Mg/Ca-Temperature Calibration of Polar Benthic foraminifera species for reconstruction of bottom water temperatures on the Antarctic shelf

Benthic foraminifera Mg/Ca is a well-established bottom water temperature (BWT) proxy used in paleoclimate studies. The relationship between Mg/Ca and BWT for numerous species has been determined using core-top and culturing studies. However, the scarcity of calcareous microfossils in Antarctic shelf sediments and poorly deﬁned calibrations at low temperatures has limited the use of the foraminiferal Mg/Ca paleothermometer in ice proximal Antarctic sediments. Here we present paired ocean temperature and modern benthic foraminifera Mg/Ca data for three species, Trifarina angulosa , Bulimina aculeata and Globocassidulina subglobosa, but with a particular focus on Trifarina angulosa. The core-top data from several Antarctic sectors span a BWT range of (cid:1) 1.7 to +1.2 (cid:1) C and constrain the relationship between Mg/Ca and cold temperatures. We compare our results to published lower-latitude core-top data for species in the same or related genera, and in the case of Trifarina angulosa, produce a regional calibration. The resulting regional equation for Trifarina angulosa is Temperature ( (cid:1) C) = (Mg/Ca (cid:1) 1.14 ± 0.035)/0.069 ± 0.033). Addition of our Trifarina angulosa data to the previously published Uvigerina spp . dataset provides an alternative global calibration, although some data points appear to be oﬀset from this relationship and are discussed. Mg-temperature relationships for Bulimina aculeata and Globocassidulina subglobosa are also combined with previously published data to produce calibration equations of Temperature ( (cid:1) C) = (Mg/Ca-1.04 ± 0.07)/0.099 ± 0.01 and Temperature ( (cid:1) C) = (Mg/Ca-0.99 ± 0.03)/0.087 ± 0.01, respectively. These reﬁned calibrations highlight the potential utility


INTRODUCTION
Satellite observations indicate that the West Antarctic Ice Sheet (WAIS) is losing mass at an accelerating rate (Rignot et al., 2019). Mass loss is primarily driven by basal melting of ice shelves by warm Circumpolar Deep Water (CDW), which upwells at the continental shelf edge and moves across the shelf through bathymetric troughs to grounding lines (Pritchard et al., 2012). CDW is characterised by temperatures >3.5°C above in-situ freezing point and salinities >34.6 and tends to flow onto the shelf at depths >300 m (Jacobs et al., 2011). Ice-shelf thinning reduces buttressing of inland ice and results in flow acceleration, drawdown of the interior ice sheet (Pritchard et al., 2012), and sea-level rise (Edwards et al., 2019).
Of particular concern is the future behaviour of ice streams draining into the Amundsen Sea Embayment (ASE), such as Pine Island and Thwaites glaciers, which contain $1.4 m of sea level-equivalent ice (Vaughan et al., 2006). Accurately predicting the future response of glaciers in this region to ongoing warming (e.g., DeConto and Pollard, 2016) requires a comprehensive understanding of the potential influence of CDW on Antarctic glaciers over coming decades and centuries (Colleoni et al., 2018). Recent work suggests that the flux of CDW is controlled by the strength and intensity of westerly winds at the shelf edge (Jenkins et al., 2018) which in turn is driven by tropical climate dynamics (Steig et al., 2012;Dutrieux et al., 2014;Jenkins et al., 2018;Holland et al., 2019). Indeed, sediment cores recovered from beneath the floating part of Pine Island Glacier indicate that it started to retreat in the 1940 s , possibly as a response to prolonged El Niñ o conditions (Steig et al., 2012). Furthermore, Dutrieux et al. (2014) demonstrated that oceanic melting of Pine Island Glacier decreased by 50% between 2011 and 2012 due to a strong La Niñ a event, which weakened the typically cyclonic wind stress and curtailed the westerlies. With the realisation that CDW forcing might change on a variety of timescales from decades (Jenkins et al., 2018) to centuries , there is a growing need to reconstruct its presence/absence over Antarctica's continental shelves over a range of timescales. However, because regional physical oceanographic data are limited to the last two decades, it is unclear whether the CDW has been a persistent feature on the ASE shelf, or if the advection and/or temperature of CDW on the shelf has varied in the past. To provide context for recent observations and improve understanding of the future behaviour of ice sheets in Antarctica, detailed records of past seawater temperatures are required from Antarctica's continental shelves (Shevenell et al., 2011;Hillenbrand et al., 2017).
Previous attempts to reconstruct CDW advection onto the Antarctic shelf utilised a range of proxies that are related to seawater temperatures, such as diatom and foraminifera assemblages (Leventer et al., 2002;Peck et al., 2015;Minzoni et al., 2017;Majewski et al., 2017;Hillenbrand et al., 2017) and stable oxygen and carbon isotopes of both planktic and benthic foraminifera tests (Shevenell and Kennett, 2002;Peck et al., 2015;Hillenbrand et al., 2017). However, each of these proxies has its own limitations. For example, stable oxygen isotope (d 18 O) in foraminifera vary with both temperature and d 18 O seawater, which proximal to Antarctica is influenced primarily by ice volume and glacial meltwater (Shackleton, 1967;Meredith et al., 2008), whereas stable carbon isotopes are influenced by biological productivity and water mass nutrient concentration (see references in Hillenbrand et al., 2017). Furthermore, despite some successes reconstructing upper ocean temperatures on the Antarctic shelf using biomarkers (TEX 86 ;Shevenell et al., 2011;Kim et al., 2012;Etourneau et al., 2013), only a single study in Prydz Bay, East Antarctica, attempted to reconstruct BWT on the Antarctic shelf (Rathburn and De Deckker, 1997). Consequently there is a dearth of information on past BWT from ice proximal settings around Antarctica, where these records are urgently needed.
Benthic foraminifera Mg/Ca paleothermometery potentially offers a powerful tool for reconstructing past temperatures in ice-proximal settings around Antarctica. Mg/Catemperature calibrations have been developed for different benthic foraminifera species (e.g. Rosenthal et al., 1997;Elderfield et al., 2006;Marchitto et al., 2007;Bryan et al., 2008;Evans et al 2015) in a wide range of environments (Barrientos et al., 2018). However, there is a distinct lack of core-top calibration data for low temperature polar environments. Low temperature Mg/Ca calibrations from the Arctic and Antarctic for common high-latitude benthic foraminifer species (e.g., Trifarina angulosa, Globocassidulina subglobosa and Bulimina aculeata) are therefore urgently required to facilitate the interpretation of down-core benthic Mg/Ca records from the polar-regions. The lack of Mg/Ca calibration data from Antarctica stems from two fundamental issues: First, calcareous foraminifera are often either poorly preserved or not abundant in many shelf regions around Antarctica. Second, the number of sites where both BWTs were directly measured by CTD casts and where surface sediments were collected is limited.
To address this data gap, we generated Mg/Ca temperature data from three species of benthic foraminifera occupying a range of BWT around Antarctica: Trifarina angulosa, Globocassidulina subglobosa, and Bulimina aculeata (Fig. 1). However, we focus primarily on the species Trifarina angulosa for the following reasons: (1) it is common in modern and Holocene sediments on the Antarctic continental shelf, and particularly in our main study area in the Amundsen Sea (e.g., Hillenbrand et al., 2017;Majewski et al., 2017); (2) it is a shallow infaunal species (Mackensen et al., 1990), which is genetically related to Uvigerina spp., with Uvigerina peregrina being more closely related to Trifarina angulosa (Trifarina earlandi) than to some other Uvigerina species (Schweizer et al. 2005, Schweizer et al., 2011. Published Mg/Ca-temperature calibrations were produced for Uvigerina spp. (summarised in Elderfield et al., 2010), to which we can compare our Trifarina angulosa records; and (3) shallow infaunal species are less likely to be influenced by the effects of carbonate ion concentration on the Mg/Ca ratios of their tests than epifaunal species (Elderfield et al., 2010). Changes in carbonate ion saturation state are known to have a secondary effect on Mg/Ca values in some foraminifera (e.g. Elderfield et al., 2006, Bryan andMarchitto, 2008).
However, changes in environmental conditions over time may lead to considerable changes in the down-core abundance or presence of certain foraminifera species, implying that continuous down-core records are difficult to obtain . Therefore, we also investigate two other calcareous benthic foraminifera that are common on the Antarctic shelf: Globocassidulina subglobosa and Bulimina aculeata.
Globocassidulina subglobosa is an important species that is common on the West Antarctic continental shelf, for example both on the West and the East side of the Antarctic Peninsula (Majewski and Pawlowski, 2010) and in the Amundsen Sea (Majewski, 2013;Hillenbrand et al., 2017). The species has been shown to dominate biofacies in fjords on the western Antarctic Peninsula (Ishman, 1990) and is sometimes associated with ice-proximal settings (Majewski and Anderson, 2009) and sub-ice-shelf settings (Murray and Pudsey, 2004). Live Globocassidulina subglobosa have been found at 0-1 cm depth in the sediments and, thus, can be considered as shallow infaunal (Murray and Pudsey, 2004).
Bulimina spp. Mg/Ca has recently been calibrated to BWT over a wide temperature range (3-13°C), but not one that includes typical Antarctic margin BWTs of $À1 to +2°C (Grunert et al., 2018). On Antarctica's continental shelves, Bulimina aculeata has been associated with nutrient-rich, muddy sediments associated with quiescent CDW-dominated environments (Ishman and Domack, 1994;Shevenell and Kennett, 2002;Majewski, 2013). As such, Mg/Ca-temperature data from this species may be the most useful for determining past variations in the advection of CDW onto the Antarctic continental shelf. Bulimina aculeata is thought to be shallow infaunal and occurrs in the uppermost 0-1 cm of the seabed sediments (Mackensen et al., 1990).
By examining the Mg/Ca-temperature relationship for these three species, we hope to assess the utility of this proxy in Antarctic ice-proximal settings, and ultimately, enable future studies to reconstruct past BWT in climatically sensitive regions such as the Amundsen Sea.

Sample site locations and collection
We selected samples from several locations in different sectors around Antarctica (Fig. 2) to encompass a range of BWTs from -1.75°C to + 1.23°C. Samples from the eastern Pacific sector span the warm water regime (CDW), whilst the samples from the East Antarctic shelf in the Weddell Sea and on the Sabrina Coast (close to Totten Glacier) span the colder water regime. This sampling strategy achieves the widest temperature range for calibration as possible with available material.
The samples (Table 1) can be geographically categorised into five continental shelf regions: the Amundsen Sea, Bellingshausen Sea, western Antarctic Peninsula, Weddell Sea, and the Sabrina Coast. All samples were taken by box core except the Sabrina Coast samples, which were recovered with a large volume multiple-corer. This strategy assures that benthic foraminifera tests were collected from undisturbed seafloor surface sediments and are most likely of modern age. In contrast, foraminifera specimens taken from piston, gravity or kasten core tops are sometimes of older age due to core-top disturbance and/or seafloor surface sediment loss during core recovery. The samples span a water depth range of 61-1789 m but most sites (with the exception of two) are shallower than 715 m (Table 1).
All our samples were processed in a similar way, with the exception that the foraminifera from the eastern Weddell Sea and the Sabrina Coast were Rose Bengal stained; their stained protoplasm indicates that the individuals were alive/recently dead at the time of sampling. The surface sediment samples were washed with deionised water through a 63 mm sieve and dried for 24 hours at 40°C in an oven; the Sabrina Coast samples were washed on the ship, 24 hours after collection and staining. The >63 mm fraction was drysieved and specimens of Trifarina angulosa, Bulimina aculeata, and Globocassidulina subglobosa were picked from the 250-355 mm fraction, where possible (approximately 30 specimens per sample), except for the Sabrina Coast samples, which were picked from the 150-355 lm size fraction due to low total foraminifer abundance. The potential influence of test size was assessed for Trifarina angulosa by picking tests from the 150-250 mm fraction and comparing their Mg/Ca ratios to those of tests from the 250-355 mm fraction. All specimens of Trifarina angulosa were the Costate (ribbed) morphotype (Fig. 1a), except for the Sabrina Coast samples, which also included the ''Hispid" (spiky) morphotype. This was again necessary to obtain enough material for analysis.
For some samples, the BWT at the same site was measured by a CTD cast deployed at the same time (Table 1). However, for other sites no concurrent CTD profiles were available. To determine BWTs for these sites, the temperature at the same water depth from the nearest available CTD cast was used. The CTD data used and distance from the sample site are given in Table 1. The CTD data were collected in austral spring/summer, which most likely is also the main period when the foraminifera would be calcifying due to increase in food supply and reduction in sea-ice cover (Mackensen et al., 1993, Skirbekk et al., 2016.

Foraminifera cleaning and analysis
Foraminifera specimens were gently crushed between two glass plates and visible contaminants were removed using a fine paintbrush. The fragments then underwent a cleaning procedure adapted from Boyle and Keigwin (1985/86). This includes: (i) the removal of clay by ultrasonication of test fragments 3x in 18.2 MX . cm deionised water and 3x in methanol; (ii) a reductive step to remove any metal oxides using a solution of hydrous hydrazine and citric acid in ammonia; (iii) an oxidative step using a solution of hydrogen peroxide in sodium hydroxide to remove organic matter; and (iv) a dilute nitric acid leach using Optima pure nitric acid. Between the mechanical clay removal step and the reductive step, samples were examined under a binocular microscope and any obvious noncarbonate particles removed. The cleaned samples were dissolved in ultra-pure nitric acid (0.1 M HNO 3 ), centrifuged, and a 'concentrate' aliquot was transferred to a new sample vial to prevent leaching from any remaining contaminants. This aliquot was then split for analysis by ICP-OES and HR-ICP-MS. This was done so that we could follow the method of Misra et al. (2014), using the HR-ICP-MS to quantify B/Ca at low concentrations. All other trace metal data is from analysis by ICP-OES following the method of de Villiers et al. (2002). Long-term instrumental precision of   (Greaves et al., 2008;Rosenthal et al., 2004). The cleaning efficiency was monitored by measuring Mn/Ca and Fe/Ca. All samples were below the thresholds set out in Barker et al. (2003) Fig. 1), and thus no samples were eliminated from the results and discussion.

RESULTS
3.1. Trifarina angulosa Fig. 3a shows Trifarina angulosa Mg/Ca ratios vs. BWT for samples from the Weddell Sea, western Antarctic Peninsula, Sabrina Coast and the Bellingshausen and Amundsen seas ( Table 2). We observe a weak positive linear relationship; R 2 value = 0.15, P = 0.21, number of samples = 12. The R 2 value is slightly improved if the Sabrina Coast samples are removed (R 2 = 0.36, P = 0.07, number of samples = 10). Possible reasons for this are discussed in Section 4.1.

3.1a. Trifarina Angulosa (250-355 lm)
Addition of our new dataset to the previously published Uvigerina spp. calibration (Elderfield et al., 2010;Bryan and Marchitto, 2008) (Fig. 4, Equations 4 and 5 (without Sabrina Coast data)) leads to a slight decrease in the R 2 value (from 0.88 to 0.83) and negligible impact on the P value (P < 0.0001) (see Table 2). Omission of the Sabrina Coast data (Equation 5) has little impact on the R 2 value (0.82) indicating that all our data fall within the variability of the Uvigerina spp. calibration.

3.1b. Trifarina angulosa (150-250 mm)
Supplementary Fig. 2 shows Mg/Ca ratios vs. BWT from the same sites as the 250-355 mm samples, except for Sabrina Coast, as there was insufficient material to analyse this size fraction separately. When fit to a linear regression, the 150-355 mm samples have an R 2 value of 0.08 and a P   Fig. 3) no consistent offset in the Mg/Ca data is observed between the two size fractions.

Bulimina aculeata (250-355 mm)
Four surface sediment samples had a sufficient number of Bulimina aculeata tests for trace metal analysis, covering a small BWT range of +0.92 to +1.23 C (Fig. 5). The addition of these data to a published Bulimina spp. calibration data set (Grunert et al., 2017, Equation 6) results in a minor change in the T-Mg/Ca slope and intercept (Equation 7; R 2 = 0.64, P < 0.0001).

Trifarina angulosa
The relationship between Mg/Ca ratios and BWT presented here for Trifarina angulosa, whilst positive and comparable to that seen in the published records of Uvigerina spp. (Elderfield et al., 2010 and references therein), shows large variability, which results in a relatively low R 2 value for the regression between Mg/Ca and BWT. We first consider the impact of different BWT calibrations on downcore data before discussing the limitations of our study and possible explanations for the scatter in our Trifarina angulosa data.
Because the gradient and intercept of the calibration equations vary, the various equations (Table 2, Equations 1-2, 4-5) result in relatively large differences between calculated temperature. This is demonstrated in Fig. 7, where Equations 1, 2 and 4 are applied to pilot down-core data from a core recovered from the outer western Amundsen Sea shelf (Smith et al., 2011). This offset is similar to the  variability seen within the different datasets that comprise the most recently published Uvigerina spp. calibration (Elderfield et al., 2010). Whilst the various calibrations create large differences in reconstructed BWTs, the amplitude of down-core changes in BWT appears to be similar. Of the three calibrations used, Equation 2 produces BWTs (+0.2 to +1.8°C) that are closest to the measured BWT range of seawater temperature in the Amundsen Sea (À1.5 to +1.4°C; note that CDW has a temperature range of +0.2 to +1.4°C). This might imply that regional calibrations offer the most promise in reconstructing palaeo BWTs. However, until these calibrations are applied to a wide range of down-core data, it remains unclear if this is the case for all sectors of the Antarctic continental margin. One limitation of our study that cannot be easily overcome is the lack of modern Trifarina angulosa samples from Antarctic margin sites with a wide BWT range. This is partially related to a general lack of calcareous microfossils in modern sediments from the Antarctic continental margin, although further suitable locations (e.g., Hauck et al., 2012) may be (re-)sampled during future fieldwork campaigns. Furthermore, low concentration of calcareous microfossils often relates to sediment dilution, owing to the high-sedimentation rates in glaci-proximal settings. Targeting more distal settings with lower sedimentation rates, or shallower basins might increase the chances of finding microfossil-rich sediments in sufficient quantities to perform trace-metal geochemistry. Secondly, the ambient seawater temperature range for Trifarina angulosa around Antarctica is relatively narrow, so any regional core top calibration is inevitably constrained and cannot be as broad as that for Uvigerina spp. (Elderfield et al., 2010). Laboratory cultures of benthic foraminifera are challenging, and calibrations based on cultured specimens of Trifarina angulosa grown under conditions outside of their natural temperature range may result in BWT-Mg/Ca relationships that are not applicable to down-core material.
Environmental variables, such as carbonate saturation state (Elderfield et al., 2006, Bryan andMarchitto, 2008), and the Mg/Ca content of seawater (Evans and Mü ller, 2012) may exert secondary influences on the Mg/Ca ratio of foraminifera shells. However, Mg/Ca ratios of shallow infaunal species, such as Uvigerina spp., Trifarina angulosa, and Oridorsalis umbonatus, are generally considered less sensitive to changes in the bottom water carbonate saturation state than those of epifaunal species (Elderfield et al., 2010, Brown et al., 2011, Mawbey and Lear, 2013 as they inhabit sediments with pore-waters that are thought to equilibrate rapidly with calcium carbonate (Elderfield et al., 2006). This notwithstanding, it has been observed for the species Oridorsalis umbonatus that large changes in carbonate saturation state in seawater may penetrate the sediments, creating so called ''dissolution events" that could affect the Mg/Ca ratios in shallow infaunal foraminifera shells (Mawbey and Lear, 2013). Furthermore, measurements of pore-water carbonate saturation state revealed its spatial variability, and a relationship between pore-water and bottom water carbonate ion saturation state (Weldeab et al., 2016). Therefore, Mg/Ca data from regions that may be subject to large changes in carbonate saturation state, such as those episodically or periodically bathed by relatively unmodified CDW, should be treated with caution. B/Ca is a possible proxy for carbonate ion saturation state (Rae et al., 2011) and was measured on the samples of this study (Supplementary Figs. 4 and 5). However, the B/Ca ratios of Trifarina angulosa were very low ($10 lmmol/mol) and, compared to the analytical blank ($10 lmol/mol vs. $2 lmol/mol), not analytically robust enough for an interpretation (cf. Hillenbrand et al., 2017). The measured low B/Ca ratios are consistent with those of Uvigerina sp. (Doss et al., 2018), and other shallow infaunal species, such as Oridorsalis umbonatus (Brown et al., 2011).
Seawater Mg/Ca ratios (Mg/Ca sw ) are generally thought to have a potential influence on Mg/Ca ratios in foraminifera (Evans et al., 2012). The long residence times of Mg and Ca in the oceans ($14 Myr and $1 Myr, respectively (Evans et al., 2012)) imply that global changes in Mg/Ca sw only become significant over long timescales and are therefore of negligible importance to our surface sediment study. However, some of the sample sites in our study are relatively shallow (Table 1) and are perhaps subject to changes in Mg/Ca sw on a local scale, which might have caused the variability of Mg/Ca in foraminiferal calcite observed by us. Localised seawater chemistry variability could be excluded as a driving mechanism for foraminiferal Mg/Ca in a future study through trace metal measurements on seawater samples and foraminifera shells from sediment samples collected at the same sites. Work by Lear et al. (2015) on the shallow infaunal benthic foraminifera species Oridorsalis umbonatus suggests that its Mg/Ca ratio is only weakly affected by Mg/Ca sw due to the foraminiferal biological regulation of Mg/Ca uptake. It is plausible that this is the case for other benthic foraminifera species, too, which would limit the effect of Mg/Ca sw on the measured Mg/Ca values. This warrants investigation and could influence the choice of species selected to evaluate Mg/Ca in foraminifera from around the Antarctic continental margin.
Finally, it has been previously demonstrated that shell size can have an important influence on stable carbon and oxygen isotope (Birch et al., 2013) and trace metal composition (Friedrich et al., 2012) of planktic foraminifera. The effect of test size on trace metal concentrations in benthic foraminifera is also possible, although less well understood (Franco-Fraguas et al., 2011, Diz et al., 2012, and Hintz et al., 2006. A potential size effect is supported by our data because the 150-250 mm Mg/Ca ratios of Trifarina anguolsa do not show a correlation with BWT ( Supplementary  Fig. 2). To mitigate this potential effect, we recommend using a larger size fraction. In our study, we could compare Mg/Ca data for 10 out of 12 samples, which had enough specimens in both the 150-250 lm and the 250-355 lm size fractions. From the remaining two samples (MC45 and MC61 from the Sabrina Coast), we had to analyse tests from the 150-355 lm fraction. These two samples are excluded from the final calibration (Fig. 3b), Table 2 equation 2, as they fall outside the 95% confidence interval of the calibration. The equation including the Sabrina Coast samples (Fig. 3a, Equation 1) is however included in Table 2. These two samples included the ''Hispid" (spiky) morpho-type of Trifarina angulosa, whereas our other samples contained exclusively the Costate (ribbed) morphotype. Investigation into whether an offset between these two morphotypes exists will be possible in the future, when more samples become available, which might improve the Mg/Ca-BWT calibration.

Bulimina aculeata
The addition of new Bulimina aculeata Mg/Ca-BWT data to the calibration of Grunert et al. (2018) neither reduces the Mg/Ca-BWT relationship (R 2 = 0.64 in both instances) nor modifies the existing calibration significantly ( Table 2, Equations 6 to 7). With an Mg/Ca value of 1.0 mmol/mol, use of the two equations results in a BWT difference of only 0.1°C. This finding supports the conclusion that this calibration may be used across different species of Bulimina spp. and can be extended to BWTs of 3°C.

Globocassidulina spp.
The combined Mg-Ca BWT data sets for Globocassidulina spp. from this study and previously published studies (Kristjánsdó ttir et al., 2007;Tisserand et al., 2013;Barrientos et al., 2018) indicate a robust positive relationship between Mg/Ca and BWT. As with the other two species presented herein, the variability within the dataset does indicate that factors other than BWT and analytical error may influence foraminiferal Mg/Ca ratios.

CONCLUSIONS
The aim of this study was to investigate the applicability of Mg/Ca benthic foraminiferal palaeothermometry to the climatically sensitive continental shelf of Antarctica. Of the three different species/genera studied Bulimina aculeata has the highest Mg/Ca sensitivity to BWT, followed by Globocassidulina spp. Trifarina angulosa also displays a positive relationship with temperature, but is less sensitive than the other two species. In this respect, it is advantageous to have multiple calibrations available in the paleothermometry toolkit as the availability of species varies across different oceanographic and environmental settings. This is particularly pertinent to Antarctic shelf areas where preservation of calcareous foraminifera is patchy at best.
The positive correlations between Mg/Ca ratios of the studied benthic foraminifera shells and BWT, whilst indicative of a strong relationship between the two, show considerable variance. Such variability indicates that Mg/Ca ratios in these foraminifera species may also depend upon other environmental parameters, which have not been evaluated in this study. Sampling of seafloor surface sediment and bottom (and pore-) water samples from the same sites and combined (geo-) chemical measurements on these samples are the next step forward in unravelling the role of these other factors. Mg/Ca-BWT calibrations, such as those produced here, require cautious use, demand further ground-truthing through application to down-core records, and should be used in conjunction with other proxies, such as those for seawater temperature, alkalinity and salinity. Furthermore, additional studies are required to determine if regional-only or global calibration datasets are most appropriate for down-core studies. For Trifarina angulosa, we observe that our regional calibration produces BWTs that are most comparable to measured seawater temperatures in the Amundsen Sea.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.