Reconciling single-chamber Mg / Ca with whole-shell δ 18 O in surface to deep-dwelling planktonic foraminifera from the Mozambique Channel

Most planktonic foraminifera migrate vertically through the water column during life, meeting a range of depth-related conditions as they grow and calcify. For reconstructing past ocean conditions from geochemical signals recorded in their shells, it is therefore necessary to know vertical habitat preferences. Species with a shallow habitat and limited vertical migration will reflect conditions of the surface mixed layer and short-term and mesoscale (i.e. seasonal) perturbations therein. Species spanning a wider range of depth habitats, however, will contain a more heterogeneous, intra-specimen variability (e.g. Mg /Ca and δO), which is less for species calcifying below the thermocline. Obtained single-chamber Mg /Ca ratios are combined with singlespecimen δO and δC of the surface-water inhabitant Globigerinoides ruber, the thermocline-dwelling Neogloboquadrina dutertrei and Pulleniatina obliquiloculata, and the deep dweller Globorotalia scitula from the Mozambique Channel. Species-specific Mg /Ca, δC and δO data combined with a depth-resolved mass balance model confirm distinctive migration and calcification patterns for each species as a function of hydrography. Whereas single-specimen δO rarely reflects changes in depth habitat related to hydrography (e.g. temperature), measured Mg /Ca of the last chambers can only be explained by active migration in response to changes in temperature stratification. Foraminiferal geochemistry and modelled depth habitats shows that the singlechamber Mg /Ca and single shell δO are in agreement with each other and in line with the changes in hydrography induced by eddies.


Introduction
Most planktonic foraminifera inhabit the upper 200 m of the water column, with the exceptions of some species living as deep as 1000 m (e.g.Hemleben et al., 1989).The average depth habitat of individual species and the range of water depths at which they are found reflect their ecology (e.g.feeding behaviour), ontogeny and seasonal preferences.Stable oxygen isotope values (δ 18 O) and Mg / Ca ratios (Shackleton, 1974;Fairbanks et al., 1980;Ortiz et al., 1996;Elderfield and Ganssen, 2000) have been used to reconstruct upper water column conditions using species with a known depth range (e.g.Ravelo and Fairbanks, 1992;Patrick and Thunell, 1997;Faul et al., 2000;Cléroux et al., 2013).For many species, however, application of Mg / Ca as a seawater temperature proxy is complicated by depth migration as a function of ontogeny.Previous studies revealed major Mg / Ca heterogeneity within foraminiferal shells (e.g.Eggins et al., 2003;Hathorne et al., 2009;Kunioka et al., 2006; J. Steinhardt et al.: Reconciling single-chamber Mg / Ca with whole-shell δ 18 O Jonkers et al., 2012), which was attributed to a combination of vertical migration during their life and vital effects.Nevertheless, species-specific patterns of vertical migration and hence depth of calcification determine what part of the water column can be reconstructed.
Field observations show that most foraminiferal species do not occupy a single depth but rather calcify at a range of depths.Many species migrate vertically as they grow, and therefore the chemical composition (e.g.Mg / Ca and δ 18 O) of their shells changes with age.Fairbanks et al. (1982) and Field (2004) suggested that foraminifera may modify their habitat depth depending on hydrographic condition and food supply.However, little is known about the exact controls on depth habitat and shell features (e.g.formation of crusts) or the termination of shell growth.A better understanding of the vertical calcification pattern of different species is needed to reconstruct past changes in vertical structure of the water column by using geochemical proxies, e.g. for temperature (δ 18 O and Mg / Ca).Through use of geochemical signals of species with different and well-constrained calcification depths (Emiliani, 1954;Mulitza et al., 1997), changes in water column conditions can be resolved.
Using core-top samples from the Indian Ocean, Birch et al. (2013) report δ 13 C and δ 18 O measurements made on several species of planktonic foraminifera across a range of tightly constrained size windows.From size-controlled δ 18 O calcite trajectories they inferred depth habitats, using modern vertical temperature profiles.However, by using multiple core-top specimens, this data set encompasses not only vertical changes in the water column structure but also inter-and intra-annual changes therein, which are both known to vary substantially in this region (e.g.McClanahan, 1988;Damassa et al., 2006;Hastenrath et al., 1993).In this study we use sediment trap samples, allowing analyses of specimens that lived during a confined time interval, and link in situ hydrographic changes (i.e.temperature) more directly to their shell chemistry.
Single-chamber Mg / Ca compositions from specimens with contrasting calcification depths (the surface dweller Globigerinoides ruber (d'Orbigny, 1839), the thermoclinedwelling species Neogloboquadrina dutertrei (d'Orbigny, 1839) and Pulleniatina obliquiloculata (Parker et al., 1865) and the deep dweller Globorotalia scitula; Brady, 1882) reflect temperatures throughout the upper 500 m and were shown to reliably reflect short-term hydrographic changes (Steinhardt et al., 2014).Mesoscale eddies such as those observed in the Mozambique Channel (MC) induce variations in temperature and salinity.Anticyclonic (anticlockwise) eddies in the MC are characterized by a warm water core and are associated with elevated sea surface heights and large vertical isopycnal excursions.Foraminifera living in the mixed layer of the MC are affected by eddy-induced changes; this is reflected by the geochemistry of G. ruber and N. dutertrei (Steinhardt et al., 2014), resulting in higher Mg / Ca ratios and more depleted δ 18 O cc values.These short-term changes in vertical water column temperature and δ 18 O sw distribution should influence shell δ 18 O and Mg / Ca throughout the different ontogenetic stages for any species migrating during its life.Alternatively, foraminifera may respond to altered hydrographic conditions by changing their calcification depth.Here we present combined single-specimen δ 18 O and single-chamber Mg / Ca measurements for different species, providing a composite of thermocline and sub-thermocline conditions.Since single-chamber Mg / Ca values cannot be compared one to one with whole-shell δ 18 O values, we evaluate our results using a mass balance model for depth-related carbonate addition of four species of planktonic foraminifera.

Oceanographic setting
In the oligotrophic Mozambique Channel (MC, Fig. 1), sea surface temperatures (SST) vary seasonally and with eddyinduced transport (Fallet et al., 2011).The SSTs range from 25 to over 30 • C with an annual mean of 27.6 • C; the seasonal change in temperatures is associated with the monsoon system.With the onset of austral summer, rainfall increases, caused by the seasonal migration of the Intertropical Convergence Zone (ITCZ), and sea surface salinities decrease slightly from 35.2 in winter to 34.9 in summer (Fallet et al., 2010).The calcite compensation depth in the western Indian Ocean is below 3000 m and hence does not result in dissolution of foraminiferal calcite at the depth of the trap's location (2225 m).Southward migration of anticyclonic mesoscale eddies, originating at 10 • S north of the Comoros Islands, affects the hydrography in the MC (Fig. 1).Eddies pass through the MC at a mean frequency of about four to seven per year (at a southward propagation speed of 3-6 km d −1 ) before joining the Agulhas Current.An eddy passage is associated with vertical movement of isopycnals, which can occasionally exceed 40 m per day in the upper layer (Ullgren et al., 2012).The formation of mesoscale eddies in the Mozambique Channel is related to variability in the South Equatorial Current (SEC) transport (Backeberg and Reason, 2010;Fig. 1).The main water masses contributing to the upper part of the MC include the Tropical Surface Water (TSW), Subtropical Surface Water (STSW) and Indonesian Throughflow Water (ITFW).The warm, fresh surface water (TSW) forms at equatorial latitudes and is transported westward within or north of the SEC (New et al., 2007).In the proximity of the western margin, where the SEC bifurcates, warm surface waters are transported poleward, either east of Madagascar or through the MC (e.g.Gründlingh, 1995;Swallow et al., 1988).The STSW is characterized by relatively high salinities and a subsurface maximum, with salinities of 35.2-35.5, at approximately 200 m below sea surface, at which depth the surface water subducts below the fresher TSW (Wyrtki, 1973).On the bottom right is a map of sea level anomaly showing the passing of an anticyclonic eddy over the trap location (star).The AVISO sea level anomaly map was produced using the AVISO live access server (http://las.aviso.altimetry.fr/las/getUI.do).EACC: East African Coastal Current; SEC: South Equatorial Current; AC: Agulhas Current.

Sediment trap and mooring array
Within the Long-term Ocean Climate Observations (LOCO) programme, an array of eight moorings across the narrowest part of the Mozambique Channel has provided continuous measurements of current velocities, temperatures and salinities at fixed depths since November 2003 (Ullgren et al., 2012).Sediment traps of the type Technicap PPS 5 were deployed at 16.8 • S and 40.8 • E in the central MC (Fig. 1), equipped with an automated sampling carousel of 24 cups and a baffled collecting area of 1.0 m 2 .The trap was positioned 250 m above the channel floor at 2250 m water depth.Between November 2003 and February 2009, a total of four sediment trap deployments took place, each programmed to a sampling interval of 17, 21 or 23 days.Prior to deployments, sample cups were filled with an HgCl 2 -poisoned and boraxbuffered solution of seawater collected from the deployment depth (Lončarić et al., 2007).Sediment trap samples were wet-split and sieved, and foraminiferal shells were cleaned according to the protocol of Barker et al. (2003), modified by Fallet et al. (2009; see also Fallet et al., 2010;Steinhardt et al., 2014, for a detailed description of the procedure here).
Using sediment trap material allows for the chemistry of the shells to be linked to recorded in situ conditions from the moorings and from real-time satellite-derived observations.Therefore we are able to link short-term changes in hydrography (i.e.eddies) to the differences in shell chemistry.Calculated back trajectories, based on a high-resolution INALT01 model (Durgadoo et al., 2013), show that specimens ending up in the sediment trap all originate from the area under influence of the eddy variability (Steinhardt et al., 2014).We selected the sediment trap intervals during which the complete sediment cup collection took place under either full eddy or full non-eddy conditions (for a full description see Steinhardt et al., 2014;Supplement).For the selected collecting intervals, temperature and salinity observations from the mooring (lmc5a) are compiled, and daily means were used to calculate eddy and non-eddy depth-resolved temperature profiles.

Temperature and salinity data
For this study, we used temperatures recorded at 110, 200 and 400 m water depth by a CTD deployed on mooring lmc5A (16.8 • S, 41.1 • E, Fig. 1), which is closest to the sediment trap site.Moored salinity and temperature data, collected during the selected intervals of eddy and non-eddy conditions (Table S1 in the Supplement), were spline-fitted in Analyseries 1.1.668 K to achieve metre-wise data resolution.Sea surface temperatures were retrieved from the 4 km daytime MODIS/AQUA data set around trap site (16-17 • S and 40-41 • E) for the period of the selected collecting intervals (http://podaac-tools.jpl.nasa.gov/las).Surface salinity data are not available for the complete deployment period and therefore CTD-based salinity-depth profiles taken during the deployment/recovery cruises were used instead (Ullgren et al., 2012).Based on the trend observed in the moored salinity data at 110 m water depth (Ullgren et al., 2012;less  imum surface salinities to represent eddy surface salinities and maximum surface salinities to represent non-eddy conditions.Since salinity mooring data were not available for depths between 400 and 1525 m, we have chosen two more "anchor points" at 700 and 1000 m water depth from the CTD depth profiles in order to better capture the Red Sea Water (RSW) advection at these depths and to achieve a more accurate salinity-fitting curve for the upper 1000 m.

Planktonic foraminiferal species and ontogeny
We selected four species from the sediment trap samples according to differences in depth habitats as reported in previous studies.Globigerinoides ruber (white) is a shallow, surface mixed-layer-dwelling species, occupying the upper 50 m of the water column, and is commonly used to reconstruct SST (Hemleben et al., 1989).To minimize a potential biases in δ 18 O and Mg / Ca associated when combining different morphotypes (Steinke et al., 2005), we used only G. ruber sensu stricto, which was by far the most abundant in these samples (Fallet et al., 2010).
Measurements on G. ruber were usually performed on specimens in the 250-315 µm size fraction.In a limited number of samples, abundances of this species were low in this size fraction, and geochemical analyses were therefore performed on specimens from a larger size fraction (315-400 µm).Analyses on N. dutertrei, P. obliquiloculata and G. scitula were generally done in the size range > 315 µm, with additional measurements in the 250-315 µm size fraction depending on the specimen's abundance within a sample.All specimens show excellent preservation, and none show any signs of diagenesis (based on scanning electron microscopy).Recently, Fallet et al. (2012) showed that shell-size-normalized weights of three species of planktonic foraminifera from the same sediment trap location do not differ from those of the surface sediment samples below this trap.Absence of dissolution is also reported by Birch et al. (2013), who described planktonic foraminifera from surface sediments at ∼ 3000 m water depth, in the northern part of the Mozambique Channel, as being glassy and excellently preserved.

Mg / Ca and stable isotope analyses
The Mg / Ca ratios of single chambers used in this study have previously been published (Steinhardt et al., 2014) and were determined by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) at Utrecht University (Reichart et al., 2003; for summary of the results see Table 1).Subsequently, specimens were analysed for whole-shell δ 18 O and δ 13 C after removal from the laser ablation stub with ethanol and inspection for possible contaminations.Measurements were performed at the Universitat Autònoma de Barcelona on a Thermo Finnigan MAT253 mass spectrometer coupled to a Kiel IV device for CO 2 sample gas preparation.External reproducibility (1σ ) of δ 13 C standards NBS19 and IAEA-CO was 0.04 ‰, and for δ 18 O it was 0.08 ‰.
Single shells from part of the sample set were analysed using a Thermo Finnigan Delta Plus mass spectrometer equipped with a GasBench II preparation device at the VU University Amsterdam.Single specimens were loaded into round-bottom vials, which were subsequently flushed with He.The samples were then reacted with phosphoric acid (H 3 PO 4 ) injected into the vial, producing CO 2 gas, which is transported in a helium stream to the mass spectrometer.Traps are used to remove residual H 2 O from the sample gas and the CO 2 is separated from other possible contaminant gases on a PoraPLOT Q GC column.Reproducibility (1σ ) of δ 13 C standards NBS19 and was 0.07 and 0.12 ‰ for δ 18 O.Values measured on the Kiel IV and the GasBench II are comparable and species-specific δ 18 O cc are in good agreement (Table 2).Measurements with the GasBench II have a somewhat wider standard deviation that is inherent to continuous flow mass spectrometry.In total, 391 single shell stable isotope values were obtained.Values deviating by more than twice the standard deviation from the average of the total data set were regarded as outliers (n = 23) and removed from the data set (Table S3).
The δ 18 O sw , expressed on the SMOW scale, is converted to the Pee Dee Belemnite (PDB) scale by subtracting 0.27 ‰ (Hut, 1987).Various δ 18 O-temperature equations have been proposed and discussed in detail in other studies (Bemis et al., 1998;Regenberg et al., 2009), without a clear consensus on the most appropriate equation.Here, we integrated calcification depth for each species calculated by matching the foraminiferal calcite δ 18 O cc with the calculated calcite δ 18 O calc following Eq.( 1) from Kim and O'Neil (1997) for the temperature-dependent fractionation of calcite by inorganic precipitation (assuming calcification in equilibrium with the ambient seawater).
We extracted δ 18 O sw values from the southern Indian Ocean for the upper 2000 m (4.5-120.2• E; 0-32.9 • S, N = 154) from the Global Seawater Oxygen-18 Database (see Table S1, http://data.giss.nasa.gov/o18data/).Additionally, we included in situ δ 18 O sw measurements from the MC, near the sediment trap location (41.08 • E; 16.74 • S), in order to determine the regional relationship between δ 18 O sw and salinity Table 1.Average Mg / Ca ratios (Steinhardt et al., 2014), δ 18 O and δ 13 C with standard errors (SE) and corresponding standard deviations (SD).Mg / Ca-based temperature are based on species-specific temperature equations.The equation developed by Fallet et al. (2010) was applied for G. ruber.The equations developed by Anand et al. (2003) were applied to N. dutertrei and P. obliquiloculata.For G. scitula, Anand's equation for G. hirsuta was applied following the example of Fallet et al. (2011).Calculate δ 18 O-based temperatures are based on the equation of Kim and O'Neil (1997).
This linear relationship (Eq. 3) is subsequently used to estimate δ 18 O sw values based on salinities measured in the proximity of the trap by moored temperature-salinity (T -S) sensors during eddy and non-eddy conditions for depths ranging from 0 to 1000 m.Seawater temperature and estimated δ 18 O sw profiles for eddy or non-eddy conditions are used to compare the δ 18 O data depending on the time interval sampled by the sediment trap.We used averaged δ 18 O sw from the depth range suggested by previously measured single-chamber Mg / Ca analyses (Steinhardt et al., 2014) to calculate the δ 18 Oderived calcification temperature for all species, following the temperature equation of Kim and O'Neil (1997): (3) The temperature equation of Kim an O'Neil (1997) is the most general calibration; it allows for comparison of interspecific differences that are automatically accounted for when using species-specific calibrations.

Calcification temperatures
The calculated multi-specimen δ 18 O-based temperature from Eq. ( 3) and the single-chamber Mg / Ca are positively and exponentially correlated (Fig. 4).Variability in this relationship is highest at higher (> 25 • C) temperatures.Mg / Caderived calcification temperatures, for G. ruber are on average 28.1 ± 2.8 • C, based on the calibration of Fallet et al. (2010) for this species in this region.Calcification temperatures for N. dutertrei and P. obliquiloculata are 22.5 ± 3.7 • C and 21.6 ± 3.1 • C, respectively, both based on species-specific calibrations from Anand et al. (2003).Mg / Ca ratios of G. scitula were transformed into temperatures using the equation for G. hirsuta (Anand et al., 2003), resulting in an average temperature of 14.4 ± 3.4 • C (Fig. 5).Calcification temperatures based on δ 18 O result in markedly different values, ranging from 29.4 ± 1.3 • C for G. ruber to 24.4 ± 2 • C for N. dutertrei, 22.5 ± 1.1 • C for P. obliquiloculata and 10.4 ± 3.9 • C for G. scitula (Table 1).Since P. obliquiloculata and G. scitula showed significant differences for δ 18 O cc between eddy and non-eddy conditions, we separately calculated temperatures for eddy and non-eddy conditions.Mean δ 18 O from Eq. ( 3) for eddy intervals yield 22.8 ± 0.9 • C for P. obliquiloculata and 7.9 ± 2.1 • C for G. scitula.For non-eddy intervals, calcification temperatures are 22.5 ± 1.2 • C for P. obliquiloculata and 11.8 ± 4.1 • C for G. scitula (Fig. 5).

Single-specimen isotope temperatures
The average single-specimen δ 18 O cc of G. ruber reflects SSTs of 27.0 ± 2.2-28.4± 2.1 • C (based on sediment trap calibrations from Fallet et al., 2010, andWilke et al., 2009, respectively), which is close to the satellite-derived annual mean SST of 27.6 • C (Fallet et al., 2010).When applying the equation of Kim and O'Neil (1997) for conversion of δ 18 O cc into temperature, SST is considerably higher (29.4 ± 1.3 • C).This discrepancy may be caused by the fact that the calcitewater calibration of Kim and O'Neil (1997) is based on inorganic precipitation experiments free of vital effects and therefore may be offset compared to the temperature-δ 18 O cc relationship of biogenic carbonates.Nevertheless this temperature estimate based on Kim and O'Neil (1997) is in good agreement with the average temperature of 28 ± 1.1 • C during the investigated intervals.The inter-test variability of this species can be explained by the high temperature variability at the sea surface, as well as differences in symbiont activity.The shallow depth habitat of G. ruber in the MC is in line with previous studies showing that this species is confined to the photic zone (e.g.Deuser et al., 1981;Lončarić et al., 2006;Peeters and Brummer, 2002) due to the light requirement of its symbionts.Based on its relatively nar-  row preferred depth habitat, this species is a suitable tracer for (sub)tropical surface-water (0-100 m, mixed layer) conditions (e.g.Deuser, 1987;Anand et al., 2003;Field, 2004;Fallet et al., 2010).Birch et al. (2013) show that shell size of specimens of G. ruber is not correlated with δ 18 O cc , confirming that this species occupies a narrow calcification depth during its life.In addition to its shallow living depth, G. ruber is known to occur in some areas relatively equally throughout the year (e.g.Deuser, 1987;Mohtadi et al., 2006;Tedesco et al., 2007), whereas in other areas, including the MC, it occurs at highest densities during summer months (e.g.Tolderlund and Bé, 1971;Duplessy et al., 1981;Ganssen and Sarnthein, 1983;Deuser and Ross, 1989;Eguchi et al., 2003;Lončarić et al., 2006;Fallet et al., 2010).This seasonal preference results in SSTs that are slightly biased towards summer conditions when using fossil specimens of this species.
Based on an average δ 18 O cc -derived temperature of 24.3 ± 2 • C (Table 2), following the equation of Kim and O'Neil (1997), calcification depths of N. dutertrei are in the range of 20-130 m (Fig. 6), with an average depth of 58 m.For eddy conditions, the average calcification depth is approximately 80 m; for non-eddy conditions it is approximately 37 m.Average Mg / Ca-based temperature of 22.5 ± 4 • C is in relatively good agreement with the average δ 18 O cc -derived temperature (Table 2).The difference between Mg / Ca-and δ 18 O-based temperatures are smaller than the 1.2 • C uncertainty associated with the Mg / Ca calibration (Anand et al., 2003).Previous studies using N. dutertrei from Indian Ocean core-top samples and Mozambique Channel sediment traps have reported similar depth ranges between 40 and 150 m (Kiefer et al., 2006) and similar average depths of 80 m (Fallet et al., 2011), respectively.Both of these studies used pooled specimens for their stable isotope analysis and hence provided the population's average calcification depth.Moreover, pooling of specimens from sediment core samples (Kiefer et al., 2006) does not allow for resolving short-term variability in calcification temperatures as do single specimens (e.g.seasonality).The inferred calcification depth for N. dutertrei is in line with its characterization as an intermediate deep-dwelling species, living preferentially in the seasonal thermocline (e.g.Fairbanks et al., 1982;Curry et al., 1983;Eguchi et al., 2003;Farmer et al., 2007), coinciding with a deep chlorophyll maximum (Fairbanks et al., 1980;Ravelo et al., 1990).Overall, the living depth of this species is confined to the upper 200 m (Farmer et al., 2007;Kroon and Darling, 1995).Variability in Mg / Ca within single-specimen shell walls of N. dutertrei from the Timor Sea suggested temperatures between 12 and 23 • C, implying migration through the entire thermocline (Eggins et al., 2003).However, most calcification seems limited to a much smaller depth interval, and the extremes in Mg / Ca might reflect upper and lower depth limits occupied by this species.Moreover, banding of Mg / Ca in shell calcite has been viewed in terms of discrete calcification events (Elderfield et al., 1996;Erez, 2003).Plankton tow studies (Fairbanks et al., 1980) showed oxygen isotope equilibrium calcification for N. dutertrei and P. obliquiloculata.
The δ 18 O cc -based calcification depths for P. obliquiloculata reported here (48-125 m, with an average of 74 m;  = 0.47 using F-1/2 Mg / Ca from G. ruber, F-0 for N. dutertrei, P. obliquiloculata and G. scitula (black circles).F-1 for N. dutertrei, P. obliquiloculata and G. scitula (red circles) and F-2 for N. dutertrei, P. obliquiloculata and G. scitula (blue circles).Mg / Ca data from Steinhardt et al. (2014).Note that the correlation coefficient also indicates that approximately 60 % of the observed variability is not due to temperature alone.(e.g. between 60 and 80 m; Mohtadi et al., 2009).Indeed, in plankton tows from the central equatorial Pacific the largest abundance of adult P. obliquiloculata with a terminal cortex was found below 60 m (Watkins et al., 1996).All specimens used in this study had the distinctive smooth outer cortex that envelops the final whorl in the adult as well as an arched aperture (Watkins et al., 1996).Non-corticated P. obliquiloculata ("juveniles") are confined mostly to the mixed layer (Watkins et al., 1996), indicating migration to greater depths at the time of cortex formation during the terminal stage of its life cycle (Erez and Honjo, 1981;Hemleben et al., 1989;Ravelo and Fairbanks, 1992).
The average δ 18 O cc for G. scitula yields a calcification temperature of 10.4 ± 3.9 • C, suggesting that this species calcifies between 290 and 1100 m (Fig. 6), with an average depth of approximately 500 m.This overlaps with the depth range indicated by the Mg / Ca temperatures of 14.4 ± 3.4 • C derived from the last few chambers added, suggesting that these shells formed at a depth between about 250 and 350 m for non-eddy and eddy conditions, respectively.The δ 18 O ccbased estimates, however, do not consider possible vital effects that were previously suggested for this species (e.g.Kahn and Williams, 1981).If taken into account, this would lower the temperature and depth habitat estimates by some 4 • C and 500 m, respectively.Birch et al. (2013) support previous findings of a distinct positive correlation between δ 18 O and size in G. scitula (e.g.Friedrich et al., 2012), which is linked to a substantial ontogenetic vertical migration through the water column.The largest individuals have been inferred to live below the thermocline, consistent with the supposed absence of symbionts in this species.This is in line with our observations, showing higher inter-specimen variability in δ 18 O cc for G. scitula than in the other species.

Habitat depth vs. calcification depth
Planktonic foraminifera collected by sediment traps might record δ 18 O cc signals comprising calcification at various depths and thus document an apparent average calcification depth by integrating the entire calcification history of the specimen.Given changes in seawater temperature with water depth, even minor changes in the upper or lower range of the depth at which planktonic species calcify can have a profound effect on the average δ 18 O cc and reconstructed temperature.Since evidence is accumulating that some species have a flexible calcification range (e.g.due to seasonality or local hydrography; Lončarić et al., 2006;Wilke et al., 2009), interpretation of down-core stable isotope data in terms of thermal structure may be challenging.Therefore, it is crucial to accurately quantify the impact of environmental factors on depth preferences of planktonic foraminifera.Contrasting eddy and non-eddy conditions, a short-term feature, allow us to disentangle seasonal and other short-term local hydrography changes and their effect on foraminiferal calcification depth.
While Mg / Ca-based temperatures of G. ruber and N. dutertrei record eddy-induced changes in upper water column stratification (Steinhardt et al., 2014), δ 18 O-based temperatures are relatively similar for both species (Fig. 6).Using the palaeo-temperature equation (Eq.1; Kim and O'Neil, 1997) and fitting δ 18 O calc with δ 18 O cc , we find that G. ruber calcifies on average at the sea surface (down to 7 m during non-eddy conditions and down to 18 m under eddy conditions, Fig. 6).N. dutertrei calcifies on average between 12 and 120 m during eddy conditions (average calci-fication depth 81 m) and between 17 and 58 m under noneddy conditions (average 37 m).During eddy conditions, P. obliquiloculata calcifies between 89 and 124 m (average 107 m), whereas it calcifies at shallower depth, between 20 and 77 m (average calcification depth 60 m) during non-eddy conditions.The largest changes in calcification depth in this study are inferred from G. scitula.From a calcification range between 500 and 1100 m and an average calcification around 716 m during eddy conditions it shifts to a calcification range from 168 to 745 m and an average calcification depth of 343 m (Fig. 6).
Conversely, δ 18 O-based temperatures are significantly different for P. obliquiloculata and G. scitula, while the Mg / Ca-based temperature of the last formed chambers of P. obliquiloculata indicates similar calcification temperature (Table 1).Mg / Ca-inferred calcification temperatures, representing the depth occupied at the later stages of the foraminifer's life, were similar between eddy and non-eddy conditions.Nonetheless, temperature mooring data show a steep temperature gradient, coinciding with the habitat depth of G. scitula, and thereby revealing a wide range of calcification depths for this species, changing significantly with deepening of the thermocline during eddy passage.
Inferred higher variability in calcification temperature for G. ruber presented in this study compared to observed satellite SST likely results from the spatial resolution employed here.Inter-individual differences in depth migration add to the variability in isotopes and element / Ca ratios when measuring single specimens.Potential effects of ontogeny on stable isotope composition are minimized by using narrow size fractions, as confirmed by the lack of ontogenetic trends with shell size in our measurements.Russell and Spero (2000) concluded that natural variability in oxygen isotopes is species-specific.From measurements of single-specimen δ 18 O cc of G. ruber shells from sediment traps in the eastern equatorial Pacific, they show that, over a 1.5-3day period, the standard deviation of δ 18 O results in a temperature variability of ± 0.87 • C. Such variability could explain between 12 and 38 % of the variability in δ 18 O-based temperatures in our samples.Another cause of natural variability might be differences in depth at which an individual calcifies.In laboratory cultures, the addition rate of new chambers in G. sacculifer ranges from 1.6 to 6.2 days (Bé and Spero, 1981), while chamber formation in G. hirsuta and G. truncatulinoides takes about 5 to 6 h (Bé et al., 1979).Considering that our sample duration ranges between 17 and 21 days, δ 18 O variability is likely to be affected by other parameters (e.g.temperature).Therefore, the observed variability in δ 18 O-based temperatures caused by species-specific natural variability in δ 18 O cc (e.g.Russell and Spero, 2000) during the time it takes to add new chambers, which might be calcified under different conditions or water depth.

Reconciling δ 18 O and Mg / Ca-derived calcification depths
Mg / Ca-derived temperatures indicate that calcification depths of N. dutertrei range between 42 and 169 m (average depth: 81 m) under non-eddy conditions and between 13 and 196 m (average depth: 98 m) during eddy conditions (Steinhardt et al., 2014).Thus, the shoaling in average calcification depth from 98 m during eddy conditions to 81 m during non-eddy conditions indicated by the whole-shell δ 18 O cc is less as than inferred from Mg / Ca, derived from the calcification of the last chambers.A more pronounced trend is present in Mg / Ca of P. obliquiloculata, shifting from depths of between 70 and 90 m (average 75 m) during noneddy conditions to depths of between 147 and 244 m (average 150 m) during eddy conditions (Steinhardt et al., 2014).The Mg / Ca-derived shift is hence larger than the shift inferred from δ 18 O cc (eddy: 107 m; non-eddy: 60 m).Mg / Caderived calcification temperatures for N. dutertrei and P. obliquiloculata are hence cooler and indicative of deeper calcification of the final chambers compared to that of the whole shell (based on δ 18 O cc ).Calcification temperatures derived from Mg / Ca for G. scitula (Fig. 5) indicate an opposite trend, shifting between approximately 200 and 460 m (average 330 m) during eddy conditions to shallower depths between approximately 120 and 420 m (average 270 m) during non-eddy conditions (Steinhardt et al., 2014).Although the δ 18 O cc suggests calcification somewhat deeper than the Mg / Ca data, both Mg / Ca-and δ 18 O-derived calcification depth indicate a shoaling for this species during non-eddy conditions.Furthermore, the average δ 18 O-derived calcification temperature of 10.4 ± 3.9 • C is in good agreement with previously published results for this species (Fallet et al., 2011;Birch et al., 2013).We refrain from correcting for a vital effect, as this would lower δ 18 O-derived calcification temperature to values unrealistically lower than the Mg / Ca-derived calcification temperatures for the last chambers.The observed remaining offset between single-specimen δ 18 O and single-chamber Mg / Ca in G. scitula suggests either that (1) there is a vital effect resulting in more enriched (i.e.positive) δ 18 O values than when this species would precipitate its shell in isotopic equilibrium with seawater, that (2) a more shallow calcification depth during formation of the final chamber, (3) that crust carbonate adds significantly to the total shell mass, or that (4) the Mg / Ca calibration for G. hirsuta (Anand et al., 2003) might be different from that of G. scitula.Following the vital effect correction of Kahn and Williams (1981), calcification temperature is 6.4 ± 3.9 • C, which is equivalent to an average calcification depth for G. scitula between 600 and deeper than 1100 m.This is in agreement with a suggested depth habitat within the upper 1000 m for this species (Schiebel et al., 1995;Ortiz et al., 1996;Itou et al., 2001).In our opinion the last two explanations are most likely; however, irrespective of the underlying mechanism, it is clear that the majority of the test carbonate precipitated at a depth greater or comparable to that of the ontogenetic carbonate of the final chambers.
The range of uncertainties related to a species' average calcification depth results from the relatively large natural interspecimen variability in Mg / Ca.Since we focus on relative differences within species between hydrographic conditions, the uncertainty in calcification temperature resulting from errors in the applied Mg / Ca-temperature calibration does not affect the absolute temperature differences between the eddy and non-eddy conditions.Instead, uncertainties in the calculated difference in calcification depths between species will be caused by the inter-specimen variability in Mg / Ca.

Cumulative calcification model
We used a conceptual oxygen isotope mass balance model (Wilke et al., 2006(Wilke et al., , 2009)), applying the temperature fractionation from inorganic calcite precipitation of Kim and O'Neil (1997) to our measured δ 18 O cc .The model equation describing foraminiferal migration as a function of depth used here is known as the cumulative form of the Weibull function (Weibull, 1939).It is a continuous probability function (Eq.4), relating the shell mass (M) to depth (z) using two constants (α and β) determining the shape of this relationship: Since shell size of planktonic foraminifera is thought to increase with water depth (Hemleben and Bijma, 1994;Peeters and Brummer, 2002), shell mass must also increase with depth.The isotopic composition of a single shell thus represents the weighted sum of equilibrium calcite precipitated over a depth range of the productive zone (i.e.where primary calcite formation takes place).Based on Eq. ( 5), the expected stable isotope composition of a specimen for a discrete water depth interval can be cal-culated as follows: Given the δ 18 O eq profile in the water column and the measured δ 18 O cc of the planktonic foraminifera, it is possible to model the mass development (growth pattern) by using the determined Mg / Ca calcification depth of the last chambers, indicating the base of the calcite production zone.The Mg / Ca-based temperature of the F-1 chamber was used to delimit 95 % of the calcite production.In Eq. ( 5), δ 18 O eq, i denotes the interval-averaged δ 18 O of equilibrium calcite for the specified depth interval.For convenience, shell mass at the sea surface was taken as zero and modelled δ 18 O cc was done by adapting the variables "α" and "β" in Eq. ( 5).
Increasing the value of "α" results in a growth curve with a narrow calcification range.Higher values for "β" result in a deepening of the growth curve, thereby determining the position of the base of the productive zone.In contrast to Wilke's (Wilke et al., 2006(Wilke et al., , 2009) approach, we have determined the calcification temperatures of the last three to four chambers, which were used to constrain the base of the calcification range and hence constrained values for "β".
In this model, it is assumed that shell growth always follows the same function, which is continuous and does not differ between species.Offsets between δ 18 O cc and δ 18 O sw from expected equilibrium ("the vital effect") are assumed to be constant over the temperature range in which the species calcifies.We have adapted δ 18 O sw in metre steps as calculated from in situ salinity measurements, which where interpolated for the upper 2000 m.We have used expected δ 18 O eq values of eddy and non-eddy conditions to compare depth distributions for all four species of planktonic foraminifera.
Calcification depths inferred from the cumulative δ 18 O model (Fig. 7) match previously published calcification depths and associated temperatures for each of the species relatively well (e.g.Cléroux et al., 2008Cléroux et al., , 2013;;Wilke et al., 2009;Fallet et., 2010Fallet et., , 2011;;Birch et al., 2013).In three species, measured δ 18 O cc values reflect shallower calcification depths than do single-chamber Mg / Ca-based calcification depths, which is consistent with the general model of migration to greater depth during growth.In the case of the deep-dwelling G. scitula, however, δ 18 O-based calcification depth is below that of the final chambers as derived from Mg / Ca temperatures.If a temperature correction for δ 18 O-based calcification temperatures of G. scitula is not applied, calcification depth based on δ 18 O cc can deviate up to 300 m from the Mg / Ca-based depths.This would suggest that the majority of the previously formed calcite was precipitated deeper in the water column.The model shows that species modulate their calcification pattern depending on the hydrographical conditions they live in (e.g.eddy or non-eddy conditions).For G. ruber, our results show that this species seems to be an exclusive surface dweller, and hence an ap- plication of the cumulative calcification model only confirms that the majority of the calcite is formed at the sea surface.
For the thermocline-dwelling species N. dutertrei, we find that this species calcifies most of its calcite in a narrow depth range.Our model indicates that calcification during eddy conditions is more intense in the deeper part of the thermocline (α = 8.8, β = 85), whereas calcification during non-eddy conditions is more equally distributed over the entire thermocline (α = 1.9, β = 47).It is worth noting that N. dutertrei appears to intensify its calcification efforts during eddy conditions deeper in the thermocline, matching well with the deepening of the isopycnals and hence a narrower range of optimal calcification conditions (Steinhardt et al., 2014).This calcification response is also reflected in more enriched δ 13 C values during eddy conditions.For P. obliquiloculata, modelled α and β values are relatively high, particularly during eddy conditions (α = 5.25, β = 133, compared to α = 3.1, β = 63 for non-eddy conditions).This indicates that most of the calcification in P. obliquiloculata takes place at a water depth of around 125 m during eddy conditions, and around 50 m during non-eddy conditions.The range at which G. scitula calcifies is well below the seasonal thermocline, reflected by high values for α and β (Fig. 7), and does not vary considerably during eddy and noneddy conditions.
In general, we conclude that temperature changes within the thermocline induced by eddies mostly affect nonsymbiotic species.Also, changes in cumulative calcite addition with depth seem to be species-specific.We modified the model by including Mg / Ca-based temperatures (following the species-specific equations of Anand et al., 2003)  lows for expected δ 18 O eq to be predicted for different species and shell sizes (Spero et al., 1997;Bijma et al., 1999;Itou et al., 2001;Peeters et al., 2002).The extended version of the model does not distinguish between calcite deposited during chamber formation (primary calcite) and calcite added as a result of wall thickening due to gametogenic calcite or the precipitation of crust (secondary calcite; Bé, 1980;Duplessy et al., 1981;Lohmann, 1995;Jonkers et al., 2012).Secondary calcification might play an important role of deeperdwelling species such as G. scitula, which could explain the offset (about 1 ‰) between δ 18 O model and δ 18 O cc .This suggests that relatively more calcite is formed deeper in the water column, or that secondary calcite is precipitated with a fundamentally different calcification mechanism.

Carbon isotopes -testing the calcification model
The δ 13 C values found in planktonic foraminifera are primarily a function of the carbon isotope composition of the dissolved inorganic carbon (DIC) in seawater (e.g.Urey, 1947;Epstein and Mayeda, 1953;McCorkle et al., 1990), which changes with water depth (e.g.Fairbanks et al., 1980;Curry and Crowley, 1987).Therefore, we can use the cumulative mass balance model output of the mass added per metre to calculated δ 13 C expect as the weighted sum of the δ 13 C DIC (Wilke et al., 2006).Depth-resolved carbon isotope composition (δ 13 C DIC ) available from locations closest to our study site (locations between 37 and 43  ture (adult) specimens (Brummer et al., 1986(Brummer et al., , 1987)).Comparison of water column δ 13 C DIC data (Fig. S1 in the Supplement) from several stations near the MC reveals that absolute values and depth range over which values decrease are similar at the different sites.To verify the depth-related calcification model, we compare measured δ 13 C cc with model-based δ 13 C expect values (Fig. 8).
Carbon isotope values become more negative from surface-dwelling G. ruber towards deeper-dwelling P. obliquiloculata near the upper thermocline.Conversely, the δ 13 C of Globorotalia scitula increases with depth.Low temperatures and reduced food availability have been suggested to result in relatively low metabolic rates in deep-dwelling species, so that their δ 13 C likely approaches δ 13 C DIC values (Birch et al., 2013).This suggests the involvement of biological controls on the δ 13 C of the different genera (Globigerinoides, Neogloboquadrina, Pulleniatina and Globorotalia).All δ 13 C expect values are higher than the measured δ 13 C cc .
Our cumulative mass balance shows that the majority of the carbonate of G. ruber is formed in surface waters (Fig. 7).Equal δ 13 C expect values for eddy and non-eddy conditions are the result of similarly enriched δ 13 C DIC in the mixed layer.The measured differences in δ 13 C cc (Fig. 8) are likely a consequence of the deepening thermocline during passage of an eddy, carrying nutrient-depleted waters (Kolasinski et al., 2012).Anticyclonic eddies are characterized by accumulation of warm, nutrient-poor and chlorophyll-depleted water in the centre, which implies that δ 13 C DIC is also more iso- Red symbols represent values for eddy conditions; blue symbols represent values for non-eddy conditions.Thick grey arrows indicate intra-species trends between non-eddy and eddy conditions; thin arrows indicate inter-specific trends.P. obliquiloculata does not calcify in isotopic equilibrium with dissolved CO 2 , but the deviation from isotopic equilibrium is a linear function of temperature (Mulitza et al., 1999), and hence there is no projected δ 13 C exect ; this is indicated by the dotted lines.The open diamond indicates δ 13 C exect for P. obliquiloculata non-eddy conditions.topically enriched.Still, local nutrient enrichment potentially occurs at the outer edge as a result of high turbulence along the isopycnal slope (e.g.Falkowski et al., 1991;Lévy, 2003).The strong response of the Mg / Ca and δ 18 O of N. dutertrei during eddy conditions (deeper calcification) is also reflected by more depleted δ 13 C cc values.Remineralization of organic matter at greater depth cause enrichment of δ 13 C DIC , resulting in the incorporation of lighter carbon isotopes into the shell of N. dutertrei during eddy conditions.Based on samples from a sediment trap in Cape Basin, Wilke et al. (2009) showed that the species N. dutertrei is an accurate recorder of the δ 13 C DIC .This is in agreement with previous findings (Mulitza et al., 1999), showing that the carbon isotopic composition of N. dutertrei exhibits a constant and temperatureindependent offset from δ 13 C DIC of ∼ 0.5 ‰ over a wide temperature range.This difference is in line with the offset in our data set between δ 13 C expect and δ 13 C cc of N. dutertrei (0.6 ‰).
The δ 13 C of the symbiont-barren G. scitula significantly deviates from those of the shallower-dwelling species as a result of a decrease in δ 13 C DIC with water depth (Figs.S1  and S2).The more depleted δ 13 C cc of G. scitula may also be a consequence of a lower metabolism of this species (Vergnaud-Grazzini, 1976;Kahn, 1977Kahn, , 1979;;Berger et al., 1978;Erez, 1978) compared to that of G. ruber and N. dutertrei.At high metabolic activity, more isotopically lighter carbon is incorporated, and since lower temperatures usually reduce metabolic rates, species inhabiting deeper water depths may incorporate relatively heavier carbon isotopes.Minor changes in δ 13 C cc for G. scitula during eddy compared to non-eddy conditions are in line with the minor response in calcification depth for this species.Similar to previous conclusions, this suggests that Mg / Ca-inferred temperature differences between N. dutertrei and G. scitula are good indicators of eddies passing (Steinhardt et al., 2014).In addition, the δ 13 C cc differences between these species might very well help to reconstruct eddy frequency in this area.The depthintegrated difference between δ 13 C of N. dutertrei and G. scitula changes from 0.25 to 0.05 ‰.
In the comparison of δ 13 C expect and δ 13 C cc for P. obliquiloculata there is a discrepancy between eddy and noneddy conditions (Fig. 8).Similar to N. dutertrei, this species is mostly associated with the thermocline (Anand et al., 2003;Cléroux et al., 2008;Sadekov et al., 2009).Our cumulative calcification model showed a slightly deeper calcification depth for N. dutertrei and a minor eddy response in the calcification range (Fig. 7).However, δ 13 C values indicate a significant difference between eddy and non-eddy conditions.Mulitza et al. (1999) showed that P. obliquiloculata does not calcify in isotopic equilibrium with dissolved CO 2 and that the deviation from isotopic equilibrium is a linear function of temperature (Fig. 8).While the mean of the δ 13 C cannot be used to infer the actual calcification depth, they argue that the spread and skewness of the individual δ 13 C measurements should still be representative of the range of calcification depths and habitat preferences within the thermocline.
Also, changes in the carbonate ion concentration with depth potentially play an important role in the observed differences between species and between eddy and non-eddy conditions (Figs. S1 and S2).Since the carbonate ion profile is expected to change in accordance with thermocline deepening when an eddy passes we refrained from correcting for this.The observed offsets between species, however, suggest that carbonate ion does play a role there.The deeper living species show an increasing offset with respect to the 1 : 1 line (Fig. 8).The exception is P. obliquiloculata, which responds to temperature rather than δ 13 C DIC carbonate ion changes (Mulitza et al., 1999).
Overall the relations observed here indicate that interpretation of the foraminifera vertical distribution in the upper water column can be unravelled by coupling various geochemical methods in order to retrieve calcification temperature at different stages in a foraminifera's life cycle.This in turn can be used to develop new proxies for the thermal and nutrient structure of the upper part of the water column.

Conclusions
Documenting changes in upper ocean stratification is essential for understanding past climatic conditions from sediment cores and is commonly estimated by determining the difference in δ 18 O between thermocline and surface-dwelling planktonic foraminifera (Spero et al., 2003;Cléroux et al., 2007;Farmer et al., 2007;Lin and Hsieh, 2007;Steph et al., 2009).We conducted stable isotope measurements on four species of planktonic foraminifera (G.ruber, N. dutertrei, P. obliquiloculata and G. scitula) from selected sediment trap samples, representing eddy and non-eddy conditions in the MC.
Using single shell δ 18 O cc paired to single-chamber LA-ICP-MS Mg / Ca measurements we applied a cumulative mass balance model in order to compare growth patterns of the various planktonic species during eddy and non-eddy conditions.The results indicate that most of the species have somewhat different calcification patterns during eddy and non-eddy conditions.Only Mg / Ca values of G. scitula suggest higher calcification temperatures than inferred from δ 18 O.Furthermore, the results of the δ 18 O cumulative mass balance model agree with previous findings that thermoclinedwelling N. dutertrei and deep-dwelling G. scitula are suitable recorders of eddy-induced hydrographic changes (Steinhardt et al., 2014).The combination of various proxies (e.g.Mg / Ca, δ 18 O and δ 13 C) can thus provide a useful set of geochemical proxies to reconstruct the thermal and nutrient structure of the upper part of the water column.
All species analysed have unique offsets from ambient seawater δ 13 C.However, comparison of species-specific isotopic trajectories with water column δ 13 C reveals that ambient δ 13 C DIC may be recorded by the species used in this study.The δ 13 C of N. dutertrei and G. scitula shows eddyrelated changes in their offsets and can potentially aid in unravelling eddy-related changes in the nutrient structure.
The Supplement related to this article is available online at doi:10.5194/bg-12-2411-2015-supplement.

Figure 1 .
Figure1.Hydrography of the southwestern Indian Ocean and location of the sediment trap (star) within the mooring array (top right).On the bottom right is a map of sea level anomaly showing the passing of an anticyclonic eddy over the trap location (star).The AVISO sea level anomaly map was produced using the AVISO live access server (http://las.aviso.altimetry.fr/las/getUI.do).EACC: East African Coastal Current; SEC: South Equatorial Current; AC: Agulhas Current.

Figure 3 .
Figure 3. Eddy (red circles) and non-eddy (blue circles) comparison of δ 13 C PDB and δ 18 O PDB for the analysed species.Black circles indicate average δ 13 C. Grey lines indicate standard deviation (SD); black capped lines are indicative of standard error (SE).
Fig. 6) are in close agreement with those reported previously www.biogeosciences.net/12

Figure 6 .
Figure 6.Apparent calcification depths of species are generally shallower during non-eddy conditions.Apparent calcification depths for eddy (red) and non-eddy conditions (blue) calculated from single-specimen δ 18 O cc using in situ temperature and δ 18 O sw .Calcification depth was determined by matching the measured foraminiferal δ 18 O cc with the δ 18 O eq , using the equation of Kim and O'Neil (1997).We used δ 18 O sw from the species calcification depth.The right panel is a close-up of the top left part of the left panel (upper 200 m).

Figure 8 .
Figure 8. Inter-species differences between expected δ 13 C values, based on the cumulative mass balance model, and measured δ 13 C values of G. ruber, N. dutertrei, P. obliquiloculata and G. scitula.Dashed line indicates the 1 : 1 line of measured and expected δ 13 C.Red symbols represent values for eddy conditions; blue symbols represent values for non-eddy conditions.Thick grey arrows indicate intra-species trends between non-eddy and eddy conditions; thin arrows indicate inter-specific trends.P. obliquiloculata does not calcify in isotopic equilibrium with dissolved CO 2 , but the deviation from isotopic equilibrium is a linear function of temperature(Mulitza et al., 1999), and hence there is no projected δ 13 C exect ; this is indicated by the dotted lines.The open diamond indicates δ 13 C exect for P. obliquiloculata non-eddy conditions.

Table 2 .
Average measurements of δ 18 O and δ 13 C with standard errors (SE) and corresponding standard deviations (SD) performed at the Universitat Autònoma de Barcelona on a Thermo Finnigan MAT253 mass spectrometer coupled to a Kiel IV device for CO 2 sample gas preparation (BCN) and a Thermo Finnigan Delta Plus mass spectrometer equipped with a GasBench II preparation device at the VU University Amsterdam (VU).Measurements of N. dutertrei, P. obliquiloculata and G. scitula are comparable and species-specific values are in good agreement.

,
Figs.2 and 3).The relationship between temperature and δ 18 O cc is generally described with more depleted δ 18 O cc values, indicating higher temperatures and thereby shallower calcification depths.Thus, each species has a distinct whole-shell δ 18 O signature, reflecting their different mean calcification depth.
18O cc range from −3.50 to 2.65 ‰.Although the values measured on individual specimens clearly overlap, each species has a different average δ 18 O cc and δ 13 C cc (Figs.2 and 3).The δ 18 O cc values are most depleted for G. ruber, somewhat more enriched in comparison to G. ruber for P. obliquiloculata and N. dutertrei, and most enriched in G. scitula (Table 1SD: ±0.24 ‰); more noticeable positive values are found for G. scitula with 1.47 ± 0.14 ‰ (SD: ±0.87 ‰; Fig.3).No significant trend between size fractions and stable isotopes was observed for any of the analysed species over the size range we used, as confirmed by ANOVA tests (Kruskal-Wallis one-way analysis of variance on ranks) www.biogeosciences.net/12/2411/2015/Biogeosciences, 12, 2411-2429, 2015 2416 J. Steinhardt et al.: Reconciling single-chamber Mg / Ca with whole-shell δ 18

Table 3 .
Results for δ 18 O and δ 13 C with standard errors (SE) and corresponding standard deviations (SD) under eddy and non-eddy conditions for G. ruber, N. dutertrei, P. obliquiloculata and G. scitula.