Volume 33, Issue 11 p. 1151-1168
Research Article
Free Access

Ventilation of Northern and Southern Sources of Aged Carbon in the Eastern Equatorial Pacific During the Younger Dryas Rise in Atmospheric CO2

Samantha C. Bova

Corresponding Author

Samantha C. Bova

Department of Marine and Coastal Sciences, Institute of Earth, Ocean, and Atmospheric Sciences, State University of New Jersey Rutgers, New Brunswick, NJ, USA

Correspondence to: S. C. Bova,

[email protected]

Search for more papers by this author
Timothy D. Herbert

Timothy D. Herbert

Department of Earth, Environmental, and Planetary Sciences, Institute at Brown for the Study of Environment and Society, Brown University, Providence, RI, USA

Search for more papers by this author
Mark A. Altabet

Mark A. Altabet

School for Marine Science and Technology, University of Massachusetts, New Bedford, MA, USA

Search for more papers by this author
First published: 11 October 2018
Citations: 12

Abstract

Atmospheric carbon dioxide (CO2) levels rose by ~90 ppmv during the last deglaciation, but the source of this carbon remains unknown. One popular hypothesis suggests carbon accumulated in the deep Southern Ocean, becoming increasingly radiocarbon (14C) depleted during the last glacial period, and was released into Antarctic Intermediate Water (AAIW) and, subsequently, the atmosphere during deglaciation. Detection of extremely 14C depleted carbon in the intermediate-depth tropical oceans during periods of atmospheric CO2 rise was initially considered the smoking gun for the hypothesis, but attempts to reproduce the anomalies closer to the Southern Ocean source have largely failed. Here we present new 14C records from four cores recovered at intermediate depths in the eastern equatorial Pacific (EEP). In the context of additional geochemical records, including benthic foraminiferal stable isotopes and sedimentary nitrogen isotopes, we demonstrate that the extreme 14C anomalies observed during the last deglaciation do not reflect the radiocarbon content of AAIW. We show that although AAIW likely transported modestly 14C depleted carbon to the EEP subsurface, the extreme 14C signatures might reflect a distinct source of aged carbon arriving from the north, suggesting the North Pacific helped transport deep ocean carbon to the atmosphere during the last deglaciation. In the EEP, additional local hydrothermal inputs of 14C-free carbon near the Galapagos Islands magnified the already low-14C signatures. Finally, regardless of arrival route (North/South Pacific or hydrothermal), 14C-depleted carbon was released from the EEP subsurface during a late deglacial pulse of renewed upwelling, likely contributing to the Younger Dryas rise in atmospheric CO2.

Key Points

  • Two distinct sources of aged carbon were detected in the eastern equatorial Pacific during the last deglaciation
  • Antarctic Intermediate Water was not the source of extremely radiocarbon depleted carbon during the last deglaciation
  • The eastern equatorial Pacific was an important source for the Younger Dryas rise in atmospheric carbon dioxide

Plain Language Summary

Atmospheric carbon dioxide (CO2) is an important greenhouse gas that helps propagate major changes in Earth's climate. The last deglaciation (11,000–18,000 years BP) is the most recent natural example of rapid CO2 rise comparable in magnitude, though not in duration, to the human perturbation. Thus, understanding what drove atmospheric CO2 rise during this transition period may help us better understand the carbon system and predict its response to future change. It has been previously hypothesized that the deep ocean stores carbon away from the atmosphere during glacial intervals, thereby cooling global climate, but it is not yet clear what drives its release or by what pathway it returns to the atmosphere during deglaciation. In this paper, we use radiocarbon measurements on the shells of deep-dwelling foraminifera (single-celled protists with shells) to detect the release of glacially stored deep ocean carbon, which should have an aged or low radiocarbon signature, in the eastern equatorial Pacific. The data presented here provide evidence for the release of this aged deep ocean carbon in the eastern equatorial Pacific during the last deglaciation and also reveal an influx of radiocarbon-free geologic carbon to the study site, likely sourced from nearby hydrothermal systems.

1 Introduction

Atmospheric carbon dioxide (CO2) levels are an excellent indicator of Earth's climate state. Ice core records from Antarctica demonstrate a tight coupling between global temperatures and atmospheric CO2 over the past 800 kyr (Luthi et al., 2008; Monnin et al., 2001; Petit et al., 1999; Siegenthaler et al., 2005), with shifts in atmospheric CO2 accounting for close to half of the radiative forcing required to move between glacial and interglacial states (Weaver et al., 1998; Yoshimori et al., 2001). Since these changes were first observed nearly 40 years ago (Berner et al., 1978), studies of oceanic radiocarbon have demonstrated that the deep ocean is capable of storing and releasing enough carbon to explain glacial-interglacial CO2 changes (e.g., Broecker, 1982; Broecker & Peng, 1989; Knox & McElroy, 1984; Sarmiento & Toggweiler, 1984; Sarnthein et al., 2013; Skinner et al., 2015). The requirements for carbon storage within the deep ocean are twofold: (1) a source of carbon to the deep ocean via the solubility, soft tissue, and/or carbonate pumps and (2) isolation of deep ocean waters from the surface ocean to prevent gas exchange with the atmosphere. Evidence for these conditions should be evident in the proxy record as large chemical gradients between surface and deep waters and the buildup of 13C-depleted and 14C-depleted, nutrient-rich, and oxygen-poor waters in the subsurface.

Aged, low-oxygen, and nutrient-rich deep water masses have been detected during the Last Glacial Maximum (LGM) at numerous sites throughout the global ocean (Basak et al., 2018; Cook & Keigwin, 2015; de la Fuente et al., 2015, 2017; Galbraith et al., 2007; Jacobel et al., 2017; Keigwin, 2004; Keigwin & Lehman, 2015; Lund et al., 2011; Okazaki et al., 2012; Robinson et al., 2005; Ronge et al., 2016; Sikes, Cook et al., 2016; Skinner & Shackleton, 2004; Skinner et al., 2010, 2015, 2017, 2014). LGM radiocarbon signatures from intermediate and deep water masses in the southwestern and south central Pacific provide evidence for aged carbon within Pacific deep water (PDW; Ronge et al., 2016), while I/Ca measurements on planktonic foraminifera indicate the presence of low-oxygen waters in the Southern Ocean subsurface (Lu et al., 2016). Radiocarbon signatures within PDW in the North Pacific are depleted as well, suggesting widespread reduction in the ventilation of PDW relative to today (Cook & Keigwin, 2015; Galbraith et al., 2007; Lund et al., 2011). Occurrences of old glacial deep waters are also observed in the South Atlantic and North Atlantic (Skinner & Shackleton, 2004; Skinner et al., 2010, 2014), with moderately aged deep water in the Drake Passage (Burke & Robinson, 2012; Chen et al., 2015). Although the magnitude of the 14C anomaly in the deep ocean is not consistent at all sites, glacial water mass ventilation ages are consistently older at depth in the deep Atlantic, the Southern Ocean, and the deep North Pacific relative to today (e.g., Zhao et al., 2018), supporting the presence of a deep ocean glacial carbon reservoir responsible for the drawdown of atmospheric CO2.

During the last deglaciation, the aged deep carbon pool disappeared, likely fueling the two-part rise in atmospheric CO2 during Heinrich Stadial 1 (HS1: 14.6–17.8 ka) and the Younger Dryas (YD: 11.7–12.9 ka; Marchitto et al., 2007; Marcott et al., 2014; Monnin et al., 2001; Ronge et al., 2016; Sikes, Cook et al., 2016; Skinner et al., 2010, 2014). Enhanced upwelling around the Antarctic continent triggered by poleward shifts in the position of the westerly winds, reduced sea ice extent, salinity-driven buoyancy differences in the Southern Ocean, or some combination of the three (e.g., Anderson et al., 2009; Keeling & Stephens, 2001; Sigman et al., 2010; Toggweiler et al., 2006; Watson et al., 2015) reinvigorated deep ocean circulation, potentially ventilating aged, 14C-depleted carbon, and low-oxygen waters to the Southern Ocean surface. Because carbon equilibration between surface waters and the atmosphere is slow, on the order of a year, the majority of upwelled carbon would have been resequestered via Antarctic Intermediate Water (AAIW) formation and spread throughout the intermediate depth ocean, while retaining a 14C-depleted radiocarbon signature, returning to the atmosphere predominantly in low-latitude upwelling regions, such as the eastern equatorial Pacific (EEP; e.g., Takahashi et al., 2002).

Strongly depleted 14C concentrations have been documented at three locations influenced by AAIW in the modern ocean (Arabian Sea, offshore of the Baja Peninsula, and the Galapagos Platform), coincident with the two-part rise in atmospheric CO2 and periods of relative warmth in Antarctica, suggesting the water mass was an important conduit for the release of CO2 to the atmosphere (Basak et al., 2010; Bryan et al., 2010; Lindsay et al., 2015, 2016; Marchitto et al., 2007; Stott et al., 2009). Efforts to corroborate these findings at sites close to AAIW formation zones, however, find comparatively small changes in water mass ventilation ages, with all those that exhibit strongly 14C depleted radiocarbon signatures located at the furthest reaches of AAIW's influence, within the tropical Pacific and Indian Oceans (Figure 1; Bryan et al., 2010; Burke & Robinson, 2012; De Pol-Holz et al., 2010; Kennett & Ingram, 1995; Mangini et al., 2010; Marchitto et al., 2007; Rose et al., 2010; Siani et al., 2013; Sortor & Lund, 2011; Stott et al., 2009). This is puzzling because if AAIW was the conduit for old carbon, the most 14C-depleted signatures should be observed closest to the Southern Ocean source, before the signature is diluted by better ventilated surface and subsurface waters.

Details are in the caption following the image
Map depicting the formation regions and extent of Pacific intermediate waters. Arrows show the likely flow directions at intermediate depth (Bostock et al., 2010). Stars indicate the location of benthic radiocarbon records that span the last deglaciation in the Pacific. Pale yellow stars = this study; seafoam green stars = previous studies (Burke & Robinson, 2012; Davies-Walczak et al., 2014; De Pol-Holz et al., 2010; Lindsay et al., 2015, 2016; Marchitto et al., 2007; Rose et al., 2010; Siani et al., 2013; Stott et al., 2009; Zhao & Keigwin, 2018). Figure adapted from Bostock et al. (2010, 2013). AAIW = Antarctic Intermediate Water; NEqPIW = Northern Equatorial Pacific Intermediate Water; NPIW = North Pacific Intermediate Water; SEqPIW = Southern Equatorial Pacific Intermediate Water.

In this study, we evaluate the radiocarbon and ventilation histories of intermediate and surface water masses in the EEP over the past 25 kyr using five independent proxies sensitive to water mass ventilation and circulation. Four new paired planktonic and benthic foraminiferal radiocarbon records from the Northern Peru Margin, at 373- and 1,023-m water depth (CDH 23: 03°44.95′S, 81°08.05′W and CDH 26: 03°59.16′S, 81°18.52′W, respectively), and the Galapagos Platform, at 617- and 595-m water depth (GGC 43: 01°15.13′S, 89°41.07′W, 617 m and CDH41: 01°15.94′S, 89°41.88′W, 595 m), are the foundation of this study and provide a direct measure of water mass age, or time since the water mass was at the surface ocean, over a depth transect in the EEP upper water column. Supporting data sets include bulk sediment nitrogen isotopes (δ15N), the abundance of C37 alkenones in the sediment (C37total; Bova et al., 2015), and benthic foraminiferal δ13C and δ18O records (Bova et al., 2015). Although at present, all four of our study sites are bathed predominantly by southern sourced mode and intermediate waters (the hypothesized conduits of aged carbon from a deep Southern Ocean reservoir, Figure 1), these data demonstrate that the configuration of intermediate waters in the Pacific at the LGM and deglaciation was different from that observed in the modern ocean. Based on this new configuration, we are able to (1) reconcile seemingly discordant geochemical records from the tropical and South Pacific; (2) fingerprint two distinct sources of 14C-depleted waters to the EEP subsurface and surface ocean during the early deglaciation; and (3) trace the release of aged carbon from the EEP subsurface, primarily during the second deglacial pulse in atmospheric CO2 during the YD.

2 Materials and Methods

2.1 Area of Study

Four sediment cores recovered from three intermediate water depths (373, 595, 617, and 1,023 m) in the EEP were recovered on the 2009 cruise of the R/V Knorr. These cores sample subantarctic mode water (SAMW) and equatorial Pacific intermediate water (EqPIW) a mixture of AAIW with contributions from PDW and North Pacific Intermediate Water (NPIW; Figure S1 in the supporting information; Bostock et al., 2010, 2013; Fiedler & Talley, 2006; Kalansky et al., 2015). Today, CDH 23 (03°44.95′S, 81°08.05′W, 373 m) and CDH 26 (03°59.16′S, 81°18.52′W, 1023 m), recovered along the Northern Peru Margin, are bathed predominantly by southern sourced waters, SAMW and AAIW, respectively. At 600 m on the Galapagos Platform, GGC 43 (01°15.13′S, 89°41.07′W, 617 m) and CDH 41 (01°15.94′S, 89°41.88′W, 595 m) lie near the boundary between SAMW and AAIW.

SAMW and AAIW are sourced primarily from the Southern Ocean (Bostock et al., 2010, 2013). SAMW forms within the subantarctic zone during late winter convective overturning and travels a circuitous route from the Southern Ocean into the western equatorial Pacific before finally flowing eastward across the Pacific into the EEP (Bostock et al., 2010, 2013; Qu et al., 2013; Rintoul & England, 2002; Toggweiler et al., 1991). EqPIW is composed primarily of AAIW with some contributions from NPIW and PDW. AAIW forms closer to the continent than SAMW, south of the subantarctic front, in three primary formation regions: (1) southwest and east of New Zealand, (2) west of the East Pacific Rise, and (3) west of the Drake Passage (Figure 1; Bostock et al., 2013). Intermediate waters in the EEP subsurface are sourced primarily from the southeast Pacific site west of the Drake Passage, but its geochemical signature is distinct from the SE AAIW endmember. EqPIW is significantly more saline, nutrient rich, and lower in oxygen and exhibits an older Δ14C signature than Drake Passage AAIW, which requires mixing between AAIW and an older, more nutrient-rich water mass, most likely PDW (Figure S1). NPIW is not a significant source of intermediate waters south of ~2°N latitude in the EEP today, but some studies suggest greater southward penetration of NPIW during the last glacial period (e.g., Max et al., 2014, 2017; Rippert et al., 2017).

2.2 Radiocarbon Measurements

Sixty paired planktonic and benthic radiocarbon measurements were made on four cores from three sites in the EEP (Data Set S1). We use a splice between two cores from the Galapagos Platform to extend the record from 600 m into the last glacial period. In total, we made 33 paired measurements from the Galapagos cores (GGC 43 and CDH 41), 15 from CDH 23, and 12 from CDH 26. Planktonic 14C dates, measured on Neogloboquadrina dutertrei (CDH 23 and CDH 26) and Globogerinoides ruber (GGC 43 and CDH 41), were collected as part of a previous study (Bova et al., 2015) and paired here with benthic 14C ages to calculate age offsets between surface and deep water masses. Benthic 14C dates were measured on 1.4 to 7.9 mg of mixed benthics (Bolivina spp., Hanzawaia concentrica, and Uvigerina spp.) at the Peru Margin sites and Uvigerina spp. at the Galapagos (Data Set S1). Particles and other debris were removed from each sample with a fine hairbrush before sonicating repeatedly in deionized water. All radiocarbon dates were analyzed at the National Ocean Sciences Accelerator Mass Spectrometry facility at Woods Hole Oceanographic Institute.

2.3 Nitrogen Isotopes

We produced an ~25-kyr bulk sediment δ15N record from site CDH 26. Samples were taken every 16 cm from the top 25.4 m of CDH 26, providing a measurement approximately every 140 years. Bulk sediment was homogenized, and an ~25-mg aliquot was prepared for δ15N analysis by weighing and directly loading into tinfoil boats prior to combustion. All δ15N measurements were made in the Altabet lab at the University of Massachusetts at Dartmouth. Details of analytical protocol and instrumentation were published by Higginson and Altabet (2004a). All samples were measured in replicate and averaged. δ15N data are reported relative to atmospheric N2.

3 Results

3.1 Radiocarbon

3.1.1 Reservoir Ages and Core Chronologies

Reservoir age variations at the four sites were estimated by (1) adjusting modern reservoir age estimates near the core sites as a function of changing atmospheric pCO2 concentrations (Galbraith et al., 2015); (2) applying nearby, previously published reservoir age reconstructions (de la Fuente et al., 2015; Umling & Thunell, 2017; Zhao & Keigwin, 2018); and (3) chronostratigraphically aligning benthic foraminiferal δ18O records to Antarctic and Greenland ice core records and chronologies. Age model 1 is the simplest approach and allows reservoir ages to vary from their modern value of ~500 years (Druffel, 1981; Etayo-Cadavid et al., 2013; Taylor & Berger, 1967) solely as a function of atmospheric pCO2 levels (Galbraith et al., 2015). This approach assumes variations in ocean circulation, and regional upwelling had no impact on surface water 14C signatures and thus the surface ocean reservoir age. Model simulations indicate that global reservoir ages varied as a linear function of 1/pCO2, reaching maximum values at the LGM of +250 years (Galbraith et al., 2015).

Age models 2 and 3 allow for variable reservoir age estimations driven by changes in ocean circulation, a scenario that seems likely during a period of rapid global change, such as the last deglaciation, in the EEP, a highly dynamic upwelling region. This scenario is further supported by two prior reservoir age reconstructions from the EEP, which both show dramatic variations in surface reservoir ages during the last deglaciation (de la Fuente et al., 2015; Umling & Thunell, 2017). We applied reservoir age reconstructions from site TR163-23 (Umling & Thunell, 2017) to construct age model 2 for our four EEP cores by linearly interpolating the reservoir ages between uncalibrated planktonic 14C age measurements. One drawback to this approach, however, is the assignment of reservoir ages where no direct calendar age constraints exist. The reservoir age reconstruction from site ODP 1240 (de la Fuente et al., 2015) was also applied to our data but was not significantly different from age model 2 during the Holocene and deglaciation and was thus not shown in the main text of this article (see Figure S5).

Age model 3 was constructed by chronostratigraphically aligning the benthic foraminiferal δ18O records from core CDH 26 and GGC 43 to ice core δ18O records from Antarctica and Greenland. This exercise is similar to that used to reconstruct reservoir ages at EEP sites TR163-23 and ODP 1240. However, these correlations were assigned using planktonic, rather than benthic, foraminiferal δ18O records (i.e., de la Fuente et al., 2015; Umling & Thunell, 2017). Our correlations therefore rely on an additional assumption that high-latitude temperature anomalies were transported rapidly to our core sites. Following this approach, we calculated reservoir ages as the difference between the interpolated atmospheric 14C age (based on IntCal13, Reimer et al., 2013) from the tie points and the measured planktonic 14C ages. To avoid specious reservoir age estimates outside of the aligned interval, we only estimate reservoir ages at measured planktonic 14C dates positioned within the aligned interval (between tie points). An alternative method is to calculate the difference between interpolated planktonic 14C age at each tie point and the atmospheric 14C age at the calendar age of each tie point, but differences are negligible when this method is used instead.

Core CDH 26, recovered at 1,023 m along the North Peru Margin, was aligned to the European Project for Ice Coring in Antarctica Dome C δ18Oice record (Jouzel et al., 2007) in the interval between 9 and 18 ka (Figure 2a). We were unable to define tie points between the two records during the LGM and Holocene due to a lack of identifiable features in the ice core record and reduced resolution in the benthic δ18O record during these intervals. The uppermost (~500 years) and lowermost (~750 years) reservoir age estimates calculated from tie points to the ice core record appear similar to variations in reservoir age expected solely as a function of varying atmospheric pCO2 levels (Galbraith et al., 2015). Outside of the ice core constrained interval we therefore adjust the observed modern reservoir age of ~500 years (Druffel, 1981; Etayo-Cadavid et al., 2013; Taylor & Berger, 1967) as a function of atmospheric pCO2 variations. Our ice-core-aligned chronology is substantiated by a new reservoir age record from the Panama Basin based on the 14C age of coexisting twigs, whose 14C content should reflect the atmospheric value, and planktonic foraminifera (Zhao & Keigwin, 2018; Figure S3). These wood-based reservoir age estimates do not rely on correlation to the ice cores and therefore are free of many of the assumptions relied upon to construct the CDH 26 chronology, foremost among them an assumed synchronicity between SEqPIW properties and high latitude climate. Thus, the similarity between our reservoir age estimates and those from wood 14C ages in the Panama Basin gives us increased confidence in our age-depth model. To construct the age model for nearby site CDH 23, we transferred the reservoir age estimates from CDH 26 by linearly interpolating reservoir ages between uncalibrated planktonic 14C dates. Core CDH 23 does not span the full deglacial interval and is therefore difficult to confidently align.

Details are in the caption following the image
Chronostratigraphic alignment of cores CDH 26 and GGC 43 to ice core chronologies. Dashed lines and X's indicate the location of tie points. Solid circles along the x-axis indicate the location of measured planktonic 14C dates. (a) Tie point alignment of the core CDH 26 benthic foraminiferal δ18O record to the European Project for Ice Coring in Antarctica (EPICA) Dome C δ18Oice record (Jouzel et al., 2007). (b) Tie point alignment of the core GGC 43 benthic foraminiferal δ18O record to the GISP2 δ18Oice record (Grootes et al., 1993). The benthic foraminiferal data are reported as δ18O versus the Vee Pee Dee Belemnite (VPDB) standard.

Core GGC 43, recovered from 617 m on the Galapagos Platform, was correlated with the GISP2 δ18Oice record below 215 cm (~11.8 ka; Figure 2b) and to the European Project for Ice Coring in Antarctica Dome C δ18Oice record (Jouzel et al., 2007) below 180 cm (Figure S2). Visual inspection of the benthic foraminiferal δ18O record from core GGC 43 suggests a better correspondence to millennial-scale features in the Greenland, rather than Antarctic, ice core record. Reservoir ages calculated from the proposed alignment to the GISP2 δ18Oice record are similar to nearby reconstructions just to the north (TR163-23) and east (ODP 1240) of our core site (de la Fuente et al., 2015; Umling & Thunell, 2017), which provides additional support for the proposed correlation with Northern Hemisphere (NH) climate events during the last deglaciation (Figure S3). In contrast, the alignment to the Antarctic ice core record produces very low reservoir age estimates during the deglaciation of ~67 years, far lower than those predicted along the Peru Margin or from the effect of changing atmospheric CO2 (Figure S4). Given the good correspondence between reservoir age estimates from the GISP2 δ18O alignment with nearby reconstructions (Figure S3), we utilize the tie points to GISP2 to construct age model 3 for core GGC 43. Outside of the ice core aligned interval (above 215 cm) in GGC 43, we utilize reservoir age estimates from nearby site, TR163-23 (Umling & Thunell, 2017). TR163-23 is relatively close to our site geographically, and, as previously noted, reservoir estimates from the site align well with those from calculated at site GGC 43 (Umling & Thunell, 2017; Figure S3). Because core CDH 41 does not span the full deglacial interval, we applied the reservoir age estimates from GGC 43 to planktonic 14C dates measured in core CDH 41 by linearly interpolating reservoir ages between uncalibrated planktonic 14C dates.

Using both the chronostratigraphic tie points and reservoir age-corrected planktonic foraminiferal 14C dates calibrated to the IntCal13 radiocarbon curve (Reimer et al., 2013), we constructed three different chronologies for each core using the Bayesian age-depth modeling program Bchron (Figure 3; Haslett & Parnell, 2008). For all age models, planktonic foraminiferal dates measured from samples at 20.25, 30.00, 40.25, 215.25, and 235.25 in core CDH 41 were identified as outliers by the software. These dates lie outside of the Bchron calculated 1 sigma uncertainty envelope in Figure 3d. We plot all three age models in Figure 3 of the main text. We suggest that age model 3 (red), based on our chronostratigraphic alignment to the ice cores, represents our best estimate for each of the four cores. Our various approaches for constructing core chronologies are nearly identical for the Peru Margin sites CDH 23 and CDH 26 but lead to significant offsets in the Galapagos core (GGC 43 and CDH 41) age-depth relationships. The largest discrepancies between the two age models occur at the Galapagos sites, where reconstructed reservoir ages are much larger during the deglaciation and LGM (Figure S3). Because reservoir ages are highly uncertain for the Galapagos cores, we always plot data from cores GGC 43 and CDH 41 on both age model 1 (pCO2-adjusted reservoir ages) and age model 3 (ice core aligned; similar to age models 2a and 2b).

Details are in the caption following the image
Age-depth models constructed using the Bchron software package (Haslett & Parnell, 2008) for cores (a) CDH 26, (b) CDH 23, (c) GGC 43, and (d) CDH41. Five 14C dates from core CDH 41, which lie outside the 1 sigma uncertainty envelope (solid lines or gray shaded area), were identified as outliers. Three age-depth models are shown for each core: Age model 1 (reservoir age corrections from pCO2 adjustment only, green), age model 2 (reservoir age corrections from TR163-23, Umling and Thunell, 2017, gray shaded area), and age model 3 (reservoir age corrections from alignment to ice core chronologies, red). See main text for more details on the construction of these age models. EPICA = European Project for Ice Coring in Antarctica.

3.1.2 Ventilation Ages

We employ three different 14C-based measures of water mass age and ventilation: (1) benthic foraminiferal Δ14C, (2) the difference between benthic and planktonic 14C ages, and (3) the difference between benthic foraminiferal 14C and the contemporaneous atmospheric 14C age (B-atm). Benthic foraminiferal Δ14C estimates the radiocarbon activity of the water mass in which the individual foraminifer grew. We calculate this value from the foraminiferal Accelerator Mass Spectrometry (AMS) 14C date and its estimated calendar age using the equation of Adkins and Boyle (1997), which accounts for decay since the foraminifera were alive. The calendar age is the largest source of uncertainty in the calculation. Benthic-planktonic 14C ages reflect the age offset between surface and subsurface waters at the core site, with higher values indicating reduced subsurface water mass ventilation. Benthic-planktonic 14C ages, however, do not account for possible variations in surface reservoir ages and therefore may not reflect true deep water ventilation ages. We use B-atm, calculated as the benthic-planktonic 14C age plus the surface reservoir age, to fully account for radiocarbon disequilibrium of EEP intermediate and mode waters from the atmosphere. The relatively high sediment accumulation rates at all core sites ensures that the benthic and planktonic populations found within a depth horizon were living approximately contemporaneously and therefore limits the likelihood for bioturbation to significantly impact the measured benthic-planktonic 14C age and B-Atm.

Patterns in Δ14C at all water depths are similar to trends in benthic-planktonic 14C age and B-Atm, with more negative Δ14C values corresponding to larger benthic-planktonic 14C age and B-Atm offsets (Figure 4). Depleted Δ14C values (larger benthic-planktonic 14C age/B-Atm) indicate reduced radiocarbon activity and thus reduced ventilation. Holocene Δ14C values from all water depths track atmospheric Δ14C values, exhibiting a near constant 100‰ offset. Δ14C signatures at 600- and 1,023-m water depth exhibit substantially larger offsets from the atmosphere and larger benthic-planktonic 14C age and B-Atm offsets during the deglaciation and last glacial period (prior to 11.8 ka), relative to their present-day offset and Holocene average. The Galapagos records (~600 m) exhibit the lowest Δ14C values and largest benthic-planktonic 14C age and B-Atm of the sites, with minor enrichments occurring between 15.2 and 13.6 ka before plunging back toward glacial values at the end of the Antarctic Cold Reversal (ACR) (~13.1 ka). Across the YD, Δ14C values in the 600-m records at the Galapagos Platform increase from −262‰ to −89‰ in ~1.5 kyr, while benthic-planktonic 14C age offsets decrease from 2,750 to 150 years. B-Atm offsets at ~600- and 1,023-m water depth are largest at the end of the last glacial period and shift to approximately modern values across the YD interval. At 373 m, benthic-planktonic 14C ages are relatively stable, and Δ14C values exhibit a near constant offset from the atmosphere over the past 14 kyr.

Details are in the caption following the image
Radiocarbon data from 373-, 595-, 617-, and 1,023-m water depth in the eastern equatorial Pacific. (a) Benthic foraminiferal Δ14C concentrations at the four core locations plotted relative to the atmosphere. Error bars reflect the interquartile range in the age-depth model and are angled because the calculation for Δ14C is contingent on calendar age. (b) Benthic-planktonic 14C ages at the same four sites. Error bars reflect interquartile range in the age-depth model in the x direction and the sum of the measurement errors for both the planktonic and benthic 14C dates in the y direction. (c) B-atm ages at the four eastern equatorial Pacific core locations. Error bars reflect interquartile range in the age-depth model in the x direction and the sum of the measurement errors for both the planktonic and benthic 14C dates in the y direction. Solid lines: 14C data plotted using age model 3, based on chronostratigraphic alignment to the ice core chronologies. Dashed lines: 14C data plotted using age model 1, in which reservoir ages are adjusted only as a function of changes in atmospheric pCO2 (Holocene: ~500 years, LGM: ~750; e.g., Galbraith et al., 2015).

3.2 Benthic Foraminiferal Stable Isotopes

Benthic foraminiferal stable isotope records from the three sites were measured on the infaunal species Uvigerina peregrina. Benthic foraminiferal δ18O records were discussed extensively in Bova et al. (2015), though the age-depth models have been modified since the original publication leading to differences in event timing. Benthic foraminiferal δ13C records from the three sites reflect the combined influence of a number of factors, including water mass ventilation, air-sea exchange in the water mass source region, local rates of primary productivity, and pore water chemistry (Bova et al., 2015). Here we combine the records from the Galapagos Platform (CDH 41: 595 m and GGC 43: 617 m) to form a complete spliced record that spans the last 25 kyr. Benthic foraminiferal stable isotope measurements from GGC 43 extend down to 15.7 kyr, with all measurements older than 15.7 kyr coming from site CDH 41. Benthic foraminiferal δ13C records, from the Peru Margin cores, exhibit larger amplitude variations and starkly different trends over the past 25 kyr, relative to the spliced record from the Galapagos Platform (see Figure S6; Siani et al., 2013). At site CDH 26, at 1,023-m water depth, we observe decreasing δ13C values at the end of the last glacial period and during the early deglaciation, followed by a period of relatively enriched foraminiferal δ13C between ~15.8 and 13.9 ka. δ13C values at site CDH 26 decrease again following the mid-deglacial peak in δ13C before rising steadily across the YD interval. In contrast, benthic δ13C values recorded at 600 m on the Galapagos Platform exhibit a smaller and more gradual decrease during the late deglaciation, before steadily increasing beginning ~12.3 ka and continuing toward present day.

3.3 Bulk Sediment δ15N

Sedimentary nitrogen isotopes reflect the isotopic composition of organic matter deposited in the sediments, which may record the preformed δ15N signature of the source water (e.g., Francois et al., 1997), the degree of local NO3 consumption (e.g., Altabet & Francois, 1994), and/or the extent of denitrification in the subsurface (e.g., Altabet et al., 1995). Although bulk sediment δ15N records are more sensitive to secondary isotopic alteration that can occur during sinking and burial than microfossil bound N isotope proxies, bulk sediment records from continental margins where sedimentation rates are high and thus the exposure time of organic matter to oxygen is low are generally considered robust (e.g., Robinson et al., 2012). During the last glacial period we observe a peak value of 6.37‰ at the LGM (21.4 ka; Figure 5). Post-LGM, δ15N values decrease rapidly until 20 ka and then remain stable at ~5.3‰ until ~16.6 ka. During the latter half of HS1 (16.6 to 14.2), we observe a rapid increase in sedimentary δ15N of nearly 1.5‰. The rise is composed of two distinct pulses centered at 16.2 and 14.7 ka, respectively, separated by a 1-kyr period of relatively stable values (16 to 15.1 ka). Sedimentary δ15N remains high until 12.9 ka and then decreases by ~0.6‰ just across the YD interval. Values remain stable until ~9 ka and then steadily decrease toward present day.

Details are in the caption following the image
Bulk sediment δ15N (brown, reversed scale) and the C37total index for phytoplankton production (green; Bova et al., 2015) as measured in core CDH 26 from the North Peru Margin. Nitrogen isotopes and the C37total are anticorrelated over the last 25 kyr, suggesting a common driver of variability in the two proxy records.

3.4 Alkenone Abundances

The abundance of C37-alkenones in the sediment (C37total) provides a qualitative measure of haptophyte production, which is linked to total phytoplankton productivity (e.g., Brassell, 1993). Along the northern Peru Margin (CDH 26), we observe large changes in the C37total, which are negatively correlated to trends in bulk sediment nitrogen isotopes (Figure 5). We observe maximum alkenone abundances during the last glacial period, reaching peak values ~19.5 ka, before decreasing sharply at the onset of the HS1. Values remain low during the last deglaciation (10–18 ka), with two small peaks during the latter half of HS1 (14.2–16.6 ka) and the YD (11.6–13 ka). The C37total productivity index indicates Holocene productivity was higher than that observed during the deglaciation but not as high as during the LGM.

4 Discussion

In this study, we present four new radiocarbon records from the intermediate depth EEP (Figure 4). The records from 595- and 617-m water depth on the Galapagos Platform were generated in an effort to replicate the exceptional record produced at site VM21-30 by Stott et al. (2009) from this location, which documented some of the lowest radiocarbon activity recorded anywhere in the global oceans during the last deglaciation. Two additional records from the Gulf of Guayaquil along the northern Peru Margin, which bracket the Galapagos sites at 373- and 1,023-m water depth, were produced to constrain the vertical extent of the extreme radiocarbon values. The results from these three sites were surprising in a number of different ways. First, the Δ14C records from ~600-m water depth on the Galapagos Platform are similar to the VM21-30 Δ14C record (Stott et al., 2009) during the LGM and early deglaciation but are not nearly as depleted after 16 ka. Beginning ~16 ka our record appears to instead track the less depleted (though still anomalously low) intermediate depth Δ14C record from off the western coast of Baja California (23.5°N, 111.6°W; Lindsay et al., 2015, 2016; Marchitto et al., 2007). Second, the radiocarbon records from the Peru Margin exhibit far more modest decreases in 14C concentrations during the LGM and deglaciation relative to the records from the Galapagos Platform and off Baja, consistent with records produced along the pathway of AAIW to the EEP (Burke & Robinson, 2012; De Pol-Holz et al., 2010; Rose et al., 2010; Siani et al., 2013) and in the Panama Basin (Zhao & Keigwin, 2018). Our challenge is therefore to decipher (1) what drives the observed differences between deglacial intermediate water records both within the EEP and the wider Pacific and (2) the relevance of the intermediate depth 14C records, if any, to the rise in atmospheric CO2 during the last deglaciation.

4.1 Deciphering Divergent Geochemical Histories of Intermediate Waters in the EEP

The equatorial Pacific is a crossroads for water masses formed in the northern and southern subtropical to polar latitudes (Figure 1; Bostock et al., 2010, 2013; Fiedler & Talley, 2006). At intermediate depths, geochemical tracers (salinity, oxygen, silicate, alkalinity, dissolved inorganic carbon, δ13CDIC, and Δ14C) fingerprint the influence of at least three water masses, principally AAIW with contributions from NPIW and PDW. The mixture is referred to as EqPIW (Bostock et al., 2010). PDW contributions to the EEP subsurface are essential to explaining the significantly older 14C ages detected in EqPIW relative to AAIW and NPIW in the modern ocean (Bostock et al., 2010, 2013), though the process by which PDW rises to mix with overlying intermediate waters is unclear (Figure S1). NPIW and AAIW arrive in the EEP principally from the north and south, respectively, following the gyre circulation (Figure 1). As a result, we observe a north-south asymmetry at ~2°N in nonconservative geochemical tracers, principally silicate, likely arising from greater NPIW contributions to the north and greater AAIW contributions to the south (Bostock et al., 2010; Sarmiento et al., 2004). Today there is no North Pacific source to thermocline waters east of the Galapagos, and a sharp boundary at the equator between South and North Pacific contributions is identified based on tritium and 3He sections from the World Ocean Circulation Experiment (Fiedler & Talley, 2006; Jenkins, 1996). Thus, EqPIW comes in two flavors, Northern EqPIW (NEqPIW) and Southern EqPIW (SEqPIW), partitioned between the equator and 2°N in the modern ocean. At present, all four of our study sites lie south of the latitudinal divide between NEqPIW and SEqPIW and are thus bathed predominantly by southern sourced mode and intermediate waters (Figure 1).

The latitudinal boundary between NEqPIW and SEqPIW is not stationary and has likely shifted over the past 25 kyr in response to changes in intermediate water properties (Basak et al., 2010; Max et al., 2014, 2017; Pena et al., 2013; Rippert et al., 2017). Vertical displacements driven by changes in buoyancy and/or formation rates of intermediate and abyssal waters across glacial-interglacial cycles have been documented and likely drove variations in the relative contributions of the three water masses across the EEP (Herguera et al., 2010; Jaccard & Galbraith, 2013; Jaccard et al., 2009; Ronge et al., 2015; Sikes, Elmore et al., 2016). At Ocean Drilling Program site 1240, located east of the Galapagos on the equator, carbon isotopes measured on thermocline dwelling planktonic foraminifera appear to track carbon isotope variations within NPIW at the end of the last glacial period and early deglaciation. These data are used to argue for greater penetration of NPIW to the EEP during the late glacial (Max et al., 2017; Rippert et al., 2017). Lastly, though sparsely sampled, neodymium isotopes at the same site indicate broad swings in water mass contributions to the EEP thermocline across this interval (Pena et al., 2013).

Given the modern heterogeneity in EEP intermediate water properties (Bostock et al., 2010, 2013) and the evidence of variable contributions of North Pacific water masses to the EEP over the last glacial cycle (Max et al., 2017; Rippert et al., 2017), we suggest that it is plausible that disparities between radiocarbon and other geochemical records at the Galapagos Platform and the northern Peru Margin over the past 25 kyr reflect, at least in part, the influence of distinct water mass combinations at each location. We argue that the geochemical divide, today located between the equator and 2°N in the EEP, shifted southward during the LGM and deglaciation, bathing the Galapagos core sites at ~1°S with substantially more northern sourced water relative to today. The Galapagos sites are positioned further north and west of the Peru Margin core sites, while also lying ~400 m shallower than CDH 26 and ~200 m deeper than CDH 23. We suggest the disparities in the geochemical records from the Galapagos sites and those along the Peru Margin reflect a combination of vertical and lateral variations in water mass contributions across the EEP. Thus, the geochemical records (δ18O, δ13C, and Δ14C) from the Peru Margin dominantly reflect variations in the physiochemical properties of Southern Ocean mode and intermediate waters, while the same records from Galapagos core locations reflect the properties of water masses formed in the North Pacific: NPIW and PDW (Figure 7).

At 3.7°S, the sites recovered along the Peru Margin were bathed predominantly by Southern Ocean water masses over the past 25 kyr. Although the shallower record, recovered at 370-m water depth, spans just the last 14 kyr, geochemical records from 1,000 m can be traced directly to an AAIW source in the Southern Ocean over the full 25-kyr record (Bova et al., 2015; Figure 2). Benthic oxygen isotopic records indicate a close correspondence of seawater temperature and salinity with climatic change over the Antarctic Continent (Figure 2). Bova et al. (2015) demonstrate that temperatures at 1,000-m water depth responded synchronously with Antarctic warming at the onset of the last deglaciation. Carbon isotope records can also be tied directly to records of AAIW from the SE Pacific (Figure S6; Siani et al., 2013). AAIW records exhibit three diagnostic features: (1) a trend toward more depleted carbon isotope values 20 to 18 ka, (2) a sharp rise in foraminiferal δ13C during the latter half of HS1 followed by a period of relatively stable values, and (3) a second sharp decrease in δ13C during the ACR. Although uncertainties in our age model preclude the identification of a precise timing of these events in our records, the carbon isotopic record from 1,000-m water depth along the Peru Margin appears to resolve all three of these features, despite measurement on the infaunal species U. peregrina (Figure S6).

In contrast, the benthic foraminiferal stable isotopic records from 600 m on the Galapagos Platform are not consistent with a Southern Ocean source and instead likely reflect substantially greater inputs of NPIW and/or PDW (Bova et al., 2015). Benthic foraminifer δ18O values correspond best to NH, rather than Southern Hemisphere (SH), millennial-scale events during the last deglaciation, and our inferred alignment to the GISP2 ice core based on this correspondence generates reservoir age estimates similar to two nearby reservoir age reconstructions (Figures 2 and S3; de la Fuente et al., 2015; Umling & Thunell, 2017). We further emphasize that the onset of the initial deglacial decrease in the Galapagos benthic δ18O record is delayed relative to the deeper Peru Margin site (Bova et al., 2015) and the rise in Antarctic temperatures as documented by ice cores (Jouzel et al., 2007). Given that SSTs in both the SAMW and AAIW source regions rose shortly after 19 ka, in phase with Antarctica and each other (Caniupan et al., 2011; Ho et al., 2012; Kaiser et al., 2007), the ~1-kyr delay in benthic δ18O at the Galapagos site (600 m) relative to the North Peru Margin (1,000 m) cannot result from differences in water mass contributions from the south but instead must reflect variations in water mass contributions from the north. Finally, benthic foraminiferal carbon isotopes are also distinct from those observed along the margin; δ13C values are more enriched and comparatively stable during the LGM and deglaciation and do not exhibit any of the AAIW diagnostic features (Figure S6).

The observed radiocarbon histories from the three sites provide our final and most definitive evidence for the influence of distinct water masses at intermediate depths at 3.7°S along the northern Peru Margin and ~1°S on the Galapagos Platform (Figure 4). A synthesis of previously published radiocarbon records spanning the last deglaciation reveals characteristic radiocarbon histories by water mass (Figure 6). Benthic foraminiferal radiocarbon records recovered from Pacific AAIW/SEqPIW maintain a relatively constant 14C offset from the atmosphere over the past 30 kyr, with slight decreases in 14C activity during the LGM and early deglaciation (Burke & Robinson, 2012; De Pol-Holz et al., 2010; Rose et al., 2010; Siani et al., 2013). The Peru Margin records (this study), from 373 and 1,023 m, compare well and are plotted with these records. The new 14C record from the intermediate depth Panama Basin also aligns well with Pacific AAIW/SEqPIW radiocarbon records, which suggests SEqPIW penetrated farther north along the continental margin (Zhao & Keigwin, 2018; Figure 7). In contrast, intermediate depth sites bathed by NEqPIW exhibit substantial reductions in intermediate water 14C content, with the offset between intermediate water and atmospheric Δ14C values rising by 100 to nearly 350‰ during the YD and HS1 (Lindsay et al., 2015, 2016; Marchitto et al., 2007). The Galapagos Platform records compare best with these records. Our Δ14C record from 600-m water depth on the Galapagos Platform aligns well with the intermediate depth Δ14C record from off the western coast of Baja California (23.5°N, 111.6°W) over the last ~16 ka (Lindsay et al., 2015, 2016; Marchitto et al., 2007). Minimum Δ14C values of approximately −250‰ is observed in the Baja and the Galapagos records during the YD interval.

Details are in the caption following the image
Δ14C evolution over the past 30 kyr by water mass relative to the atmosphere (Reimer et al., 2013). Radiocarbon records from SEqPIW/AAIW depths in the Pacific Ocean maintain a relatively constant offset with the atmospheric Δ14C concentrations over the past 30 kyr. The North Peru Margin record 1,023-m water depth compares well to the AAIW Δ14C history. Sites bathed by NEqPIW exhibit negative Δ14C anomalies during the deglaciation, similar to what we observed at 600-m water depth on the Galapagos Platform. The radiocarbon history of PDW in the South and equatorial Pacific is characterized by low Δ14C concentrations during the last glacial period, which recover to near modern values by the end of HS1 (de la Fuente et al., 2015; Ronge et al., 2016; Sikes et al., 2016; Umling and Thunnell, 2017). Gray bars indicate periods of warming in Antarctica and rising atmospheric CO2 (Ahn & Brook 2014; Monnin et al., 2001, 2004; Marcott et al., 2014). For the Galapagos cores presented in this study (GGC 43 and CDH 41) solid lines show the 14C data plotted using age model 3 and dashed lines show the 14C data plotted using age model 1. AAIW = Antarctic Intermediate Water; NEqPIW = Northern Equatorial Pacific Intermediate Water; PDW = Pacific Deep Water; SEqPIW = Southern Equatorial Pacific Intermediate Water.
Details are in the caption following the image
Hypothesized shifts in extent of NEqPIW and SEqPIW, such that the Galapagos sites are bathed by (a) NEqPIW during the LGM and deglaciation and (b) SEqPIW during the Holocene. Locations of our core sites (yellow stars). Locations of previously published intermediate water radiocarbon records (seafoam green stars; Marchitto et al., 2007; Stott et al., 2009; Lindsay et al., 2015, 2016; De Pol-Holz et al., 2010; Rose et al., 2010; Burke & Robinson, 2012; Siani et al., 2013; Davies-Walczak et al., 2014; Zhao & Keigwin, 2018). AAIW = Antarctic Intermediate Water; NEqPIW = Northern Equatorial Pacific Intermediate Water; NPIW = North Pacific Intermediate Water; SEqPIW = Southern Equatorial Pacific Intermediate Water.

During the LGM and early deglaciation (~16 to 30 ka), however, our Galapagos record is significantly more 14C depleted than the record from off Baja and, in fact, compares well to the even more anomalous Δ14C record from nearby Galapagos site VM21-30 (Lindsay et al., 2015, 2016; Marchitto et al., 2007; Stott et al., 2009). In fact, prior to 16 ka, intermediate waters near the Galapagos Platform at 600-m water depth as recorded at GGC 43, CDH 41, and VM21-30 were even more 14C depleted than PDW in the EEP (ODP site 1240, 2,921 m; TR163-23, 2,730 m; de la Fuente et al., 2015; Keigwin & Lehman, 2015; Umling & Thunell, 2017). This observation eliminates enhanced vertical mixing of PDW as a possible source of the low-14C carbon observed within the three Galapagos records. The only records that are more 14C depleted than those from the Galapagos come from the South Pacific between ~2,500- and 3,600-m water depth (PS75/100-4 and PS74/059-2) and are attributed, at least in part, to an influx of 14C-free hydrothermal CO2 (Ronge et al., 2016; Figure 6). Given the proximity of the Galapagos core sites to active hydrothermal systems, it therefore seems plausible that the extreme depletion in radiocarbon detected at the Galapagos sites is also a signature of geologic carbon (Stott & Timmermann, 2011; Zhao & Keigwin, 2018). Benthic radiocarbon values at site VM21-30 remain anomalously low over the entire ~25-kyr record, which suggests sustained active venting of 14C-free geologic carbon nearby to the core site. Our core locations (CDH 41 and GGC43) appear less affected by hydrothermal 14C after ~16 ka, instead tracking the still low, but comparatively enriched, 14C record from off the Baja Peninsula. Notably, this similarity is maintained when any of the three age models presented above are applied to the data (Figure 4). This shift suggests the Galapagos vent site was located closer to site VM21-30 than to our core locations and either less active after 16 ka or subsurface currents shifted the plume of 14C-depleted geologic carbon mostly away from our sites.

Synchronized local hydrothermal venting off Baja and on the Galapagos, leading to comparable variations Δ14C, in both magnitude and trends, at the two locations is difficult to imagine. Rates of hydrothermal release and proximity of the cores to the vent site would need to be serendipitously alike. Alternatively, Lindsay et al. (2015) suggest that the two sites are connected via the California Undercurrent, which transports 14C-depleted carbon from the equatorial subsurface to the upwelling waters off Baja today, an assertion supported by neodymium isotope data (Basak et al., 2010). However, transport of the 14C-depleted signature of Galapagos hydrothermal inputs to Baja seems suspect, given the diverging 14C histories at GGC 43/CDH41 and VM21-30 (Stott et al., 2009), which are only 5 km away. We therefore instead suggest that the Δ14C signatures recorded off Baja and, after 16 ka, at Galapagos core sites GGC 43 and CDH 41 reflect widespread changes in NEqPIW 14C content and a potentially sizable 14C-depleted carbon pool. We speculate that the aged carbon was sourced from the North Pacific via the transfer of 14C-depleted deep ocean carbon to the intermediate water layer and/or large-scale hydrothermal CO2 fluxes.

Benthic foraminiferal stable isotope and radiocarbon records from the northern Peru Margin and the Galapagos Platform indicate the influence of distinct water masses at the two locations during the LGM and deglaciation and two distinct sources of 14C-depleted carbon (Figure 7). We demonstrate that the Peru Margin records were dominantly influenced by Southern Ocean water masses, SAMW and AAIW, over the last 25 kyr. At ~600 m on the Galapagos Platform, the relative contribution of Southern Ocean water masses, which bathe the core site today, was reduced relative to North Pacific waters during the LGM and deglaciation. As such, the isotopic records presented here provide insight to both the carbon content and circulation of Southern Ocean intermediate waters (CDH 23 and CDH 26) and NPIWs (GGC43 and CDH41) and their possible distinct roles in the rise of atmospheric CO2 during the last deglaciation. We emphasize that the extreme 14C anomalies observed off Baja (Lindsay et al., 2015; 2016; Marchitto et al., 2007) and on the Galapagos Platform (this study; Stott et al., 2009) during the last glacial termination do not reflect the radiocarbon content of AAIW.

4.2 Relevance of Pacific Intermediate Water Radiocarbon Histories to the Deglacial Rise in Atmospheric CO2

Despite significant offsets in the absolute magnitude of the 14C anomalies detected in NEqPIW and SEqPIW in the EEP, trends in intermediate water ventilation (B-atm) within NEqPIW (GGC 43 and CDH 41, ~600 m) and SEqPIW (CDH 26, ~1,000 m) are similar (Figure 8e). The observed correspondence suggests ventilation of NEqPIW and SEqPIW, and their 14C-depleted carbon loads were controlled by a common mechanism; both records exhibit maximum offsets from the contemporaneous atmosphere at the LGM and abruptly decrease to near modern offsets at the end of the YD. A second, smaller ventilation event during HS1 is apparent in the Galapagos B-atm estimates and may also occur along the Peru Margin, as the resolution of the B-atm record from CDH 26 makes this time interval difficult to resolve.

Details are in the caption following the image
Assessing the role of the eastern equatorial Pacific (EEP) versus the Southern Ocean in atmospheric CO2 rise during the last deglaciation. Gray bars denote the YD and HS1. Dashed lines highlight the presence of a plateau in the HS1 rise in atmospheric CO2 rise. (a) Atmospheric CO2 from West Antarctic Ice Sheet (WAIS) divide ice core (Marcott et al., 2014); (b) boron isotope-inferred oceanic ΔpCO2 from EEP core ODP 1238 (age model adjusted from the original publication by applying reservoir age estimates from site CDH 26; Martinez-Boti et al., 2015); (c) sedimentary δ15N from core CDH 26; (d) surface reservoir ages from core CDH 26 (green) plotted on a separate axis, GGC 43 (purple), TR163-23 (blue; Umling & Thunell, 2017), and ODP 1240 (de la Fuente et al., 2015; solid lines: age model 3, dashed lines: age model 1); (e) NEqPIW (CDH 41: purple, GGC 43: pink) and SEqPIW (green) ventilation ages (B-atm); (f) boron isotope-inferred oceanic ΔpCO2 from site PS2498-1 in the Atlantic sector of the Southern Ocean (Martinez-Boti et al., 2015); and (g) Δ14C signature of PDW in southwest Pacific cores PS75/100-4 and PS75/059-2 (Ronge et al., 2016). EqPIW = Equatorial Pacific Intermediate Water.

In the EEP, the efficiency of exchange between intermediate and surface waters depends strongly on the background stratification state of the EEP as well as locally forced ocean-atmosphere interactions that dictate patterns of upwelling in the region (e.g., Fedorov & Philander, 2000). Stratification indices, based on proxy gradients between surface and intermediate water foraminiferal δ18O-, δ13C-, and δ18O-derived temperature estimates, from the same set of four cores presented here, indicate periods of reduced vertical stratification during the YD and HS1 (Bova et al., 2015). In the 2015 paper, these data were interpreted as being consistent with the canonical view of carbon release from the EEP during the two-part rise in atmospheric CO2; “a decrease in stratification across the upper water column led to enhanced vertical mixing, which provided a pathway for carbon-rich intermediate waters to invade surface waters and exchange with the atmosphere during the YD and HS1” (Bova et al., 2015). Additionally, periods of reduced stratification in the EEP were noted to correspond well to periods of enhanced CO2 outgassing as inferred from boron isotopic measurements at nearby site ODP 1238, recovered just ~300 km northwest of cores CDH 26 and CDH 23 (Figure 8b; Martinez-Boti et al., 2015). Boron isotope measurements on planktonic foraminifera are a direct tracer for oceanic CO2 degassing due to the dependence of boron speciation on pH and a significant isotopic fractionation between the two aqueous forms of boron (e.g., Foster & Rae, 2016). In the context of the present study and the newly observed trends in EEP intermediate water ventilation, the inferred periods of reduced stratification, enhanced vertical mixing, and heightened CO2 outgassing from EEP proxy gradients (Bova et al., 2015) and boron isotopes (Martinez-Boti et al., 2015) are generally consistent with our new radiocarbon data for improved intermediate water ventilation during the YD and HS1 (reduced B-atm), with periods of reduced stratification occurring during periods of decreasing B-atm, or improved ventilation. However, we suggest that the inferred stratification patterns from Bova et al. (2015) likely overemphasize the magnitude of the reduction in EEP stratification during HS1 relative to the YD. Our radiocarbon data from the same sites indicate improved, but still quite poorly ventilated, 14C-depleted waters at 600- and 1,000-m water depth in the EEP during HS1.

Productivity and nutrient utilization proxies provide additional insight on the exchange of nutrient-rich intermediate waters with the surface ocean. In the EEP, intervals of enhanced upwelling or entrainment of subsurface waters into the surface mixed layer are characterized by greater nutrient availability and greater primary production (e.g., Pennington et al., 2006). At site CDH 26, we measured the abundance of C37 alkenones in the sediment (C37total) as a proxy for haptophyte productivity, which is linked to rates of total phytoplankton productivity (e.g., Brassell, 1993). These data generally indicate higher phytoplankton productivity at the LGM and Holocene and a broad trough during the deglaciation. At the end of the LGM, productivity drops off sharply and is coupled to an increase in bulk sedimentary nitrogen isotopes, which suggests a sudden reduction in nutrient supply to surface waters.

Sedimentary nitrogen isotopes reflect the isotopic composition of organic matter deposited in the sediments, which may record the preformed δ15N signature of the source water (e.g., Francois et al., 1997), the degree of NO3 consumption (e.g., Altabet & Francois, 1994), and/or the extent of denitrification or oxygen availability in the subsurface (e.g., Altabet et al., 1995). Core site CDH 26 lies just north of the eastern South Pacific oxygen minimum zone, outside the region of strong subsurface denitrification (Fuenzalida et al., 2009), and, relative to sedimentary δ15N records from cores collected further south along the Peruvian and Chilean Margins within the heart of the oxygen minimum zone (OMZ) (De Pol-Holz et al., 2006, 2007; Higginson & Altabet, 2004b), the δ15N values in our records are lower and exhibit smaller amplitude variations over the last deglaciation. We therefore infer comparatively limited or even no strong impact of subsurface denitrification on our δ15N record. Although variations in the preformed δ15N signature of source waters may impact our δ15N record to a degree, trends in reconstructed productivity are anticorrelated with bulk sediment δ15N, with higher δ15N reflecting periods of reduced coccolithophorid production (Figure 5). This relationship suggests local NO3 consumption is the dominant influence on bulk sediment δ15N during the LGM and deglaciation. We therefore infer more upwelling and consequently greater nutrient supply to the EEP surface ocean, during periods of high productivity and low δ15N (Figure 8c).

The observation of increased upwelling during the LGM and Holocene and a broad pulse of less upwelling during the deglaciation is counter to the canonical view of upwelling-driven carbon release from the EEP during the deglaciation (e.g., Anderson et al., 2009; Martinez-Boti et al., 2015). This observation is corroborated by reservoir age estimates from this study (GGC 43 and CDH 26) as well as from two nearby locations in the EEP (TR163-23, Umling & Thunell, 2017; ODP 1240, de la Fuente et al., 2015; Figure 8d). All four reservoir age reconstructions exhibit broadly similar trends (Figures 8d and S3), consistent with a decrease in upwelling during HS1 and the Bølling-Allerød (BA)/ACR and an increase in upwelling across the YD. Reservoir ages are lower (higher) under reduced (strengthened) upwelling conditions when less (more) aged subsurface water reaches the surface ocean. A final line of evidence for reduced deglacial upwelling comes from the vertical temperature gradient between the sea surface (Globigerinoides ruber SSTMg/Ca) and the deep thermocline (Neogloboquadrina dutertrei subSSTMg/Ca) as calculated from foraminiferal Mg/Ca temperature estimates at site M772-059-1 recovered a few nautical miles from site CDH 26 along North Peru Margin (Nürnberg et al., 2015). These data indicate a generally deep, though highly variable thermocline from 17.3 to 13 ka, consistent with reduced upwelling during HS1 and the BA/ACR in the EEP.

We note, however, that within the deglacial trough in upwelling are two small millennial-scale peaks in the C37total, occurring ~16.5 to 14.8 ka and ~13 to 11.6 ka, that are coupled to decreasing δ15N signatures, with the more prominent of the two occurring during the YD (Figures 5, 8c). These intervals likely reflect short-lived returns to upwelling favorable conditions in the EEP and correspond well to the identified periods of improved intermediate water ventilation on the Galapagos Platform and the Peru Margin (B-atm), reduced upper ocean stratification (δ18O, δ13C, and temperature gradients; Bova et al., 2015), and CO2 outgassing (δ11B; Martinez-Boti et al., 2015). Thus, despite generally low upwelling conditions during the deglaciation, these relatively small increases in upwelling were apparently enough to ventilate the EEP subsurface and drive CO2 outgassing from the region. In fact, during the YD, the EEP may have been a significant source of CO2 to the atmosphere, possibly contributing to the second sustained deglacial rise in atmospheric CO2. Boron isotopes indicate peak CO2 outgassing rates from the EEP surface ocean during the YD (Martinez-Boti et al., 2015) coupled to a prolonged renewal in intermediate water ventilation and the release of 14C-depleted carbon at both the Galapagos and North Peru Margin locations. A similar, though more muted event, also occurred during the latter half of HS1. However, this event transpires during a plateau in atmospheric CO2 levels (Marcott et al., 2014), which suggests the region either did not release a significant amount of CO2 to the atmosphere or was compensated by CO2 uptake elsewhere (as was suggested to explain the early Holocene peak in EEP outgassing; Martinez-Boti et al., 2015).

The evidence presented here for renewed subsurface ventilation and CO2 outgassing during the YD is unique to the EEP and stands in contrast to Southern Ocean records (e.g., Burke & Robinson, 2012; Ronge et al., 2016; Skinner et al., 2010, 2015; Figures 8f and 8g), which demonstrate that Southern Ocean deep stratification disappeared by the start of the BA/ACR, ~14.6 ka. At the end of the last glacial period, just prior to the onset of HS1 and deglaciation, the strong vertical chemical gradients observed in the glacial Southern Ocean began to erode, signifying a reduction in ocean stratification and movement of aged carbon toward the surface ocean. Benthic foraminifer Δ14C activity along a depth transect from 835 to 4,339 m converges toward more positive values, indicative of improved ventilation (Ronge et al., 2016; Sikes, Cook et al., 2016). For example, at 2,498-m water depth near the base of the glacial reservoir, Δ14C dropped by 419‰ between 20 and 14.4 ka, approaching near modern values by the end of HS1 (Ronge et al., 2016; Figure 8g). Similar trends are also observed in deep ocean radiocarbon records from both the southeast Pacific (Burke & Robinson, 2012) and South Atlantic (Skinner et al., 2010) sectors of the Southern Ocean. At the sea surface, boron isotopes measured on planktonic foraminifera at site PS2498-1 in the Atlantic sector of the Southern Ocean corroborate the distinct lack of a YD carbon ventilation signal in the Southern Ocean; although the record only extends back to ~15.5 ka, surface ocean ΔpCO2 at the Southern Ocean site peaks at ~15 ka during HS1 and drops to near zero during the YD, indicating that although the site may have been a source of CO2 to the atmosphere during HS1, it was not during YD (Figure 8f; Martinez-Boti et al., 2015). We therefore suggest that the YD rise in atmospheric CO2 was sourced outside of the Southern Ocean and driven by renewed upwelling and vertical mixing in the EEP that ventilated 14C-depleted carbon from EqPIWs.

5 Northern Versus Southern Ocean Contributions to EEP CO2 Outgassing

Our radiocarbon records from the EEP spanning the last 25 kyr provide critical new insights to the glacial-interglacial CO2 conundrum and the interpretation of deglacial intermediate depth radiocarbon records. In the context of additional geochemical records, they demonstrate that the extreme 14C anomalies observed off Baja (Lindsay et al., 2015, 2016; Marchitto et al., 2007) and on the Galapagos Platform (this study; Stott et al., 2009) during the last glacial termination do not reflect the radiocarbon content of AAIW. Instead, we observe a detectable but far more modest imprint of an aged AAIW in the subsurface EEP at the end of the last glacial period and early deglaciation. Modestly depleted Δ14C records from AAIW in the southeast Pacific and within SEqPIW in the EEP closely track one another at the end of the last glacial period and early deglaciation, while benthic foraminifer δ13C values at 1,000-m water depth in the EEP (CDH 26) track variations in δ13C at intermediate depths in the SE Pacific (Siani et al., 2013). A portion of this Southern Ocean-derived low-14C carbon may have escaped to the atmosphere during the second half of HS1, but it appears SEqPIW remained relatively poorly ventilated until the YD, thus trapping most of the aged carbon in the subsurface until the YD ventilation event.

AAIW, however, was not the only conduit of aged carbon to the EEP during the last deglaciation. A second source of 14C-depleted carbon, as documented at cores CDH 41 and GGC 43, arrived in the EEP from the north via NPIW (extremely 14C depleted). Although the origin of this aged carbon, whether derived from geologic carbon or prolonged isolation of a North Pacific water mass, remains uncertain, the fingerprint of the North Pacific aged carbon source can be traced from the Galapagos Platform to the Baja Peninsula and perhaps as far north as the Gulf of Alaska. A record of benthic-planktonic age differences from intermediate depths in the Gulf of Alaska spanning the last 17 kyr is interpreted to reflect the transfer of aged carbon from the deep ocean to NPIW during the YD (Davies-Walczak et al., 2014). The authors point out, however, that the venting of the aged water mass did not occur near the core site due to the presence of a strong halocline and downwelling along the continental margin, making a downstream outgassing location, such as the EEP, a likely candidate. Although benthic-planktonic age differences from intermediate depths in the Gulf of Alaska and even deeper sites in the northeast Pacific (Galbraith et al., 2007; Lund et al., 2011) are not as large as those observed within NEqPIW records, we note that this metric does not take into account possible variations in surface reservoir ages, which could be large in these locations. Furthermore, although ultimately dismissed by the authors, age model variations for the Gulf of Alaska site based on an alignment to the Greenland ice core chronology would lead to significant B-atm offsets during the YD as well as HS1. Alternatively, an admixture of hydrothermal CO2 to the already 14C-depleted water mass could bring the NEqPIW Δ14C records in line with NPIW as documented in the Gulf of Alaska. Thus, considering the uncertainties, we hypothesize that the North Pacific played an important role in transferring 14C-depleted carbon form the deep ocean to NPIW during the last deglaciation, thereby contributing to the extreme 14C anomalies observed off Baja (Lindsay et al., 2015, 2016; Marchitto et al., 2007) and on the Galapagos Platform and ultimately contributing to the YD rise in atmospheric CO2 through outgassing in the EEP.

Acknowledgments

We thank Luke Skinner and two other anonymous reviewers for helpful comments that greatly improved this manuscript. Support for this work was provided by funds from NSF-OCE grant 1003387. Radiocarbon data are provided as a supporting information data set to this publication. All data sets (radiocarbon, benthic δ13C, benthic δ18O, and sedimentary δ15N) are archived online at the World Data Service for Paleoclimatology (https://www.ncdc.noaa.gov/data-access/paleoclimatology-data).