Ice Age‐Holocene Similarity of Foraminifera‐Bound Nitrogen Isotope Ratios in the Eastern Equatorial Pacific

Bulk sediment δ15N records from the eastern tropical Pacific (ETP) extending back to the last ice age most often show low glacial δ15N, then a deglacial δ15N maximum, followed by a gradual decline to a late Holocene δ15N that is typically higher than that of the Last Glacial Maximum (LGM). The lower δ15N of the LGM has been interpreted to reflect an ice age reduction in water column denitrification. We report foraminifera shell‐bound nitrogen isotope (FB‐δ15N) measurements for the two species Neogloboquadrina dutertrei and Neogloboquadrina incompta over the last 35 ka in two sediment cores from the eastern equatorial Pacific (EEP), both of which have the typical LGM‐to‐Holocene increase in bulk sediment δ15N. FB‐δ15N contrasts with bulk sediment δ15N by not indicating a lower δ15N during the LGM. Instead, the FB‐δ15N records are dominated by a deglacial δ15N maximum, with comparable LGM and Holocene values. The lower LGM δ15N of the bulk sediment records may be an artifact, possibly related to greater exogenous N inputs and/or weaker sedimentary diagenesis during the LGM. The new data raise the possibility that the previously inferred glacial reduction in ETP water column denitrification was incorrect. A review of reconstructed ice age conditions and geochemical box model output provides mechanistic support for this possibility. However, equatorial ocean circulation and nitrate‐rich surface water overlying both core sites allow for other possible interpretations, calling for replication at non‐equatorial ETP sites.

one in the ETNP and one in the ETSP. Within the ODZs, denitrification reduces nitrate to N 2 , providing the oxidant for the respiration of organic matter. This water column denitrification is a major driver of fixed N loss from the ocean (Deutsch et al., 2001), and it can also produce nitrous oxide, a potent greenhouse gas, as a side-product (Pierotti & Rasmussen, 1980). Moreover, reconstruction of the history of denitrification may provide insight into past changes in ocean circulation and the strength of the global ocean's biological carbon pump (Altabet et al., 1995(Altabet et al., , 2002Ganeshram et al., 1995;Martinez et al., 2006;Robinson et al., 2007).
During denitrification, 14 N-nitrate is preferentially lost from the ocean water column, elevating the residual nitrate in δ 15 N (Cline & Kaplan, 1975;Liu & Kaplan, 1989) (with δ 15 N = [( 15 N/ 14 N) sample /( 15 N/ 14 N) atm N2 − 1]*1000‰). This 15 N-enriched nitrate is then transported laterally in the subsurface, spreading this isotopic signal through the mid-depth ocean. The high-δ 15 N nitrate is also supplied to the surface by upwelling, resulting in the production of high-δ 15 N organic matter. This organic N is then exported to depth, possibly after significant lateral transport in the surface. Regeneration back to nitrate in the subsurface represents an additional mechanism for spreading the high-δ 15 N signal of denitrification both laterally and vertically ).
A second N isotope signal that pertains to the EEP (as well as to the central equatorial Pacific) is that of partial nitrate consumption in surface waters (Altabet, 2001;Altabet and François, 1994). The preferential consumption of 14 N nitrate by phytoplankton produces low-δ 15 N organic matter above the upwellings and leaves the residual nitrate enriched in 15 N. The high-δ 15 N nitrate is then advected away from the upwelling, with continued consumption leading to the production of high-δ 15 N organic matter (Altabet & François, 1994). This pattern is reflected in the δ 15 N of bulk organic matter and foraminifera shell-bound organic matter on the seabed (Costa et al., 2016;Farrell et al., 1995). Sediment records of the N isotopes in the EEP are thus expected to record changes in both water column denitrification and the spatial pattern of surface nitrate consumption.
A third potential influence is the δ 15 N of shallow subsurface nitrate imported into the EEP by the equatorial system. Nitrate δ 15 N data are lacking in the EEP east of the Galapagos Islands. West of Galapagos, the δ 15 N of nitrate in the EUC proper is ∼7.1‰, similar to that of Subantarctic Mode Water, which is the ultimate nitrate source to much of the tropical Pacific thermocline . Strengthening in the flow of the EUC has the potential to lower the δ 15 N of the nitrate supply to the EEP relative to that supplied from the denitrification zones. However, the caveat must be added that the EUC flow is interrupted by the Galapagos, and thus its nitrate isotopic signature may not persist east of the islands. This must be clarified by future nitrate isotope studies.
A water mass known as the subsurface equatorial "thermostad" just below the EUC as well as the eastward-flowing Southern Subsurface Counter Current (SSCC) have a nitrate δ 15 N that can reach below 6‰ Rafter & Sigman, 2016). This lower nitrate δ 15 N results from the remineralization of lowδ 15 N sinking N associated with the upwelling and incomplete consumption of nitrate that upwells along the equator and its southern margin. Thus, rather than being a distinct signal of nitrate source from the west, the lower nitrate δ 15 N of the thermostad and SSCC is an additional aspect of the signal of incomplete nitrate consumption in the equatorial Pacific. Accordingly, its likely effect is to amplify the δ 15 N lowering of sinking N due to incomplete nitrate consumption in the region.
Existing bulk sediment δ 15 N records from the ETNP, ETSP, and EEP typically show low glacial δ 15 N that is followed by a deglacial δ 15 N rise (De Pol-Holz et al., 2006;Emmer & Thunell, 2000;Farrell et al., 1995;Ganeshram et al, 1995Ganeshram et al, , 2000Hendy et al., 2004;Hendy & Pedersen, 2006;Higginson & Altabet, 2004;S. S. Kienast et al., 2002;Pichevin et al., 2010;Pride et al., 1999;Robinson et al., 2007Robinson et al., , 2009Thunell & Kepple, 2004). This δ 15 N increase is widely interpreted as a deglacial rise in water column denitrification across the ETP, an interpretation supported by a comparable deglacial δ 15 N rise in the Arabian Sea of the Indian Ocean (Altabet et al., 2002;Ganeshram et al., 2000). Typically, in the ETSP, the δ 15 N rise is observed to begin at ∼17-18 ka, in step with southern hemisphere warming (De Pol-Holz et al., 2006;Higginson & Altabet, 2004;Robinson et al., 2007), whereas the δ 15 N rise in the ETNP (and downstream of it) occurs with the first abrupt northern hemisphere warming at ∼14-15 ka, associated with the Northern Hemisphere's Bolling-Allerod Warm Period (Chang et al., 2014(Chang et al., , 2008Emmer & Thunell, 2000;S. S. Kienast et al., 2002;Pride et al., 1999). The deglacial rises in water column denitrification in the ETNP and ETSP, with their respective timings, are often interpreted as the result of a tendency for warming to enhance water column suboxia and denitrification, through the decreased solubility of O 2 and/or slower ventilation of intermediate waters and the thermocline under warmer climate (Galbraith et al., 2004). A deglacial rise in nutrient supply to the thermocline has also been implicated, the concept being that increased thermocline nutrients would drive higher lower latitude export production and thus more oxygen demand for respiration in the thermocline (Robinson et al., 2005). Millennial scale variations during the deglaciation are largely consistent with these interpreted mechanisms (Robinson et al., 2007).
In many ETP records, the later deglaciation and Holocene is characterized by a subsequent decline in bulk sediment δ 15 N. In some records, the decline is minor; in others, the decline almost completely compensates for the more abrupt early deglacial δ 15 N rise. There have been multiple plausible interpretations of this decline: as part of a whole-ocean trend in nitrate δ 15 N, a late deglacial decline in water column denitrification, and/or a rise in N 2 fixation (Deutsch et al., 2004;Galbraith et al., 2013;Thunell & Kepple, 2004). The deglacial δ 15 N maximum is often accompanied by deglacial maxima in redox sensitive trace metals and/or sediment lamination (Pride et al., 1999;Y. Zheng et al., 2000). These parallel changes support the interpretation of the deglacial δ 15 N maximum as due to an associated temporal peak in water column denitrification. However, as with the δ 15 N records, the Holocene evolution of trace metal concentrations and lamination varies among sediment cores, declining back to baseline in some but not in others (Chang et al., 2014;Muratli et al., 2010;Y. Zheng et al., 2000). Finally, the deglacial δ 15 N maximum itself has significantly different temporal structure in different regions and sediment cores (De Pol-Holz et al., 2006;Emmer & Thunell, 2000;Ganeshram et al., 2000;Hendy & Pedersen, 2006;Higginson & Altabet, 2004;S. S. Kienast et al., 2002;Pichevin et al., 2010;Robinson et al., 2009Robinson et al., , 2007Thunell & Kepple, 2004).
In the EEP, the LGM-to-Holocene δ 15 N rise in the first sediment δ 15 N records was initially interpreted as a lower degree of nitrate consumption during the ice ages (Farrell et al., 1995). However, the change was very similar to that reconstructed from the ETP (Ganeshram et al., 1995); as described above, that change has been interpreted as deriving from an increase in water column denitrification. The LGM-to-Holocene sediment δ 15 N rise in the EEP is now most often interpreted as a shared ETP/EEP signal, reflecting a rise in the δ 15 N of subsurface nitrate as a result of an LGM-to-Holocene increase in water column denitrification . Further west in the EEP, studies have revealed no LGM-to-Holocene rise in sediment δ 15 N but rather a deglacial maximum, which is also apparently observed in the western equatorial Pacific . This sediment δ 15 N signal might be linked to the deglacial δ 15 N maximum that is apparent in some records from the ETP. Differencing of western and eastern equatorial Pacific sediment δ 15 N records yields a signal with orbital periodicities, dominantly of precession; this has been interpreted as a changing degree of nitrate consumption that results from insolation-driven EEP upwelling changes . Interpretations aside, sediment δ 15 N records from the EEP appear to hold signals of both an LGM-to-Holocene sediment δ 15 N rise and, further westward, a deglacial δ 15 N maximum.
The existing δ 15 N records mentioned above are all based on N isotopic measurement of the bulk sediment (δ 15 N bulk ). However, the primary N isotopic signals acquired in the surface ocean may be obfuscated in δ 15 Nbulk . Two causes for this are (1) diagenetic alteration during sinking in the water column and early burial in the sediment (Altabet and François, 1994), and (2) the contribution of organic and inorganic N derived from terrestrial or distal marine (e.g., shelf) sources (Meckler et al., 2011;Schubert & Calvert, 2001). To avoid these biases, recent work has focused on measuring the δ 15 N of organic matter bound within and protected by the tests of microfossils, such as foraminifera and diatoms (Ren et al., 2009;Robinson et al., 2004).
Here, we report measurements of foraminifera shell-bound δ 15 N (FB-δ 15 N) on the two species Neogloboquadrina dutertrei and Neogloboquadrina incompta (formerly named Neogloboquadrina pachyderma (dextral); Darling et al., 2006) in two sediment cores extending back to the last ice age. These cores (ME0005-24JC and ME0005-27JC) fall within the nutrient-rich EEP ( Figure 1). They both have abundant foraminifera and a rich set of previous measurements. Of particular interest, each core shows a LGM-to-Holocene increase in bulk sediment δ 15 N similar to that observed at other EEP and ETP sites, which has most often been interpreted as indicating that ETP water column denitrification was reduced during the LGM (De Pol-Holz et al., 2006Ganeshram et al., 1995Ganeshram et al., , 2000Robinson et al., 2007).

Materials and Methods
Sediment cores ME0005-24JC and ME0005-27JC (hereafter abbreviated as ME-24 and ME-27) were retrieved from the EEP. ME-24 is located north of Carnegie Ridge (0°1.3′ N, 86°27.8′ W, 2941 m), while ME-27 is located south of Carnegie Ridge (1°51.2′ S, 82°47.2′ W, 2203 m) ( Figure 1). The age models for both sediment cores have been updated by Dubois et al. (2014) and are based on (1) radiocarbon ages measured on the planktonic foraminifera N. dutertrei by accelerator mass spectrometry (S. S. Kienast et al., 2007;Kusch et al., 2010), (2) correlation of benthic foraminifera oxygen isotopes to the LR04 stack, and (3) the identification of the Los Chocoyos Ash Layer (Drexler et al., 1980) in the sediment cores. In core ME-27, we have included three additional 14 C dates on N. dutertrei at 0.5 cm, 80.5 cm, and 125.5 cm depth from Mekik (2014). All radiocarbon ages were calibrated with Calib 7.1 and the marine calibration curve MARINE13 (Reimer et al., 2013;Stuiver & Reimer, 1986), assuming a reservoir age of 467 years as given in Dubois et al. (2014). Ages were linearly interpolated between the stratigraphic tie points, which are given in Supplementary Table 1. Foraminifera-bound δ 15 N was measured with the "persulfate-denitrifier" technique (Knapp et al., 2005;Ren et al., 2009;Straub et al., 2013). In brief, ∼3-5 mg of foraminifera (N. dutertrei and N. incompta from the 300-600 µm size fraction) were picked, were cut open with a scalpel (rather than being crushed, to minimize sample loss), and underwent a chemical cleaning (Ren et al., 2009;Straub et al., 2013). The organic N bound within the calcite was then released by dissolution with HCl and converted to nitrate in a basic potassium persulfate solution (Nydahl, 1978). The nitrate concentration of the solution was determined by chemiluminescence (Braman & Hendrix, 1989), and an aliquot of the nitrate solution equivalent to 5 nmol N was converted to nitrous oxide (N 2 O) by denitrifying bacteria (Sigman et al., 2001). The N isotopic composition of the N 2 O was measured with a custom continuous-flow system for N 2 O extraction and purification on-line to a Thermo MAT253 stable isotope mass spectrometer and referenced to air N 2 using the international nitrate standards IAEA-N3 and USGS-34 (Weigand et al., 2016). The FB-δ 15 N data were then corrected for the contribution of the oxidation procedural blank with an in-house aminocaproic acid standard of known isotopic composition.
After the blank correction, we found linear offsets in FB-δ 15 N of up to 0.5-0.6‰ among the individual batches analyzed. The origin of this linear offset may be an additional blank introduced during the cleaning process that was not monitored by the procedural blanks at the time of these analyses. Since we analyzed almost all samples in duplicate across separate batches of analyses, we were able to correct for the artifact by applying offsets to whole batches ( Supplementary Figures 1 and 2). Replicate FB-δ 15 N analyses for each depth were then averaged, yielding species-specific FB-δ 15 N records for both N. dutertrei and N. incompta in each sediment core. The offset correction does not impact the interpretation of the data.

Results
In the δ 15 N bulk records measured in ME-24 and ME-27 by , δ 15 N bulk is lowest (4-5‰) during the last ice age, increases to 6-7‰ during the deglaciation, and then slightly increases (ME-24) or decreases (ME-27) during the Holocene (Figure 2). FB-δ 15 N is almost always higher than δ 15 N bulk, a common finding when the foraminifera species analyzed lacks dinoflagellate endosymbionts (Martinez-Garcia et al., 2014;Ren et al., 2012;Smart et al., 2018). In terms of downcore change, FB-δ 15 N is notably different from δ 15 N bulk at both core sites. Most importantly, unlike δ 15 N bulk , FB-δ 15 N during the last ice age (6-8‰) is similar to or higher than that of the late Holocene (6-7.5‰). At both sites, FB-δ 15 N gradually increases during deglaciation and shows a deglacial maximum during the HS1 and early Bolling-Allerod (BA). At ME-27, FB-δ 15 N then gradually declines throughout the YD and the Holocene. At ME-24, FB-δ 15 N declines abruptly during the second half of the BA and is at a minimum during the YD, peaks again during the early Holocene, and then declines into the middle Holocene. Another difference between the two core sites is that at ME-27, FB-δ 15 N of N. dutertrei and N. incompta are very similar, while at ME-24, FB-δ 15 N of N. dutertrei is on average 0.6‰ higher than FB-δ 15 N of N. incompta. Despite this difference, the two species show the same down-core changes within each core. STUDER ET AL.

Disagreement Between δ 15 N bulk and FB-δ 15 N
The similarity among the δ 15 N bulk records from ME-24 and ME-27 and other cores in the region (Figure 3) appears consistent with a coherent change in nutrient dynamics in the EEP. Furthermore, the location close to the continent and in proximity to two upwelling systems promises high primary productivity that in turn leads to rapid biogenic fluxes and sediment accumulation, which are expected to minimize the influence of diagenetic alteration at the seafloor (Ganeshram et al., 2000;Thunell & Kepple, 2004). However, the disagreement with the trends in FB-δ 15 N indicates a bias in either δ 15 N bulk or FB-δ 15 N.
More ground-truthing of FB-δ 15 N is called for, especially in dynamic environments such as upwelling systems. Observed species-associated offsets point to influences from symbiosis and possibly also from feeding habits, seasonality, and depth habitat (Costa et al., 2016;Ren et al., 2012;Smart et al., 2018Smart et al., , 2020. Moreover, the two species measured here, N. dutertrei and N. incompta, show an offset in ME-24 of 0.6‰ on average but have no clear offset in ME-27 ( Figure 2). N. dutertrei are non-spinose and pelagophyte symbiont-bearing, whereas N. incompta are non-spinose and asymbiotic (Bird et al., 2018;Takagi et al., 2019). Modern studies indicate that asymbiotic and/or non-spinose species have a distinctly higher FB-δ 15 N than spinose, dinoflagellate-bearing species, by as much as 4‰ (Ren et al., 2012;Smart et al., 2018). These studies indicate that pelagophyte symbiont-bearing species, including N. dutertrei, cluster with asymbiotic species such as N. incompta, consistent with their similar FBδ 15 N in ME-27. We suggest that the δ 15 N offset between the two species in ME-24 speaks to a stronger affinity of N. incompta for upwelling, which may cause N. incompta to be biased toward periods and locations of maximal upwelling, resulting in a lower δ 15 N for its food source. Above ME-27, the strong flow of surface waters from the South and their pervasive nutrient-richness may preclude seasonal or small-scale spatial variations in food source δ 15 N, leading to more similar FB-δ 15 N for the two species.
However, a glacial-to-interglacial bias in the FB-δ 15 N records is unlikely, lacking a clear mechanism. Supporting this view, the two species show similar changes within each core, including the pronounced δ 15 N minimum during late deglaciation in ME-24. Thus, we question the primary nature of the bulk sediment δ 15 N records, in particular, their glacial-to-interglacial δ 15 N increase. We point to two processes that may bias the bulk sediment δ 15 N records to indicate an artificially low δ 15 N for sinking N during the last ice age.
First, given the proximity of the South American continent, significant input of organic and inorganic nitrogen from terrestrial sources to the core sites may be occurring. Linear regression of bulk sediment total nitrogen (TN) against total organic carbon (TOC) yields a non-zero intercept (Figure 4a), suggesting the presence of clay-bound N with a low C/N, such as clay-bound ammonium and amino acids (Meckler et al., 2011;Muller et al., 1977;Schubert & Calvert, 2001). If the intercept is entirely due to inorganic N, then it accounts for ∼23 ± 14% of the average TN in ME-24 and ME-27. Inorganic N, likely present as ammonium bound in clay minerals, is reported to have a low δ 15 N, for example, of 2-4‰ (Schubert & Calvert, 2001). The δ 15 N of terrestrial organic N probably depends strongly on the region and environment of the source. Nevertheless, given the relatively high δ 15 N of marine organic matter in this region, the combined contributions of terrestrial organic and inorganic N would likely work to lower δ 15 N bulk . δ 15 N bulk is indeed mostly lower than FB-δ 15 N at both ME-24 and ME-27, even after subtracting 1-2‰ from FB-δ 15 N to account for STUDER ET AL.
10.1029/2020PA004063 6 of 16 the typical elevation of asymbiotic and/or non-spinose foraminifera δ 15 N relative to the δ 15 N of the nitrate consumed in the euphotic zone (Ren et al., 2012;Smart et al., 2018).
Second, weaker diagenetic alteration during the last ice age may have caused or contributed to the glacial-to-interglacial δ 15 N bulk difference. The majority of studies indicate that bulk sediment δ 15 N increases during oxic organic matter diagenesis (Robinson et al., 2012 and references therein). Oxic diagenesis appears to be strongly affected by oxygen exposure time, which is in turn controlled by bottom water oxygen concentrations, sediment burial rate, and the flux of organic matter to the sediment (Hartnett et al., 1998). Compared to the Holocene, ice age bottom water oxygen concentrations were lower in the deep EEP and ETP, including in the depth range of ME-24 (2941 m) and ME-27 (2203 m) (R. F. Anderson et al., 2019;Bradtmiller et al., 2010;Hoogakker et al., 2018;Jacobel et al., 2020;Loveley et al., 2017). This may have worked to minimize diagenetic loss, and thus, δ 15 N elevation, of bulk sedimentary N. Consistent with this possibility, in both ME-24 and ME-27, N content and N flux were higher during the last glacial and HS1 periods relative to the Holocene S. S. Kienast et al., 2007). Moreover, there is a correlation between the difference between δ 15 N bulk and FB-δ 15 N and the N content of the sediments, with a greater difference between δ 15 N bulk and FB-δ 15 N during the last glacial (Figure 4b). This is as expected if glacial-age sediment underwent less oxic diagenetic N loss. The slope of correlation in δ 15 N against ln(% TN) space provides a Rayleigh model-based estimate of the net isotope effect of diagenesis of 3.0 ± 0.5‰ (Figure 4c) that is consistent with previous estimates (Lehmann et al., 2002;Moebius, 2013). However, we caution that, as discussed above, this trend may alternatively (or additionally) result from exogenous N inputs being more important during the LGM.

Interpretation of EEP δ 15 N Change: Denitrification versus Surface Nitrate Consumption
At EEP sites, both the δ 15 N of the nitrate supply and the degree of nitrate consumption may influence the δ 15 N of export production. Accordingly, δ 15 N bulk records have been scrutinized for changes in both. Differences among EEP records and between EEP and ETP records during deglaciations and interglacials have STUDER ET AL.  Cross-plot of δ 15 N bulk -FB-δ 15 N against bulk sediment TN of sediment cores ME-24 (circles) and ME-27 (triangles), color-coded for the last ice age and HS1 periods (blue) and the BA period through the Holocene (red). The FB-δ 15 N values shown are the two-species averages for each depth. Nitrogen content and N flux (not shown) in ME-24 and ME-27 were higher during the last ice age and HS1 relative to the interglacial, suggesting that a higher fraction of the N flux to the seabed was preserved. This enhanced preservation, in turn, may explain the greater ice age difference between δ 15 N bulk and FB-δ 15 N as the result of reduced diagenetic elevation of bulk sedimentary δ 15 N (i.e., reduced diagenetic N loss) during the last ice age (e.g., due to the lower deep ocean oxygen (Hoogakker et al., 2018)). (c) Following from (b), but with the following modifications. The y-axis gives an estimate of the difference between δ 15 N bulk and the δ 15 N of nitrate consumed (NO 3 − cons.) in the euphotic zone. The latter was calculated from the FB-δ 15 N of N. dutertrei, from which is subtracted the average difference between the FB-δ 15 N of N. dutertrei and the δ 15 N of euphotic zone nitrate consumed, as observed in a ground-truthing study (1.3 ± 0.6‰; Ren et al., 2012). The N. incompta data are not used in (c), due to a lack of surface sediment calibration for this species. The x-axis is the natural logarithm of the bulk sediment TN percentage. Assuming the Rayleigh model, the slope of the correlation of δ 15 N with ln(%TN) provides an estimate of the net isotope effect for diagenetic N loss of 3.0 ± 0.5‰. In this interpretation, the tendency for δ 15 N bulk to be lower than the δ 15 N of nitrate consumed in the euphotic zone suggests the background input of low-δ 15 N exogenous N to the sediments, as supported by the data in (a). been recognized, and these have been interpreted as arising from changes in the degree of nitrate consumption in the EEP Robinson et al., 2009). However, the deglacial δ 15 N bulk rise and the higher δ 15 N bulk during interglacials are largely shared between the EEP and ETP records (as well as among EEP records), leading to the view that the δ 15 N of the nitrate supply in the EEP changes in parallel with the ETP, with water column denitrification being the major driver of these changes .
Modern ocean data should provide some insight into the degree to which the EEP should share changes in the δ 15 N of the nitrate supply with the ETP. Data from 95°W suggest that the subsurface nitrate being upwelled along the equator has a δ 15 N of ∼7.1‰ or slightly lower Rafter & Sigman, 2016). At this longitude, the lowest δ 15 N of shallow subsurface nitrate is associated with the SSCC, which also underlies the highest nitrate concentrations at the surface. The low nitrate δ 15 N of the SSCC likely derives from the remineralization of low-δ 15 N sinking N under the region of upwelling (Rafter & Sigman, 2016). These modern observations do not support a dominant role for the intensity of the EUC or other zonal subsurface flows on the δ 15 N of subsurface nitrate, at least this far east in the EEP. Rather, the biogeochemical dynamics of the region, both the water column denitrification of the ETP and the incomplete nitrate consumption in the EEP surface, appear to most strongly influence the δ 15 N of the nitrate being upwelled in the EEP. On the basis of prior studies of equatorial circulation and biogeochemical dynamics, a similar interpretation was put forward by Robinson et al. (2009). Thus, EEP sites should detect a significant reduction in ETP denitrification during the LGM. Such an influence is even more likely east of the Galapagos, due to the proximity to the suboxic zones and the interference of the islands with the eastward subsurface flows along the equator. In summary, modern ocean data support prior interpretations of shared δ 15 N bulk changes among EEP and ETP records as deriving from changes in ETP denitrification.
Thus, based in part on the evidence above, the deglacial rise and Holocene elevation of bulk sediment δ 15 N from ME-24 and ME-27 have been interpreted as dominantly reflecting reduced water column denitrification during the LGM relative to the Holocene , consistent with the previous conclusion of such a change on the basis of non-equatorial ETP records (Emmer & Thunell, 2000;Ganeshram et al., 2000Ganeshram et al., , 1995S. S. Kienast et al., 2002;Robinson et al., 2009). From this perspective, the FB-δ 15 N records, lacking such an LGM-to-Holocene δ 15 N difference, would appear to cast doubt on this overarching interpretation of an LGM reduction in ETP denitrification. We address this question next.

LGM-to-Holocene History of ETP Water Column Denitrification
As discussed in the prior section, the low glacial δ 15 N bulk observed at both ETP and EEP sites has been interpreted as a reduction in water column denitrification during the last ice age Emmer & Thunell, 2000;Ganeshram et al., 2000Ganeshram et al., , 1995S. S. Kienast et al., 2002;Robinson et al., 2009). However, not all δ 15 N bulk records in the ETP indicate an LGM-to-Holocene δ 15 N rise. For example, δ 15 N bulk records from the Gulf of Tehuantepec indicate a deglacial maximum in δ 15 N but no glacial-to-interglacial change, consistent with the FB-δ 15 N measurements (Figure 3b; Hendy & Pedersen, 2006;Pichevin et al., 2010;Pride et al., 1999;Thunell & Kepple, 2004). More broadly, in specific records from both further North and South in the ETP, bulk sediment δ 15 N is observed to return to glacial values over the Holocene (Figure 3a,c; Emmer & Thunell, 2000;Higginson & Altabet, 2004;S. S. Kienast et al., 2002;Pride et al., 1999). Thus, the FB-δ 15 N data are not alone in casting doubt on an LGM-to-Holocene rise in δ 15 N in the ETP and EEP. We propose that the "true" δ 15 N signal for the ETP is a deglacial maximum, with the low glacial δ 15 N observed at some ETP sites being a consequence of higher exogenous N input and/or reduced diagenetic alteration during the ice ages.
Suggested explanations for the gradual Holocene decline in δ 15 N bulk have ranged widely, including a decline in ETP water column denitrification, a rise in N 2 fixation, and/or a decline in mean ocean nitrate δ 15 N (Deutsch et al., 2004;Galbraith et al., 2013;Thunell & Kepple, 2004). In other proxies that are sensitive to suboxia, the dominant signal appears to be of a deglacial maximum in ETP suboxia, rather than a secular rise in suboxia from the last glacial to the interglacial (Hendy & Pedersen, 2006;Hoogakker et al., 2018;Y. Zheng et al., 2000). These data favor the interpretation of a deglacial maximum in ETP water column denitrification, with the possibility that denitrification rates were comparable between the LGM and the late Holocene. However, other records suggest a more durable shift to suboxia (Chang et al., 2014;Muratli et al., 2010).
Addressing this question further is warranted from multiple perspectives. First, the history of ETP water column denitrification is central to the operation of the global ocean N budget over glacial cycles. Second, denitrification reflects thermocline suboxia, the extent of which is an important aspect of the ocean's environmental conditions. Third, the extent of suboxia is impacted by ocean circulation and the cycling of carbon and nutrients; thus, it provides an important constraint on changes in these processes over the glacial cycles. Next, we consider the plausibility that ETP suboxia was similar between the LGM and the latter Holocene, in comparison to the prevailing interpretation that suboxia was reduced during the LGM.

Suboxia in the EEP and ETP During the Last Ice Age
The interpretation of a reduction in ETP water column denitrification during ice ages has been argued to be consistent with expectations given other characteristics of the ice age ocean (Galbraith et al., 2004). First, the water ventilating the thermocline was colder during the LGM, increasing preformed O 2 concentrations. Second, arguments have been made for faster thermocline ventilation during the ice ages (Galbraith et al., 2004). Third, enhanced nutrient consumption in polar and subpolar surface waters would have worked to reduce the preformed nutrient concentration of the thermocline, tending to reduce low latitude biological productivity and the rate of oxygen consumption from the respiration of the resulting sinking organic matter (Figure 5; Keir, 1988;Robinson et al., 2005;Sigman and Boyle, 2000). Finally, North Atlantic overturning has alternated between North Atlantic Deep Water (NADW) formation during interglacials and Glacial North Atlantic Intermediate Water (GNAIW) formation during glacials (Oppo & Lehman, 1993). In most configurations of their downstream circulations, switching from NADW to GNAIW works to raise mid-depth O 2 concentrations (Boyle, 1988;Hain et al., 2010;Sigman et al., 2003).
However, countering these changes is the evidence that the ocean's biological pump was more efficient during the ice ages. Broecker (1982a, b) first proposed a stronger global ocean biological pump as the cause of the lowering of atmospheric CO 2 during the last ice age. He recognized that such a strengthening in the biological pump should also have caused a decline in the average O 2 concentration of the ocean interior. Indeed, there is now strong evidence for lower O 2 in the glacial deep ocean (R. F. Anderson et al., 2019;Bradtmiller et al., 2010;François et al., 1997;Galbraith and Jaccard, 2015;Hoogakker et al., 2018  . Schematic cross section of the low latitude upper ocean today and during the last glacial period, highlighting the counteracting effects on the oxygen-deficient zones (ODZs) of the eastern tropical Pacific during the last ice age. The colored arrows and lines indicate the nutrient (i.e., phosphate and nitrate) concentration of waters supplied by advection (unidirectional arrows) or by mixing (bidirectional arrows). The gray shading is darker for lower oxygen concentration, with the two ODZs (of the ETNP and ETSP) shown as a single ODZ for simplicity. NPIW, North Pacific Intermediate Water; SAMW, Subantarctic Mode Water; AAIW, Antarctic Intermediate Water. The FB-δ 15 N data reported in this study raise the possibility that water column denitrification rates in the ETP were similar during the last glacial period and today. Colder temperatures and lower preformed nutrient concentrations of intermediate waters (SAMW and AAIW) would have worked against the development of ODZs during the last ice age. However, the lower O 2 concentration of the deep ocean and its mixing with and upwelling into the tropical thermocline would have fostered ODZ development. It is possible that these opposing mid-depth and deep ocean forcings offset one another to yield similar ODZ extent in the LGM and Holocene (see text). The implied lack of change in export production (C org ) is uncertain but appears appropriate for the easternmost equatorial Pacific (Costa et al., 2017(Costa et al., ). et al., 2020. This glacial decline in deep ocean O 2 , through mixing and diffuse interior upwelling, would also have worked to lower the O 2 concentration in overlying mid-depth waters ( Figure 5).
To gain insight into these competing influences on mid-depth O 2 , we turn to output from a geochemical box model that has been used to investigate the potential causes of the ice age reduction in atmospheric CO 2 (Hain et al., 2010). Here, we focus on the model-indicated sensitivities of O 2 in the model's deep and mid-depth South and North Pacific boxes ( Figure 6). Considering first the solubility effects of an ice age ocean that was 3°C colder and 1 PSU saltier than modern, the ocean interior oxygen concentration increases modestly, by 14 µmol/kg at mid-depths and 22 µmol/kg in the deep ocean (Garcia & Gordon, 1992). The combined effects of Subantarctic iron fertilization (which reduced the nutrient content of Subantarctic surface water and Subantarctic Mode Water; Martinez-Garcia et al., 2014;Robinson et al., 2005) (Garcia and Gordon, 1992) caused by cooling from 14˚C-to-11°C and 4˚C-to-1˚C in the mid-depth and deep Pacific, respectively, combined with a whole-ocean salinity increase from 34.7 to 35.7. (Center left) Simulated changes in O 2 utilization when glacial shoaling of the AMOC (GNAIW replacing NADW circulation; see Sigman et al., 2003) is combined with more complete surface nutrient consumption in the Subantarctic (SAZ) responding to iron fertilization, reducing the preformed nutrient concentration of the waters ventilating the mid-depths (see Martinez-Garcia et al., 2014;Robinson et al., 2005). (Center right) Simulated changes in O 2 utilization in response to two changes in the Antarctic Zone: reduced exchange between the deep Southern Ocean and the Antarctic surface combined with more complete Antarctic surface nutrient consumption (Ai et al., 2020;François et al., 1997;Kemeny et al., 2018;Sigman et al., 2021;Studer et al., 2015;Wang et al., 2017). (Right) Sum of O 2 solubility change and simulated O 2 utilization change when all model forcings are combined to represent the LGM (Hain et al., 2010). O 2 utilization is calculated using O 2 :P regenerated of −170:1 (L. A. Anderson and Sarmiento, 1994). The indication of this analysis is that there were counteracting influences on the O 2 concentration of the mid-depth Pacific during the LGM. On the one hand, AMOC shoaling, Subantarctic iron fertilization and cooling would have increased O 2 concentrations in the mid-depth and deep Pacific. On the other hand, the above-described Antarctic changes would have made the global biological pump more efficient, raising the concentration of regenerated nutrients and causing a major decline in the O 2 concentration in most of the global ocean interior, especially the deep ocean but also the mid-depth Pacific through vertical exchange and/or deep upwelling. The net result of these combined changes is a strong decline in O 2 concentration in the deep Pacific and relatively minor predicted O 2 changes in the mid-depth Pacific during the LGM. "All combined" does not equal a simple summation of the component changes, due to interactions among the components. For example, the effect of the switch from NADW to GNAIW on ocean interior O 2 depends on the degree of nutrient drawdown at the Antarctic surface. For further context, description of the model scenarios described here, and detailed analysis of simulated carbon cycle and atmospheric CO 2 changes, see Hain et al. (2010). and O 2 utilization in the Pacific that translates to an additional and more substantial rise in mid-depth O 2 . However, further strengthening of the ocean's biological pump through an increase in the degree of nutrient consumption in Antarctic surface waters as well as a decline in Antarctic overturning (Ai et al., 2020;François et al., 1997;Sigman et al., 2021;Studer et al., 2015;Wang et al., 2017) leads to a buildup of regenerated nutrients and a strong decline in deep ocean O 2 , consistent with available O 2 reconstructions (see above) and indicative of a more efficient global biological pump (Hain et al., 2014). The circulation communicates some of this O 2 decline to the mid-depth Pacific boxes as well. Combining this full list of ice age changes, the net result is little change in mid-depth Pacific O 2 concentration ( Figure 6). In short, increasing the efficiency of the global ocean's biological pump so as to lower atmospheric CO 2 to ice age observations tends to offset other influences on mid-depth O 2 , calling for little difference in mid-depth O 2 between the LGM and the Holocene.
These large-scale impacts having been considered, we must recognize that changes in ODZ extent and volume may be partially decoupled from average mid-depth O 2 concentration. In addition to being affected by the preformed O 2 and nutrient concentrations and the rate of newly forming intermediate waters, the ODZs may be sensitive to the spatial pattern of thermocline ventilation and the wind-driven upwelling of thermocline waters into the surface, the former delivering O 2 and the latter causing focused O 2 consumption. However, there is as yet no clarity as to how these smaller scale processes would have changed between the LGM and the Holocene.

Deglacial Features
In the two cores studied, both bulk sediment and foraminifera-bound δ 15 N show deglacial features (Figure 2); we focus here on the shell-bound data. Our first concern is the overall deglacial δ 15 N maximum shared by the two records (Figure 2), which is also observed in many low latitude bulk sediment δ 15 N records within and beyond the ETP (Figure 3). Previous work suggests that the δ 15 N maximum includes both a global thermocline signal and an ETP-specific deglacial maximum in water column denitrification (e.g., De Pol-Holz, 2006;Deutsch et al., 2004;Galbraith et al., 2013;Hendy & Pedersen, 2006;Y. Zheng et al., 2000). These and other studies propose a range of plausible explanations for these global and/or local deglacial maxima. However, while the FB-δ 15 N records reported here are not completely clear on the point, there is a suggestion that the δ 15 N rise begins before the deglaciation. If so, this adds to a body of data suggesting that the EEP "deglacial" signal actually includes a change preceding deglaciation (M. Kienast et al., 2006). One possibility is a response to orbitally (e.g., precession) driven insolation change .
Another possible origin for the putative deglacial peak in ETP suboxia, which may also explain its early onset, involves apparently distinct timings for different Southern Ocean changes. In the Subantarctic, the indicators of iron fertilization appear to begin their declines early in the deglaciation (Martinez-Garcia et al., 2014;Wang et al., 2017). In addition, ocean warming began quite early in the deglacial sequence (Bereiter et al., 2018). In contrast, the Antarctic surface "isolation" of the LGM appears to decrease somewhat later and more gradually, continuing through the deglaciation and Holocene (Ai et al., 2020;Wang et al., 2017). As outlined above, the Antarctic deglacial change should have caused a deglacial decline in suboxia while the other two (Subantarctic and temperature) deglacial changes should have caused a deglacial rise in suboxia ( Figure 6). Accordingly, the relative timing of these components may be appropriate to have caused a deglacial/early Holocene peak in suboxia and thus in ETP/EEP nitrate δ 15 N. Evaluation of this hypothesis awaits additional records with strong chronological controls.
While our two records both share an overall deglacial δ 15 N maximum, there are significant differences between the FB-δ 15 N records from ME-24 and ME-27. The clearest distinction is the FB-δ 15 N minimum of ME-24 centered on the Younger Dryas. The two sediment cores are separated by only 2° of latitude and 4° of longitude ( Figure 1). However, ME-24 is located immediately above the equator, where the δ 15 N of this subsurface nitrate may have a greater sensitivity to changes in denitrification in the ETNP Rafter & Sigman, 2016). ME-27, on the other hand, is located under the influence of the coastal upwelling off Peru as well as in the broad region of southern hemisphere-biased open ocean upwelling of the EEP (Figure 1) and may thus be more sensitive to changes in denitrification in the ETSP (Dubois et al., , 2014Robinson et al., 2009).
The Younger Dryas FB-δ 15 N minimum in ME-24 is accompanied by similar δ 15 N minima in bulk sediment records from the ETNP (Figure 3a; Emmer & Thunell, 2000;Hendy et al., 2004;S. S. Kienast et al., 2002), also coinciding with metal redox indicators of higher ETNP thermocline oxygen at this time (Cartapanis et al., 2011;Chang et al., 2014;Hendy & Pedersen, 2005;Y. Zheng et al., 2000). We thus tentatively interpret the δ 15 N minimum as the result of lower nitrate δ 15 N in the subsurface waters of ME-24, resulting from a reduction in ETNP denitrification. Consistent with this interpretation, the lack of such a minimum in ME-27 matches its corresponding lack in bulk sediment δ 15 N records from the ETSP margin of Peru and Chile (Figure 3c; Higginson & Altabet, 2004).
As for the physical cause of this putative YD decline in ETNP denitrification, one possibility is a southward shift in the Intertropical Convergence Zone (ITCZ) during the YD, driven by northern hemisphere cooling at the time (Lea et al., 2003). Such a shift may have reduced upwelling along the EEP as a result of reduced cross-equatorial southeast trade winds (Koutavas & Lynch-Stieglitz, 2005), decreasing denitrification in the ETNP. However, this fails to explain why a similar event did not occur during HS1, when there was also abrupt northern hemisphere cooling (Bond et al., 1993). Indeed, it appears that HS1 was associated with higher, not lower, upwelling in the EEP (M. Kienast et al., 2006), which has been explained as the result of intensified northeast trade winds during HS1 when the ITCZ appears to have been at its southernmost position (S. S. Kienast et al., 2013;Timmermann et al., 2007). These conflicts encourage consideration of larger scale causes for the putative YD decline in ETNP denitrification. In this category, a cooling-associated acceleration of North Pacific thermocline ventilation has been suggested for the YD (Mikolajewicz et al., 1997).
In our interpretation of the deglacial changes, we have assigned primary signal to bulk sediment δ 15 N records. Further above, the disagreement of bulk sediment and foraminifera-bound δ 15 N in ME-24 and ME-27 was interpreted to argue against a primary origin for the glacial/interglacial δ 15 N difference in bulk sediment. We do not see these interpretations as contradictory. In some cases, down-core changes in bulk sediment δ 15 N will dominantly reflect the primary isotopic signal of the N sinking out of the surface ocean; in other cases, bulk sediment δ 15 N and its changes may be heavily altered by exogenous N or changes in diagenesis. The latter case is most likely when substantial changes in sedimentary environment occur, such as between glacial and interglacial intervals (Meckler et al., 2011).

Conclusions
In two sediment cores from the EEP, we measured foraminifera-bound δ 15 N back to ∼35 ka during the last ice age. FB-δ 15 N values are similar during the Holocene and the last glacial, unlike bulk sedimentary δ 15 N, which is lower during the Last Glacial Maximum (LGM) relative to the Holocene. Our FB-δ 15 N data do not appear to agree with previous bulk sediment-based inferences of reduced water column denitrification during the last ice age. Instead, our FB-δ 15 N data imply similar rates of water column denitrification during the Holocene and the last glacial and thus no dramatic LGM-to-Holocene change in suboxia in the ETP. Mechanisms have previously been identified for reduced ETP suboxia during the ice ages (e.g., Galbraith et al., 2004). We suggest that these were offset by the documented ice age decline in oxygen concentration in the deep Pacific waters underlying the Pacific thermocline. Mixing with, and upwelling from, these deeper less-well oxygenated waters would have helped to maintain the suboxic zones during the ice ages ( Figure 5).
We caution that the observation of lower bulk sediment δ 15 N during the LGM extends both northward and southward along the eastern Pacific margin (De Pol-Holz et al., 2006;Galbraith et al., 2004;Ganeshram et al., 2000;Hendy et al., 2004;S. S. Kienast et al., 2002;Robinson et al., 2007), and our data do not directly test the veracity of these higher latitude bulk sediment δ 15 N records. Moreover, in part due to our lack of spatial coverage, we have been unable to assess the possible role of variations in the degree of nutrient consumption in the reconstructed δ 15 N changes, and yet, core-top bulk sediment δ 15 N and FB-δ 15 N data indicate that this is a potentially significant variable in the nutrient-rich EEP (Costa et al., 2016;Farrell et al., 1995). In addition, the rate at which undercurrent and countercurrent waters are imported from the west could also change. As discussed above, modern ocean data argue against this as a major driver of δ 15 N change, but it cannot be ruled out. Finally, mean ocean nitrate δ 15 N may have changed from the LGM to the deglaciation, and then through the Holocene (Galbraith et al., 2013). In net, the potential for overlapping of signals leaves open the possibility that the LGM-Holocene similarity in EEP FB-δ 15 N does not reflect similar extents of suboxia during the LGM and Holocene. Generating additional FB-δ 15 N records from the ETP suboxic zones will not address all uncertainties; nevertheless, this is the obvious next step.