Geochemistry, Geophysics, Geosystems

The relationship between shell size and Mg/Ca, Sr/Ca, δ18O, and δ13C of species of planktonic foraminifera

Authors


Abstract

[1] The relationship between shell size and Mg/Ca, Sr/Ca, δ18O, and δ13C of planktonic foraminiferal calcite has been investigated using seventeen species in six different size fractions. Mg/Ca increases and Sr/Ca decreases with increasing size, except for two globorotaliid species which show the opposite trend. The changes in Mg/Ca broadly follow δ18O calcification temperatures except that surface and near-surface dwelling species show larger changes in Mg/Ca than can be accounted for by differences in calcification temperature derived from δ18O. The increases in Mg/Ca and decreases in Sr/Ca vary linearly with the well-established increase in δ13C with size. This is consistent with smaller individuals calcifying faster than larger individuals and larger individuals forming calcite that more closely reflects seawater temperature and composition. It appears that variations in calcification rate affect Mg and Sr shell chemistry. The observations may account for part of the temporal variability in foraminiferal Sr/Ca within a single size fraction that has been attributed to changes in seawater Sr/Ca.

1. Introduction

[2] Two issues have dominated paleoceanographic research using Mg/Ca and Sr/Ca ratios of the calcium carbonate shells of foraminifera: The first of these is the refinement and applications of foraminiferal Mg/Ca thermometry [e.g., Nürnberg et al., 1996; Rosenthal et al., 1997, 2000; Hastings et al., 1998; Lea et al., 1999; Mashiotta et al., 1999; Elderfield and Ganssen, 2000; Lear et al., 2000; Toyofuku et al., 2000]. Studies of foraminifera in culture [e.g., Lea et al., 1999] and from core tops [e.g., Elderfield and Ganssen, 2000] have shown that the Mg/Ca ratio is related to growth temperature with an exponential temperature dependence. Although other paleothermometers have been developed, the potential of using Mg/Ca is that it provides a species specific temperature estimate and can be used in conjunction with measurement of δ18O to estimate the oxygen isotopic composition of seawater, providing information on past salinity and ice volume. It seems clear from the previous work referred to earlier that an exponential temperature dependence of about a 10% increase in Mg/Ca°C exists in planktonic foraminifera.

[3] The second issue is the question of whether records of foraminiferal Sr/Ca provide information on past changes in seawater Sr/Ca [e.g., Graham et al., 1982; Stoll and Schrag, 1998; Martin et al., 1999; Stoll et al., 1999; Elderfield et al., 2000]. Assuming that the Sr/Ca ratios of foraminifera are related to seawater Sr/Ca through a constant partition coefficient, estimates have been made of changes in ocean composition both on Cenzoic timescales [Graham et al., 1982] and on Quaternary timescales [Stoll and Schrag, 1998; Martin et al., 1999].

[4] Although work to date has led to important advances in the use of these tracers, uncertainties exist in both and there is a need to evaluate secondary factors that affect Mg/Ca and Sr/Ca in foraminiferal calcium carbonate. These factors are thought to include shell growth rate, pH and salinity, and species-specific effects [Delaney et al., 1985; Elderfield et al., 1996; Rathburn and De Decker, 1997; Rosenthal et al., 1997; Isuka, 1988; Lea et al., 1999; Toyofuku et al., 2000].

[5] Here we present results of a study of Mg/Ca and Sr/Ca ratios in 17 species of planktonic foraminifera in six size fractions from a single sample recovered near the top of a northeast Atlantic sediment core, together with ancillary data on δ18O and δ13C. The results allow us to address two general issues. The first is that studies on multiple species of planktonic foraminifera from a single site have the potential to reconstruct the hydrography of the upper ocean using species that inhabit surface waters and those that lay down a significant proportion of their calcite in deeper waters. The second is that the choice of shell size in planktonic foraminifera is a practical issue of importance in the design of methodological procedures for paleoceanographic work. It is well established, for example, that increases occur in δ13C of planktonic foraminifera as a function of size [e.g., Berger et al., 1978; Oppo and Fairbanks, 1989]. The difference in δ13C between size fractions in a single sample is greater than glacial-interglacial δ13C amplitudes [Oppo and Fairbanks, 1989]. Size-related differences in important paleoceanographic tracers do not preclude their use, however, provided such information is known and understood. As the work is based on ∼10–13% of the seafloor, we cannot claim it is representative of all situations, but the results should nevertheless be instructive.

2. Samples and Methods

[6] BOFS core 31 K was recovered from the north Atlantic at 19°00'N, 20°10'W, water depth 3200 m [Chapman et al., 1996]. Seventeen species of planktonic foraminifera were used, three of which are represented by two varieties (Table 1). Samples were picked, where possible, of 20 individuals of each species in six size fractions: 212–250 μm, 250–300 μm, 300–350 μm, 350–425 μm, 425–500 μm, and >500 μm from 10–12 cm in the core (of Holocene age). Not all species of foraminifera were present in all size fractions, and in some cases, few individuals were available for analysis. The smallest-sized individuals were 212 μm, and therefore no juveniles were analyzed. We also emphasize that sizes of foraminifera were not measured, and the sizes quoted are size fractions defined by sieving. Sedimentation rates were of the order of 2.5 cm ka−1 in the Holocene so the sampling interval corresponds to ∼800 years. Carbonate preservation was evaluated by Chapman et al. [1996] who concluded from fragmentation index measurements that the foraminiferal assemblage is well preserved for faunal estimation of sea surface temperature.

Table 1. Species of Planktonic Foraminifera Analyzed
Globigerina bulloidesGloborotalia inflata
Globigerinella calidaGloborotalia menardii
Globigerinella siphoniferaGloborotalia scitula
Globigerinita glutinataGloborotalia truncatulinoides (r)
Globigerinoides conglobatusGloborotalia truncatulinoides (l)
Globigerinoides ruber (w)Globorotalia tumida
Globigerinoides ruber (p)Neogloboquadrina dutertrei
Globigerinoides sacculiferNeogloboquadrina pachyderma (r)
Globigerinoides trilobusOrbulina universa
Globorotalia crassaformisSphaeroidinella dehiscens

[7] The δ18O and δ13C were obtained using a VG Prism mass spectrometer. Prior to cleaning for elemental analysis, the foraminifera tests were gently crushed between glass plates to ensure that all the chambers were opened. The cleaning procedure involved washing with water and methanol to remove clays etc., hot alkaline oxidative cleaning using buffered hydrogen peroxide and a short (30 s) leach with 0.001 M nitric acid followed by dissolution in 400 μL of 0.075 M nitric acid. The samples were analyzed using a Varian Vista simultaneous ICP-AES. A micronebuliser and concentric spray chamber were used, which allowed the analyses to be carried out on between 200 and 400 μL of sample. At the time of the analyses we had not evaluated the long-term precision of the method but a conservative estimate was of relative standard deviation, rsd, <0.4%. Subsequent development to the ICP-AES methodology [de Villiers et al., 2002] has increased its accuracy to comparable to its precision (<0.3%), but the accuracy of the data presented here far exceeds the variability seen. A significant issue in this respect is the matrix effect on Mg/Ca and Sr/Ca resulting from variations in Ca2+ in the analyte solution. Under the conditions employed at a Ca concentration of ∼20 ppm, the error associated with offsets from a Ca concentration of 20 ppm is as much as 0.7% Mg/Ca/ppm Ca. At ∼60 ppm, the apparent optimum sample Ca concentration, the error associated with offsets in Ca concentration from 60 ppm is ∼0.02–0.05%/ppm Ca [de Villiers et al., 2002]. In comparison, the ranges in Mg/Ca are from 12 to 36% of the mean value over the size range. All results are listed in Table A1.

Table A1. Chemical Data for Planktonic Foraminifera From BOFS core 31 K
Size Fraction, μmδ18O, ‰δ13C, ‰Mg/Ca, mmol/molSr/Ca, mmol/mol
Globigerina bulloides
212–250−0.50−1.212.981.41
250–300−0.37−0.893.431.39
300–350−0.24−0.933.121.394
350–425−0.20−0.813.251.33
425–500    
>500    
 
Globigerinella calida
212–250−0.1600.042.841.36
250–3000.0000.313.181.33
300–3500.1700.283.441.26
350–4250.0300.454.051.28
425–500    
>500    
 
Globigerinella siphonifera
212–250−0.100.692.751.31
250–300−0.280.952.491.31
300–3500.000.843.511.27
350–4250.001.073.651.40
425–500    
>500    
 
Globigerinita glutinata
212–250−1.05−1.582.551.43
250–300    
300–350    
350–425    
425–500    
>500    
 
Globigerinoides conglobatus
212–250  2.451.33
250–300    
300–350    
350–425    
425–500    
>500    
 
Globigerinoides ruber (white)
212–250−1.160.743.051.46
250–300−1.030.973.431.46
300–350−1.001.113.541.44
350–425−1.281.343.721.43
425–500−1.351.463.881.42
>500    
 
Globigerinoides ruber (pink)
212–250−1.531.614.131.41
250–300−1.591.773.511.44
300–350−1.541.673.651.47
350–425−1.591.363.461.46
425–500−1.581.303.871.47
>500−1.260.813.541.43
 
Globigerinoides sacculifer
212–250−0.830.653.011.39
250–300−0.811.413.131.40
300–350−0.691.753.051.36
350–425−0.772.053.171.36
425–500−0.642.193.371.38
>500−0.792.553.541.34
 
Globigerinoides trilobus
212–250−0.790.842.701.38
250–300−0.791.352.891.40
300–350−0.921.72  
350–425−0.792.013.211.35
425–500−0.802.283.411.35
>500−0.852.373.791.34
 
Globorotalia crassoformis
212–250    
250–3000.930.722.18 
300–3501.580.901.621.30
350–4251.500.961.581.37
425–5001.620.961.791.40
>5001.711.021.651.44
 
Globorotalia inflata
212–2500.630.172.081.34
250–3000.610.072.101.36
300–3500.180.072.141.37
350–4250.430.132.131.34
425–5000.570.262.351.33
>500    
 
Globorotalia menardii
212–250−0.171.302.271.34
250–300−0.101.572.441.35
300–3500.261.462.301.31
350–4250.261.662.221.32
425–5000.281.752.201.36
>5000.301.812.481.35
 
Globorotalia scitula
212–2501.850.242.131.46
250–300    
300–350    
350–425    
425–500    
>500    
 
Globorotalia truncatulinoides (r)
212–250    
250–300    
300–350    
350–4250.980.872.271.43
425–500    
>500    
 
Globorotalia truncatulinoides (l)
212–2500.770.422.501.41
250–3000.810.582.511.41
300–3500.940.722.361.40
350–4250.940.942.311.42
425–5001.111.052.171.39
>5001.021.062.061.43
 
Globorotalia tumida
212–250    
250–3000.711.432.161.30
300–350    
350–4250.821.932.001.31
425–5000.921.911.651.30
>5000.891.901.791.30
 
Neogloboquadrina dutertrei
212–250    
250–300    
300–350    
350–4250.101.712.121.35
425–5000.121.712.581.36
>500−0.152.202.581.38
 
Neogloboquadrina pachyderma (r)
212–2500.01−0.872.181.36
250–300−0.02−0.802.061.31
300–350    
350–425    
425–500    
>500    
 
Orbulina universa
212–250 0.64  
250–300−0.171.224.411.41
300–350−0.271.345.471.43
350–425−0.281.625.601.38
425–500−0.381.896.091.39
>500−0.451.976.271.33
 
Sphaeroidinella dehiscens
212–250    
250–300    
300–350    
350–425    
425–500    
>500−0.031.832.291.34

3. Results

[8] The oxygen isotopic compositions of the globorolotaliids, except for G. inflata, increase with increasing size fraction, tending to reach a plateau value, for example at 300–350 μm for G. menardii and G. truncatulinoides (l) (Figure 1a). Of the remaining species for which sufficient data are available (Figure 1b) G. bulloides and G. calida show the same trend as the globorolotaliids. In the case of O. universa, N. dutertrei, and perhaps G. ruber (w), δ18O decreases with increasing size fraction. In the remaining cases, there are no systematic changes in changes in δ18O.

Figure 1.

Oxygen and carbon isotopic compositions versus size fraction of foraminifera: (a) δ18O of globorolotaliids, (b) δ18O of other species, (c) δ13C of globorolotaliids, and (d) δ13C of other species. Straight lines connect adjacent points. Figures 1a and 1b show δ18O calcification temperature. Both show temperature estimates based on foraminiferal δ18O (δc) using the paleotemperature equation of Shackleton [1974] assuming δ18Oseawater (δw = 0.70‰ (see text). Figure 1a also shows (as extreme right hand scale) temperature based on correlation between temperature (T) and δw obtained from water column profiles at sites close to core 31 K (see Figure 3): δw = −0.0928 + 0.043 T (r = 0.92), resulting in the δw independent expression T = 20.40 − 5.42δc.

[9] All the species examined for which there are sufficient data show an increase in δ13C with increasing size fraction (Figures 1c1d), with a wide range in δ13C, 0.1–1.7‰ over the size range. The size effect on δ13C is greatest for the spinose foraminifera and least for the globorotaliids, as found by Ravelo and Fairbanks [1995].

[10] There are eight individual species for which near-complete elemental data are available: G. bulloides, G. calida, G. ruber (w), G. sacculifer (w s), G. sacculifer (w/o s), G. crassaformis, G. inflata, G. truncatulinoides (l), and O. universa (Figure 2). For all species illustrated with two exceptions, Mg/Ca increases with increasing size (Figure 2a). The ranges in Mg/Ca are large, from 12 to 36% of the mean value over the size range. The behavior of Mg/Ca of G. truncatulinoides and G. crassaformis is opposite and decreases with increasing size, by ∼20% of the mean. This is not a pattern seen for the globorotaliids as a whole, however, as G. inflata Mg/Ca increases slightly with size fraction (data for the other globorotaliids are incomplete). Note also that Mg/Ca of different species are similar in the smallest size fraction and diverge in larger size fractions. The Mg/Ca ratio from which values diverge is 2.65 mmol/mol (SD = 0.45; n = 12). O. universa has higher Mg/Ca but follows this general trend. In all species, the change in Sr/Ca is opposite to that for Mg/Ca: Sr/Ca decreases with increasing size (Figure 2b) except for G. crassaformis and, to a lesser extent, G. truncatulinoides for which Sr/Ca increases with increasing size (Figure 2c).

Figure 2.

Mg/Ca and Sr/Ca versus size fraction of foraminifera (a) Mg/Ca and (b, c) Sr/Ca. Straight lines connect adjacent points.

4. Discussion

4.1. Temperature

[11] The first issue we wish to address is what temperatures may be deduced from the foraminiferal data and how do they compare with the modern hydrography. Once this is established we will turn to what controls size variability in Mg/Ca and Sr/Ca. The most obvious explanation for Mg/Ca might be that it reflects the temperatures of the different size fractions. In order to examine this, we have estimated temperatures from foraminiferal δ18O for comparison with Mg/Ca ratios. Changes in foraminiferal δ18O may occur for reasons other than temperature, for example, as has been observed as a function of carbonate ion concentration [Spero et al., 1997]. Here we make the initial assumption that changes in foraminiferal δ18O reflect changes in calcification temperature and consider other factors later.

4.1.1. Calcification temperatures based on δ18O

[12] The calculation of δ18O calcification temperatures from foraminiferal δ18O (δc) is complex when comparing different size fractions in a sediment. Apart from choice of temperature equation and the possibility of nonequilibrium effects between species, the particular difficulty is in estimation of δ18Oseawater (δw) for different size fractions, which may calcify at different water depths or times of year. Oceanographic data for nearby stations (Figure 3a) show that δw in surface waters is ∼0.70‰ (varying between 0.55 and 0.85‰). Additionally, both δw and temperature (T) decrease with depth at the core site (Figure 3) and are thus related as δw = −0.0928 + 0.043 T (r = 0.92). Therefore we have used two methods to derive δ18O calcification temperatures, both employing the paleotemperature equation of Shackleton [1974] and ignoring possible between-species calibration differences [see Bemis et al., 1998].

Figure 3.

Water column profiles of temperature and δ8O at sites close to core 31 K (Pierre et al., 1994, from G. A. Schmidt, G. R. Bigg, and E. J. Rohling Global Seawater Oxygen-18 Database available at http://www.giss.nasa.gov/data/o18data/): (a) δ18O versus water depth, (b) temperature versus water depth, and (c) temperature versus water depth together with δ18O calcification temperatures estimated for size fraction data for five species of foraminifera shown at appropriate depths. (d) δ13C versus water depth (based on phosphate data with correction of surface water values for anthropogenic CO2) depth together with δ13C for size fraction data for five species of foraminifera shown at depths from Figure 3c.

[13] Method 1 assumes a constant δw = 0.7 ± 0.15‰, making the assumption that the foraminifera calcify in surface waters. Method 2 exploits the link between δw and water temperature to derive calcification temperature independently of δw, making the assumption that differences in calcification temperature are related to habitat depth.

[14] The increase in δ18O with increasing size fraction seen for the globorolotaliids (apart from G. inflata) presumably reflects increased calcification of larger more mature individuals at greater depths and hence lower temperatures [e.g., Erez and Honjo, 1981]. The temperature estimates must take this into account and results in a 2°C larger temperature range overall for these species compared with the assumption of constant δw (Figure 1a). Calcification temperatures of the smallest size fraction are in the range ∼15°–21°C and of the largest size fraction in the range 11°–19°C. G. bulloides shows a ∼1.5°C decrease over the size fraction range and O. universa ∼1.5°C increase, based on constant δw (Figure 1b). Of the other species (Figure 1b) the ranges in calcification temperatures are small and most are <1°C, within the uncertainty of the method.

[15] Therefore, δ18O calcification temperatures show either no resolvable temperature variability (nine species), a decrease in temperature with increasing size fraction (the globorolotaliids, except G. tumida, and G. bulloides), and an increase in temperature with increasing size fraction (O. universa).

4.1.2. Comparison of calcification temperatures and upper ocean temperatures

[16] Modern sea surface temperatures from Levitus Atlas data at the site of Core BOFS 31K vary between ∼17° and 25°C (Figure 4). Temperature data (for February and March) for nearby stations (Figure 3b) range from 21°C at surface to 12.5°C at 300 m and 6.5°C at 1000 m.

Figure 4.

Mg/Ca compared with δ18O and δ18O calcification temperature as in Figure 1 with δ18Oseawater linked to depth and thus temperature for the globorolotaliids (see text) and a constant δ18Oseawater = 0.70‰ for other species: (a) all data; the curve is an exponential fit through all data except for O. universa and G. calida and is Mg/Ca = 0.89 exp(0.05T) (r = 0.91). (b) Exponential fits of the form Aexp(0.1T) through data for separate species (black lines r > 0.75; red lines r < 0.75). Horizontal bars are shown for some species to illustrate an uncertainty in temperature of ±1°C, the approximate values associated with the assumption of using a constant value for δ18Oseawater. For illustrative purposes a second set of data are shown for O. universa where δ18O have been corrected for the carbonate ion effect using the δ18O–δ13C relationship of Spero et al. [1997]; the horizontal grey line centered at Mg/Ca = 2.62 mmol/mol shows the approximate Mg/Ca of the smallest size fractions. (c) Data except for O. universa compared with two calibration curves.

[17] Calculated calcification temperatures for the globorotaliids (Figure 1a) fall with the range 11°–21°C (Method 2). When these are compared with the modern temperature distribution, it can be seen that they correspond to waters depth of calcification down to about 450 m (Figure 3c) with calcification depths in the sequence: G. menardii < G. inflata < G. tumida < G. truncatulinoides < G. crassaformis.

[18] Calcification temperatures for the other species (Figure 1b) fall with the range 19°–27°C (Method 1), the same range but 2°C higher than Levitus data (Figure 4). Sea surface temperature estimates for 10–12 cm within Core BOFS 31K based on the FA20 transfer function are 17°C (summer) and 25°C (winter) and the alkenone temperature is 21°C [Chapman et al., 1996]. Ravelo et al. [1990] found that core top faunas in the region of Core BOFS 31 K correlate strongly with sea surface temperature seasonality rather than with thermocline depth or subsurface seasonality. Because the temperatures based upon δ18O values for different size fractions within each nongloborotaliid species are very similar, it would seem that a seasonality effect within size fractions is small.

4.1.3. Calcification temperatures based on Mg/Ca

[19] The most obvious explanation as to what controls size variability in Mg/Ca might be that it reflects the temperatures of the different size fractions, which vary as a result of changing habitat during the evolution of the foraminifera from small to large individuals. When Mg/Ca and δ18O data are compared (Figure 4a), it is seen that, with the exception of O. universa and G. calida, the data show a negative correlation and, when δ18O data are converted to calcification temperature and compared with Mg/Ca, they fall on an exponential curve of the type typical of published Mg thermometry calibrations [Rosenthal et al., 1997; Lea et al., 1999; Elderfield and Ganssen, 2000] but with a weaker temperature dependence: Mg/Ca = 0.89 exp(0.05T) (r = 0.91).

[20] The weaker temperature dependence is an artefact obtained by considering all size fractions of all species together. When data for individual species of the globorolotaliids are examined separately (Figure 4b), the increase in δ18O with increasing size fraction, interpreted as changes in calcification depth, are associated with changes in Mg/Ca consistent with an exponential temperature dependence of ∼10% per °C together with small offsets between some species Most published studies of different species have given a similar temperature sensitivity of 9 ± 1% [Lea et al., 1999].

[21] However, the dominant effect for the other species is of significant differences in Mg/Ca between size fractions of a species but of insignificant changes in δ18O (Figure 4b). In particular, the changes in Mg/Ca with size fraction for two species, G. calida and G. sacculifer (without s) are not correlated with changes in temperature (uncertainties are shown in temperature arising from the range in δw = 0.7 ± 0.15‰). O. universa is similar but may be biased by one data point (see Figure 4 caption).

[22] Given the narrow ranges in δ18O calcification temperatures for the size fractions, it is not possible to define distinct calibration equations for each species. Taken overall, all the data (except for O. universa) fall within curves, which encompass calibration constants commonly cited in the literature: preexponential constants of 0.35 and 0.55 and exponential constants of 0.09 and 0.10 (Figure 4c). Therefore it is possible that foraminiferal size is a reason for the variability in literature data and that calibrations might be improved if a better control over size is exercised.

[23] Perhaps the most striking observation from the relationship between Mg/Ca and δ18O calcification temperatures is that the smallest size fraction exhibits a narrow range in Mg/Ca (Figure 2a) yet a proportionally wider range in δ18O. The mean Mg/Ca for the 212–250 μm size fraction of 2.65 mmol/mol (SD = 0.45) translates to Mg-based calcification temperatures (for example, using Mg/Ca = 0.35exp0.1T) of 20.2 ± 1.7°C, whereas the δ18O-based calcification temperatures in this size fraction range from ∼16.5° to 24.5°C.

4.2. Disequilibrium Effects

[24] Disequilibrium precipitation of calcite is commonly observed for δ13C. The pattern seen here of δ13C increasing with size (Figures 1c1d) is consistent with a considerable body of earlier work and has been interpreted in three ways: that with increasing size (1) kinetic isotope fractionation decreases, (2) incorporation of metabolic CO2 decreases, and (3) incorporation of photosynthetic CO2 increases [e.g., Erez and Honjo, 1981; Spero and Williams, 1998; Oppo and Fairbanks, 1989; Ravelo and Fairbanks, 1995; Spero and Lea, 1996; Kohfeld et al., 2000]. More recent work (not involving size fractions) has shown that foraminiferal δ13C and δ18O both decrease with increasing seawater carbonate concentration [Spero et al., 1997].

4.2.1. Comparison of foraminiferal and ocean water δ13C

[25] The nongloborolotaliid species mostly calcify in surface and near-surface waters. The δ13C in the largest size fraction of the majority of these species is in the range of ∼1.5–2.5‰. Comparison of foraminiferal δ13C (Figure 1d) and surface water δ13C values of ∼2.0–2.3‰ (Figure 3d) shows that the species move closer to equilibrium with seawater δ13C with increasing size. G. bulloides and G. calida have distinctly lower δ13C ratios but follow this trend. It is possible that the results may in part reflect temporal/spatial variability in dissolved δ13C within the upper water column through the productivity seasonal cycle. This could account for a significant part of the observed shift in foraminiferal δ13C not connected with “vital” effects.

[26] The globorolotaliids tend to calcify in colder deeper water than the other species. This is accompanied by increasing δ13C with size (Figure 1c), which reflects a move to values closer to equilibrium with seawater δ13C at the depth of calcification (Figure 3d).

4.2.2. Comparison of Mg/Ca and δ13C

[27] All of the thirteen species for which there are sufficient data show a linear correlation between Mg/Ca and δ13C (Figure 5a). In general, all species have Mg/Ca of about 2.6 mmol/mol for the smallest size fraction (Figure 2a) and lowest δ13C (Figure 5a), which increases or decreases with increasing δ13C. O. universa has higher Mg/Ca but follows this general trend.

Figure 5.

Comparison between (a) Mg/Ca and δ13C; the horizontal grey line centered at Mg/Ca = 2.62 mmol/mol shows the approximate Mg/Ca of the smallest size fractions. (b) Sr/Ca and δ13C.

[28] This relationship with δ13C implies that, initially, most of the species examined lay down calcite in the surface or near surface waters with a somewhat similar Mg/Ca ratio (equivalent to a partition coefficient, DMg, of ∼0.5 × 10−3). During their growth, the globorolotaliids tend to add further calcite in colder deeper water. This is accompanied by increasing δ13C, which reflects a move to values closer to equilibrium with seawater δ13C at the depth of calcification (Figure 3d). The decreases in Mg/Ca are, as discussed previously, also accompanied by increasing δ18O (Figure 4b) and reflect a response to the decreased temperature of calcification.

[29] The other species, whose calcification temperatures place them within the upper 50 m or so of the water column, also show increases in δ13C with increasing size fraction and that they move to values closer to equilibrium with average surface seawater δ13C of ∼2.0–2.3‰. In the case of these surface and near-surface dwelling species, the increases seen for Mg/Ca are not accompanied by decreasing δ18O, and therefore the changes in Mg/Ca do not appear to be a temperature response. The results imply either that δ18O or Mg/Ca data are affected by nonequilibrium effects. Partition coefficients for foraminifera are much smaller than those predicted by thermodynamics or those observed in inorganic experiments and natural inorganic carbonate cements [see Cicero and Lohmann, 2001]. Consequently, theoretically or experimentally derived temperature dependence of Mg substitution in calcite may not be directly relevant to foraminifera. The fact that Mg/Ca increases (toward inorganic values) with increasing size, as does δ13C, implies that Mg uptake is a nonequilibrium process.

4.2.3. Comparison of Sr/Ca and δ13C

[30] Because Mg/Ca increases with size fraction and Sr/Ca decreases (Figure 2), the pattern for Sr/Ca compared with δ13C is generally the opposite to that for Mg. Of the nine species shown in Figure 2, seven show a negative correlation of Sr/Ca and δ13C (Figure 5c). It would seem that, in a manner analogous but opposite to Mg/Ca, species lay down calcite initially with high Sr/Ca ratios (but with a wider relative (but not absolute) range, ∼1.35–.45 mmol/mol, equivalent to DSr of ∼0.16) which decrease during growth.

[31] This is consistent with ideas of a growth rate or calcification rate control on Sr/Ca. Sr/Ca data on polyspecific coccolith separates from core top sediments show a strongly correlation with coccolithophorid growth and calcification rates [Stoll and Schrag, 2000]. More recent culture experiments of coccolithophorids [Stoll et al., 2002; Rickaby et al., 2002] support this view that the prime control on Sr partitioning in coccoliths is calcification rate related to coccolithophorid productivity.

[32] A calcification rate control on Sr/Ca also has been observed for inorganic calcite [Lorens, 1981; Tesoriero and Pankow, 1996]. At high carbonate precipitation rates, Sr uptake occurs under kinetic control such that the calcite cannot discriminate effectively against Sr, resulting in higher DSr at higher calcification rates. Furthermore, experiments have shown that calcite growth leads to differential partitioning in trace element incorporation and that the responses of Mg and Sr to calcification rate and growth mechanism are opposite [Parquette and Reeder, 1995], just as seen here.

[33] It has also been suggested that lighter δ13C values associated with high seawater [CO32−] are associated with higher calcification rates [Spero et al., 1997]. In this work, lighter δ13C corresponds to lower Mg/Ca and higher corresponds to Sr/Ca. As foraminifera within the larger size fractions calcify closer to C isotopic equilibrium with seawater than those within the smaller size fractions, it is consistent with a general view that smaller individuals calcify faster than larger ones. This is presumably because smaller individuals grow faster but to a smaller size because they live under somehow less favorable conditions or because larger individuals represent more mature versions of smaller individuals and contain an initial fast growing phase and a later slower growing phase.

4.3. Implications

[34] The major implication of the size fraction results for Mg thermometry using planktonic foraminifera is the obvious one that work should be carried out using a single size fraction and that the calibration used should be based upon that same size fraction. It may be that narrower size ranges and that measurement of size, rather than sieving within a size range, would improve precision. It may also be more appropriate to use the optimum size of a species rather than the same size for all species. The largest size appears to be the most reliable and perhaps has the least interspecies temperature sensitivity, although this requires confirmation. It also follows that a cause of variability within one size fraction may be because of a calcification rate control.

[35] Another implication concerns Sr/Ca. Because the temperature sensitivity of Sr/Ca in foraminifera appears to be small [Lea et al., 1999; Elderfield et al., 2000], the question has arisen as to whether foraminiferal Sr/Ca can be used to define past changes in seawater Sr/Ca [Martin et al., 1999; Stoll et al., 1999; Elderfield et al., 2000]. In the same way that coccolith Sr/Ca may be useful to define coccolithophorid productivity, it is possible that Sr/Ca of planktonic foraminifera, even within one size fraction, may be affected by changing foraminiferal calcification rates. Records though time of the distribution of size or mean size, for particular species and their chemistry would be of particular interest to understand this further. It should also be possible to test this idea using analysis of materials from core top transects across productive regions and from sediment trap time series. This may account for part of the temporal variability in foraminiferal Sr/Ca within a single size fraction that has been attributed to changes in seawater Sr/Ca.

Acknowledgments

[36] We thank Mervyn Greaves for invaluable help in the laboratory. Stephen Barker, Peggy Delaney, Stephanie de Villiers, Heather Stoll, Laurent Labeyrie, Ann Russell, and Howie Spero provided helpful comments on this work and earlier versions of the manuscript. This work was supported by grants from NERC (GR3/JIF/05a and GR3/13108) and the CMI.

Ancillary