Hydrothermal carbon release to the ocean and atmosphere from the eastern equatorial Pacific during the last glacial termination

Arguably among the most globally impactful climate changes in Earth’s past million years are the glacial terminations that punctuated the Pleistocene epoch. With the acquisition and analysis of marine and continental records, including ice cores, it is now clear that the Earth’s climate was responding profoundly to changes in greenhouse gases that accompanied those glacial terminations. But the ultimate forcing responsible for the greenhouse gas variability remains elusive. The oceans must play a central role in any hypothesis that attempt to explain the systematic variations in pCO2 because the Ocean is a giant carbon capacitor, regulating carbon entering and leaving the atmosphere. For a long time, geological processes that regulate fluxes of carbon to and from the oceans were thought to operate too slowly to account for any of the systematic variations in atmospheric pCO2 that accompanied glacial cycles during the Pleistocene. Here we investigate the role that Earth’s hydrothermal systems had in affecting the flux of carbon to the ocean and ultimately, the atmosphere during the last glacial termination. We document late glacial and deglacial intervals of anomalously old 14C reservoir ages, large benthic-planktic foraminifera 14C age differences, and increased deposition of hydrothermal metals in marine sediments from the eastern equatorial Pacific (EEP) that indicate a significant release of hydrothermal fluids entered the ocean at the last glacial termination. The large 14C anomaly was accompanied by a ∼4-fold increase in Zn/Ca in both benthic and planktic foraminifera that reflects an increase in dissolved [Zn] throughout the water column. Foraminiferal B/Ca and Li/Ca results from these sites document deglacial declines in [ CO 3 2 − ] throughout the water column; these were accompanied by carbonate dissolution at water depths that today lie well above the calcite lysocline. Taken together, these results are strong evidence for an increased flux of hydrothermally-derived carbon through the EEP upwelling system at the last glacial termination that would have exchanged with the atmosphere and affected both Δ14C and pCO2. These data do not quantify the amount of carbon released to the atmosphere through the EEP upwelling system but indicate that geologic forcing must be incorporated into models that attempt to simulate the cyclic nature of glacial/interglacial climate variability. Importantly, these results underscore the need to put better constraints on the flux of carbon from geologic reservoirs that affect the global carbon budget.


Abstract
Arguably among the most globally impactful climate changes in Earth's past million years are the glacial terminations that punctuated the Pleistocene epoch. With the acquisition and analysis of marine and continental records, including ice cores, it is now clear that the Earth's climate was responding profoundly to changes in greenhouse gases that accompanied those glacial terminations. But the ultimate forcing responsible for the greenhouse gas variability remains elusive. The oceans must play a central role in any hypothesis that attempt to explain the systematic variations in pCO 2 because the Ocean is a giant carbon capacitor, regulating carbon entering and leaving the atmosphere. For a long time, geological processes that regulate fluxes of carbon to and from the oceans were thought to operate too slowly to account for any of the systematic variations in atmospheric pCO 2 that accompanied glacial cycles during the Pleistocene. Here we investigate the role that Earth's hydrothermal systems had in affecting the flux of carbon to the ocean and ultimately, the atmosphere during the last glacial termination. We document late glacial and deglacial intervals of anomalously old 14 C reservoir ages, large benthic-planktic foraminifera 14 C age differences, and increased deposition of hydrothermal metals in marine sediments from the eastern equatorial Pacific (EEP) that indicate a significant release of hydrothermal fluids entered the ocean at the last glacial termination. The large 14 C anomaly was accompanied by a ∼4-fold increase in Zn/Ca in both benthic and planktic foraminifera that reflects an increase in dissolved [Zn] throughout the water column. Foraminiferal B/Ca and Li/Ca results from these sites document deglacial declines in [ -CO 3 2 ] throughout the water column; these were accompanied by carbonate dissolution at water depths that today lie well above the calcite lysocline. Taken together, these results are strong evidence for an increased flux of hydrothermally-derived carbon through the EEP upwelling system at the last glacial termination that would have exchanged with the atmosphere and affected both Δ 14 C and pCO 2 . These data do not quantify the amount of carbon released to the atmosphere through the EEP upwelling system but indicate that geologic forcing must be incorporated into models that attempt to simulate the cyclic nature of glacial/interglacial climate variability. Importantly, these results underscore the need to put better constraints on the flux of carbon from geologic reservoirs that affect the global carbon budget.

Introduction-recent discoveries and motivation for this study
There have been several hypotheses put forth over the years to account for the systematic variations in atmospheric pCO 2 that accompanied glacial/interglacial greenhouse gas variations, the answer remains elusive. This likely reflects the fact that no single mechanism acting alone appears capable of explaining all aspects of carbon cycle behavior that accompanied the glacial cycles (Fischer et al 2010). For example, recent discoveries indicate marine carbon budgets underestimate the amount of carbon entering the oceans from hydrothermal systems. This is because over the past two decades reservoirs of liquid and hydrate CO 2 have been discovered in rocks and sediments that blanket the margins of some active marine hydrothermal vents (Sakai et al 1990, Inagaki et al 2006, Lupton et al 2006, Lupton et al 2008. In the back arc basin of the Okinawa Trough a 'lake' of liquid CO 2 accumulates beneath a cap of CO 2 -hydrate (Inagaki et al 2006, Nealson 2006) and at shallower ocean depths pure CO 2 gas is observed to emanate directly from sediments that blanket the margins of hydrothermal vents (figure 1). Pools of liquid CO 2 have been discovered in the Aegean Sea (Camilli et al 2015). Geologic carbon reservoirs such as these are not explicitly included in current marine carbon budgets. In fact, quantifying how carbon from sediment and rock-hosted reservoirs contributes to the overall marine carbon budget is not yet possible because there are too few observations and virtually no in situ measurements of the CO 2 flux from these types of reservoirs (Burton et al 2013). Yet, in one remarkable study, Lupton et al (2008) were able to quantify the flux of liquid CO 2 from a sediment/rock-hosted reservoir at a site on Mariana trench in the western Pacific by observing the rate that bubbles of pure liquid CO 2 emanate from a small opening on the seafloor. They estimated the carbon flux from this single opening to be 0.1% of the entire Mid Ocean Ridge flux. When Lupton et al (2008) documented the gas and liquid phases in the western Pacific, they pointed out that of the hundreds of vents thus far investigated, only a few were found to have a separate CO 2 phase, concluding this must be a rare circumstance and perhaps limited to volcanic arcs. However, investigations targeting CO 2 fluxes from the cool flanks of active hydrothermal sites has not been a priority and observations are therefore sparse (Burton et al 2013). With only about one-third of the global ocean spreading ridges surveyed, vast portions of the sea floor have not been investigated at all (Beaulieu et al 2015). Even if only a small percentage of the unsampled hydrothermal systems contain separate gas or liquid CO 2 phases it could change the global marine carbon budget substantially (Burton et al 2013). Importantly, where these reservoirs have been found at shallowintermediate depths, the flux from the sediment and rock reservoirs is modulated by a density-stratified boundary layer at the sediment/water interface where a thin layer of CO 2 saturated-seawater and CO 2 hydrate forms (Inagaki et al 2006, Lupton et al 2006. The boundary layer is sensitive to temperature changes. Reservoirs of carbon like this can exist at deep water depths but may be less sensitive to temperature changes. Stott and Timmermann (2011) were motivated by the observations described above to consider whether the large deglacial radiocarbon Δ 14 C excursions documented from biogenic carbonates in the eastern Paci-  2009, Bryan et al 2010) could be indicative of past episodes of geologic carbon release from hydrothermal sources. The magnitude of the deglacial Δ 14 C excursions in the eastern equatorial Pacific (EEP) core VM21-30 are too large to be explained by ventilation of a previously isolated deep-water mass. Furthermore, similarly large Δ 14 C anomalies are not found at other intermediate depth sites bathed by the same water mass (De Pol-Holz et al 2010, Ronge et al 2016). Given this, Stott and Timmermann (2011) reasoned that the magnitude and duration of the excursion in the EEP would have required a sustained input of 14 C-dead carbon from a localized source, the most likely being a hydrothermal source in the EEP. There are several volcanic provinces located within the EEP that could have potentially contributed carbon-rich hydrothermal fluids, including the Galapagos Hot Spot (Christopher et al 2013, Harpp andWhite 2018), the Galapagos Spreading Center, the Cocos Ridge, the Carnegie Ridge and perhaps the submerged margins of the Central American Arc. If a substantial release of geologic carbon to the upper ocean occurred in the EEP from one or more of these volcanic sources, it has important implications for reconciling why there was a large decrease in atmospheric Δ 14 C during the glacial termination when other cosmogenic radionuclides did not (Muscheler et al 2004, Petrenko et al 2016 because the EEP is a primary conduit for exchange of carbon from the ocean to the atmosphere (Takahashi et al 2009).
The original Stott and Timmermann (2011) hypothesis met with considerable skepticism because few paleoceanographers were aware of the recent discoveries of geologic reservoirs of CO 2 in the ocean or that hydrothermalism may have varied on glacial/ interglacial timescales. But over the past several years other studies have begun to consider ways that geologic carbon could have influenced glacial/interglacial CO 2 and climate variability (Lund and Asimow 2011, Broecker et al 2015, Tolstoy 2015, Lund et al 2016, Ronge et al 2016. The seafloor discoveries and the hypotheses summarized above pose challenges to the paleoceanography community to effectively evaluate whether geologic processes contributed significantly to the systematic variations in atmospheric pCO 2 that accompanied each glacial/interglacial cycle over the past million years. In the present study we evaluated further the suggestion by Stott and Timmermann that the large Δ 14 C anomalies in the EEP were due to a release of hydrothermal carbon during the last glacial termination. We emphasize that the current study does not attempt to identify the specific site of hydrothermal carbon release. There are multiple sites in the EEP where hydrothermal and volcanic activity is known. And while there have been various studies of active hydrothermalism in the EEP, there is a significant amount of area in the EEP that has not been explored. The Galapagos island chain itself is an active volcanic center as is the EPR. The flanks of the Central American Arc are also known to emit large amounts of CO 2 through the sediments that blanket the margins of the arc (Salazar et al 2001). Hence, there are large gaps in knowledge about where CO 2 is emitted in the EEP. We also emphasize that this study is not an attempt to identify a specific phase of CO 2 released to the ocean at the glacial termination. That topic is beyond the scope of this effort. The primary intent here is to further evaluate the suggestion by Stott and Timmermann (2011) that hydrothermal carbon was released to the ocean in the EEP during at the last glacial termination and hydrothermal carbon contributed to the very large Δ 14 C anomalies observed in marine sediments. The Stott and Timmermann hypothesis made several predictions about carbonate chemistry and carbonate preservation in the EEP during the last glacial termination. An evaluation of those predictions here is one step in an effort to evaluate if and how geologic sources of carbon contributed significantly to the glacial/ interglacial carbon cycle variability.

Methods
A complete description of the analytical methods used in this study is provided in the supplementary section. The Stott and Timmermann hypothesis predicts that if there was a release of hydrothermal carbon of sufficient quantity to produce the large 14 C age anomalies in the EEP it would have changed the carbonate chemistry and affected calcite and aragonite preservation. It may have also left a trace metal fingerprint of hydrothermal elements on the sediments. We demonstrate that large benthic-planktic 14 C age differences originally documented in the VM21-30 core by Stott et al (2009) are observed in other cores from the EEP (figure 2; table 1). We also demonstrate that the surface ocean reservoir age varied significantly across the EEP during the glacial termination with the largest increase in surface ocean reservoir age of more than 6000 years, just south of the equator within the zone of upwelling.
Variability in the ocean carbonate system and carbonate preservation has been evaluated using scanning electron microscopy to inspect carbonates for signs of dissolution. The weight and abundance of Orbulina universa, a planktic foraminifera, is quantified because changes in the shell weight of individual O. universa has been shown to correlate positively with carbonate ion concentration [ We quantified changes in accumulation of hydrothermal elements in oxide overgrowths precipitated on the carbonate sediments in the EEP by analyzing uncleaned foraminifera from the late glacial and deglacial sections of two cores. This is not an effort to   figure 3(A)) and a negative δ 13 C excursion centered on the deglaciation ( figure 3(B)). This δ 13 C excursion was documented previously in planktic foraminifera (Spero andLea 2002, Bostock et al 2004) and in atmospheric CO 2 extracted from Antarctic ice cores where it has a well-established chronology (Schmitt et al 2012) (figure S1 is available online at stacks.iop.org/ERL/14/025007/mmedia). In the ice core record the deglacial δ 13 C excursion began at 17.2 kyBP ±∼500 years and the first δ 13 C minima was at 15.8 kyBP ±∼500 years (Schmitt et al 2012). The magnitude of this initial δ 13 C decrease in the EEP foraminiferal δ 13 C records is ∼−0.4‰, as it is in the ice core record. There was a second δ 13 C minimum at 12.3ky BP in the ice core record but this minima is not distinguishable in our cores, even though it is wellresolved in the higher resolution records from the Isotope Datum 1 is identified in each core with the arrow. Note the close similarity of the δ 18 O stratigraphies with depth between VM21-30 and ODP1242A, suggesting similar sedimentation rates between the two sites. (b) Benthic and planktic δ 13 C ‰ with isotope Datums 2 and 3 indicated by arrows. The gap in Cibicidoides mundulus δ 13 C at ODP1242A from 151 cm to 181 cm is due to absence of this species in those intervals.
western equatorial Pacific (Stott et al 2004(Stott et al , 2009). Here we use the mid-point of the δ 13 C excursion as a chronologic datum, which has a mean calendar age of 13.5 ky BP (figure S1).
The δ 13 C excursions (Datums 2 and 3) are also observed in the EEP benthic stratigraphies ( figure 3(B)). The early-deglacial δ 13 C excursion in the Cibicidoides record from ODP1242A stands out distinctly against the long-term δ 13 C increase that characterizes the later deglacial section ( figure 3(B)). In the Uvigerina record from VM21-30 there is longer-term decrease in δ 13 C that extends through the entire deglacial interval. We know of no other intermediate depth cores in the Pacific that record a comparable Cibicidoides δ 13 C excursion. It appears this benthic δ 13 C excursion is a local signal.
The depth and age of each of the isotope datums are listed in table 2. Large reservoir age corrections are required to bring 14 C-based calendar ages (particularly in VM21-30) into agreement with the stable isotope datums. For example, the onset in benthic δ 18 O decrease marking the beginning of deglaciation (Datum 1) occurs at the same depth in both ODP1242A and VM21-30 ( figure 3(A)). The beginning of the deglacial δ 13 C excursion (Datum 2) also occurs at the same depth in both cores ( figure 3(B)). In fact, the late glacial and deglacial stable isotope stratigraphies of both cores are very similar and this implies they share a similar sediment accumulation rate. This is important because the deglacial 14 C stratigraphies of these two cores are very different. In figure 4 the calendar age for each of the three datums is plotted versus its depth in ODP 1242A and VM21-30. The planktic radiocarbon ages shown in figure 5 were converted to calendar age using Calib 7.1 (Reimer et al 2013) with ΔR of 425±200 years. Increasing ΔR from 0 to 425 years brings the 14 C calendar ages in the ODP1242A record into agreement with the datum ages. This implies that reservoir ages in the EEP near ODP 1242A were at least 400 years older during the deglaciation than during the late Holocene (Benway et al 2006). However, even more dramatic changes in reservoir age are evident in the VM21-30 record (figure 4) where the offset between the stable isotope datums and the 14 C calendar ages is as much as 6000 years. For example, the initial benthic δ 18 O decrease (Datum 1) in VM21-30 occurs at 250 cm. The 14 C-based calendar age for this horizon (with ΔR=425 years) is 22 kyBP, nearly 5000 years older than the age of this same δ 18 O horizon in the ODP1242A core and in other benthic foraminiferal records from the Pacific (Stern and Lisiecki 2014). The oldest surface reservoir ages at the VM21-30 site was during the latest glacial and early deglacial. The accumulation rate of VM21-29 is much lower than the other two core and hence, it is more difficult to place the exact horizon of the isotope datums in this core.
3.2. Benthic-planktic 14 C age differences Stott et al (2009) documented large benthic-planktic 14 C age differences (B-P ages) in the glacial and deglacial section of VM21-30. The B-P ages are >4000 years in most intervals of the core between 100 and 300 cm, spanning late glacial and deglacial sections (figure 5). Anomalously large B-P 14 C ages are also observed in ODP 1242A and VM21-29 (figure 5). The B-P 14 C ages are highest in the late glacial section of ODP 1242A core and began to increase at about the same time as at the VM21-30 site. However, at the ODP 1242A site, the large B-P ages did not persist through the deglaciation.
In VM21-29 the B-P 14 C ages are also anomalously large between ∼200 and 75 cm, which corresponds to the late glacial and deglacial section (figure 5). The B-P 14 C ages from these intervals are between 2000 and 5000 years whereas the Holocene B-P 14 C age differences are similar to modern. Putting the data from all three cores together suggests that there were large increases in benthic-planktic age difference across the EEP during the latest glacial and deglacial, but the largest changes occurred at the equator, near VM21-30 and VM21-29, during the deglaciation.

Carbonate preservation
A visual inspection of the core samples documents variable changes in the preservation and abundance of benthic and planktic foraminifera in the late glacial and deglacial sections of all three cores (table S7). Aragonitic pteropods are present throughout late glacial and Holocene, although their abundance and preservation varies between samples. At 254.8 cm in VM21-30 for example, the abundance of aragonite is very low but pteropods are intact and whole. Yet, in this same sample some of the calcitic benthic and planktic foraminifera are partially dissolved (figure 6). At the 268.5 cm horizon pteropods are very abundant and benthic shells are well-preserved but the B-P 14 C ages are approximately 4000 years.
The VM21-29 core also exhibits variable carbonate preservation and abundances in the late glacial and deglacial sections (table S7). The aragonite and carbonate are abundant and well-preserved between 170 and 150 cm, which corresponds to the latest glacial. Directly above this interval, at the onset of the deglacial, aragonite abundance drops significantly and some of the benthic foraminifera are partially dissolved. Above 100 cm (midway through the deglacial section) carbonate preservation is good and aragonite   is abundant. From these observations, it appears that only the early deglacial section of VM21-29 exhibits lower carbonate preservation whereas in the VM21-30 core the variable carbonate preservation and abundances persists from the late glacial through the entire deglacial section. Visual comparison of ODP1242A samples with those from VM21-30 and VM21-29 is difficult because the ODP samples contain a much higher clay content and the samples are indurated. However, several important observations stand out. In this core the calcite foraminifera are abundant, intact and well-preserved below 441 cm, which corresponds to the late glacial. Foraminiferal abundances drop, and preservation declines above 323 cm. This is associated with increased numbers of fragmented planktic foraminifera. There is a marked decline in foraminiferal abundance and the presence of partially dissolved benthic foraminifera at 281 cm. The lowest foraminiferal abundances and dissolved benthic foraminifera are observed between 281 and 161 cm, which is the interval with largest B-P 14 C age offsets.
To summarize, intervals of poor preservation and lower carbonate abundance occur in the latest glacial and early deglacial section of each core, and in VM21-30 extend through the deglacial section. The most pronounced decline in preservation occurs during the early deglacial section of all three cores.

Orbulina universa shell weights and abundance as a proxy for carbonate ion variability
The abundance of O. universa drops dramatically across the glacial termination ( figure 7). A similarly striking change in abundance is also observed in the VM21-29 core (figure 7). In the late glacial samples of VM21-29 the abundance of O. universa is ∼600 specimens per gram and then drops to between 100 and 200 specimens per gram in the deglacial. In both VM21-30 and VM21-29, the abundance of O. universa returns to higher numbers in the late deglacial and Holocene sections.
The reduced abundance in O. universa in the latest glacial and deglacial samples is accompanied by a dramatic drop in the average shell weight of individual O. universa specimens (figure 7). In VM21-30 the weights drop by approximately 50% in the latest glacial and deglacial samples. The lowest shell weights are also observed in the deglacial samples of the VM21-29 core (figure 7). In the VM21-30 core, the late deglacial and Holocene shell weights do not return to the higher values seen in the glacial samples, whereas in the VM21-29 core the late deglacial weights increase to more than double the weight of the early to mid-deglacial specimens (figure 7).
In  (Lovenduski et al 2015). Interannual variations in [CO 3 2 ] accompanies changes in upwelling associated with ENSO, but those changes are small in comparison to the ∼50% change implied by the O. universa shell weights. We also acknowledge that the magnitude of carbonate ion change within the water column implied by the abundance drop and the shell weight drop seen in the VM21-30 and VM21-29 records may be exaggerated due to dissolution on the seafloor, which would bias an explicit estimate of carbonate ion change. No calibration yet exists for N. dutertrei, but the range of B/Ca values in this record agree well with those reported for this species by Foster (2008). These values are consistently lower than those for N. incompta but N. dutertrei is known to form a gametogenic crust deeper in the water column and is commonly thought have a deeper habitat than N. incompta. Therefore, the N. dutertrei results may reflect lower [ -CO 3 2 ] conditions at deeper water depths. Here too we acknowledge that release of excess carbon to the water column was likely discontinuous (see discussion on dissolution estimates) and because of this, it is also likely that the B/Ca results could be biased due to the dissolution of or reduced abundance of specimens that lived at times of highest carbon flux. Hence, we may not record a precise estimate of carbonate ion decline at times of highest carbon release. This would account for the fact that the carbonate ion change implied by the O. universa data appears much larger than is implied by the B/Ca data.
The Li/Ca values of G. ruber and N. dutertrei decrease by about 25% across the entire deglacial section of the VM21-30 core (figure 8). We rule out a temperature effect in this ∼25% decrease in Li/Ca because if this decline were caused by rising temperature, it would require a surface and thermocline warming of ∼6°C (Hall and Chan 2004), whereas alkenone and Mg/Ca records indicate sea surface temperatures (SSTs) warmed during the glacial termination by no more than ∼1°C-2°C in the EEP (Koutavas et al 2002, Koutavas andLynch-Stieglitz 2003, Koutavas and Sachs 2008) ( figure S2). In addition, much of

Zn/Ca from VM21-30
The planktic and benthic records of Zn/Ca exhibit very different deglacial trends relative to Li/Ca or B/ Ca. All species exhibit a broad maximum in Zn/Ca during the deglacial section of VM21-30 (figure 8). The Zn/Ca values among the foraminifera are also quite similar ( figure 9), although the G. ruber Zn/Ca are ∼5-10 μmol mol −1 higher throughout most of the record. The relatively high values of Zn/Ca in G. ruber during the deglacial section does not appear to reflect contamination, or residual oxyhydroxide coatings on the foraminifera because other elemental data used for monitoring contamination remain low (for Al/Ca, Mn/Ca, and Fe/Ca, r 2 values are <0.29, <0.29, and <0.24, respectively, for all species).
Because the incorporation of Zn into foraminiferal calcite depends on both seawater [Zn] and on  the B/Ca records leads us to conclude that the deglacial increase in Zn/Ca represents changes in seawater [Zn] rather than [ -CO 3 2 ]. The increased Zn/Ca corresponds with the large decrease in Δ 14 C implying that both share a common source.

Hydrothermal metal precipitates
Hydrothermal plume particles from vent fluids are highly enriched in metals. These particles are scavenged from and precipitated out of seawater near the vents. Therefore, variations in metal precipitates in sediments and foraminiferal coatings should reflect hydrothermal activity nearby. In contrast to measuring metal precipitate fluxes directly from sediments, measuring the elemental composition of foraminifera and their coatings has the advantage of circumventing complicating issues such as dilution by non-hydrothermal inputs and sediment focusing. And because we do not have independent measurements of a constant flux proxy such as 230 Th (Lund et al 2016, Costa et al 2017 we have instead normalized the elemental values to the foraminiferal calcium. In this way we document relative changes in abundance of hydrothermal metals rather than actual flux estimates. In these data we find that Fe/Ca and Zn/Ca values of the uncleaned foraminifera are ∼5-20x higher than that of cleaned foraminifera from the same core ( figure 9, table S6).
In the ODP1242A record ( figure 9(A)), the highest El/Ca values occur in the glacial section; this is most pronounced for Fe/Ca and V/Ca. The higher Fe/Ca and V/Ca values also coincide with the largest B-P ages in this core. There is a strong correlation between Fe/Ca and V/Ca (figure 10), which agrees with modern studies that find scavenging of V by colloidal ferrihydrite in equal measure (Feely et al 1994, Cave et al 2002, German et al 2002, Dunk and Mills 2006. The Cu/Ca and Co/Ca values are more variable than Fe and V but the highest Cu/Ca values also occur late glacial section of ODP 1242A ( figure 9(A)). The Zn/Ca values in ODP1242A glacial samples do not appear significantly higher in the late glacial section in association with the anomalous radiocarbon ages as is the case in the VM21-30 core. The only explanation we can suggest is that the Zn content of hydrothermal fluids near the ODP sites were different than those near the VM-21-30 sites.
In the VM21-30 core the highest El/Ca values are centered in the deglacial section ( figure 9(B)), and deglacial maxima occur in both Zn/Ca (6X higher) and in Cu/Ca (3X higher) compared to values in the glacial section ( figure 9(A)). These maxima in VM21-30 El/Ca also coincide with the largest B-P ages, as they do in the ODP1242A core. The Cu/Ca and Co/Ca values more variable with less pronounced enrichment in the LGM and early deglacial section of VM21-30. The Cu/Ca values in VM21-30 are one to two orders of magnitude lower than at ODP1242A and Co/Ca values are largely invariant. Still, some coherence between Cu/Ca with Fe/Ca and V/Ca can be explained by the co-precipitation of these elements with iron oxyhydroxides (Dunk and Mills 2006). And it is clear from a comparison between the ODP1242A and VM21-30 results that the flux of these hydrothermally derived metals varied across the EEP, with higher amounts of metal accumulation in the late glacial near the ODP1242A site and later lower amounts during the deglacial near the VM21-30 site.

Discussion
In the late 1970s the plate tectonics theory gained strong support from oceanographic data collected by the Deep-Sea Drilling Program and other initiatives that acquired sea floor observations, including magnetic reversals and biostratigraphic data that documented increasing ages of the sediments overlying basaltic crust away from the mid-Atlantic mountain chain. The discovery of 'thermal springs' along the Galapagos Rift during the late 1970s (Corliss et al 1979) still stands as one of the most transformative discoveries of that era. That initial discovery at the Galapagos Rift did not quantify the actual flux of CO 2 emanating from the 'thermal springs'. But the measurements of alkalinity and pH indicated those springs carried a significant amount of CO 2 (Corliss et al 1979). It has taken many decades since that original discovery to obtain what is still a very sparse database of observations from the boundaries of active tectonic plates where much of the geologic activity and hydrothermalism occurs. That sparse observational database leaves considerable uncertainty about how vents contribute to the overall carbon budget of the oceans. It is in this long-term context of exploration that the discovery of liquid and solid CO 2 reservoirs stands-out as an important advance. These discoveries, together with geologic observations of variable hydrothermalism during the Pleistocene glacial cycles make it clear that processes that regulate the flux of carbon from the Earth's interior to its exterior are not spatially uniform nor are they necessarily slow and continuous. Thus, there remains much to learn about what regulates the flux and accumulation of hydrothermal carbon. The question of whether the accumulation and variable flux of geologic carbon to the ocean from these sources influenced the global carbon cycle on glacial/interglacial time scales remains an incipient hypothesis to be tested. Towards that goal, we evaluated whether there was increased flux of hydrothermal carbon to the ocean during the last glacial termination. The findings presented here are compelling evidence that there was increased flux of hydrothermally-derived carbon to the ocean in the EEP at the last glacial termination. This has important implications for the history of atmospheric pCO 2 and Δ 14 C because the EEP is a primary conduit for exchange of carbon between the ocean to the atmosphere.
The results do not put constraint on the amount of hydrothermal carbon released to the surface ocean in the EEP during the last glacial termination but the magnitudes of change in radiocarbon, δ 13 C and trace elements implies a large amount of carbon was released over thousands of years that would have contributed to the higher ΔpCO 2 documented in boron isotope records (Martinez-Boti et al 2015). The dissolution of carbonates at 600 meters in the EEP alone is strong evidence that there was large flux of carbon into the upper ocean that altered the carbonate chemistry. The O. universa shell weight data also imply a drop of surface ocean [CO 3 2 ] that may have been as much as 50%. The planktic foraminiferal B/Ca and Li/Ca data presented here also indicate a large increased flux of carbon into the surface ocean that lowered the carbonate saturation. Together with boron isotope results, our findings indicate geologic carbon would have been a significant fraction of the carbon released from the ocean to the atmosphere at the last glacial termination.
In the modern ocean the VM21-29 and VM21-30 sites are above the calcite lysocline and close to the aragonite lysocline in the EEP. The variable dissolution evident in benthic foraminifera implies that the carbonate saturation in the EEP intermediate waters dropped sometimes during the late glacial and deglaciation. This implies episodic upwelling of 14 C-depleted, CO 2 -rich waters in the EEP during the late glacial and early deglacial. This is consistent with large range of 14 C ages and very old reservoir ages among planktic foraminifera within the late glacial and deglacial samples.
The large B-P 14 C age anomalies and increased surface ocean reservoir ages together with the increased accumulation of hydrothermal metals is evidence that the source of excess carbon that affected the carbonate changes was from a hydrothermal source and not from respired metabolic carbon. The large enrichment of Zn in both the foraminiferal calcite and in oxide coatings, along with the enrichment and covariation of Fe and V indicates there was a hydrothermal source nearby. The oxide coatings would have formed after deposition at the sea floor. However, the elevated Zn/Ca in the cleaned planktic and benthic foraminifera of VM21-30 in the deglacial section indicates there was a large enrichment of dissolved Zn throughout the water column. The most likely source of this Zn would be from hydrothermal systems. However, it is not possible to infer from what hydrothermal system the Zn came. The fact that Zn is elevated throughout the water column over the deglacial implies there was a persistent supply. And this input of Zn coincided with the input of 'old' carbon. At the present time there is no known occurrence of liquid CO 2 at the hydrothermal vents in the EEP or on the margins of the Galapagos that might act as a supply of these metals during the deglacial. But it is worth noting that supercritical CO 2 (SFC-CO 2 ) is a highly efficient solvent for extraction of heavy metals (Lin et al 2014). Therefore, if there was SCF-CO 2 formed at depth beneath the hydrothermal or volcanic systems in the EEP, it could act as an efficient source for Zn and other metals. This is obviously a question that will require additional research. But the elevated Zn/Ca and 'very old' carbon recorded by carbonates in the EEP points to a geologic source of carbon.
The results presented here point to a large amount of hydrothermal carbon released at times during the last glacial termination and that carbon upwelled to the surface, but the changes in carbonate chemistry were likely also responding to changes in the upwelling intensity (Koutavas et al 2002, Koutavas and Lynch-Stieglitz 2003, Koutavas and Sachs 2008 that indicates stronger upwelling. The stronger upwelling documented in previous studies and the geochemical anomalies documented here coincided with elevated ΔpCO 2 in the EEP (Martinez-Boti et al 2015). It therefore appears that there were two factors affecting the flux of carbon to the surface ocean and atmosphere from the EEP during the glacial termination: episodic release of geologic carbon to the intermediate waters, and intensified upwelling. Because the EEP is one of the primary conduits for exchange of carbon from the ocean to the atmosphere our findings require reconsideration of the prevailing view about where the ocean released excess carbon to the atmosphere at the glacial termination. The results also underscore the need to investigate other sites for evidence of geologic carbon release.
The present study does not put a constraint on the mechanism responsible for the increased flux of hydrothermal carbon at the glacial termination. The two prevailing ideas, sea level forcing Asimow 2011, Huybers and and temperature induced changes in hydrothermal carbon storage (Stott and Timmermann 2011) require additional evaluation. In that effort it will be important to conduct benthic carbon flux measurements on the flanks of active hydrothermal systems to quantify how these carbon sources contribute to the ocean's carbon budget. It will also be important to investigate trace metal concentrations at sites where there is known sources of SFC-CO 2 such as in the western tropical Pacific (Lupton et al 2008). Earth System Models may also offer opportunities to evaluate how point sources of carbon to the atmosphere such as the upwelling system in the EEP influence the global carbon cycle, including the carbonate system. Such experiments could then be compared with the growing database of observations that include proxies for carbonate ion and pH.

Conclusions
In this study we present evidence of increased carbon flux to the upper ocean within the upwelling region of the EEP at the glacial termination. The geochemical measurements support the suggestion that hydrothermal carbon contributed significantly to that increased carbon flux. When combined with results of previous research, it also appears that the increased flux of geologic carbon to the EEP was accompanied by enhanced upwelling strength in one of the ocean's primary conduits for exchange of carbon with the atmosphere. We conclude that the prevailing view that glacial/interglacial pCO 2 variability that characterized the glacial/interglacial cycles was controlled by the storage and subsequent ventilation of respired marine metabolic carbon must be modified to include geological processes that affect the flux of carbon from hydrothermal sources. We further conclude that there is vital importance in learning what regulates the flux of hydrothermal carbon to the ocean. This will require research that explores where and how geologic carbon is stored in the ocean basins and measurements of the actual fluxes of carbon from these sources that can better constrain the marine carbon budget. This research will not only advance knowledge about one of the major scientific challenges in Climate Science but will also provide important constraints on how the marine carbon cycle may be affected by warmer global temperatures in Earth's future.