Refining benthic foraminiferal Mg/Ca-temperature calibrations using core-tops from the western tropical Atlantic: Implication for paleotemperature estimation


  • Amandine A. Tisserand,

    Corresponding author
    1. Uni Research AS, Bergen, Norway and Bjerknes Centre for Climate Research
    2. Department of Earth Science, University of Bergen, Bergen, Norway
    Search for more papers by this author
  • Trond M. Dokken,

    1. Uni Research AS, Bergen, Norway and Bjerknes Centre for Climate Research
    Search for more papers by this author
  • Claire Waelbroeck,

    1. LSCE/IPSL, Laboratoire CNRS-CEA-UVSQ, Gif Sur Yvette, France
    Search for more papers by this author
  • Jeanne-M. Gherardi,

    1. Uni Research AS, Bergen, Norway and Bjerknes Centre for Climate Research
    2. LSCE/IPSL, Laboratoire CNRS-CEA-UVSQ, Gif Sur Yvette, France
    Current affiliation:
    1. LSCE/IPSL, Laboratoire CNRS-CEA-UVSQ, Gif Sur Yvette, France
    Search for more papers by this author
  • Vincent Scao,

    1. Uni Research AS, Bergen, Norway and Bjerknes Centre for Climate Research
    2. LSCE/IPSL, Laboratoire CNRS-CEA-UVSQ, Gif Sur Yvette, France
    Current affiliation:
    1. LSCE/IPSL, Laboratoire CNRS-CEA-UVSQ, Gif Sur Yvette, France
    Search for more papers by this author
  • Christophe Fontanier,

    1. Laboratory of Recent and Fossil Bio-Indicators, UMR CNRS 6122 LPGN, Angers University, Angers, France
    2. Laboratoire Environnements, Paléoenvironnements Océaniques et Continentaux, UMR CNRS 5805, University of Bordeaux 1, Talence, France
    Search for more papers by this author
  • Frans Jorissen

    1. Laboratory of Recent and Fossil Bio-Indicators, UMR CNRS 6122 LPGN, Angers University, Angers, France
    Search for more papers by this author


[1] Benthic foraminiferal Mg/Ca has been shown to have great potential as a proxy for reconstructing deep water temperatures. However, the exact relationship between Mg uptake in benthic foraminifera and temperature is still ambiguous, and further exploration and refinement is much needed to reduce uncertainties associated with the method. Here, we present new core-top Mg/Ca data from benthic foraminiferal species from the lower part of the thermocline in the western tropical Atlantic (northern Brazilian margin). This area is unusual in that the changes in carbonate chemistry along the transect are very small, making it an ideal region for isolating and studying the role of temperature in the incorporation of Mg into the benthic shells. Our results show that benthic foraminiferal Mg/Ca largely reflects temperature in this area. Our data are combined with previously published data to produce new and improved Mg/Ca calibration equations for a number of benthic foraminiferal species within the Atlantic Ocean. Our study provides the first C. wuellerstorfi Mg/Ca data for the 4–6°C temperature range and indicates that C. wuellerstorfi Mg/Ca is strongly controlled by temperature. As a result, the newly established C. wuellerstorfi calibration over the entire 0–6°C temperature range is significantly improved with respect to previously published C. wuellerstorfi calibrations limited to the coldest part between 0 and 4°C. Other benthic species (Cibicidoides kullenbergi, Globocassidulina subglobosa, Uvigerina peregrina, and Oridorsalis umbonatus) have also been studied, although these results are less conclusive.

1 Introduction

[2] Over the past decade, Mg/Ca ratios measured in benthic foraminiferal shells have been used to estimate the temperature of the deep seawater in which the foraminifera calcified [Billups and Schrag, 2002; 2003; Bryan and Marchitto, 2008; Elderfield et al., 2006; 2010; 2012; Lea et al., 2000; Lear et al., 2000; Marchitto and deMenocal, 2003; Martin et al., 2002; Skinner et al., 2003; Sosdian and Rosenthal, 2009; Katz et al., 2011]. Independent temperature estimates from Mg/Ca ratios, in combination with the oxygen isotopic composition of the foraminiferal calcite (δ18Ocalcite), can help constrain the oxygen isotopic composition of seawater (δ18Oseawater), which itself reflects changes in the local hydrography [Duplessy et al., 1991; Waelbroeck et al., 2002]. The use of Mg/Ca paleothermometry on planktonic foraminiferal for reconstructing sea surface temperatures is now widespread [Anand et al., 2003; von Langen et al., 2005; Lea et al., 1999; Mashiotta et al., 1999; Nürnberg et al., 1996; Russell et al., 2004], but extensive studies during the last decades have also suggested that the same method may be applied to benthic species, in order to generate records of bottom water temperature (BWT) in a wide range of temperatures [Elderfield et al., 2006; Healey et al., 2008; Kristjánsdóttir et al., 2007; Lear et al., 2002; Marchitto et al., 2007; Martin et al., 2002; Rosenthal et al., 2006; 2011; Skinner et al., 2003; Yu and Elderfield, 2008].

[3] However, using benthic foraminifera Mg/Ca to reconstruct BWT is complicated: glacial-interglacial temperature variations in intermediate and deep waters are small compared to those seen at the surface, and as a result, the calibrations have to be far more tightly constrained in order to minimize uncertainties.

[4] A number of core-top and culture calibration studies on epifaunal and infaunal benthic species have been performed over the last 15 years (Table 1 and Figure 1 and references therein) and show positive correlations with temperature (e.g., Figure 1). Previously published benthic foraminifera Mg/Ca-temperature calibrations based on a range of species from the Cibicidoides genus cover a large temperature range, from −1.1 to 18.4°C (Table 1 and references therein). Previous studies have shown that the relationship describing Mg incorporation in shell calcite as a function of temperature is species specific [Elderfield et al., 2006; Lear et al., 2002; Rosenthal et al., 1997]. Calibrations based on multispecies Mg/Ca ratios thus include uncertainties due to interspecies differences. It is notable that existing Atlantic datasets contain a relatively large number of Cibicidoides spp. measurements from waters colder than 4°C (see histogram in Figure 1) but only a few measurements below 1°C or above 4°C. As a result, existing calibrations of Cibicidoides spp. are poorly constrained outside of this small 1 to 4°C window. Other studies that focus exclusively on Cibicidoides wuellerstorfi (a common deep and intermediate water species [Murray, 1991]) cover the colder part of the temperature range (<4°C) and suggest that the slope of the Mg/Ca-temperature calibration is steeper than that of the mixed Cibicidoides spp. calibration (Table 1 and Figure 1). The first possibility proposed to explain this difference was that C. wuellerstorfi Mg/Ca has a higher sensitivity to temperature compared to mixed Cibicidoides spp. Mg/Ca. However, the difference between C. wuellerstorfi and the mixed species calibration has led many to highlight seawater carbonate saturation as a potentially significant control on Mg incorporation into the test of this particular species [Lear et al., 2004; Martin et al., 2002]. Elderfield et al. [2006] showed that samples from the Norwegian Sea (subzero temperature and high carbonate ion saturation state inline image) give anomalous ratios (large range in Mg/Ca with little change in temperature), which led them to propose and establish the first empirical relationship between the inline image and Mg/Ca in C. wuellerstorfi, based on core-top samples. Since then, an increasing number of studies evaluating the effect of inline image on the Mg/Ca of benthic foraminifera have been conducted [Healey et al., 2008; Raitzsch et al., 2008; Yu and Elderfield, 2008]. Published studies show that Mg/Ca in C. wuellerstorfi and Cibicidoides kullenbergi (= Cibicidoides mundulus) shells are strongly affected by low inline image (defined as the difference between the measured carbonate ion concentration inline image and the carbonate ion concentration at calcite saturation inline imagesat), for BWT less than 4°C [Elderfield et al., 2006; Raitzsch et al., 2008; Yu and Elderfield, 2008]. Yu and Elderfield [2008] further assessed the influence of BWT and inline image on core-top samples from across the global ocean. They suggested that C. wuellerstorfi Mg/Ca ratios show a weak correlation with BWT but a strong one with inline image, whereas C. kullenbergi Mg/Ca ratios show no correlation with changes in BWT or inline image.

Table 1. Published Mg/Ca-Temperature Calibrations
ReferenceSpeciesMethodMg/Ca-Temperature RelationshipTemperature Range (°C) 
Billups and Schrag [2002]C. wuellerstorfiNo leachingMg/Ca = 0.32T + 0.761.6–5.34[1]
 C. kullenbergi    
Elderfield et al. [2006]Cibicidoides spp.OxidativeMg/Ca = 0.90 exp (0.11 BWT)−1.1–18[2]
Healey et al. [2008]C. wuellerstorfiReductiveMg/Ca = 0.781 exp (0.23 BWT)0.95–3.8[3]
Lear et al. [2002]Cibicidoides spp.ReductiveMg/Ca = 0.867 exp (0.109 BWT)0.8–18.4[4]
 including C. wuellerstorfi    
Marchitto et al. [2007]C. pachydermaReductiveMg/Ca = 0.116T + 1.25.8–18.6[5]
Martin et al. [2002]Cibicidoides spp.ReductiveMg/Ca = 1.22 exp (0.109 BWT)−1.1–18[6]
after Healey et al. [2008]C.wuellerstorfi Mg/Ca = 0.652 exp (0.28 BWT)1.8–3[7]
Raitzsch et al. [2008]C. wuellerstorfiLaser ablationMg/Ca = 0.830 exp (0.145 BWT)0.39–3.88[8]
 C. kullenbergiLaser ablationMg/Ca = 0.627 exp (0.143 BWT)0.43–14.85[9]
Rathburn and De Decker [1997]C. wuellerstorfiSonicationMg/Ca = 0.342T + 1.392.25–5.87[10]
Rosenthal et al. [1997]C. pachydermaLeachingMg/Ca = 1.36 exp (0.1 BWT)4.5–18[11]
Russell et al., [2004] (as cited by Martin et al., [2002])C. wuellerstorfiLeachingMg/Ca = 0.953 exp (0.206 BWT)0.4–2.8[12]
Yu and Elderfield [2008]C. wuellerstorfiOxidativeMg/Ca = 0.59 exp (0.28 BWT)0.98–4.26[13]
   (without carbonate ion saturation correction)  
Figure 1.

Single or multispecies of Cibicidoides spp. Mg/Ca-temperature calibration between 0 and 10°C. Numbers between brackets refer to calibrations listed in Table 1. The histogram shows the number of Mg/Ca measurement of Cibicidoides spp. (in gray) and Cibicidoides wuellerstorfi (in blue) in the Atlantic Ocean. The labeled numbers correspond to references in which the calibration equations have been used (see details in Table 1).

[5] Calibrations have also been published on infaunal species, which live within the sediment, such as Uvigerina species [Bryan and Marchitto, 2008; Elderfield et al., 2006; 2010; Lear et al., 2002; Yu and Elderfield, 2008], Hyalinea balthica [Rosenthal et al., 2011], Oridorsalis umbonatus [Healey et al., 2008; Lear et al., 2002; Rathman et al., 2004; 2008], the argonitic species Hoeglundina elegans [Rosenthal et al., 2006], and other species from the Arctic (Islandiella norcrossi/helenae, Melonis barleeanus, and Cassidulina neoteretis) [Kristjánsdóttir et al., 2007; Lear et al., 2002]. According to Elderfield et al. [2006; 2010], shallow infaunal species seem to be far less affected by changes in inline image than epifaunal species, largely because they calcify in pore waters within the sediment. The infaunal species O. umbonatus and Melonis spp. exhibit temperature sensitivities similar to those observed in benthic epifaunal species, whereas other infaunal species have a lower sensitivity. After correction for the inline image effect on Mg/Ca ratios, the temperature sensitivity of epifaunal foraminiferal species appears lower and similar to the sensitivity found for Uvigerina spp. [Raitzsch et al., 2008; Yu and Elderfield, 2008].

[6] Factors other than seawater inline image can also control Mg incorporation during calcite shell growth of benthic foraminiferal species, including salinity [Dissard et al., 2010; Kisakürek et al., 2008], pH [Russell et al., 2004], and microenvironment variations [Bentov and Erez, 2006; Erez, 2003]. Although it is known that test size might influence the Mg/Ca ratios in planktonic foraminiferal species [Elderfield et al., 2002], no studies have yet been conducted on benthic species. Although the effect of temperature on Mg/Ca ratios might, therefore, be the dominant factor for some benthic foraminiferal species, it is clearly important to assess the potential impact of other such factors.

[7] In order to address these questions, we present new benthic foraminiferal Mg/Ca results from core-top samples from a short depth transect (600–1000 m) from the western equatorial Atlantic Ocean, along the northern Brazilian margin. The core-top locations have a small temperature range of 4–6°C, but the sites are unusual in that they exhibit only very minor changes in inline image. This essentially removes any concern over the potential impact of inline image on Mg incorporation and allows us to study the key relationship of Mg/Ca to temperature. Hence, this study aims to further constrain the controls involved in benthic foraminiferal Mg incorporation (e.g., temperature versus inline image).

[8] We first examine the Mg/Ca-BWT relationships of six benthic species in order to assess the temperature sensitivity of each species. We then summarize new and published C. wuellerstorfi core-top Mg/Ca measurements from the Atlantic Ocean [Elderfield et al., 2006; Raitzsch et al., 2008; Yu and Elderfield, 2008]. Our study provides the first C. wuellerstorfi Mg/Ca data for the 4–6°C temperature range. The new C. wuellerstorfi results, together with published C. wuellerstorfi data, lead us to revise the published Mg/Ca-BWT calibration and extend the temperature range above 4°C.

2 Oceanographic Setting

[9] The study area is located on the upper continental slope off northeastern Brazil at ~4ºS (Figure 2a). This ocean margin is characterized by well-oxygenated bottom water, where organic matter fluxes are generally low. Productivity in the western tropical North Atlantic is considered to be similar to that of the oligotrophic open ocean, with a carbon production of 20−90 µmol−1 [Cooley and Yager, 2006]. Because of the oligotrophic conditions characterizing this area, some benthic foraminiferal species generally considered as deep-sea taxa, such as C. wuellerstorfi, O. umbonatus, and Globocassidulina spp., are found in fair numbers in the upper part of this transect (see section 3.1). Western tropical Atlantic surface waters are warm and saline. The upper 200 m of the water column is dominated by the North Brazilian Current, characterized by low nutrient content, with some supply from South Equatorial Current waters [Peterson and Stramma, 1991]. Below, South Atlantic Current waters from the South Atlantic subtropical gyre (200–500 m depth) help to transport heat and salt to the North Atlantic Ocean. The main thermocline is situated in the upper 500 m and spans a temperature interval from more than 25°C at the surface to less than 10°C in its lower part. Below the main thermocline, the dominant water mass is Antarctic Intermediate Water (AAIW), which is recognized by a prominent salinity minimum between ~700 and ~1200 m depth in the study area (Figure 2b). Hydrographic observations by Silva et al. [2009] indicate seasonal changes of ~2°C in the top 50 m of the water column. Below the seasonal thermocline, water properties are uniform throughout the year [Arz et al., 1998; Levitus et al., 1994].

Figure 2.

(a) Location of sediment transect cores (black circle) and profile chemistry section (red box). (b) Distribution of salinity (psu) in the modern tropical western Atlantic Ocean (generated with Ocean Data View) [Schlitzer, 2010]. The black circles show the location and water depths of core-tops used for this study (see location information in Table 2). (c) Temperature. (d) Δ[CO32−] profile calculated with the software CO2sys version 14 [Lewis and Wallace, 1998] and with data (temperature, salinity, phosphorous, and nitrate) from World Atlas Ocean 2009 [Antonov et al., 2010; Garcia et al., 2010a; 2010b; Locarnini et al., 2010] and data (alkalinity and TCO2) from Goyet et al. [2000]. (e) δ18O of seawater in the water column [this study; Ostlund et al., 1987; Karr and Showers, 2002]. AAIW: Antarctic intermediate water; u-NADW: upper North Atlantic deep water.

Table 2. Core-Top Locationsa
CoreLat. (°S)Long. (°W)WD (m)BWT (oC)SalinityΔ[CO32−]-calcite (µmol/kg)Sample Depth (cm)SpeciesAMS 14C ages (years BP)ErrorsCal. Ages (years BP)ErrorsSize FractionMg/Ca (mmol/mol)δ18O (‰)
U. PeregrinaB. InflataO. UmbonatusG. SubglobosaC. kullenbergiC. wuellerstorfiU. PeregrinaB. InflataO. UmbonatusG. SubglobosaC. kullenbergiC. wuellerstorfi
  1. aBWT from the CTD was taken during the cruise RETRO GS07-150.
GS07-150-22/1MC (I)04°16.23137°09.0515986.0634.533.20-0.5G. Sacculifer312330289749             
       core-top     -2.102.28----2.042.10----
       core-top     --2.34-----2.48----
       core-top     150–200----4.46-----1.98-
       core-top     150–250--3.103.37-2.95--2.062.66-1.66
       core-top     150–250--2.88--2.95-----1.35
       core-top     150–250--2.66---------
       core-top     250–315----4.142.41----1.911.41
       core-top     250–315-----2.53-----1.68
       core-top     315–400-----3.07-----1.77
       core-top     315–400-----3.91-----1.87
       core-top     250–400--2.321.74----2.412.63--
       core-top     250–400--2.71-----2.43---
GS07-150-21/1MC (II)04°15.98137°08.7116565.6734.4733.60–0.5G. Sacculifer263320231730             
       core-top     150–250---1.93-2.55---2.40-1.81
       core-top     200–250----5.06-----1.95-
       core-top     250–315-----2.76------
       core-top     250–315-----2.04-----1.96
       core-top     250–400---1.71-----2.28--
GS07-150-20/1MC (III)04°15.65537°08.2437005.3834.4637.50–0.5G. Sacculifer265330232824             
       core-top     150–250---1.75-2.56---2.52-0.91
       core-top     150–250-----2.23-----1.96
       core-top     200–250----4.83-----1.98-
       core-top     250–315----3.762.24----2.03-
       core-top     250–315-----2.31-----2.02
       core-top     315–400-----2.44-----1.87
       core-top     315–400-----2.30-----1.97
       core-top     250–400---1.70-----2.66--
GS07-150-26/2MC (IV)04°15.26737°07.7617504.9234.45370–0.5G. Sacculifer207820165140             
       core-top     -1.142.48----2.602.88----
       core-top     150–200----4.80-------
       core-top     200–250----4.77-----2.04-
       core-top     150–250--2.441.80-2.44---2.60-2.07
       core-top     150–250-----2.65-----2.06
       core-top     150–250-----2.55-----2.12
       core-top     250–315-----1.80-----1.97
       core-top     250–315-----2.45-----2.10
       core-top     315–400-----2.00-----1.96
       core-top     250–400--2.611.50----2.622.78--
GS07-150-19/1MC (V)04°14.86837°07.1978004.6434.4436.40–0.5G. Sacculifer7273036845             
       core-top     -1.331.86----2.862.79----
       core-top     150–200----4.50-----2.06-
       core-top     200–250----4.10-----1.98-
       core-top     150–250---1.61-2.40---2.75-2.13
       core-top     150–250-----2.11-----2.13
       core-top     250–315----3.962.32----2.102.26
       core-top     250–315-----1.88-----2.14
       core-top     315–400-----2.05------
       core-top     315–400-----2.10-----2.04
       core-top     250–400--2.261.51-----2.67--
GS07-150-18/1MC (VI)04°13.82737°05.9458984.334.4734.70-0.5G. Sacculifer6193027128             
       core-top     -1.331.70----2.702.94----
       core-top     150–200----5.18-----2.08-
       core-top     200–250----3.25-----2.16-
       core-top     150–250--2.901.65-2.08--2.492.72-2.16
       core-top     250–315----3.712.16----1.98-
       core-top     250–315-----1.95-----2.15
       core-top     315–400-----1.81-----2.19
       core-top     250–400--2.271.51----2.522.68--
GS07-150-17/2MC (V)04°12.98637°04.51810004.1734.56334–4.5G. Sacculifer13032079638             
       core-top     -1.141.74----2.862.93----
       core-top     150–200----4.29-----2.33-
       core-top     200–250----4.10-----2.27-
       core-top     150–250--3.011.73-----2.84-2.46
       core-top     250–315----4.401.91----2.362.02
       core-top     315–400-----2.09-----2.32
       core-top     315–400-----2.03-----2.14
       core-top     250–400--2.311.49----2.952.78--

[10] The CTD (conductivity-temperature-depth) cast (station 13) (04º15.403S; 37º03.083W; 1250 m depth), performed during the (R/V) G.O. Sars 2007 cruise, shows a temperature profile similar to the annual temperature profiles from the World Ocean Atlas 2009 [Antonov et al., 2010; Garcia et al., 2010a; 2010b; Locarnini et al., 2010] (Figure 2c), confirming that these profiles adequately describe the physical properties along the margin where our samples were collected. During the same cruise, seawater samples were collected near the coring sites (03º46.443S; 37º03.857W) at 300 and 500 m water depth. The oxygen isotopic ratio of these water samples is similar to the values described for nearby sites in the global seawater δ18O database edited by Schmidt et al. [1999] and available at (e.g., Figure 2e [Karr and Showers, 2002; Ostlund et al., 1987]).

[11] AAIW is characterized by low inline image and low salinity (~34.5). There are no large variations in salinity, carbonate ion saturation, and δ18Oseawater, which might have a secondary influence on the Mg/Ca ratio along the coring transect (Figures 2d and 2e). The temperature range is also rather small, but this allows us to test the sensitivity of the Mg/Ca method over a temperature range of about 2°C (between 4 and 6°C).

3 Materials and Methods

[12] Surface sediment samples from the depth transect were collected using a multicorer during the 2007 RETRO GS07-150 cruise, with the research vessel (R/V) G.O. Sars. Multicores were retrieved from seven sites located between 600 and 1000 m depth on the upper continental slope off northeastern Brazil at about 4ºS and 37ºW (Figures 2a and 2b and Table 2). The surface sediment samples analyzed in this study correspond to the top 0–0.5 cm interval of the multicore. Based on Accelerator Mass Spectrometry (AMS) 14C ages of surface sediments from gravity cores retrieved in the same area, Arz et al. [1998; 1999] indicated a late Holocene age, with sedimentation rates of 15 cm/1000 years. AMS 14C dates based on monospecific samples of Globigerinoides sacculifer in the seven multicores were measured in the Rafter Radiocarbon Laboratory in New Zealand and give a calibrated radiocarbon age of between 271±28 and 2897±49 years before present (BP) (Table 2). Radiocarbon ages were calibrated using the Calib6.01 software and the marine 09 calibration curve, with a standard reservoir correction of 400 years given by the calibration program [Reimer et al., 2009]. As multicoring does not disturb the sediment-water interface, we are confident that the analyzed samples from the depth transect are true surface sediment samples and thus representative of modern or late Holocene conditions. Aragonitic pteropod shells are present in all multicore samples, suggesting negligible post-depositional dissolution.

[13] Carbonate calculations of the system include the effects of temperature, salinity, pressure, phosphate, silicate, alkalinity, and total DIC (dissolved inorganic carbon), using the software CO2sys version 14 [Lewis and Wallace, 1998]. DIC is corrected for anthropogenic CO2. Except for temperature, which has been measured with CTD, chemistry data were estimated using the nearest stations in the Global Ocean Data Analysis Project database [Key et al., 2004].

3.1 Mg/Ca Analysis

[14] The sediment samples were washed and wet sieved at 63 and 150 µm. Samples were then dry-sieved at 250, 315, and 400 µm for C. wuellerstorfi and C. kullenbergi and at 250 µm for some of the infaunal species, to obtain 7–25 benthic foraminifer specimens per sample. Six benthic foraminiferal species (Cibicidoides wuellerstorfi, Cibicidoides kullenbergi, Globocassidulina subglobosa, Uvigerina peregrina, Oridorsalis umbonatus, and Bulimina inflata) were picked from the >150 µm fraction. U. peregrina, O. umbonatus, and G. subglobosa are shallow infaunal species. C. wuellerstorfi has an epifaunal microhabitat and is thus exposed to the ambient bottom water conditions, which here are characterized by a low organic matter flux to the sediment [Altenbach and Sarnthein, 1989]. In the Northeastern Atlantic, C. wuellerstorfi has been found between 1360 and 4280 m [Murray et al., 1991] but may be present in small numbers around 1000 m [Fontanier et al., 2013; Eberwein and Mackensen, 2006]. A recent study from New Zealand demonstrates that high relative abundances of C. wuellerstorfi have been found from upper bathyal to upper abyssal depths (400–3000 m) [Hayward et al., 2010]. The distribution of benthic species is controlled by two principal factors: the supply of organic matter and oxygen content [Jorissen et al., 1995; Levin and Gage, 1998; Van der Zwaan et al., 1999]; however, current intensity, sediment grain size, and the dominant water masses in the area are also known to affect the representation and abundance of the species [Jorissen et al., 2007]. In our study area, the presence of C. wuellerstorfi in waters shallower than 1000 m may be explained by the exceptionally low organic matter flux to the sea floor, which is typical for the continental slope off northeastern Brazil (see section 2). C. wuellerstorfi is able to live in shallow waters above 1000 m depth in this region. The very high abundance of Globocassidulina spp., which is generally considered as an oligotrophic taxon [Schmiedl et al., 1997; Mackensen et al., 1995], confirms the low export production in this region, with only intermittent organic matter input.

[15] Sample weights greater than 0.175 mg are required for each Mg/Ca analysis. In this study, measured samples generally weighed 0.300–0.400 mg (typically corresponding to ~15–20 shells). We separated specimens by size fraction to test the potential size effect on Mg/Ca ratios. We crushed foraminiferal shells between two glass plates to break the chambers and homogenized them before splitting them into sub-samples for Mg/Ca replicates and δ18O analyses. Some samples did not contain sufficient material to conduct both Mg/Ca and δ18O replicates.

[16] The crushed foraminifera were cleaned for contaminant phases following the procedure developed by Barker et al. [2003]. This method includes a clay removal step, oxidation of the organic matter, and surface leaching. The samples were analyzed for Mg/Ca, Fe/Ca, and Mn/Ca by using a Thermo Finnigan IRIS inductively coupled plasma optical emission spectrometer at Bergen GeoAnalytical Facility at the Department of Earth Sciences, University of Bergen. For every six to eight samples, known standard solutions were analyzed to correct for instrumental biases and analytical drift of the instrument.

[17] Long-term instrumental precision of elemental ratio data (determined by 60 replicate analyses of a standard solution containing Mg/Ca = 1.020 mmol/mol) is ±0.043 mmol/mol (1σ) or 0.81% (relative standard deviation (RSD)). The pooled standard deviation of the duplicate Mg/Ca determination for the epifaunal species C. wuellerstorfi (based on 63 replicates (from 7 to 25 specimens)) is equivalent to an overall average precision of 5% (pooled RSD), including artifacts from cleaning, picking, and sample heterogeneity, in addition to instrumental precision. The epifaunal species C. kullenbergi has not been replicated. For the infaunal species, only O. umbonatus has been replicated and the average precision is equivalent to 7.1% (pooled RSD). Measured Mg/Ca values show no correlation with Fe/Ca and Mn/Ca, indicating the absence of contamination due to insufficient cleaning or the presence of diagenetic coatings or silicates. Fe/Ca is <0.1 mmol/mol on all except for the C. kullenbergi Mg/Ca analyses. However, there is no trend between Mg/Ca and Fe/Ca in C. kullenbergi (r2 = 0.1).

3.2 Stable Isotopes Analysis

[18] Stable isotope analyses were conducted on the same sample material as thetrace element analyses. The foraminifera shell fragments were cleaned in alcohol in an ultrasonic bath and dried at 60°C. The isotope measurements were carried out at the stable isotope laboratory at the Department of Earth Sciences at University of Bergen. Between ~30 and 50 µg of foraminifera were analyzed for each sample on a Finnigan MAT 253 mass spectrometer coupled to an automated Kiel carbonate preparation device. The isotopic values were calibrated to Vienna Pee Dee Belemnite using the standards NBS19 and NBS18. The long-term δ18O analytical precision (1σ) of the standard is ±0.08‰.

4 Results

[19] Figure 3 shows Mg/Ca ratios and δ18O of calcite plotted versus water depth for all the infaunal foraminifera species studied. Globocassidulina subglobosa shows a nearly constant Mg/Ca value (1.79 ± 0.47 mmol/mol) along the slope for the two size fractions analyzed; it seems not to be significantly affected by test size or by temperature (Mg/Ca = 1.98±0.62 mmol/mol for the 150–250 µm size fraction; 1.59±0.12 mmol/mol for the 250–355 µm size fraction) (Figure 3a). The δ18O value of G. subglobosa tends to be lower in the upper part of the depth transect, except the upper site at 600 m, which displays higher values (Figure 3b).

Figure 3.

(a, c) Shell Mg/Ca and (b, d) δ18O for infaunal foraminiferal species plotted against water depth. The Mg/Ca axis is adjusted to the temperature axis by using an exponential relationship (adopted from Elderfield et al. [2006]).

[20] Mg/Ca ratios of O. umbonatus are systematically lower in the 250–400 µm size fraction (2.41±0.19 mmol/mol) than in the 150–250 µm size fraction (2.83±0.25 mmol/mol). This difference is particularly apparent in the two deepest sites (17/1MC - 1000 m and 18/2MC - 900 m) (Figure 3c). In the 150–250 µm size fraction, Mg/Ca values from the upper site are similar to those from the lower sites, with minimum values found at 750 and 800 m depths. The Mg/Ca values for the 250–400 µm size fraction seem in line with temperature, with Mg/Ca for the uppermost sites (750–600 m) displaying a slightly higher Mg/Ca ratio (mmol/mol) compared to that for the deeper sites. Overall, however, there is no clear relationship between Mg/Ca and temperature for O. umbonatus. There is a lack of O. umbonatus δ18O measurements in the 150–250 µm size fraction, as too few specimens were obtained to achieve reliable measurements (Figure 3d). In the 250–400 µm size fraction, there is ~0.8‰ change toward lighter values in δ18O from the deepest site to the shallow site, corresponding to ~2°C if the δ18O of the water is constant. However, the pattern of O. umbonatus δ18O changes versus depth is not regular and does not follow temperature.

[21] The size effect in U. peregrina and B. inflata has not been tested due to their limited abundances in the samples. Results for these species are therefore a composite of all size fractions (Figures 3c and 3d). Core-top data for these two species show a slight decrease in Mg/Ca with decreasing BWT. The δ18O of these species also tends to exhibit lighter δ18O at shallower sites.

[22] Figure 4 presents the Mg/Ca ratios and δ18O of calcite versus water depth for the epifaunal C. wuellerstorfi and C. kullenbergi. Cibicidoides kullenbergi Mg/Ca shows a rather large scatter, both for different water depths (when a single size fraction is considered) and for different size fractions (Figure 4a). Our C. kullenbergi Mg/Ca values are significantly higher than other published Mg/Ca values from the Atlantic Ocean [Elderfield et al., 2006; Lear et al., 2003; Raitzsch et al., 2008; Yu and Elderfield, 2008]. The temperature difference between this study area and those used in the previous studies in the Atlantic may account for some of this difference. In spite of rigorous cleaning, iron contamination is apparent in C. kullenbergi measurements, with Fe/Ca ratios 10 times higher than the recommended maximum ratio [Barker et al., 2003]. However, there is no significant correlation between Mg/Ca and Fe/Ca (r2 = 0.1). This is unlikely to be the result of inadequate cleaning since the cleaning procedure for C. kullenbergi was identical to that of C. wuellerstorfi. Also, the measured C. kullenbergi δ18O values do not show the same large scatter as the measured C. kullenbergi Mg/Ca (Figure 4b).

Figure 4.

(a, c) Shell Mg/Ca and (b, d) δ18O for epifaunal foraminiferal species plotted against water depth. The Mg/Ca axis is adjusted to the temperature axis by using an exponential relationship (adopted from Elderfield et al. [2006]).

[23] Figure 4c shows a decrease in C. wuellerstorfi Mg/Ca values with decreasing BWT. There is no apparent size effect in the dataset for the three size fractions tested (Figure 4c). These results clearly suggest a positive relationship between Mg/Ca and calcification temperature, and a higher temperature sensitivity than observed for other species in this study (Figures 3c and 4c). Cibicidoides wuellerstorfi δ18O values are also seen to increase with increasing water depth, closely following observed temperatures along the transect (Figure 4d).

5 Discussion

[24] Since the bottom waters associated with our core transect are characterized by very little variability in all properties except temperature, we except this to be the dominant control on Mg/Ca and δ18O for the different species included in this study.

5.1 Mg/Ca of Infaunal Species

[25] For most of the infaunal species, our results do not clearly indicate on how Mg/Ca and δ18O respond to changes in physical properties (i.e., temperature) within the water column. In the upper part of the transect, we observe a tendency for high Mg/Ca ratios (mmol/mol) to correlate with higher BWT (Figure 3). Infaunal foraminiferal δ18O also show slightly lower values moving upward in the water column.

[26] To extend the temperature range of our calibration, we combined different datasets from different regions of the world ocean (Figure 5) [Bryan and Marchitto, 2008; Elderfield et al., 2010; Healey et al., 2008; Lear et al., 2002; Rathman et al., 2004]. Yu and Elderfield [2008] have attributed a small difference of 0.09 mmol/mol between results obtained using the oxidative versus the reductive cleaning methods [Boyle and Keigwin, 1985; 1986]. As we have used the oxidative method [Barker et al., 2003], we applied the +0.09 mmol/mol correction to the Uvigerina species and O. umbonatus datasets to allow us to directly compare all the data [Bryan and Marchitto, 2008; Healey et al., 2008; Lear et al., 2002]. The combined Mg/Ca linear regression for Uvigerina spp. and U. peregrina measurements indicates only a weak dependence on temperature, with a slope of ~0.068±0.007 mmol/mol/°C, over a temperature range of 1–11°C (Figure 5a). As shown by Bryan and Marchitto [2008], there does not appear to be significant interspecies differences within the genus Uvigerina, in contrast to the clear differences observed between Cibicidoides species [Elderfield et al., 2006].

Figure 5.

Mg/Ca versus temperature: (a) U. peregrina and Uvigerina spp. and (b) O. umbonatus.

[27] Our O. umbonatus Mg/Ca data agree well with the dataset from Rathman et al. [2004] (Figure 5b). Most of the O. umbonatus Mg/Ca ratios from Lear et al. [2002] have lower values than those from Rathman et al. [2004] study. They are also lower than our values, despite the adjustment made to account for the different cleaning methods. The Mg/Ca data from Healey et al. [2008] show a higher sensitivity than the other data [this study; Lear et al., 2002; Rathman et al., 2004], and this cannot only be attributed to the different cleaning method used. Moreover, the O. umbonatus Mg/Ca ratios also show differences related to test size. This suggests that there may be an ontogenic effect on the O. umbonatus temperature response. Further work is required to clarify the Mg/Ca-temperature relationship for O. umbonatus and to better constrain the warmest part of the calibration (6–11°C).

[28] To summarize, our results suggest that, although the O. umbonatus Mg/Ca temperature sensitivity is potentially high, other factors such as test size [this study] and inline image from pore water [Spero et al., 1997] may substantially influence the Mg/Ca values. More studies on O. umbonatus Mg/Ca are thus necessary before Mg/Ca from this species can be used as a reliable temperature recorder.

5.2 Mg/Ca of Epifaunal Species

[29] We separate C. wuellerstorfi and C. kullenbergi because they have different behavior with respect to trace element incorporation. This will be further discussed in the following sections.

5.2.1 C. kullenbergi

[30] Although our results exhibit unusually high Mg/Ca values in Figure 4a, we examine the new C. kullenbergi Mg/Ca data together with other published Atlantic Mg/Ca studies [Elderfield et al., 2006; Raitzsch et al., 2008; Healey et al., 2008; Yu and Elderfield, 2008] (Figure 6). Again, we have applied the same +0.09 mmol/mol correction to samples cleaned with the reductive method [Elderfield et al., 2006; Healey et al., 2008; Yu and Elderfield, 2008]. We observe a large scatter in C. kullenbergi Mg/Ca through the 0–15°C temperature range (Figure 6). We obtain the following equation:

display math(1)
Figure 6.

Core-top C. kullenbergi Mg/Ca versus BWT from Atlantic sites from different sources. Published temperature calibrations (gray lines) on C. kullenbergi are shown.

[31] This calibration equation does not include the published data between 5 and 15°C (see Figure 6 and data represented in gray). By increasing the number of data points in the 4–6°C range (where data were previously seriously lacking), we show that our calibration has much steeper slopes than the calibration of Raitzsch et al. [2008] but agrees well with the exponential relationship determined by Healey et al. [2008] for the same species. Although C. kullenbergi exhibits very high Fe/Ca values and we cannot entirely exclude a potential contamination, this species Mg/Ca does appear to exhibit strong temperature sensitivity. Cibicidoides kullenbergi certainly has some potential for reconstructing temperature based on Mg/Ca, but more studies will be needed to confirm these results.

5.2.2 C. wuellerstorfi

[32] We examine our C. wuellerstorfi Mg/Ca data from the Brazilian margin alongside published Atlantic Mg/Ca data from Elderfield et al. [2006], Raitzsch et al. [2008], and Yu and Elderfield [2008] (Figure 7). The Norwegian Sea data are discounted here for the reason mentioned previously. We also exclude core M35024-6 from the Caribbean basin (17ºN, 66ºW, depth 4.71 km, T = 4.35°C, S = 34.98, inline image = 3.1 µmol/kg), since it shows anomalously low Mg/Ca values for warm temperatures (see Yu and Elderfield [2008] for further details). To avoid introducing noise due to different cleaning methods, we only considered C. wuellerstorfi Mg/Ca data obtained using the oxidative technique [this study, unpublished data; Elderfield et al., 2006; Yu and Elderfield, 2008] and laser ablation [Raitzsch et al., 2008].

Figure 7.

(a) Core-top C. wuellerstorfi Mg/Ca versus BWT from Atlantic sites from different sources. The black error bars associated to the marks represent standard errors. The dark line represents the Mg/Ca-temperature calibration for ~0–6°C temperature range. Dashed lines bracket the 95% confidence interval. (b) Core-top C. wuellerstorfi Mg/Ca versus Δ[CO32−] from Atlantic sites from different sources.

[33] Combining published C. wuellerstorfi Mg/Ca values from the 0–4°C temperature range with our Mg/Ca data from the 4–6°C range, we obtain the following relationship over the entire 0–6°C range (Figure 7a):

display math(2)

[34] Our results show that C. wuellerstorfi Mg/Ca temperature has a steeper regression slope compared to those presented for mixed Cibicidoides spp. studies [Elderfield et al., 2006], reinforcing the conclusion that the temperature sensitivity of C. wuellerstorfi is much higher than that of mixed Cibicidoides spp. (Figure 1 and calibration equation (2)). As previously discussed, Elderfield et al. [2006] suggested that the difference between C. wuellerstorfi and mixed Cibicidoides spp. calibrations (including the Norwegian Sea data) might be related to inline image effect. Also, other studies suggest that BWT does not drive C. wuellerstorfi Mg/Ca ratios at low temperatures (below 3°C) due to the effect of inline image [Yu and Elderfield, 2008; Raitzsch et al., 2008]. Yu and Elderfield [2008] therefore recommend separating the temperature signal and the inline image effect on the epifaunal benthic Mg/Ca ratios. Ideally, temperature calibrations should be produced in areas where the inline image can be controlled or is constant. As we have shown previously, the northeastern Brazilian margin is unusual in that it is characterized by a homogeneous inline image between ~500 and 1200 m (Figures 2d and 7b). Consequently, we have produced the Mg/Ca-temperature calibration extension for C. wuellerstorfi in a context where the parameter considered most likely to complicate Mg/Ca paleothermometry, i.e., inline image is controlled. When combining these new data with existing C. wuellerstorfi Mg/Ca data, we observe that the low temperature data do not appear anomalous but are consistent with the exponential curve describing the warmer temperature data. This lends strong support to the temperature control on C. wuellerstorfi Mg/Ca. However, we recommend that Mg/Ca calibration work in the future should focus on the temperature extremities of the calibration, where the data are sparse.

[35] Salinity variations are also very small over the Brazilian depth transect. It is therefore reasonable to assume that salinity is unlikely to exert a major control on Mg/Ca in our area either. However, the salinity effect on Mg incorporation into benthic foraminiferal calcite is poorly constrained in the literature [Dissard et al., 2010; Toyofuku et al., 2000] and further work is required to understand its potential impact on benthic foraminifer Mg/Ca ratios.

6 Conclusions

[36] In this study, we present new benthic foraminiferal Mg/Ca-BWT calibrations based on additional data from the western tropical Atlantic Ocean. We have produced benthic foraminiferal Mg/Ca data for the 4–6°C temperature range, using core-top samples from an area characterized by very low variability in water mass properties such as inline image and salinity. This allows for more reliable calibrations of Mg/Ca versus BWT for the benthic species studied. We have also examined the species-specific Mg/Ca-BWT sensitivity as well as test size effects on Mg/Ca ratios.

[37] Our results suggest that the temperature sensitivity of the infaunal species O. umbonatus is higher than that of the infaunal species Uvigerina peregrina. However, there may be an ontogenic effect on Mg incorporation for O. umbonatus at the deepest sites. Our results also show that C. kullenbergi Mg/Ca-temperature calibrations may have potential but need further study.

[38] Combining our new C. wuellerstorfi data with published data, we have extended the existing Mg/Ca-temperature calibration for this species over the temperature range from 4 to 6°C in the Atlantic Ocean. Cibicidoides wuellerstorfi Mg/Ca exhibits higher temperature sensitivity over the 0–6°C temperature range than mixed Cibicidoides spp. Mg/Ca. Since inline image is homogeneous and stable in this region, we can essentially remove the question of its influence above 4°C and show that temperature is indeed the major control on C. wuellerstorfi Mg/Ca.

[39] Although further work is required in other parts of the Atlantic Ocean to confirm this calibration, our results clearly indicate that among the six most commonly used benthic species, C. wuellerstorfi has the greatest potential for reconstructing BWT between 0 and 6°C and more robustly above 4°C.


[40] This work is a contribution to the RETRO project, a joint European Science Foundation (ESF)/EUROMARC, funded by Research Council of Norway (RCN), France (CNRS/INSU), Germany, and the Netherlands. The cores and in situ hydrographic conditions were recovered by (R/V) G.O. Sars during RETRO Cruise I in 2007. We appreciate the help from the captain and crew of (R/V) G.O. Sars. We gratefully acknowledge U. Ninnemann, J. Stromsøe, and R. Søraas for technical assistance. The authors would like to thank C. Andersson for constructive suggestions and E. Farmer for improving the written English. We are grateful to the anonymous reviewer, the associate editor, G. Foster, and the editor, J. Tyburczy, for their constructive comments on an earlier version of the manuscript. This is Bjerknes Centre for Climate Research contribution number A414.