Volume 36, Issue 4 e2020PA004206
Research Article
Free Access

Enhanced Carbonate Dissolution Associated With Deglacial Dysoxic Events in the Subpolar North Pacific

Calie R. Payne

Corresponding Author

Calie R. Payne

Department of Geology and Geophysics, Texas A&M University, College Station, TX, USA

Correspondence to:

C. R. Payne,

[email protected]

Search for more papers by this author
Christina L. Belanger

Christina L. Belanger

Department of Geology and Geophysics, Texas A&M University, College Station, TX, USA

Search for more papers by this author
First published: 30 March 2021
Citations: 2

Abstract

Here we use volume density (ρV) measurements as a metric of size-normalized weights for Neogloboquadrina pachyderma, a planktonic foraminifer, from upper OMZ and abyssal depth sites in the Gulf of Alaska over the past ∼20,000 years to test for covariation between carbonate preservation and OMZ intensity. We find that dissolution of N. pachyderma is most intense at the upper OMZ site where oxygenation is generally lower than at the abyssal site. We also examine Uvigerina peregrina, a benthic foraminifer, at the upper OMZ site and find that the lowest ρV measurements in both taxa occur during deglacial and early Holocene dysoxic events. We use computed tomography images to confirm that changes in ρV are related to shell thickness, observe dissolution features, and test for growth influences on ρV. Further, we use stepwise selection of multiple regression models in which coregistered environmental proxies are potential predictors of ρV and find that the best supported models retain negative associations between ρV and the concentration of redox-sensitive metals and the relative abundance of dysoxia-tolerant and opportunistic benthic foraminifera, indicating that low ρV is associated with low-oxygen conditions and pulsed availability of organic matter at the seafloor. Taken together, our results suggest the primary driver of carbonate dissolution here is related to organic carbon respiration at the seafloor. This highlights the importance of metabolic dissolution in understanding the inorganic carbon cycle and the role regions with high-organic carbon export, such as OMZs, can have as CO2 sources as metabolic dissolution intensifies.

Key Points

  • Both benthic and planktonic foraminiferal fossils have lower shell density due to dissolution during dysoxic events in the Gulf of Alaska

  • Dissolution occurs primarily at the seafloor and is associated with high benthic organic carbon respiration

  • Metabolic dissolution is an important contributor to the inorganic carbon cycle in oxygen minimum zones and may enhance this CO2 source

1 Introduction

1.1 Oxygen Minimum and Carbonate Maximum Zones in the Pacific

Oxygen minimum zones (OMZs) are regions at intermediate ocean depths with low dissolved oxygen content, the extent of which vary over time with ocean circulation, nutrient availability, and climate (Deutsch et al., 2011; Helly & Levin, 2004; Keeling et al., 2010). These zones are forecast to expand and intensify with ongoing global climate change driven by rising anthropogenic CO2 (Keeling et al., 2010; Long et al., 2016; Stramma et al., 2008). OMZs also correspond with maxima in dissolved inorganic carbon (DIC), forming corresponding carbon maximum zones (CMZs), due to high-organic carbon respiration and, potentially, enhanced carbonate dissolution driven by that respiration (Paulmier et al., 2011; Wyrtki, 1962). In addition, prolonged hypoxic to anoxic conditions may amplify the effects of benthic respiration on bottom water pH and carbonate saturation (Ω), further enhancing carbonate dissolution in OMZ sediments as they do in coastal environments (Hu et al., 2017; Wang et al., 2020). As a result of these processes, OMZs are presently the primary carbon reservoir of the subsurface ocean and can be sites of CO2 transfer from the ocean to the atmosphere with significant consequences for oceanic carbon storage and global climates (Naik et al., 2014; Paulmier et al., 2011). Examining the relationship between past intensification of OMZs and changes in marine calcification and carbonate dissolution is important for parameterizing changes in the inorganic carbon cycle that may occur with climate change.

The Pacific Ocean contains the world's largest OMZ (Paulmier & Ruiz-Pino, 2009). Oxygen content has decreased and the upper OMZ boundary has shoaled in this region since at least the 1960s in response to warming-induced declines in solubility, enhanced upper ocean stratification, reduced ventilation, and decreased North Pacific Intermediate Water (NPIW) formation (Bograd et al., 2008; Ito et al., 2017; Keeling et al., 2010; Kwon et al., 2016; Long et al., 2016; Pierce et al., 2012; Schmidtko et al., 2017; Whitney et al., 2007). DIC accumulation is also highest in the North Pacific, particularly in the Gulf of Alaska OMZ, because the North Pacific contains the oldest waters in the “conveyor belt” of oceanic circulation (Paulmier & Ruiz-Pino, 2009; Paulmier et al., 2011). In addition, ocean acidification (OA) has also intensified and led to shoaling of the aragonite and calcite saturation horizons in the open ocean since the industrial era, including in the Pacific, although OA is primarily related to the intrusion of atmospheric CO2 into the surface ocean (Feely et al., 20042008). The combination of these stressors can have severe consequences to marine ecosystems, including adverse effects on marine calcification (Breitburg et al., 2015; Gilly et al., 2013; Orr et al., 2005).

North Pacific paleoceanographic records document increases in OMZ extent and intensification, particularly during the Bolling-Allerod (B/A) and early Holocene, which correspond to intervals of abrupt warming (Belanger et al., 2020; Cannariato & Kennett, 1999; Davies et al., 2011; Jaccard & Galbraith, 2012; Moffitt et al., 2015; Ohkushi et al., 2013; Praetorius et al., 2015; Saravanan et al., 2020; Sharon et al., 2021; Shibahara et al., 2007; Zou et al., 2020). Proposed drivers of these low-oxygen events include enhanced productivity and respiration of organic carbon at intermediate depths (Davies et al., 2011; Gray et al., 2018; Preatorius et al., 20152020), which would not only reduce dissolved oxygen content, but could also promote dissolution via the production of carbonic acid during respiration (de Villiers, 2005; Boudreau & Canfield, 1993). The nutrients required for stimulating this productivity are hypothesized to come from terrestrial and coastal sources (Addison et al., 2012), meltwater (Davies et al., 2011), remobilization of sedimentary iron under hypoxic conditions (Praetorius et al., 2015), and upwelling of nutrient-enriched deep waters (Gray et al., 2018). Additional hypotheses for these low-oxygen events include decreased oxygen solubility at warmer temperatures (Praetorius et al., 20152020) and decreased intermediate water ventilation (Jaccard & Galbraith, 2012). Here, we use well-studied foraminiferal and geochemical paleoceanographic records from the Gulf of Alaska to test for covariation between changes in carbonate preservation and OMZ intensity to determine if intensification of dissolution also occurred, as expected if increased respiration of organic carbon drove intensification of the OMZ.

1.2 Foraminiferal Shell Weights

Foraminiferal size-normalized shell weights (SNW) were developed as a proxy for bottom water carbonate ion concentrations under the assumption that reduction in shell weights reflect thinning of the shell by dissolution in CO32− undersaturated environments primarily at or within the seafloor (Broecker & Clark, 20012002; Lohmann, 1995). However, dissolution after death can occur in different environments that a shell encounters within the taphonomically active zone (TAZ) including at the sediment-water interface, in sedimentary pore spaces, and, for planktonic species, in the upper water column (Petro et al., 2018). For example, dissolution can occur in the presence of supersaturated bottom waters if foraminifera encounter a zone of undersaturation within organic-rich “benthic fluff” near the sediment-water interface before burial (de Villiers, 2005). The extent of dissolution a shell experiences may, thus, reflect not only the carbonate ion concentration of the bottom waters, but may also reflect the amount of time a shell spends exposed to the TAZ after death, carbonate ion concentration in the upper water column, seafloor sedimentation rates, the carbonate content of the sediments in which the shell is deposited, and the amount of organic carbon remineralization occurring in those environments (Broecker & Clark, 2001; de Villiers, 2005; Dunne et al., 2012; Petro et al., 2018). The influence of dissolution on some SNW records is supported by evidence of surface smoothing of foraminiferal shells with lower SNWs (Barker et al., 2004; Pak et al., 2018). Increases in foraminiferal shell weights may also occur via carbonate overgrowths if CO32− is oversaturated in portions of the TAZ (Petro et al., 2018).

Foraminifera tend to more heavily calcify their tests when there is greater carbonate ion availability (Marshall et al., 2013; Russel et al., 2004; Spero et al., 1997), thus SNW may also reflect initial calcification and growth rates in foraminifera, prior to any dissolution. For example, lower carbonate ion and pH conditions leads to lower calcification, and thus lower SNW, in at least some species (Bijma et al., 2002; Marshall et al., 2013; Moy et al., 2009). In addition, other aspects of the growth habitat, including temperature, salinity, and food availability, may drive changes in foraminiferal growth and calcification, and thus influence SNW (Beer et al., 2010b; Bijma et al., 1990; Broecker & Clark, 2001; de Villiers, 2004; Pak et al., 2018; Todd et al., 2020; Weinkauf et al., 2016). In some paleoceanographic records, dissolution proxies such as fragmentation counts and external smoothing of shell features show a decoupling between dissolution and SNW (Davis et al., 2016; Pak et al., 2018; Todd et al., 2020). For example, Todd et al. (2020) used fragmentation counts as an independent measure of dissolution and found that variation in SNW could not be explained by variation in dissolution intensity and instead interpreted SNW in terms of local growth conditions. In the Santa Barbara Basin (SBB) prior to 1975, G. bulloides SNWs were highest during times with upwelling of cool waters with lower Ω and showed no external signs of dissolution, suggesting that growth responses to temperature had a greater influence on SNW than carbonate chemistry (Pak et al., 2018). Thus, while some studies suggest carbonate chemistry is the primary control on shell weights (Barker & Elderfield, 2002; Marshall et al., 2013), this relationship cannot be assumed.

Furthermore, physiological and ecological differences among species affect calcification responses to environmental changes and, thus, species may have different SNW responses to the same environmental gradients (Beer et al., 2010b; Marshall et al., 2013; Weinkauf et al., 2016). For example, growth in planktonic foraminifera with symbionts may be less affected by low-pH conditions than in nonsymbiotic species (Lombard et al., 2010; Manno et al., 2012). Studies using benthic foraminifera have also found inconsistencies in SNW changes among species indicating that species have dissimilar responses, perhaps due to the different microhabitats the species prefer during calcification (Davis et al., 2016). Thus, initial calcification may reflect a multitude of environmental and ecological factors and, thus, SNW may covary with factors other than carbonate ion saturation state at the time of calcification or after death.

Benthic and planktonic foraminifera together comprise ∼25% of modern global carbonate production (Langer, 2008), thus changes in foraminiferal calcification and preservation in the marine sedimentary record can have a significant effect on the marine inorganic carbon cycle. Planktonic foraminifera have particular importance in the transfer of inorganic carbon and alkalinity from the surface ocean to the deep sea and may account for up to 80% of the CaCO3 flux to the seafloor (Schiebel, 2002). Estimates of modern dissolution at the sediment-water interface account for only 22% of the total estimated carbonate dissolution in the oceans, thus much of the dissolution affecting foraminifera must occur in either the water column or within the sediments after burial (Sulpis et al., 2018). Disentangling the influences of initial calcification, dissolution in the water column, and dissolution at or within the seafloor on SNW is necessary for understanding the environmental correlates with past changes in calcification and dissolution and for forecasting the consequences of global climate change on the inorganic carbon cycle.

Here we present SNW records from two sediment core records from the Gulf of Alaska (GoA) covering the past ∼20,000 years; one from a site in the upper OMZ and one at abyssal depths below the OMZ (Figure 1). We compare planktonic SNW records between these sites to test the relative influence of the environmental conditions that the foraminifera experienced in life and in death. Unlike previous studies, we also examine a benthic (Uvigerina peregrina) and a planktonic (Neogloboquadrina pachyderma) species from the same sedimentary record, which enables us to assess the contributions of water column and pore water processes. In recognition that changes in SNW may reflect changes in either shell growth or the intensity of carbonate dissolution, we examine computed tomography (CT) images to evaluate changes in shell thickness, chamber sizes, and surface textures. Finally, we quantitatively relate changes in SNW to independent proxies for temperature, seasonality of organic carbon fluxes, and dysoxia using stepwise multiple regressions to determine which environmental variables are most supported as predictors of changes in SNW. This approach allows us to fully interpret changes in SNW and determine the drivers of low SNW in the GoA.

Details are in the caption following the image

Study site locations. (a) Location map of Sites U1419/EW0408-85JC and U1418/EW0408-87JC. Meridional cross sections (144°W) of (b) modern dissolved oxygen, and (c) modern carbonate saturation. The figure was created using Ocean Data View (Schiltzer, 2018). The oxygen and carbonate saturation data are derived from GLODAP v2.2020 (Olsen et al., 2020).

2 Methods

2.1 Study Site

Integrated Ocean Drilling Program (IODP) Site U1419 (59.6° N, 144.3°W, 697 m) and its colocated site-survey core EW0408-85JC are located on the slope of a 25 km wide continental shelf in the Gulf of Alaska near Kayak Island (Figure 1a). This intermediate-depth site underlies NPIW and is in the upper edge of the modern OMZ (Figure 1b). The abyssal site, IODP Site U1418 (58.8°N, 144.1°W, 3,677 m water depth) and its associated site-survey core EW0408-87JC, underlies Pacific Deep Water (PDW) and is below the modern OMZ (Figure 1b). Both sites are currently bathed in waters undersaturated in calcite (Ω < 1) and have Ω values > 2 in the upper 200 m of the water column (Figure 1c). We integrated samples from U1419 and EW0408-85JC at the intermediate-depth site and from U1418 and EW0408-87JC at the abyssal site using their respective age models, which are based on reservoir-corrected radiocarbon ages from foraminifera and tephrochronology (Davies-Walczak et al., 2014; Du et al., 2018; Praetorius et al., 2015; Walczak et al., 2020). Sedimentation rates at the intermediate-depth site average ∼50 cm/ka over the past 12 ka and are as high as 800 cm/ka during the deglacial interval (Walczak et al., 2020). At the abyssal site, radiocarbon ages also suggest very high sedimentation rates during the deglacial; sedimentation rates averaged ∼475 cm/ka from 15.5 – 17.7 ka and samples spaced 80 cm apart were undisguisable in age at ∼17.3 ka given a 220–240 years 2σ age uncertainty on the radiocarbon dates (Praetorius et al., 2015).

Changes in benthic foraminiferal faunal composition and redox-sensitive metal concentrations in sediments from both GoA sites indicate dysoxic conditions during the B/A and early Holocene at intermediate depths and a less severe decrease in oxygenation at abyssal depths during the deglacial (Belanger et al., 2020; Davies et al., 2011; Praetorius et al., 2015; Sharon et al., 2021). Paleotemperature reconstructions using alkenones from the same GoA records suggest these dysoxic events are associated with abrupt warming events (Praetorius et al., 2015). Increased productivity is also associated with these events, as evinced by high opal concentrations, abundant upwelling-associated diatoms and silicoflagellates, and increased total organic carbon, suggesting enhanced export productivity contributed to the low-oxygen conditions at the intermediate-depth site (Addison et al., 2012; Barron et al., 2009; Davies et al., 2011; Praetorius et al., 2015). In addition, benthic foraminiferal taxonomic diversity and compositional evenness increased at the abyssal site coincident with the occurrence of low-oxygen tolerant species, suggesting organic matter availability increased at the abyssal site during these low-oxygen events as well (Belanger et al., 2020). Given the attenuation of organic carbon flux with depth and the less severe decrease in oxygenation at the abyssal site, we would expect dissolution intensity to be more intense at the intermediate-depth site than the abyssal site if increased organic carbon remineralization enhances dissolution during the low-oxygen events. However, enhanced upwelling of CO2-rich water from the deep Pacific during these events, inferred from boron isotope records in the western subpolar Pacific (Gray et al., 2018) and eND records from these same GoA sites (Du et al., 2018), may also affect calcification and dissolution of foraminifera in the water column. Thus, these GoA records provide the opportunity to test whether CMZ intensification accompanied past OMZ intensification and to determine where in the environment foraminiferal calcification and preservation were primarily affected.

2.2 Measuring SNW

For calculations of SNW, we selected 32 samples from the intermediate-depth site and 26 samples from the abyssal site spanning ∼5.7–20 ka. Each sample was freeze dried, disaggregated in DI water, and wet sieved over a 63 μm sieve. From each sample, we picked 7–10 (mean: 9.4) individuals of N. pachyderma, formerly N. pachyderma var. sinistral (Darling et al., 2006), a planktonic foraminifer present at both sites that calcifies at ∼50 m water depth and adds a thick calcite crust late in ontogeny at ∼50–200 m water depth (Kohfield et al., 1996; Kuroyanagi et al., 2011). At the intermediate-depth site, we also picked 2–10 (mean: 8) individuals of U. peregrina from each sample; this benthic foraminifer calcifies in a shallow infaunal microhabitat (Tachikawa & Elderfield, 2002). All individuals included are complete, unbroken, specimens from the 250 to 125 μm size fraction and were translucent enough when wet to visually confirm they had no infilling sediment. We selected only those N. pachyderma consistent with morphotype Nps-2 of Altuna et al. (2018) and with a crystalline surface texture as in “Group A” of Reynolds and Thunell (1986), which reflects the calcite crust this taxon adds late in ontogeny (Davis et al., 2017; Kohfield et al., 1996). At the intermediate-depth site, both species are present in sufficient abundance, and with sufficient taphonomic condition, in 24 of selected samples; N. pachyderma was analyzed at an average spacing of ∼1 sample per 425 years and U. peregrina was analyzed at ∼1 sample per 500 years. At the abyssal site N. pachyderma was analyzed at a sample spacing of ∼1 sample per 525 years. All sample-level data are included in Table S1.

Given shell weight varies with shell size, weights must be normalized by size to capture changes in shell density and thickness (Beer et al., 2010a). Three methods are typically used to normalize foraminiferal weights by size: sieve-based weights (SBW), measurement-based (MBW), and area-density (ρA). SBW accomplishes size normalization by restricting the specimens that are weighed to a narrow size fraction (Broecker & Clark, 2001; de Villiers, 2004; Pak et al., 2018) and is the least time-consuming procedure. However, SBW does not account for the size variation among individuals within the sieved size fraction. MBW typically normalizes weight using a one-dimensional measurement of the specimens weighted within a narrow size fraction (Barker & Elderfield, 2002; Beer et al., 2010a; Todd et al., 2020), while others normalize weight to the relative area of two-dimensional silhouettes of specimens among samples (Davis et al., 2016). The ρA methods calculate density by dividing the weight of individuals by their silhouette areas and averaging individual densities within samples (Marshall et al., 2013). Both MBW and ρA, however, simplify the geometry of the test although using the three-dimensional volume would be ideal for size normalization of weights (Beer et al., 2010a). Here we use calculate volume density (ρV) by estimating the volume of each specimen contributing to our weight measurements using similar methods as studies of foraminiferal body size (Keating-Bitonti & Payne, 2016; Payne et al., 2012).

We weighed specimens of each species from each sample in aggregate to the nearest 0.1 μg using a Sartorius Ultramicro Balance to determine the total weight (WT). Individual shell dimensions were digitally measured from 2D images of these same specimens using a Nikon SMZ 1500 stereoscope at 30× and the software program NIS Elements BR 2.10 (Nikon Instruments Inc., 2019). To calculate shell volume, we approximated N. pachyderma as a sphere using the average of the semimajor axis and semiminor axis of a hypothetical 2D ellipse that encapsulates the shell as the radius of that sphere (Figure 2a). U. peregrina was similarly measured and we approximated shell volume as a cylinder (Figure 2b). The individual specimen volumes were then summed to determine the total sample volume (VT). Each sample weight was then size normalized to volume using the formula: ρV = WT/VT. We report ρV as ×10−7 μg/μm3 throughout, which is the foraminiferal shell density including the volume of the carbonate and interior spaces.

Details are in the caption following the image

Measurements taken on (a) Neogloboquadrina pachyderma and (b) Uvigerina peregrina. (c) Relationship between the average diameter and average weight per-individual of N. pachyderma in samples from U1419/85JC and U1418/87JC. (d) Relationship between the average length, or average width, and weight per-individual of U. peregrina in sample for U1419/85JC. Filled symbols indicate samples from which computed tomography data were also collected.

2.3 Computed Tomography and Shell Thickness Calculations

CT scanning is costly and time consuming, therefore only 2–4 individuals of each species in selected samples were scanned (Table S2). Individuals from samples with high and low ρV were selected for CT scanning to test the hypothesis that ρV is associated with changes in shell thickness (Figures 3a–3d). For U. peregrina, we used only macrospheric individuals for thickness and volume measurements. In addition, CT scans provide high-quality images of the surface of the test allowing us to assess changes in surface texture and pore sizes expected with dissolution.

Details are in the caption following the image

Time series of morphological and environmental proxy data from the intermediate site (U1419/85JC; red circles) and the abyssal site (U1418/87JC; blue squares). (a) Volume density (ρV) and (b) average volume of N. pachyderma calculated for both sites. (c) ρV and (d) average volume of U. peregrina from U1419/85JC. Filled symbols in (a–d) indicate samples from which computed tomography data on specimens was also collected. Proxies for oxygenation including (e) Mo/Al and (f) U/Al measured on sediments and (g) the proportional abundance of benthic foraminifera with known affinity for dysoxic sediments. (h) The proportional abundance of opportunistic benthic foraminifera, which reflect pulsed influx of phytodetritus to the seafloor. Stable isotope proxies including (i) δ13C and (j) δ18O measured on N. pachyderma. (k) Sea surface temperature derived from alkenone data (Preatorious et al., 2015). (l) Log number of benthic foraminifera per gram of sediment. Data for (e)–(j) and (l) were previously published in Belanger et al. (2020) and here we reproduce these data for only those samples from which we derived ρV values.

All selected individuals were micro-CT scanned at the University of Texas Computed Tomography (UTCT) facility. Twenty-one samples were imaged in 2019 using a Xradia MicroXCT 400 scanner (Table S2). Individuals were scanned at a resolution of ∼1.5 μm. In 2020, an additional 15 samples were imaged at the UTCT facility using Zeiss Versa 620 scanner (Table S2); these were scanned at a resolution of ∼0.04 µm. In most cases, specimens were scanned in pairs although smaller individuals from the same sample were scanned four at a time.

Each scan was cropped to isolate coscanned individuals using ImageJ (Schneider et al., 2012) and reoriented to reflect anatomical axes using the CT image processing software Dragonfly (Object Research Systems Inc., 2020). We then measured the shell thickness of each specimen using the thickness mesh tool in Dragonfly. This tool calculates thickness as the diameter of a hypothetical sphere that fits within the boundary of the shell. The number of points measured on each shell varies (median = 1.36 × 107; interquartile range (IQR) = 7.56 × 106 − 2.42 × 107) depending upon the size of the shell and is determined by the Lapalacian algorithm that converts the voxel-based shell volume to a triangle-based mesh (Object Research Systems Inc., 2020). This thickness mesh tool produced a minimum measurement value of ∼4.48 µm for samples scanned in 2019 and ∼4.32 µm for samples scanned in 2020, which reflects a measurement limit of each micro-CT scanner. To correct for this bias in the thickness distribution of each specimen, each data set was truncated at this lower limit and balanced by removing an equal number of maximum values (median proportion of measurements removed = 0.006, IQR = 0.004–0.028; Table S2), which affects the range of values but not the median. To test for significance differences in median shell thickness, we used pairwise Mann-Whitney tests (wilcox.test in the stats package in R; R Core Team, 2020).

2.4 Test Volumes and Growth Rates

In addition to measuring shell thickness, CT scanning enables us to measure the relative volumes of carbonate shell and interior space in individual foraminifera. We expect that the relative volume of carbonate compared to the total volume of the foraminifer (sum of carbonate and void space volumes) will be lower in specimens with low ρV. In U. peregrina, internal chambers were well preserved in all samples, which allowed us to measure the interior volumes of individual chambers and, thus, also reconstruct the growth history of individuals. In N. pachyderma, internal chambers were often absent, thus we instead measured the total interior volume of these specimens. Volume measurements were done within the CT imaging processing software 3D Slicer 4.10.2 using the segmentation editor with thresholding, level tracing, painting, grow-from-seed, fill between slices, and quantification tools (Fedorov et al., 2012).

We approximate ontogenetic changes in the internal test volume of U. peregrina in each sample using an exponential regression of the relationship between the mean cumulative chamber size and the cumulative number of chambers in megalospheric individuals. We compare the 95% confidence intervals of parameter estimates from the exponential regression to determine if the rate of volume expansion (slope parameter) or initial size (intercept parameter) are distinct between samples. If ρV is related to the rate of increase in internal test volume, we would expect higher ρV to be associated with individuals that have more, smaller volume chambers, and thus slower lower rates of volume expansion and/or initial sizes.

2.5 Predicting SNW from Morphological and Environmental Variables

To determine which environmental factors are most closely associated with changes in ρV, we use multiple regressions in which ρV is the response variable. We use coregistered environmental proxies for oxygenation (Mo/Al, U/Al and the proportional abundance of dysoxia-tolerant benthic foraminifera), organic matter flux (the proportional abundance of opportunistic taxa), and glacial-interglacial conditions (δ13C and δ18O measured on N. pachyderma) previously published in Belanger et al. (2020) as predictor variables. In addition, we include the number of benthic foraminifera per gram of sediment as a predictor; while foraminifera per gram may covary with oxygenation and organic carbon flux and tend to be highest in OMZ environments (Phleger & Soutar, 1973), it also serves as a rough measure of the amount of carbonate available to buffer individual tests from extensive dissolution. Further, we interpolated SST values for 85JC from Praetorius et al. (2015) to use as a predictor variable for multiple regressions of data at the intermediate-depth site. These temperature data were collected at higher resolution than our SNW data and the age models for U1419 and 85JC are well-integrated (Belanger et al., 2020; Praetorius et al., 2015).

We then performed backward-selecting step-wise regressions on the z score transformed environmental variables to determine the best supported combination of predictors from each species at each site based upon the relative Akaike Information Criterion (AIC), which penalizes models with more parameters if information content is not sufficiently increased by the additional parameters (Anderson et al., 2000) All multiple regressions were performed using the lm and stepAIC functions in R (R Core Team, 2020). To ensure the final supported regression models were not affected by collinearity, we calculated the variance inflation factor (VIF) using the function vif in the HH package (Heiberger, 2020). If all VIF values are less than 5, collinearity is not significantly influencing the model (Bowerman & O’Connell. 1990).

3 Results

3.1 Foraminiferal Measurements

Neogloboquadrina pachyderma weighs 2–9 μg per-individual for samples across both sites and have average diameters from ∼172 to 276 μm (Figure 2c). Weight per-individual has a positive, linear, relationship with the diameter of the specimens. Average specimen volume generally declines over time at the intermediate-depth site with local minima at 14.3 and 5.8 ka (Figure 3b). At the abyssal site, average volume is highly variable prior to ∼17 ka and generally declines after ∼13 ka (Figure 3b).

Uvigerina peregrina weighs 5–55 μg per-individual in a given sample (Figure 2d). The sample-level average diameter of U. peregrina ranges from 234 to 401 μm and average length ranges from 250 to 832 μm (Figure 2d). Weight per-individual generally increases with increasing diameter and length (Figure 2d). Average specimen volume has a local maximum at ∼14.1 and ∼10.7 ka and are generally lower during the Holocene than in older portions of the record (Figure 3d).

3.2 Temporal Changes in SNW

For N. pachyderma, ρV ranges from 3.1 to 9.2 × 10−7 μg/μm3 at the intermediate-depth site and from 5.8 to 9.8 × 10−7 μg/μm3 at the abyssal site. In general, ρV is lower at the intermediate-depth site than at the abyssal site after 14.8 kya but is higher at the intermediate site in the older portion of the record (Figure 3a). The lowest ρV values (<5.0 × 10−7 μg/μm3) occur at ∼14.0–14.3, ∼11.0, and 8.0 kya at the intermediate-depth site, but similar decreases do not occur at the abyssal site. For U. peregrina, ρV ranges from 2.71 to 6.38 × 10−7 μg/μm3 at the intermediate-depth site (Figure 3c). The lowest U. peregrina ρV values (<3.00 × 10−7 μg/μm3) occur at ∼14.1 and ∼11.0 kya, coincident with the highest volume shells (Figures 3c and 3d).

3.3 Shell Thickness and Carbonate Volume

Median shell thicknesses of N. pachyderma individuals range from ∼11 to ∼26 μm (Figure S1). Individuals from samples with high ρV have visibly thicker shells (Figure 4) than individuals with lower ρV. Individuals with lower ρV and lower median shell thickness often lack internal chambers and have visually enlarged pores (Figure 4). At the intermediate site, the greatest shell thickness occurs at ∼15 ka, where ρV is highest, and the lowest shell thickness is at ∼14 ka, where ρV is lowest (Figure 5a). Between the Holocene samples, shell thickness is greater at 9.2 ka, where ρV is greater, than at 7.8 ka. At the abyssal site, thicker tests also occur in the sample with higher ρV (Figure 5c). Similarly, the average percentage of carbonate in the total shell volume tends to be higher in samples with thicker tests and greater ρV (Figures 5a and 5c).

Details are in the caption following the image

Computed tomography (CT) scans of N. pachyderma showing differences in shell thickness and preservation. (a and b) Representative specimen from a sample with high ρV (Sample U1419 B 1H 4 W 46–48; 9.3 kyrs). (c and d) Specimen from a sample with low ρV and some loss of internal chambers (Sample U1419 B 1H 5 W 4–6; 14.0 kyrs). (e and f) Specimen from a sample with low ρV and complete loss of internal chambers (Sample U1419 B 1H 5 W 18–20; 14.3 kyrs). Cross sections (b, d, and f) show the greater loss of thinner, internal chamber walls, and greater perforation of external chamber walls in samples with low ρV. Images were produced using ORS Dragonfly v. 2020.1 for Windows (Dragonfly 2020.1).

Details are in the caption following the image

Shell thickness and percentage of carbonate by volume for CT scanned specimens. (a) N. pachyderma from U1419/85JC. (b) U. peregrina from U1419/85JC. (c) N. pachyderma from U1418/87JC. Boxplots depict shell thickness measured at greater than ∼4,000,000 points on each shell; bold horizontal line is the median, box size reflects the interquartile range, whiskers are 1.5× the interquartile range (IQR), and circles are thickness measurements outside 1.5× the IQR. Open squares show the percentage of carbonate by volume for each specimen; filled squares connected by dotted lines show median values for each sample. In (b) 10 is added to the percent carbonate by volume to avoid symbol overlap. All specimens in a given sample are pooled. ρV = volume density; n = number of specimens. For plots of individual specimen thickness, see Figures S1 and S2.

In U. peregrina, median shell thickness per-individual ranges from 13.6 to 24.2 μm (Figure S2). Individuals with lower median thickness tend to have a smoother surface texture with thinner, less prominent, costae. Chamber walls also have pits and holes in the thinnest individuals, although internal chamber walls are present (Figure 6). The thinnest tests with the lowest percentage of carbonate by volume occur at ∼14 ka, where ρV is also low, and the thickest tests with the highest percentage of carbonate by volume occur at 9.2 ka where ρV is high (Figure 5b).

Details are in the caption following the image

Computed tomography (CT) scans of U. peregrina showing differences in shell thickness. (a and b) Representative specimen from a sample with high ρV (Sample U1419 B 1H 4 W 46–48, 9.3 kyrs). (c and d) Specimen from a sample with moderate ρV (Sample U1419 B 1H 3 W 135–138, 7.8 kyrs). (e and f) Specimen from a sample with low ρV (Sample U1419 B 1H 5 W 4–6; 14.0 kyrs). Images were produced using ORS Dragonfly v. 2020.1 for Windows (Dragonfly 2020.1).

3.4 Volume Expansion in U. peregrina

In all analyzed U. peregrina specimens, the initial chamber (proloculus) is larger than the second chamber (Figure 7a), consistent with our selection of only megalospheric individuals. Chambers generally increase in size during ontogeny, however the final chamber is often smaller than the previous chamber.

Details are in the caption following the image

Chamber volumes for Uvigerina peregrina for CT imaged specimens. (a) Volume of individual chambers and (b) cumulative volume of chambers plotted in growth in growth order. In (b) volumes are plotted on a logarithmic scale and lines represent the linear regression of cumulative volume and chamber number derived from all individuals in a given sample. The legend in (b) gives the slope (b = ) with 95% confidence intervals and the R2 value of the linear regression.

Volume expansion in U. peregrina is well described by an exponential curve and model fits for each sample have R2 values > 0.92 (Figure 7b). The intercept parameter for samples from 9.3 and 15.1 ka have overlapping 95% confidence intervals (11.31 ± 0.24 and 11.41 ± 0.1, respectively), thus are statistically identical (Figure 7b). The sample from 7.8 ka has a significantly higher intercept (12.18 ± 0.22) and the sample from 14.0 ka has the highest intercept (13.05 ± 0.1; Figure 7b). This high intercept value is also reflected in the higher proloculus volume of specimens from the 14.0 ka sample (Figure 7a). In contrast, the 95% confidence intervals on the estimated exponential parameter largely overlap among samples, indicating that that rate of volume expansion is similar across samples (Figure 7b).

3.5 Environmental Predictors of ρV

The lowest ρV values at the intermediate-depth site appear contemporaneous with abrupt increases in SST, high values of redox-sensitive metals, and high relative abundances of benthic foraminifera associated with dysoxic conditions and pulsed fluxes of organic matter (Figure 3). Stepwise multiple regression model selection indicates that Mo/Al concentration, the number of foraminifera per gram sediment, and the relative abundances opportunistic benthic foraminifera are well supported predictors of SNW for both N. pachyderma and U. peregrina (Table 1). The relative abundance of benthic foraminifera tolerant of dysoxic conditions and δ13Cpachy are also supported predictors of N. pachyderma ρV whereas U/Al concentration and δ18Opachy are also supported predictors for U. peregrina ρV. All retained predictors have a negative relationship with ρV except for the number of foraminifera per gram (Table 1) and VIF values are less than 5.

Table 1. Parameters for the Best Supported Linear Regression Model Predicting ρV Using Environmental Proxy Data
Site / species Intercept Mo/Al U/Al δ13CN.pachy δ18ON.pachy Benthic foraminifera (#/g) % Opportunistic foraminifera % Dysoxic foraminifera SST R2
U1419/85JC / N. pachyderma −1.18 × 10−16 −0.60** −0.42* 0.30* −0.38* −0.47** 0.82***
U1419/85JC / U. peregrina 3.05 × 10−16 −0.60* −0.51 −0.52* 0.27 −0.71* 0.49*
U1418/87JC / N. pachyderma −0.06 −0.42* −0.38* NA 0.37**
  • Note. * p<0.05, **p<0.01, *** p<0.001.
  • – Indicates parameters that were not included in the best supported model after the stepwise regression procedure.

At the abyssal site, the best supported multiple regression has only the proportion of dysoxia-tolerant benthic foraminifera and the number of benthic foraminifera per gram of sediment as supported predictors. Both retained predictors have a negative relationship with N. pachyderma ρV (Table 1) and vif values are less than 5.

4 Discussion

Multiple factors during the life of a foraminifer and during its residence in the TAZ after death can impact shell weights and thus our ability to reconstruct changes in marine calcification and carbonate dissolution using ρV. In this study, we take a multifactorial approach that incorporates foraminifera from the upper water column and from benthic growth environments and examines foraminifera preserved in sediments from intermediate and abyssal water depths to infer the dominant processes influencing ρV in GoA. We further use independent environmental proxies measured from the same samples as the ρV calculations to determine the environmental factors most supported as predictors of ρV to support our inferences about the drivers of ρV. Thus, unlike previous studies that focus on shell weights from single species or multiple species from similar growth and preservation environments, we can deconvolve potentially confounding factors and quantitatively explore multiple potential drivers of ρV.

We also combine morphological measurements and CT scans of specimens from select samples to test whether ρV reflects relevant shell characteristics. CT scans of N. pachyderma shells from select samples with high and low ρV show that both test thickness and the relative volume of the shells occupied by carbonate are lowest in samples with low ρV. CT scans show that pores are continuous through the outermost shell in even our thickest specimens of N. pachyderma, suggesting that the thick crust is shell material rather diagenetic crust; foraminifers with diagenetic crusts lack regularly spaced pores that manifest on the exterior of the fossil (Branson et al., 2015). Further, foraminiferal shell weights are positively associated with linear shell dimensions for both N. pachyderma and U. peregrina, which suggests that shell size is the primary driver of weight variation across the broader data set. Together, these observations reinforce that ρV primarily reflects changes in the shell itself rather than in relative amounts of infilling or diagenetic encrustation. The strong relationship between shell size and weight even within narrow size fractions (such as 200–250 µm; Figure 2) also indicates that we must use measured sizes rather than sieve-based sizes before comparing weights among samples for these species.

4.1 Site Comparisons Suggest ρV is not Related to Water Column Conditions

By examining N. pachyderma morphology, SNW, and carbonate content in fossil samples from two sites at different water depths that are influenced by different water masses, we can address whether differences in SNW primarily reflect growth processes in upper water column during life or taphonomic processes after death. Given carbonate sedimentation at high latitudes is largely driven by N. pachyderma (Huber et al., 2000), changes in calcification and dissolution of this taxon can have a significant effect on carbonate burial.

While ρV in N. pachyderma oscillates around a relatively stable mean of 7.5 × 10−7 μg/μm3 at the abyssal site, ρV at the intermediate-depth site declines from its highest values in samples older than ∼15 ka to lower values in samples younger than ∼7 ka (Figure 3a). At the abyssal site, ρV is also less variable than at the intermediate-depth site and we only observe the extreme lows (<5.0 × 10−7 μg/μm3) in ρV at ∼11 ka and ∼14 ka at the intermediate-depth site (Figure 3a). However, N. pachyderma has a similar range of shell volumes in samples from both intermediate and abyssal depths which decline from glacial times through the Holocene at both sites (Figure 3b), suggesting that the sites sample morphologically similar populations. Further, CT scans of N. pachyderma shells show both the inner shell layer and the thicker outer “calcite crust” layer (Figure 4), indicating that individuals are also of similar life stage (Davis et al., 2017; Jonkers et al., 2016; Kohfeld et al., 1996). Thus, the differences in ρV in N. pachyderma we observe between sites are not dominated by differences in size or ontogenetic stage.

Alternatively, local differences in water column chemistry between the sites could impact ρV via their influence on calcification and dissolution as dead shells settle through the water column. In N. pachyderma (sinistral) from polar areas, calcification experiments demonstrate a positive correlation between calcification and Ω (Manno et al., 2012). Calcification in N. pachyderma is particularly sensitive to low pH compared to other planktonic foraminifera, perhaps because it is a nonsymbiotic species (Lombard et al., 2010). However, in the modern, the surface conditions at each site do not differ significantly in Ω, suggesting that calcification conditions are unlikely to explain the ρV differences between sites. Some estimate that 60–80% of calcium carbonate dissolution occurs in the upper ocean from 500 to 1,000 m depth (Milliman et al., 1999), thus the time of exposure to corrosive waters could play a role given the ∼3,000 m depth difference between the sites. However, others suggest that planktonic foraminifera sink too quickly for significant dissolution to occur prior to reaching the seafloor (Jansen et al., 2002) and we would expect that, if dissolution occurred primarily in the water column, the abyssal site would have relatively lower ρV, counter to the better preservation and higher ρV we observe at the abyssal site compared to the intermediate-depth site. Thus, despite boron isotope records from planktonic foraminiferal calcite that suggest a decrease in upper water column pH occurred during the B/A in the North Pacific (Gray et al., 2018), it does not appear that dissolution in the water column dominates dissolution in these records.

4.2 Covariation in Benthic and Planktonic ρV Suggest Seafloor Dissolution

Benthic and planktonic foraminifera live in fundamentally different environments but occur in the same sediments after death. Thus, by comparing temporal patterns in ρV between benthic and planktonic foraminifera from the same sediments, we can test whether changes in ρV are primarily controlled by shared seafloor conditions after death. In our GoA record, lows in ρV at the intermediate-depth site occur contemporaneously in N. pachyderma and U. peregrina at ∼11 and ∼14 ka, suggesting a common driver at the seafloor consistent with dissolution in the sediments or near the sediment-water interface. Dissolution is especially apparent in N. pachyderma in which internal chambers are absent in addition to test thinning. U. peregrina specimens also have smoother test surfaces and missing portions of chambers walls that suggest dissolution. Interestingly, a local low in average N. pachyderma shell volume corresponds with a low in ρV at ∼14 ka at the intermediate-depth site (Figure 3). This aberration may reflect greater fragmentation and loss of larger individuals as dissolution intensifies (Berger, 1970; Todd et al., 2020). The covariation of temporal patterns in ρV between benthic and planktonic species at the intermediate-depth site and the less intense dissolution at the abyssal site both support that dissolution at the seafloor is the dominant control of ρV in these GoA records.

4.3 Low ρV is Associated with Pulsed Organic Matter Export and OMZ Intensification

Increased carbonate dissolution in sedimentary pore waters can be driven by increased organic carbon remineralization because it produces carbonic acid (Archer et al., 1989; Boudreau & Canfield, 1993; de Villiers, 2005) and some models estimate that sedimentary organic carbon respiration accounts for the majority of carbonate dissolution below 1,000 m (Dunne et al., 2012). Organic carbon rain that reaches the shallower seafloor experiences less water column remineralization, thus the abyssal seafloor should receive less labile organic matter (Sarmiento & Gruber, 2006). Thus, in a system where ρV is primarily driven by organic carbon remineralization at the seafloor, we would expect intermediate depths to have lower ρV than abyssal depths, consistent with the lower N. pachyderma ρV values we observe at intermediate depths within the upper OMZ. Similarly, we would expect the most intense dissolution to occur during OMZ intensification, if increased organic carbon mineralization also underlies the decline in benthic oxygenation. Indeed, lows in ρV are visually associated with increases in the concentration of redox-sensitive metals and in the proportional abundance of benthic foraminifera that are tolerant of dysoxic conditions at the intermediate-depth site at ∼11 and ∼14 ka (Figure 3). In addition, ρV is generally lower in the Holocene portion of the record, when opportunistic benthic foraminifera are most abundant and analyses of the whole assemblage indicate suboxic conditions, in contrast to the glacial portion of the record where opportunists are less abundant and assemblages suggest better-oxygenated conditions (Figure 3; Belanger et al., 2020; Sharon et al., 2021). Further, these dissolution events do not correspond with large differences in benthic and planktonic foraminiferal 14C ages (Walczak et al., 2020), suggesting increased subsurface water mass age is not causing the increased corrosivity. The association among low ρV, high abundances of opportunistic foraminifera, and low-oxygen conditions suggests remineralization of organic matter at the seafloor is the dominant control of dissolution intensity.

We quantitatively test this qualitative association between ρV and seafloor oxygenation by comparing the relative support among multiple regression models with alternative environmental predictors for ρV. In all three ρV records we collected for the GoA, the best supported models of ρV include proxies of low-oxygen conditions as predictor variables (Table 1). Mo/Al, which is indicative of sulfidic conditions (Crusius et al., 1996; Zheng et al., 2000), is the strongest correlate with ρV for both species at the intermediate-depth site indicating a strong relationship between low ρV and the lowest oxygen conditions. At the abyssal site, where suboxic conditions develop contemporaneously with the sulfidic conditions at the intermediate site (Belanger et al., 2020), the proportion of dysoxia-tolerant benthic foraminifera is a supported correlate with ρV despite their relatively low abundances; this demonstrates that even small, or brief, decreases in oxygenation are associated with increased dissolution. Further, at the intermediate-depth site, high abundances of opportunistic foraminifera are associated with low ρV in the best supported model, as expected if remineralization of organic matter delivered to the seafloor in pulses is contributing to both the corrosive and the low-oxygen conditions. SST is not retained as a supported predictor of ρV, which suggests that while low ρV occurs during warmer times (Figure 3), the proximal driver of ρV is more closely associated with variations in oxygenation, as expected if changes in organic carbon remineralization are driving both changes. This further supports that, where intensification of the OMZ is driven by increases in organic carbon remineralization at intermediate depths, intensification of dissolution will also occur. Increased dissolution of planktonic foraminiferal tests is also observed during productivity-driven expansions of the Arabian Sea OMZ (Mungekar et al., 2020; Naik et al., 2014), thus this phenomenon is not unique to GoA.

Changes in global oceanic carbon budgets associated with the glacial-interglacial transition the GoA records span, rather than locally driven conditions, can also influence ρV. Interglacial times in the equatorial Pacific are generally associated with greater carbonate dissolution, evinced by the artificially high relative abundances of dissolution-resistant planktonic foraminifera and lower shell weights in planktonic foraminifera such as N. dutertrei (Farrell & Prell, 1989; Mleneck-Vautravers, 2018; Thompson & Saito, 1974), consistent with the somewhat lower ρV values in the Holocene than in the glacial that we observe in GoA. However, highs in CaCO3 content in sediments are often centered on transitions from glacials to interglacials (Farrell & Prell, 1989) whereas in this GoA record ρV lows occur in the deglacial and the early Holocene. Thus, these known glacial-interglacial preservation cycles do not appear to dominate ρV in our GoA record. Further, while δ18ON.pachy was retained as a predictor in the best supported model of U. peregrina ρV, δ18ON.pachy has a negative relationship with ρV. This could imply that warmer, or fresher, conditions contribute to better preservation when other environmental predictors are simultaneously considered. However, this is counter to the expectation of greater dissolution due to the introduction of fresh meltwater to the site during the B/A and early Holocene (Praetorius et al., 2020), which can reduce Ω due to the lower availability of Ca2+ (Evans et al., 2014). Alternatively, δ18ON.pachy may reflect changes in upwelling, given N. pachyderma's preference for upwelling conditions in the modern GoA (Reynolds & Thunell, 1986), and thus the negative relationship between δ18ON.pachy and U. peregrina ρV may indicate poorer preservation during upwelling of cooler waters with lower Ω in addition to the effects of higher organic carbon remineralization.

Changes in sedimentation rate may also affect our ρV records. Low sedimentation rates would keep foraminiferal shells in the TAZ for longer, enhancing taphonomic loss in corrosive environments. However, if in situ carbonate production is high, low sedimentation can also lead to higher concentrations of carbonate available to buffer corrosive pore waters, which would, in turn, decrease the amount of dissolution experienced by any one individual. In the multiple regression, we used the number of benthic foraminifera per gram of sediment, which reflects sedimentation rates, the rate of production of new benthic foraminifera, and the amount of taphonomic loss, as a potential predictor of ρV. At the intermediate-depth site, the concentration of benthic foraminifera in the sediments is retained in the best supported model where it is positively related to ρV. This positive relationship suggests that the high availability of carbonate grains decreased per-individual carbonate loss and is apparent even though benthic foraminiferal concentrations are high when faunal and geochemical proxies indicate oxygen is lowest (Figure 3). At the abyssal site, however, the number of foraminifera per gram of sediment is negatively associated with ρV suggesting that extended exposure to corrosive waters during times of lower sedimentation rates may play a role in the extent of dissolution at that site.

4.4 The Role of Growth in ρV

We also consider the alternative hypothesis that environmentally driven differences in growth rates account for the association between ρV and proxies for low-oxygen and high-organic flux conditions. N. pachyderma growth rates vary with temperature and tend to be highest at low temperatures (∼5 °C), unlike other planktonic species where growth rate increases with temperature (Lombard et al., 2010). In other planktonic foraminifera, such as Globigerina bulloides, fast growth can lead to thinner tests (Pak et al., 2018), however in N. pachyderma up to 90% of its shell carbonate is added late in ontogeny as a gametogenic crust (Kohfield et al., 1996; Davis et al., 2017), which limits the potential for growth rates to influence final shell weights. Further, low ρV is also not consistently associated with larger volume shells in the N. pachyderma record as we might expect if low ρV was driven by fast growth (Figure 3). Thus, we do not see evidence for growth-driven ρV in N. pachyderma, The U. peregrina samples with the greatest shell diameters and the greatest shell lengths have moderate shell weights and are outliers in the general trend of increasing weight with increasing size (Figure 2d). These samples, predictably, have the lowest ρV values. This indicates that, for U. peregrina, there is an association between morphological differences and ρV. Size and growth rate in marine benthic organisms can vary along oxygen gradients and will increase or decrease with oxygenation depending on the physiological preferences of the species (Belanger et al., 2020; Glock et al., 2019; Keating-Bitonti & Payne, 2017). In U. peregrina from GoA, shell volumes are greatest where Mo/Al and the relative abundance of dysoxia-tolerant foraminifera are highest (Figure 3), similar to results from SBB (Davis et al., 2016), which indicates an association between larger U. peregrina shells and lower oxygen environments. Foraminifera capable of using nitrate as their electron acceptor in metabolic reactions may grow faster in low-oxygen environments and Uvigerina spp. do store nitrate and perform denitrification (Glock et al., 2019; Piña-Ochoa et al., 2010). However, our analysis of ontogenetic changes in individual chamber volumes of U. peregrina in the present study show no evidence for significant differences in the rate of test expansion among samples. Instead, U. peregrina specimens from the low-oxygen, low-ρV sample for which we have test volume data has a significantly larger proloculus than specimens from the other three samples, suggesting that growth from a larger initial embryonic size compounded to result in larger size at death. Thus, while changes in growth conditions in the upper water column did not dominate ρV in N. pachyderma, morphological differences related to changes in the benthic environment may have enhanced the differences in ρV we record in U. peregrina.

4.5 Implications for OMZs as a CO2 Source

OMZs can become local sources of CO2 to the atmosphere due to their high DIC concentrations if vertical mixing intensifies (Palmier et al., 2011). The addition of DIC to the intermediate-depth ocean by carbonate dissolution at the seafloor would have enhanced the intensity of the CMZ in GoA. This contribution of intermediate-depth DIC by dissolution may have played an important role in maintaining CO2 outgassing from the North Pacific during the B/A, which was driven by upwelling and overturning mechanisms that brought CO2-rich waters to the surface ocean (Du et al., 2018; Gray et al., 2018). Including the effect of enhanced organic carbon export on carbonate dissolution in models of the global carbon cycle may help account for the ∼80 ppm rise in CO2 during deglaciation (Marcott et al., 2014). If modern increases in organic carbon export contribute to carbonate dissolution at the seafloor as they did in the deglacial, this mechanism could enhance the potential of OMZs to affect future global climates.

5 Conclusions

Understanding the mechanisms underlying paleoceanographic records of dissolution is essential for determining the role dissolution plays in oceanic carbon storage and regulation of the global carbon cycle. We combine benthic and planktonic foraminiferal ρV records from intermediate and abyssal depths with independent environmental proxy data to recognize past dissolution events, determine the primary environment in which dissolution occurred, and differentiate among the potential drivers of changing carbonate preservation. We find that low-ρV in GoA is primarily driven by dissolution at the seafloor and occurs prominently at intermediate depths where faunal and geochemical proxies indicate dysoxic conditions, supporting a common role for organic carbon export in both the enhanced dissolution and decreased benthic oxygenation in the upper OMZ. Thus, these ρV records do not reflect acidification of the upper water column, but instead emphasize the importance of metabolically driven dissolution related to increased organic carbon export and its subsequent remineralization in seafloor environments to carbonate preservation. This dissolution mechanism may become increasingly important to understanding the modern role of OMZs as CO2 stores and sources because organic carbon export to the deep sea is predicted to increase in high latitudes with ongoing global climate change (Sweetman et al., 2017). Our results further caution against using paleoceanographic records of dissolution to infer upper water column chemistry during times of high and changing organic carbon export, however we show that differentiating among dissolution mechanisms is possible with a multitaxon and multisite approach. Recognizing these different dissolution processes is important for interpreting changes in the inorganic carbon cycle from paleoceanographic records as well as for forecasting the role OMZs and high productivity regions will have in future CO2 rise.

Acknowledgments

This work was supported by NSF 1502746 and 1801511 to Christina L. Belanger. We would like to thank J. Maisano and M. Colbert at the University of Texas High-Resolution X-ray Computed Tomography Facility for performing the CT scans and teaching the authors how to analyze the images. We also thank D. Bapst at TAMU for assistance with R analyses.

    Data Availability Statement

    Data are archived DRYAD at https://doi.org/10.5061/dryad.b8gtht7c3. CT images are archived at https://www.morphosource.org/projects/000346774.