverge during MIS 2 (Figs. 2B and 2C). From these δ11B and pH records, we calculate that surface water in the Caribbean and western Atlantic (ODP 999 and GeoB1523) remained close to equilibrium with the atmosphere (purple line in Fig. 2D) with respect to CO2 for the past 30 k.y., albeit with a minor and consistent offset comparable to that observed today at both sites (20–30 µatm; Figs. 1D and 2D). This contrasts with GeoB1105 in the eastern equatorial Atlantic, where we reconstruct a significant excess of pCO2sw compared to the contemporaneous atmosphere (by >100 µatm), indicating that this region was a strong source of CO2 to the atmosphere during MIS 2 and MIS 3. Therefore, in marked contrast to MIS 1 (Fig. 2), a significant east to west gradient in ΔpCO2 existed during the last glacial that was well correlated with the contemporaneous SST gradient (Figs. 2, 3A, and 3B). Notably, this strong cross-basin gradient in calculated pCO2sw is also evident in δ11B alone, as reflected in the large differences between δ11B of G. ruber (to 1.5‰) from the three sites during the last glacial (e.g., ODP 999 versus GeoB1105; Fig. 2B) in comparison to correspondingly small δ11B gradients during MIS 1.

Our SST reconstruction across the equatorial Atlantic (Fig. 2A) shows a glacial-age cooling that was stronger in the east (3–4 °C) than the west (1–2 °C) (e.g., MARGO Project Members, 2009; Pflaumann et al., 2003), likely attributable to enhanced eastern upwelling of cold water owing to stronger glacial-age southeast trade winds (Mix and Morey, 1996; Wolff et al., 1999; MARGO Project Members, 2009). Intensified glacial upwelling in the east is also consistent with an increased east to west tilt in the equatorial Atlantic thermocline (Fig. 3D; Wolff et al., 1999), a shallow nutricline in the east (Molfino and McIntyre, 1990), and a consequent increase in primary productivity (Fig. 3C; Schmidt et al., 2003) and opal fluxes (not shown; Bradtmiller et al., 2007), and an increase in ΔpCO2 (Fig. 3A) at eastern Site GeoB1105.

In the eastern equatorial Atlantic today, during a particularly strong upwelling season, SSTs cool considerably and ΔpCO2 can locally reach as high as 80–100 µatm at 12°W (Bakker et al., 2001), compared to 25–50 µatm during other years in the climatology of Takahashi et al. (2009) (Fig. 1D). The simplest explanation of our reconstructions for the past 30 k.y. is that, during the last glacial, the upwelling of cold, CO2-rich water with significant CO2 excess compared to the atmosphere was the norm in the eastern Atlantic, rather than the exception. In the modern ocean, only those regions where nutrient utilization is inefficient (due to micronutrient limitation) are significant sources of CO2 to the atmosphere (Fig. 1D). A number of reconstructions indicate that the dust-related Fe flux to the eastern equatorial Atlantic during MIS 2 and 3 was likely even higher than today (by 2–5× ; Kohfeld and Harrison, 2001; Bradtmiller et al., 2007; Mahowald et al., 1999). In addition, diatom productivity increased throughout the equatorial Atlantic, suggesting an enhanced Si supply, probably from Southern Ocean sources (Bradtmiller et al., 2007). It is therefore very unlikely that the strong CO2 outgassing in the eastern equatorial Atlantic that we have shown during MIS 2 and 3 was due to micronutrient (Fe) or Si limitation (as in the modern eastern equatorial Pacific). Instead, it is probable that the strength of upwelling was sufficiently high to overwhelm the observed increase in productivity to maintain elevated surface nutrient (and pCO2sw) levels, as happens today during strong upwelling seasons (Bakker et al., 2001). Given the excess CO2 (associated with the reconstructed ΔpCO2sw) at GeoB1105 and the δ13C gradient between ODP 999 and GeoB1105, we calculate a glacial age (between 18 and 25 ka) surface water [PO43–] of ∼0.7 μM at GeoB1105, comparable to the eastern equatorial Pacific today (Fig. 1D; for details of this calculation, see the Data Repository). If the [PO43–] of the thermocline was similar to today (∼1.5 µM), and given the threefold increase in primary production (Schmidt et al., 2003), this estimate of surface water [PO43–] suggests a fivefold increase in the rate of upwelling (for full details, see the Data Repository). Because the [PO43–] of deep water upwelled in the eastern Atlantic during the last glacial likely decreased (due to the dominance of Glacial North Atlantic Intermediate Water at mid-depth; Straub et al., 2013), this is likely to be a minimum estimate.

As the CO2-rich surface waters in the east were advected westward during the last glacial toward the western (GeoB1523) equatorial Atlantic and Caribbean (ODP 999), ΔpCO2 rapidly decreased (to ∼20–30 µatm; Fig. 3A), probably through biological utilization of CO2, as in the modern equatorial Pacific (Quay, 1997). While primary productivity at GeoB1523 was not higher during MIS 2 (Fig. 3C), it is apparent from the regional compilations of Kohfeld et al. (2005) and Bradtmiller et al. (2007) that export production was enhanced throughout much of the eastern and western equatorial Atlantic (Bradtmiller et al., 2007, their figure 2c). We therefore conclude that the enhanced biological productivity in the equatorial Atlantic during the last glacial (e.g., 3× higher rates of Corg burial and primary productivity during MIS 2 compared to MIS 1; Fig. 3C) was primarily driven by enhanced upwelling in the eastern part of the basin (e.g., Schmidt et al., 2003) and the lateral transport of unused nutrient from the site of enhanced upwelling (and incomplete nutrient consumption) toward the central and western equatorial Atlantic. Our data from the eastern Atlantic demonstrate that this enhanced upwelling (and the presumed incomplete consumption of nutrients) increased the excess CO2 in the eastern Atlantic surface, thereby enhancing local CO2 outgassing. However, by analogy with the modern Pacific equatorial upwelling system, our data from the western equatorial Atlantic and Caribbean (Fig. 3A) suggest that the excess CO2 upwelled in the east was ultimately consumed by intensified downstream biological carbon export, fuelled by the unused nutrients that were laterally advected alongside the excess CO2. For this reason, the coupled intensification of upwelling in the east and a stronger biological pump in the wider equatorial Atlantic region likely had little or no effect on glacial atmospheric CO2 levels.

By reconstructing the patterns of pCO2sw across the equatorial Atlantic over the past 30 k.y. we demonstrate that the equatorial Atlantic underwent an at least fivefold increase in the rate of upwelling. Although this upwelling intensified regional productivity (Kohfeld et al., 2005; Schmidt et al., 2003), our reconstructed higher than modern ΔpCO2 for the eastern equatorial Atlantic indicates that export production did not keep pace with the strengthened upwelling of nutrients, giving rise to significant CO2 outgassing (Fig. 3A). We thus find that the low-latitude Atlantic was not a significant contributor to the drawdown of atmospheric CO2 concentrations during the last glacial, providing support for the dominance of high-latitude processes in glacial-interglacial CO2 change.

This work was supported by Natural Environment Research Council (NERC) grant NE/D00876X/2 to Foster and NERC grant NE/I006346/1 to Sexton and Foster. We thank Chris Coath for maintaining the mass spectrometers at the University of Bristol (UK) and James Rae and Mathis Hain for discussion on an earlier draft of this manuscript. We also thank Stephanie Henson for producing panel B of Figure 1, Barbara Donner for supplying samples from Site GeoB1105, and Karen Kohfeld, Jimin Yu, and two anonymous reviewers for thorough reviews that significantly improved the manuscript.

1GSA Data Repository item 2014351, data table, details of methods, and a figure showing phosphate concentration over the past 30 k.y. at GeoB1105, is available online at, or on request from or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.