Seawater pH reconstruction using boron isotopes in multiple planktonic foraminifera species with different depth habitats and their potential to constrain pH and pCO2 gradients

Abstract. Boron isotope systematics of planktonic foraminifera from core-top sediments and culture experiments have been studied to investigate the sensitivity of δ11B of their calcite tests to seawater pH. However, our knowledge of the relationship between δ11B and pH remains incomplete for several taxa. Thus, to expand the potential scope of application of this proxy, we report data for 7 different species of planktonic foraminifera from sediment core-tops. We utilize a method for the measurement of small samples of foraminifera and calculate the δ11B-calcite sensitivity to pH for Globigerinoides ruber, Trilobus sacculifer (sacc or w/o sacc), Orbulina universa, Pulleniatina obliquiloculata, Neogloboquadrina dutertrei, Globorotalia menardii and Globorotalia tumida, including for unstudied coretops and species. The sensitivity of δ11Bcarbonate to δ11Bborate (eg. Δδ11Bcarbonate/Δδ11Bborate) in core-tops is close to unity. Deep-dwelling species closely follow the core-top calibration for O. universa, which is attributed to respiration-driven microenvironments, likely caused by light limitation for symbiont-bearing foraminifera. These taxa have diverse ecological preferences and are from sites that span a range of oceanographic regimes, including some that are in regions of air-sea equilibrium and others that are out of equilibrium with the atmosphere. Our data support the premise that utilizing boron isotope measurements of multiple species within a sediment core can be utilized to constrain vertical profiles of pH and pCO2 at sites spanning different oceanic regimes, thereby constraining changes in vertical pH gradients and yielding insights into the past behavior of the oceanic carbon pump.


Seawater pH reconstruction using boron isotopes in multiple planktonic foraminifera species with different depth habitats and their potential to constrain pH and pCO(2) gradients Maxence Guillermic, Sambuddha Misra, Robert Eagle, Alexandra Villa,
Knowledge of seawater pH, in conjunction with constraints on one other carbonate system parameter (total alkalinity (TA), DIC (dissolved inorganic carbon), , [CO 2− 3 ]), can be utilized to constrain aqueous pCO 2 . Application of empirical calibrations for boron isotope ratio, determined for select species of foraminifera from core tops and laboratory cultures, has resulted in accurate reconstructions of pCO 2 utilizing downcore samples from sites that are currently in quasi-equilibrium with the atmosphere at present. Values of pCO 2 reconstructed from planktonic foraminifera boron isotope ratios are analytically indistinguishable from ice core CO 2 records (Foster et al., 2008;Henehan et al., 2013;Chalk et al., 2017).
The last decade has produced several studies aiming at reconstructing past seawater pH using boron isotopes to constrain atmospheric pCO 2 in order to understand the changes in the global carbon cycle (Hönisch et al., 2005(Hönisch et al., , 2009Foster et al., 2008Foster et al., , 2012Foster et al., , 2014Seki et al., 2010;Bartoli et al., 2011;Henehan et al., 2013;Martínez-Boti et al., 2015a, b;Chalk et al., 2017). In addition to reconstructing atmospheric pCO 2 , the boron isotope proxy has been applied to mixedlayer planktonic foraminifera at sites out of equilibrium with the atmosphere to constrain past air-sea fluxes Martínez-Boti et al., 2015b). A small body of work has examined whether data for multiple species in core-top (Foster et al., 2008) and down-core samples could be used to constrain vertical profiles of pH through time (Palmer et al., 1998;Pearson and Palmer, 1999;Anagnostou et al., 2016).
Here we add to the emerging pool of boron isotope data in planktonic foraminifera from different oceanographic regimes, including data for species that have not previously been examined. We utilize a low-blank (15 to 65 pg B), high-precision (2 SD on the international standard JCp-1 is 0.20 ‰, n = 6) δ 11 B carbonate analysis method for small samples (down to ∼ 250 µg CaCO 3 ), modified after Misra et al. (2014a), to study multiple species of planktonic foraminifera. The studied sediment core tops span a range of oceanographic regimes, including open-ocean oligotrophic settings and marginal seas. We constrain calibrations for different species, and we compare results to published work (Foster et al., 2008;Henehan et al., 2013Henehan et al., , 2016Martínez-Boti et al., 2015b;Raitzsch et al., 2018). We also test whether these data support the application of boron isotope measurements of multiple species within a sediment core as a proxy for constraining vertical profiles of pH and pCO 2 .

Planktonic foraminifera as archives of seawater pH
Planktonic foraminifera are used as archives of past environmental conditions within the mixed layer and thermocline, as their chemical composition is correlated with the physiochemical parameters of their calcification environment (Ravelo and Fairbanks, 1992;Elderfield and Ganssen, 2000;Dekens et al., 2002;Anand et al., 2003;Sanyal et al., 2001;Ni et al., 2007;Henehan et al., 2013Henehan et al., , 2015Henehan et al., , 2016Howes et al., 2017;Raitzch et al., 2018). The utilization of geochemical data for multiple planktonic foraminifera species with different ecological preferences to constrain vertical gradients has been explored in several studies. The framework for such an approach was first developed using modern samples of planktonic foraminifera for oxygen isotopes, where it was proposed as a tool to constrain vertical temperature gradients and study physical oceanographic conditions during periods of calcification (Ravelo and Fairbanks, 1992).
Because planktonic foraminifera species complete their life cycle in a particular depth habitat due to their ecological preference (Ravelo and Fairbanks, 1992;Farmer et al., 2007), it is theoretically possible to reconstruct water column profiles of pH using boron isotope ratio data from multiple taxa (Palmer and Pearson, 1998;Anagnostou et al., 2016). The potential use of an analogous approach to reconstruct past profiles of seawater pH was first highlighted by Palmer and Pearson (1998) on Eocene samples to constrain water depth pH gradients. However, in these boron isotope-based studies, it was assumed that boron isotope offset from seawater and foraminiferal carbonate was constant, which is an assumption not supported by subsequent studies (e.g., Hönisch et al., 2003;Foster et al., 2008;Henehan et al., 2013Henehan et al., , 2016Raitszch et al., 2018;. Furthermore, boron isotope ratio differences between foraminifera species inhabiting waters of the same pH make the acquisition of more core-top and culture data essential for applications of the proxy.

Boron systematics in seawater
Boron is a conservative element in seawater with a long residence time (τ B ∼ 14 Myr) (Lemarchand et al., 2002). In seawater, boron exists as trigonal boric acid B(OH) 3 and tetrahedral borate ion B(OH) − 4 (borate). The relative abundance of boric acid and borate ions is a function of the ambient seawater pH. At standard open-ocean conditions (T = 25 • C and S = 35), the dissociation constant of boric acid is 8.60 (Dickson, 1990), implying that boron mainly exists in the form of boric acid in seawater. The fact that the pK B and seawater pH (e.g., ∼ 8.1, NBS) values are similar implies that small changes in seawater pH will induce strong variations in the abundance of the two boron species (Fig. 1).
Boron has two stable isotopes, 10 B and 11 B, with average relative abundances of 19.9 % and 80.1 %, respectively. Variations in B isotope ratio are expressed in conventional delta (δ) notation: where positive values represent enrichment in the heavy isotope 11 B and negative values enrichment in the light isotope 10 B, relative to the standard reference material. Boron isotope values are reported versus the NIST SRM 951 boric acid standard (Cantazaro et al., 1970). B(OH) 3 is enriched in 11 B compared to B(OH) − 4 with a constant offset between the two chemical species, within the range of physiochemical variation observed in seawater, given by the fraction factor (α). The fractionation (ε) between B(OH) 3 and B(OH) − 4 of 27.2 ± 0.6 ‰ has been empirically determined by Klochko et al. (2006) in seawater. Note that Nir et al. (2015) calculate this fractionation, using an independent method, to be 26 ± 1 ‰, which is within the analytical uncertainty of the Klochko et al. (2006) value. We use a fractionation of 27.2 ‰ determined by Klochko et al. (2006) in this study.
At present, culture and core-top calibrations have been published for several planktonic species including Trilo-batus sacculifer, Globigerinoides ruber, Globigerina bulloides, Neogloboquadrina pachyderma, and Orbulina universa (Foster et al., 2008;Henehan et al., 2013Henehan et al., , 2015Sanyal et al., 1996Sanyal et al., , 2001. Although the boron isotopic composition of several species of foraminifera is now commonly used for reconstructing surface seawater pH, for other species, there is a lack of data constraining the sensitivity of boron isotopes in foraminiferal carbonate and borate ions in seawater.

Origin of biological fractionations in foraminifera
Perforate foraminifera are calcifying organisms that maintain a large degree of biological control over their calcification space, and thus mechanisms of biomineralization may be of significant importance in controlling the δ 11 B of the biogenic calcite. The biomineralization of foraminifera is based on seawater vacuolization (Erez, 2003;de Nooijer et al., 2014) with parcels of seawater being isolated by an organic matrix, thereby creating a vacuole filled with seawater. Recent work has also demonstrated that even if the chemical composition of the reservoirs is modified by the organism, seawater is directly involved in the calcification process with vacuoles formed at the periphery of the shell (de Nooijer et al., 2014). Culture experiments by Rollion- Bard and Erez (2010) have proposed that the pH at the site of biomineralization is elevated to an upper pH limit of ∼ 9 for the shallow-water, symbiont-bearing benthic foraminifera Amphistegina lobifera, which would support a pH modulation of a calcifying fluid in foraminifera. The extent to which these results apply to planktonic foraminifera is not known, although pH modulation of calcifying fluid may influence the δ 11 B of planktonic foraminifera.
δ 11 B in foraminifera is primarily controlled by seawater pH but also depends on the pH alteration of microenvironments due to calcification, respiration, and symbiont photosynthesis. δ 11 B carbonate should therefore reflect the relative dominance of these processes and may account for speciesspecific δ 11 B offsets. Theoretical predictions from  and foraminiferal data from Hönisch et al. (2003) explored the influence of microenvironment pH in the δ 11 B signature of foraminifera. Their work also suggested that for a given species there should be a constant offset observed  Foster et al., 2010), and dissociation constant α = 1.0272 (Klochko et al., 2006). between the boron isotope composition of foraminifera and borate ions over a large range of pH, imparting confidence in utilizing species-specific boron isotope data as a proxy for seawater pH.
Comparison of boron isotope data for multiple planktonic foraminiferal species indicates that taxa with high levels of symbiont activity such as T. sacculifer and G. ruber show higher δ 11 B values than the δ 11 B of ambient borate (Foster et al., 2008;Henehan et al., 2013;Raitzsch et al., 2018). The sensitivities δ 11 B carbonate / δ 11 B borate (hereafter referred to as the slope) of existing calibrations suggest a different species-specific sensitivity for these species compared to other taxa (Sanyal et al., 2001;Henehan et al., 2013Henehan et al., , 2015Raitzsch et al., 2018). For example, Orbulina universa exhibits a lower δ 11 B than in situ δ 11 B values of borate ions (Henehan et al., 2016), consistent with the species living deeper in the water column characterized by reduced photosynthetic activity.
It is possible that photosynthetic activity by symbionts might not be able to compensate for changes in calcification and/or respiration, leading to an acidification of the microenvironment. It is interesting to note that for O. universa the slope determined for the field-collected samples is not statistically different from unity (0.95 ± 0.17) (Henehan et al., 2016), while culture experiments report slopes of ≤ 1 for multiple species including G. ruber (Henehan et al., 2013), T. sacculifer (Sanyal et al., 2001), and O. universa (Sanyal et al., 1999). More core-top and culture calibrations are needed to refine those slopes and understand if significant differences are observed, which is part of the motivation for this study.

Planktic foraminifera depth and habitat preferences
The preferred depth habitat of different species of planktonic foraminifera depends on their ecology, which in turn is dependent on hydrographic conditions. For example, G. ruber is commonly found in the mixed layer (Fairbanks and Wiebe, 1980;Dekens et al., 2002;Farmer et al., 2007) during the summer (Deuser et al., 1981), whereas T. sacculifer is present in the mixed layer until mid-thermocline depths (Farmer et al., 2007) during spring and summer (Deuser et al., 1981(Deuser et al., , 1989. Specimens of P. obliquiloculata and N. dutertrei are abundant during winter months (Deuser et al., 1989), with an acme in the mixed layer (∼ 60 m) for P. obliquiloculata and at mid-thermocline depths for N. dutertrei (Farmer et al., 2007). In contrast, O. universa tends to record annual average conditions within the mixed layer. Specimens of G. menardii calcify within the seasonal thermocline (Fairbanks et al., 1982;Farmer et al., 2007;Regenberg et al., 2009), and in some regions in the upper thermocline (Farmer et al., 2007), and record annual temperatures. G. tumida is found at the lower thermocline or below the thermocline and records annual average conditions (Fairbanks and Wiebe, 1980;Farmer et al., 2007;Birch et al., 2013). Although the studies listed above showed evidence for species-specific living depth habitat affinities, recent direct observations showed that environmental conditions (e.g., temperature, light) were locally responsible for the variability in the living depth of certain foraminifera species in the eastern North Atlantic (Rebotim et al., 2017).  Thomson et al., 2000. b Wilson et al., 2012. c Age for core top of site 806B from Lea et al. (2000). . Red open circles represent the sites used for in situ carbonate parameters from the GLODAP database (Key et al., 2004).

Localities studied
Core-top locations were selected to span a broad range of seawater pH, carbonate system parameters, and oceanic regimes. Samples from the Atlantic Ocean (CD107-A), Indian Ocean (FC-01a and FC-02a), Arabian Sea (FC-13a and FC-12b), and Pacific Ocean (WP07-01, A14, and Ocean Drilling Program 806A and 807A) were analyzed; characteristics of the sites are summarized in Tables 1 and S7 and Figs. 2 and 3. Atlantic site CD107-a (CD107 site A) was cored in 1997 by the Benthic Boundary Layer program (BENBO) (Black, 1997 -cruise report RRS Charles Darwin cruise 107). Arabian Sea sites FC-12b (CD145 A150) and FC-13a (CD145 A3200) were retrieved by the Charles Darwin in the Pak-istan margin in 2004 (Bett, 2004 -cruise report no. 50 RRS Charles Darwin cruise 145). A14 was recovered by a box corer in the southern area of the South China Sea in 2012. Core WP07-01 was obtained from the Ontong Java Plateau using a giant piston corer during the Warm Pool Subject Cruise in 1993. Holes 806A and 807A were retrieved on Leg 130 by the Ocean Drilling Program (ODP). The top 10 cm of sediment from CD107-A has been radiocarbon dated to be Holocene <3 kyr (Thomson et al., 2000). Samples from multiple box cores from Indian Ocean sites were radiocarbon dated as Holocene <7.3 kyr (Wilson et al., 2012). Samples from western equatorial Pacific site 806B, close to site WP07-01, are dated to between 7.3 and 8.6 kyr (Lea et al., 2000). Arabian Sea and Pacific core-top samples were not radiocarbon dated but are assumed to be Holocene.

Sample cleaning
Briefly, picked foraminifera were gently cracked open, clay was removed with successive ultrasonication steps in MQ water and methanol, and then they were checked for coarsegrained silicates. The next stages of sample processing and chemical separation were performed in a class 1000 clean lab equipped with boron-free HEPA filters. Samples were cleaned using full reductive and oxidative cleaning (Boyle, 1981;Boyle and Keigwin, 1985;Barker et al., 2003). Samples from the South China Sea (sites A14, E035) presented high Mn and high Fe. Due to potential Fe-Mn oxide and hydroxides the reductive cleaning was used. Previous comparisons of cleaning methods have shown there is no impact of the reductive step on B/Ca (Misra et al., 2014b), but there is an impact of the reductive step on Mg/Ca (Barker et al., 2003 and others). Nevertheless, it is possible that Fe-Mn oxide and hydroxides can result in non-negligible Mg and B contamination. Because this study was designed to investigate boron proxies and in order to be consistent in methodology, the reductive cleaning was used at all sites. Cleaned samples selected for this study did not yield high Mn concentrations (see Supplement for discussion on contamination).
A final leaching step with 0.001 N HCl was done before dissolution in 1 N HCl. Hydrochloric acid was used to allow complete dissolution of the sample including Fe-Mn oxide and hydroxides if present. Each sample was divided into two aliquots: an aliquot for boron purification and one aliquot for trace element analysis.

Reagents
Double-distilled HNO 3 and HCl acids (from Merck ® grade) and a commercial bottle of ultrapure-grade HF were used at Brest. Double-distilled acids were used at Cambridge. All acids and further dilutions were prepared using doubledistilled 18.2 M cm −1 MQ water. Working standards for isotope ratio and trace element measurements were freshly diluted on a daily basis with the same acids used for sample preparation to avoid any matrix effects.

Boron isotopes
Boron purification for isotopic measurement was done utilizing the microdistillation method developed by Gaillardet et al. (2001) for Ca-rich matrices by Wang et al. (2010) and adapted at Cambridge by Misra et al. (2014a). A total of 70 µL of carbonate sample dissolved in 1 N HCl was loaded on a cap of a clean fin legged 5 mL conical beaker upside down. The tightly closed beaker was put on a hot plate at 95 • C for 15 h. The beakers were taken off the hot plate and were allowed to cool for 15 min. The cap where the residue formed was replaced by a clean one. Then, 100 µL of 0.5 % HF was added to the distillate.
Boron isotopic measurements were carried out on a Thermo Scientific ® Neptune+ MC-ICP-MS at the University of Cambridge. The Neptune+ was equipped with a Jet interface and two 10 13 resistors. The instrumental setup included Savillex ® 50 µL min −1 C-flow self-aspirating nebulizer, a single-pass Teflon ® Scott-type spray chamber constructed utilizing Savillex ® column components, a 2.0 mm Pt injector from ESI ® , a Thermo ® Ni "normal"-type sample cone, and 'X' type skimmer cones. Both isotopes of boron were determined utilizing 10 13 resistors (Misra et al., 2014a;Lloyd et al., 2018).
The sample size for boron isotope analyses typically ranged from 10 ppb B (∼ 5 ng B) to 20 ppb B samples (∼ 10 ng B). Instrumental sensitivity for 11 B was 17 mV ppb −1 B (e.g., 170 mV for 10 ppb B) in wet plasma at a 50 µL min −2 sample aspiration rate. Intensity of 11 B for a sample at 10 ppb B was typically 165 mV ± 5 mV, which closely matched the 170 mV ± 5 mV of the standard. Due to the low boron content of the samples, extreme care was taken to avoid boron contamination during sample preparation and reduce memory effect during analysis. Procedural boron blanks ranged from 15 to 65 pg B and contributed to less than <1 % of the sample signal. The acid blank during analyses was measured at ≤ 1 mV on 11 B, meaning a contribution <1 % of the sample intensity; no memory effect was observed within and across sessions. No matrix effect resulting from the mix HCl/HF was observed on the δ 11 B.

Trace elements
The calcium concentration of each sample was measured on an inductively coupled plasma atomic emission spectrome-ter (ICP-AES) Ultima 2 HORIBA at the Pôle spectrometrie Océan (PSO), UMR6538 (Plouzané, France). Samples were then diluted to fixed calcium concentrations (typically 10 ppm or 30 ppm Ca) using 0.1 M HNO 3 and 0.3 M HF matching multielement standard Ca concentration to avoid any matrix effects (Misra et al., 2014b). Levels of remaining HCl (<1 %) in these diluted samples were negligible and did not contribute to matrix effects. Trace elements (e.g., X/Ca ratios) were analyzed on a Thermo Scientific ® Element XR high-resolution inductively coupled plasma mass spectrometer (HR-ICP-MS) at the PSO, Ifremer (Plouzané, France).
Trace element analyses were done at a Ca concentration of 10 or 30 ppm. The typical blanks for a 30 ppm Ca session were 7 Li <2 %, 11 B <7 %, 25 Mg <0.2 %, and 43 Ca <0.02 %. Additionally, blanks for a 10 ppm Ca session were 7 Li <2.5 %, 11 B <10 %, 25 Mg <0.4 %, and 43 Ca <0.05 %. Due to strong memory effect for boron and instrumental drift on the Element XR, long sessions of conditioning were done prior to analyses. Boron blanks were driven below 5 % of signal intensity usually after 4 to 5 d of continuous analyses of carbonate samples. External reproducibility was determined on the consistency standard Cam-Wuellestorfi (courtesy of the University of Cambridge) (Misra et al., 2014b), Table S3. Our X/Ca ratio measurements on the external standard Cam-Wuellestorfi were within error of the published value all the time (Table S3), validating the robustness of our trace element data. Analytical uncertainty of a single measurement was calculated from the reproducibility of the Cam-Wuellestorfi, measured during a particular mass spectrometry session. The analytical uncertainties (2SD, n = 31, Table S3) on the X/Ca ratios are ±0.4 µmol mol −1 for Li/Ca, ±7 µmol mol −1 for B/Ca, and ±0.01 µmol mol −1 for Mg/Ca, respectively.

Oxygen isotopes
Carbonate δ 13 C and δ 18 O were measured on a GasBench II coupled to a Delta V mass spectrometer at the stable isotope facility of Pôle spectrometrie Océan (PSO), Plouzané. Around 20 shells were weighed, crushed, and had clay removed following the same method described in Sect. 3.3 (Barker et al., 2003). The recovered foraminifera were weighed in tubes and flushed with He gas. Samples were then digested in phosphoric acid and analyzed. Results were calibrated to the Vienna Pee Dee Belemnite (VPDB) scale by international standard NBS19, and analytical precision on the in-house standard Ca21 was better than ±0.11 ‰ for δ 18 O (SD, n = 5) and ±0.03 ‰ for δ 13 C (SD, n = 5).

Calcification depth determination
We utilized two different chemo-stratigraphic methods to estimate the calcification depth (CD) in this study (Tables S6  and S7). The first method (CD1), commonly used in paleoceanography, utilizes δ 18 O measurements of the carbonate (δ 18 O c ) to estimate calcification depths (referred to as δ 18 Obased calcification depths) (Schmidt et al., 2002;Mortyn et al., 2003;Sime et al., 2005;Farmer et al., 2007;Birch et al., 2013). Rebotim et al. (2017) also showed good correspondence between living depth habitat and calcification depth derived using CD1. The second method (CD2) utilizes Mg/Ca-based temperature estimates (T Mg/Ca ) to constrain calcification depths (Quintana Krupinski et al., 2017). However, we note that reductive cleaning leads to a decrease in Mg/Ca that in turn would result in a bias towards deeper calcification depths, which is not the case when we utilize non-Mg/Ca-based methodologies. In both cases, the prerequisite was that vertical profiles of seawater temperature are available for different seasons in ocean atlases and cruise reports and that hydrographic data and geochemical proxy signatures can be compared to assess the depth in the water column that represents the taxon's maximum abundance.
Because both methods have their uncertainties (in one case use of taxon-specific calibrations and in the other analytical limitations), both estimates of calcification depth were compared to published values for the basin (CD3) and where available for the same site (Table S6). To select which calcification depth to use for further calculations, we first looked at CD 1 , CD 2 , and CD 3 . If CD1 and CD2 were similar we selected this calcification depth, and if CD 1 and CD 2 were different we chose literature values, CD 3 , when available. For some less studied species, like G. tumida, G. menardii, or P. obliquiloculata, CD 3 was not always available but when available showed good correspondence with our CD 2 . Moreover due to availability of Mg/Ca-derived temperature taxonspecific calibrations, we preferentially use CD 2 for those species.
We applied (based on uncertainties of our measurements) an uncertainty of ±10 m for calcification depths >70 m and an uncertainty of ±20 m when calcification depths <70 m. Direct observations of living depths of foraminifera remain limited. However, the depth uncertainties reported here are in line with the uncertainties calculated based on direct observations in the eastern North Atlantic which give a standard error on average living depths ranging from 6 to 22 m for the same species (Rebotim et al., 2017). The decrease in Mg/Ca due to reductive cleaning was not taken into account because it has not been studied for most of the species used in this study and because the depth uncertainty applied based on δ 18 O analytical error is conservative relative to the uncertainty of a 10 % decrease in Mg/Ca equivalent that would be equivalent to ∼ 1.2 • C. The depth habitats utilized to derive in situ parameters are summarized in Table S7.
3.9 δ 11 B borate Two carbonate system parameters are needed to fully constrain the carbonate system. Following the approach of Foster et al. (2008), we used the GLODAP database (Key et al., 2004) corrected for anthropogenic inputs in order to esti-mate preindustrial carbonate system parameters at each site. Temperature, salinity, and pressure for each site are from the World Ocean Database 2013 (Boyer et al., 2013). We utilized the R© code in Henehan et al. (2016) (courtesy of Michael Henehan) to calculate the δ 11 B borate and δ 11 B borate uncertainty and derive our calibrations. Uncertainty for δ 11 B borate utilizing Henehan's code was similar to uncertainty calculated by applying 2SD of the δ 11 B borate profiles within the limits imposed by our calcification depth.
The MATLAB© template provided by Zeebe and Wolf-Gladow (2001) was used to calculate pCO 2 from TA; temperature, salinity, and pressure were included in the calculations. Total boron was calculated from Lee et al. (2010), and K 1 and K 2 were calculated from Mehrbach et al. (1973) refitted by Dickson and Millero (1987).
Statistical tests were performed utilizing GraphPad© software, and linear regressions for calibration were derived utilizing R© code in Henehan et al. (2016) (courtesy of Michael Henehan) with k (number of wild bootstrap replicates) equal to 500.

Depth habitat
The calcification depths utilized in this paper are summarized in Tables S6 and S7, including a comparison of calcification depth determination methods. The calculated calcification depths are consistent with the ecology of each species and the physical properties of the water column of the sites. Specimens of G. ruber and T. sacculifer appear to be living in the shallow mixed layer (0-100 m), with T. sacculifer living or migrating deeper than G. ruber (down to 125 m). Specimens of O. universa and P. obliquiloculata are living in the upper thermocline; G. menardii is found in the upper thermocline until the thermocline depth specific to the location; N. dutertrei is living near thermocline depths and G. tumida is found in the lower thermocline.
Data from the multiple approaches for calculating calcification depth (CD1, CD2, and CD3) imply that some species inhabit deeper environments in the western equatorial Pacific (WEP) relative to the Arabian Sea, which in turn are deeperdwelling than the same morphospecies occurring in the Indian Ocean. In some cases, we find evidence for differences in habitat depth of up to ∼ 100 m between the WEP and the Arabian Sea. This trend is observed for G. ruber and T. sacculifer, but not for O. universa.
Some differences are observed between the two methods for calcification depth determination that are based on δ 18 O and Mg/Ca (CD1 and CD2). These differences might be due to the choice of calibration. Alternatively, our uncertainties for δ 18 O imply larger uncertainties on calcification depth determinations that use this approach, compared to Mg/Cabased estimates.

Empirical calibrations of foraminiferal
δ 11 B carbonate to δ 11 B borate Results for the different species analyzed in this study are presented in Figs. 4 and 5 and summarized in Table 2; additionally, published calibrations for comparison are summarized in Table 3.

G. ruber
Samples were picked from the 250-300 µm fraction, except for the WEP sites where G. ruber shells were picked from the 250-400 µm fraction. Weight per shell averaged 11±4 µg (n = 4, SD), although the weight was not measured on the same subsample analyzed for δ 11 B and trace elements or at the WEP sites. In comparison to literature, the size fraction used for this study was smaller: Foster et al. (2008)  Our results for G. ruber (Fig. 4) are in close agreement with published data from other core tops, sediment traps, tows, and culture experiments for δ 11 B borate > 19 ‰ (Foster et al., 2008;Henehan et al., 2013;Raitzsch et al., 2018). However, the two data points from δ 11 B borate < 19 ‰ are lower compared to previous studies. Elevated δ 11 B carbonate values relative to δ 11 B borate have been explained by the high photosynthetic activity of symbionts Zeebe et al., 2003). Three calibrations have been derived (Table 3). Linear regression on our data alone yields a slope of 1.12 (±1.67). The uncertainty is significant given limited data in our study, and given this large uncertainty, our sensitivity of δ 11 B carbonate to δ 11 B borate is also consistent with the low sensitivity trend of culture experiments from Sanyal et al. (2001) or Henehan et al. (2013). The second calibration made compiling all data from literature shows a sensitivity similar (e.g., 0.46 (±0.34)) to the one recently published by Raitzsch et al. (2018) (e.g., 0.45 (±0.16), Table 3). The third linear regression made only on data from the 250-400 µm fraction from our study and from the 250-300 µm fraction from Henehan et al. (2013) Table 3). This third calibration is offset by ∼ −0.4 ‰ (p>0.05) compared to culture calibration from Henehan et al. (2013).

T. sacculifer
δ 11 B carbonate results for T. sacculifer (sacc and without sacc) (Fig. 4) are compared to published data (Foster et al., 2008;Martínez-Boti et al., 2015b;Raitzsch et al., 2018). Results for T. sacculifer are in good agreement with the literature and exhibit higher δ 11 B carbonate compared to expected δ 11 B borate at their collection location. A linear regression through our data  1.99 ± 0.01 35 ± 7 0.04 ± 0.01 a Uncertainties given in 1SD (see text). b When two measurements were carried out uncertainty was calculated with a = (1/ i (1/ ai ) 2 ); with only one measurement the error was determined on reproducibility of the AE121 standard. c Uncertainty given in 2SD, calculated on the reproducibility of Cam-Wuellestorfi (see text and Table S3, reference in Misra et al., 2014a).  Figure 4. Boron isotopic measurements of mixed-layer foraminifera plotted against δ 11 B borate . δ 11 B borate was characterized by determination of the calcification depth of foraminifera utilizing data presented in Fig. 3.  . When regressing data from the 250-400 µm fraction, our results are not significantly different from the regression through data that combine all size fractions (Fig. 4).  Table 3.

Comparison of core-top and culture data
The data for G. ruber and T. sacculifer from the core tops we measured are broadly consistent with previous published results. The calibrations between these core-top-derived estimates and culture experiments are not statistically different due to small datasets and uncertainties on the linear regressions (Henehan et al., 2013;Marinez-Boti et al., 2015;Raitzsch et al., 2018; Table 3). The sensitivities of the species analyzed are not statistically different and are close to unity.
deep-dwelling foraminifera are observed, with lower values and a smaller range for the deeper-dwelling taxa (58-126 µmol mol −1 vs. 83-190 µmol mol −1 for shallow dwellers); however, the trend for the surface dwellers can also be driven by interspecies B/Ca variability. The B/Ca data for deep-dwelling taxa exhibit a significant correlation with [B(OH) − 4 ]/[HCO − 3 ] (p<0.05) but no correlation with δ 11 B carbonate , and temperature (Fig. S3) (Fig. S3). When comparing data from all sites together, a weak decrease in B/Ca with increasing calcification depth is observed (R 2 = 0.11, p<0.05, Fig. S4). A correlation also exists between B/Ca and the water depths of the cores (not significant, Fig. S4).

Discussion
5.1 Sources of uncertainty relating to depth habitat and seasonality at studied sites 5.1.1 Depth habitats and δ 11 B borate Because foraminifera will record ambient environmental conditions during calcification, the accurate characterization of in situ data is needed not only for calibrations but also Biogeosciences, 17, 3487-3510, 2020 https://doi.org/10.5194/bg-17-3487-2020 to understand the reconstructed record of pH or pCO 2 . The species we examined are ordered here from shallower to deeper depth habitats: G. ruber > T. sacculifer (sacc) > T. sacculifer (without sacc) > O. universa > P. obliquiloculata > G. menardii > N. dutertrei > G. tumida (this study; Birch et al., 2013;Farmer et al., 2007), although the specific water depth will vary depending on the physical properties of the water column of the site (Kemle-von Mücke and Oberhänsli, 1999). We note that calculation of absolute calcification depths can be challenging in some cases as many species often transition to deeper waters at the end of their life cycle prior to gametogenesis (Steinhardt et al., 2015).
We find that assumptions about the specific depth habitat a species of foraminifera is calcifying over, in a given region, can lead to differences of a few per mill in calculated isotopic compositions of borate (Fig. 3). Hence this can cause a bias in calibrations if calcification depths are assumed instead of being calculated (i.e., with δ 18 O and/or Mg/Ca). Factors including variations in thermocline depth can impact depth habitats for some taxa. At the sites we examined, most of the sampled species live in deeper depth habitats in the WEP relative to the Indian Ocean, which in turn is characterized by deeper depth habitats than in the Arabian Sea. In the tropical Pacific, T. sacculifer is usually found deeper than G. ruber except at sites characterized by a shallow thermocline, in which case both species tend to overlap their habitat (e.g., ODP Site 806 in the WEP which has a deeper thermocline than at ODP Site 847 in the eastern equatorial Pacific, EEP) (Rickaby et al., 2005). The difference in depth habitats for T. sacculifer and N. dutertrei between the WEP and EEP can be as much as almost 100 m (Rickaby et al., 2005).

Seasonality and in situ δ 11 B borate
As discussed by Raitzsch et al. (2018), depending on the study area, foraminiferal fluxes can change throughout the year. Hydrographic parameters related to carbonate chemistry may change across seasons at a given water depth. We therefore recalculated the theoretical δ 11 B borate using seasonal data for temperature and salinity and annual values for TA and DIC for each depth at each site. The GLODAP (2013) (Key et al., 2004) database does not provide seasonal TA or DIC values.
The low sensitivity of δ 11 B borate to temperature and salinity means that calculated δ 11 B borate values for each water depth at our sites were not strongly impacted (Fig. S1 in the Supplement). Thus, these findings support Raitzsch et al. (2018), who concluded that calculated δ 11 B borate values corrected for seasonality were within the error of non-corrected values for each water depth. As Raitzsch et al. (2018) highlight, seasonality might be more important at high-latitude sites where seasonality is more marked; however, the seasonality of primary production will also be more tightly constrained due to the seasonal progression of win-ter light limitation and intense vertical mixing and summer nutrient limitation.
Data for our sites suggest that most δ 11 B borate variability we observe does not come from seasonality but from the assumed water depths for calcification. With the exception of a few specific areas such as the Red Sea (Henehan et al., 2016;Raitzsch et al., 2018), at most sites examined seasonal δ 11 B borate at a fixed depth does not vary by more than ∼ 0.2 ‰. We conclude that seasonality has a relatively minor impact on the carbonate system parameters at the sites we examined.
There are several studies indicating that the δ 11 B signatures in foraminiferal calcite reflect microenvironment pH (Jørgensen et al., 1985;Rink et al., 1998;Köhler-Rink and Kühl, 2000;Hönisch et al., 2003;Zeebe et el., 2003). Foraminifera with high photosynthetic activity and symbiont density, such as G. ruber and T. sacculifer, are expected to have a microenvironment pH higher than ambient seawater and a δ 11 B carbonate higher than expected δ 11 B borate , which is the case in our study and in previous studies (Foster et al., 2008;Henehan et al., 2013;Raitzsch et al., 2018). We also observed in our study that N. dutertrei, G. menardii, P. obliquiloculata, and G. tumida record a lower pH than ambient seawater, with δ 11 B carbonate lower than expected δ 11 B borate , and we suggest the results are consistent with lower photosynthetic activity compared to the mixed-layer dwelling species. These observations, based on δ 11 B carbonate measurements, are in line with direct observations from Takagi et al. (2019) that show dinoflagellatebearing foraminifera (G. ruber, T. sacculifer, and O. universa) tend to have a higher symbiont density and photosynthesis activity while P. obliquiloculata, G. menardii, and N. dutertrei have lower symbiont density and P. obliquiloculata and N. dutertrei have the lowest photosynthetic activity. In the same study, P. obliquiloculata exhibited minimum symbiont densities and levels of photosynthetic activity, which may explain why P. obliquiloculata exhibited the lowest microenvironment pH as recorded by δ 11 B.
Based on the observations of Takagi et al. (2019), we can assume that the low δ 11 B of O. universa and T. sacculifer (without sacc) from the WEP is explained by low photosynthetic activity. It has been shown for T. sacculifer and O. universa that symbiont photosynthesis increases with higher insolation (Jørgensen et al., 1985;Rink et al., 1998) and the photosynthetic activity is therefore a function of the light level the symbionts received. This is, in a natural system, dependent on the depth of the species in the water column. For the purpose of this study, we do not consider turbidity which also influences the light penetration in the water column. In this case, photosynthetically active foraminifera living close to the surface should record microenvironment pH (thus δ 11 B) that is more sensitive to water depth changes. A deeper habitat reduces solar insolation, and as a consequence may lower symbiont photosynthetic activity, possibly reducing pH in the foraminifera's microenvironment. This is supported by the significant trend observed between 11 B and the calcification depth for G. ruber and T. sacculifer at our sites (Fig. S2), where microenvironment pH decreases with calcification depth. We observe a significant decrease in δ 11 B in the WEP for T. sacculifer (without sacc) compared to the other sites (p<0.05). Additionally, the 11 B ( 11 B=δ 11 B carbonate −δ 11 B borate ) of G. ruber and T. sacculifer (without sacc and sacc) is significantly lower in the WEP compared to the other sites (p<0.05).
T. sacculifer has the potential to support more photosynthesis due to its higher symbiont density, and higher photosynthetic activity compared to other species, which may support higher symbiont-host interactions (Takagi et al., 2019). These results would be consistent with a greater sensitivity of T. sacculifer's photosynthetic activity with changes in insolation-water depth. To test if the low δ 11 B signature of T. sacculifer (without sacc) in the WEP is related to a decrease in light at greater water depth, we have independently calculated the calcification depth of the foraminifera based on various light insolation culture experiments (Jørgensen et al., 1985) and the microenvironment pH derived from our data ( Fig. 7a and b). This exercise showed that the low δ 11 B of T. sacculifer (without sacc) from the WEP can be explained by the reduced light environment due to a deeper depth habitat in the WEP (Fig. 7b). It can also be noted that T. sacculifer exhibits the largest variation in symbiont density versus test size (Takagi et al., 2019), suggesting that lower size fraction reported for the WEP (250-400 µm) compared to the 300-400 µm at the other sites can be related to a decrease in photosynthetic activity and a lower δ 11 B. Unfortunately, no weight-per-shell data were determined on foraminifera samples to constrain whether test size was significantly different across sites. Future studies could use shell weights to test these relationships.
When the same approach of independently reconstructing calcification depth based on culture experiments is applied to O. universa, the boron data suggest a microenvironment pH of 0.10 to 0.20 lower than ambient seawater pH, which would be in line with the species living deeper than 50 m (light compensation point (Ec); Rink et al., 1998), which is consistent with our calcification depth reconstructions. The low δ 11 B carbonate of O. universa compared to T. sacculifer for the similar calcification depth at some sites (e.g., FC-02a, WP07-a) might reflect differences in photosynthetic potential between the two species, which is supported by observation of a lower photosynthetic potential in O. universa than in T. sacculifer (Tagaki et al., 2019).
Microenvironment pH based on our δ 11 B carbonate data was calculated for the rest of the species. We observed that microenvironment pH is higher in T. sacculifer > G. ruber > T. sacculifer (without sacc -WEP) > O. universa, N. dutertrei, G. menardii, G. tumida > P. obliquiloculata. These results are in line with the photosymbiosis findings from Takagi et al. (2019). Also, the higher δ 11 B data from the west African upwelling published by Raitzsch et al. (2018) for G. ruber and O. universa may reflect a higher microenvironment pH due to a relatively shallow habitat, higher insolation, and high rates of photosynthesis by symbionts. This could highlight a potential issue with calibration when applied to sites with different oceanic regimes as the δ 11 B species-specific calibrations could also be location-specific for the mixed dweller species.
Microenvironment pH for N. dutertrei, G. menardii, and G. tumida is similar to O. universa and suggests a threshold for a respiration-driven δ 11 B signature. This threshold can be induced by a change of photosynthetic activity at lower light intensity in deeper water and/or differences in symbiont density and/or by the type of symbionts at greater depth (nondinoflagellate symbionts). We also note that P. obliquiloculata, which has the lowest symbiont density and photosynthetic activity (Takagi et al., 2019), has the lowest microenvironment pH compared to other deeper-dweller species, supporting our hypothesis that respiration can control microenvironment pH. The deep-dwelling species sensitivity of δ 11 B carbonate to δ 11 B borate with values close to unity might also be explained by relatively stable respiration-driven microenvironments, as the deeper-dweller species do not experience large changes of insolation (e.g., photosynthesis), thereby making them a more direct recorder of environmental pH.  (Farmer et al., 2019) and B/Ca (Farmer et al., 2019;Gabitov et al., 2014;Kaczmarek et al., 2016;Mavromatis et al., 2015;Uchikawa et al., 2015). A recent study from Farmer et al. (2019) has proposed that in foraminifera at higher precipitation rates, more borate ions may be incorporated into the carbonate mineral, while more boric acid may be incorporated at lower precipitation rates.  (Wang et al., 2008). Theoretical depths were calculated for a decrease in microenvironment pH of pH 1 = −0.02 (e.g., WP07a); pH 1 = −0.04 (e.g., A14), pH 2 = −0.06 (e.g., 806A). Light penetration corresponding to Ec is ∼ 12 %, pH 0 ∼ 7 %, pH 1 ∼ 5 %, pH 2 ∼ 1 %, and respective calcification depths are 75, 90, 110, and 150 m. The grey band is the calcification depth calculated that explains the microenvironment pH from pH 0 to pH 2 . Dotted lines show the range of the calcification depth for T. sacculifer (without sacc) in the WEP utilized in this study.
The authors also suggest this may explain low sensitivities of culture experiments.
When combining all literature data, T. sacculifer and G. ruber have sensitivities of δ 11 B carbonate to δ 11 B borate of 0.83± 0.48 and 0.46 ± 0.34, respectively, in line with previous literature and paleo-CO 2 reconstructions. Also, if we only take into account our data, and the observation that the sensitivity of δ 11 B carbonate to δ 11 B borate is not statistically different from unity for most of the species investigated, we can speculate that for these taxa, changes in precipitation rate and contributions of boric acid are not likely to be important. If considering only the data from this study, G. ruber (1.12 ± 1.67) and T. sacculifer (1.38 ± 1.35) present higher sensitivities of δ 11 B carbonate to δ 11 B borate . We can then again speculate that the observed high values for δ 11 B carbonate at high seawater pH can be due to higher precipitation rates. We note this could also be consistent with the higher sensitivity of B/Ca signatures in these two surface dwelling species to ambient [B(OH) − 4 ]/[HCO − 3 ] relative to deeperdwelling species. Those interspecific differences still remain to be explained; however, part of this variability is likely due to changes in the carbonate chemistry of the microenvironment resulting in changing competition between borate and bicarbonate. A caveat is that we can not exclude specific biological processes and that in taxa with a nonrespiration-driven microenvironment, changes in day / night calcification ratios also impact observed values. As indicated by Farmer et al. (2019), studies of calcite precipitation rates in foraminifera may help to improve our understanding of the fundamental basis of boron-based proxies.

Evaluation of species for pH reconstructions and water depth pH reconstructions
This data set allows us to reassess the utility of boronbased proxies for the carbonate system. The main aim of using boron-based proxies relates to the reconstruction of past oceanic conditions, specifically pH and pCO 2 . Mixedlayer species (e.g., G. ruber and T. sacculifer) are potential archives for atmospheric CO 2 reconstructions. Other species can shed light on other aspects of the carbon cycle including the physical and biological carbon pumps.
There are a few main inferences we can make. When integrated with published data, the sensitivities of δ 11 B carbonate to δ 11 B borate for G. ruber and T. sacculifer are similar to those in previous studies (Martínez-Boti et al., 2015b;Raitzsch et al., 2018), which supports the fidelity of previous paleo-reconstructions that use published calibrations between δ 11 B carbonate and δ 11 B borate . The regression we have made for G. ruber supports a decrease in δ 11 B carbonate with decreasing size fractions (offset of −0.4 ‰, p>0.05) with the sensitivity of δ 11 B carbonate to δ 11 B borate not being statistically different from a higher size fraction (p<0.05). The variability in our weight per shell for our G. ruber, based on data from Henehan et al. (2013), can potentially imply a deviation down to 1 ‰ relative to the calibration line from Henehan et al. (2013), which can be in line with the maximum deviation observed in our data (∼ 1.2 ‰) and not inconsistent with a size effect explaining the offset in our calibration. Our δ 11 B carbonate data and the sensitivity to δ 11 B borate of O. universa support previous data from Henehan et al. (2016). N. dutertrei δ 11 B carbonate data span a large range of pH, al- Figure 8. Water depth pH profiles reconstructed at every site applying the monospecific calibrations derived from our results ( Table 3). The figure shows measured δ 11 B calcite , δ 11 B borate calculated according to different calibrations (see Table 3 and text), calculated pH based on δ 11 B (pH δ 11 B ), and pCO 2 calculated from pH δ 11 B and alkalinity.
Biogeosciences, 17, 3487-3510, 2020 https://doi.org/10.5194/bg-17-3487-2020 Figure 9. Evaluation of the reconstructed parameters, δ 11 B borate , pH, and pCO 2 versus in situ parameters calculated in Fig. 8 (based on δ 11 B and alkalinity). The recalculated parameters are consistent with in situ data, except for G. ruber, and this variability might be explained by the different test sizes within measured size fractions.
lowing us to derive a robust calibration with δ 11 B borate . It remains premature to assume that a unique calibration with a slope of ∼ 0.9 can be used for all deeper-dwelling species. More data are needed for P. obliquiloculata, G. menardii, and G. tumida to robustly test this assertion. In order to derive accurate reconstructions of past ambient pH and pCO 2 , accurate species-specific calibrations need to be used that are constrained by core tops or samples from similar types of settings (Figs. 8, 10, S6). Lower δ 11 B signatures in T. sacculifer (without sacc) are observed in the WEP, which may be explained by the deeper depth habitat for this taxa, as lower light levels might reduce symbiont photosynthetic activity. Also, we show that a correction is needed for T. sacculifer (without sacc) in the WEP in order to accurately reconstruct atmospheric CO 2 . When applying calibrations nos. 2 and 4 to T. sacculifer and G. ruber (compilation of all data, Table 3), our data show more variability, especially for G. ruber which leads to the larger mismatch compared to in situ parameters. The greater divergence of reconstructed values from in situ measurements is observed at site WPO7-01 for both T. sacculifer (without sacc) and G. ruber. More data would be needed to determine a proper correction for both species and a core-top study will be determinant for future downcore reconstructions, especially in the WEP. We also find that for two species the boron isotope proxy is a relatively straightforward recorder of ambient pH, with sensitivities close to unity observed for O. universa and N. dutertrei.
There is also promise in using multiple species in a sample from different hydrographic regimes to reconstruct vertical profiles of pH and pCO 2 . We are able to reproduce pH and pCO 2 profiles from multiple sites with different water column structures (Fig. 8) with those reconstructions within error of the in situ values, for most sites. In order to avoid circularity, to validate these calibrations, we recalculated ambient pH and pCO 2 by first excluding site-specific data and then recalculating species-specific calibrations, followed by application to each specific site. The comparison of the two methods, first using all the data to derive the calibration and recalculate pH and pCO 2 (circular) and second by excluding the site of interest, deriving calibrations, and calculating pH and pCO 2 (not circular), does not show significant differences and validates the robustness of the calibrations (Fig. S5). We utilized the calibrations derived from our data for G. ruber (calibration nos. 1 and 2,  Table 3), and for N. dutertrei, G. tumida, and G. menardii we utilized the calibration on the compilation of the deep dwellers (calibration no. 13, Table 3). Results are shown in Fig. 8 and evaluated in Fig. 9. For G. menardii, more data would be helpful to provide additional constraints. Results for G. ruber are the most scattered, potentially due to difference in test sizes (Henehan et al., 2013) or depth habitat. Results reaffirm the importance of working with narrow size fractions (Henehan et al., 2013), the utilization of calibrations derived from the same size fraction, or use of offsets to take into account this size fraction effect and the importance of core-top studies before paleo-application.

Conclusions and future implications
Our study has extended the boron isotope proxy with data for new species and sites. The work supports previous work showing that depth habitats of foraminifera vary depending on the oceanic regime, and this can impact boron isotope signatures. Low δ 11 B values in the WEP compared to other regions for T. sacculifer (without sacc) may be explained by a reduction in microenvironment pH due to a deeper depth habitat associated with reduced irradiance and thus photosynthetic activity.
In order to accurately develop downcore reconstructions, constraining the depth habitat using core-top studies is important, as the same species can record the seawater pH at different water depths, potentially introducing biases when comparing between different locations. Also, we speculate that a change of the thermocline depth in the past could imply variations in depth habitat and introduce biases in the reconstructions, but further work is needed to test this assertion.
The sensitivity of δ 11 B carbonate to pH is in line with previously published data for T. sacculifer and G. ruber. The sensitivity of δ 11 B carbonate to pH of O.universa (mixed dweller), N. dutertrei, G. menardii, and G. tumida (deep dwellers) is similar, but more data are needed to fully determine those sensitivities. The similarity of boron isotope calibrations for deep-dwelling taxa might be related to similar respirationdriven microenvironments.
Reconstruction of seawater pH and carbonate system parameters is achievable using foraminiferal δ 11 B, but additional core-top and down-core studies reconstructing depth profiles will be needed in order to further verify calibrations published to date. Past pH and pCO 2 water depth profiles can potentially be created by utilizing multiple foraminiferal species in concert with taxon-specific calibrations for similar settings. This approach has much potential for enhancing our understanding of the past workings of the oceanic carbon cycle and the biological pump.
Author contributions. RE and AT wrote the proposals that funded the work. AT and FC provided the samples. MG, SM, and AT contributed to the experimental design. AV helped for sample preparation. MG and SM contributed to developing the method of boron isotope analysis. MG performed the measurements with assistance from SM. MG conducted the data analysis. MG drafted the paper, which was edited by all authors. Interpretation was led by MG, AT, and SM with input from RE, AV, and FC.