Volume 32, Issue 11 p. 1174-1194
Research Article
Free Access

Stable Oxygen Isotopes and Mg/Ca in Planktic Foraminifera From Modern Surface Sediments of the Western Pacific Warm Pool: Implications for Thermocline Reconstructions

Martina Hollstein

Corresponding Author

Martina Hollstein

MARUM–Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany

Correspondence to: M. Hollstein,

[email protected]

Search for more papers by this author
Mahyar Mohtadi

Mahyar Mohtadi

MARUM–Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany

Search for more papers by this author
Yair Rosenthal

Yair Rosenthal

Department of Marine and Coastal Sciences and Department of Earth and Planetary Sciences, Rutgers State University of New Jersey, New Brunswick, NJ, USA

Search for more papers by this author
Paola Moffa Sanchez

Paola Moffa Sanchez

School of Earth and Ocean Sciences, Cardiff University, Cardiff, UK

Search for more papers by this author
Delia Oppo

Delia Oppo

Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA, USA

Search for more papers by this author
Gema Martínez Méndez

Gema Martínez Méndez

MARUM–Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany

Search for more papers by this author
Stephan Steinke

Stephan Steinke

Department of Geological Oceanography, Xiamen University, Xiamen, China

Search for more papers by this author
Dierk Hebbeln

Dierk Hebbeln

MARUM–Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany

Search for more papers by this author
First published: 12 October 2017
Citations: 49


Mg/Ca and stable oxygen isotope compositions (δ18O) of planktic foraminifera tests are commonly used as proxies to reconstruct past ocean conditions including variations in the vertical water column structure. Accurate proxy calibrations require thorough regional studies, since parameters such as calcification depth and temperature of planktic foraminifera depend on local environmental conditions. Here we present radiocarbon-dated, modern surface sediment samples and water column data (temperature, salinity, and seawater δ18O) from the Western Pacific Warm Pool. Seawater δ18O (δ18OSW) and salinity are used to calculate individual regressions for western Pacific surface and thermocline waters (δ18OSW = 0.37 × S-12.4 and δ18OSW = 0.33 × S-11.0). We combine shell δ18O and Mg/Ca with water column data to estimate calcification depths of several planktic foraminifera and establish regional Mg/Ca-temperature calibrations. Globigerinoides ruber, Globigerinoides elongatus, and Globigerinoides sacculifer reflect mixed layer conditions. Pulleniatina obliquiloculata and Neogloboquadrina dutertrei and Globorotalia tumida preserve upper and lower thermocline conditions, respectively. Our multispecies Mg/Ca-temperature calibration (Mg/Ca = 0.26exp0.097*T) matches published regressions. Assuming the same temperature sensitivity in all species, we propose species-specific calibrations that can be used to reconstruct upper water column temperatures. The Mg/Ca temperature dependencies of G. ruber, G. elongatus, and G. tumida are similar to published equations. However, our data imply that calcification temperatures of G. sacculifer, P. obliquiloculata, and N. dutertrei are exceptionally warm in the western tropical Pacific and thus underestimated by previously published calibrations. Regional Mg/Ca-temperature relations are best described by Mg/Ca = 0.24exp0.097*T for G. sacculifer and by Mg/Ca = 0.21exp0.097*T for P. obliquiloculata and N. dutertrei.

Key Points

  • Combined Mg/Ca and stable oxygen isotopes in planktic foraminifera tests from accurately dated modern surface sediments
  • Seawater oxygen isotope-salinity regressions for surface and subsurface waters in the Western Pacific Warm Pool
  • Calcification depth estimates and regional multispecies and species-specific Mg/Ca-temperature calibrations

1 Introduction

The Western Pacific Warm Pool (WPWP) is a major source of heat and water vapor to the global atmosphere with far-reaching climate impacts (e.g., Gagan et al., 2004). The area is also thought to play an essential role in the global overturning circulation, because it provides waters to the Pacific equatorial current system and the Indonesian Throughflow, (e.g., Gordon, 1986). Present climate in the WPWP is mainly controlled by the Austral-Asian monsoon system and large-scale climate phenomena such as the El Niño–Southern Oscillation (ENSO). The regional climate is strongly coupled to ocean conditions. Changes in the prevailing climate conditions affect, for example, mixed layer depth and the thermocline structure (e.g., DiNezio et al., 2011; Vecchi et al., 2006). Thus, reconstructing past hydrographic conditions and variations in the vertical structure of the water column allows to draw conclusions on the regional WPWP climate evolution.

There is an ongoing debate how the thermocline depth varied throughout the past. For example, some records indicate a thermocline deepening during the Last Glacial Maximum (e.g., Bolliet et al., 2011), others indicate a thermocline shoaling during the same period (Andreasen & Ravelo, 1997; Beaufort et al., 2001; de Garidel-Thoron et al., 2007; Regoli et al., 2015; Sagawa et al., 2012), and yet others indicate no change compared to the modern ocean (Patrick & Thunell, 1997). Many of these reconstructions are based on the calculation of differences between shell Mg/Ca-derived temperature and/or δ18O of planktic foraminifera calcifying at different depth levels to estimate vertical temperature gradients within the upper water column (e.g., Bolliet et al., 2011; de Garidel-Thoron et al., 2007; Regoli et al., 2015). Previous studies used, for example, the difference between shell Mg/Ca in G. ruber as surface indicator and P. obliquiloculata or N. dutertrei as thermocline depth indicators (e.g., Bolliet et al., 2011) or the difference between G. ruber sensu strictu and G. ruber sensu lato (here referred to as G. ruber and G. elongatus following Aurahs et al., 2011) (Regoli et al., 2015). However, to choose species and interpret such proxy records correctly, it is essential to understand how modern hydrographic conditions are reflected in foraminiferal calcite.

The choice of species to use depends on regional calcification depths, which are determined by the species preferences and local environmental conditions. However, although many paleoclimate reconstructions for the WPWP exist, precise estimates of calcification depths are sparse in this area. Published reconstructions rely on plankton tow and sediment trap studies in the central equatorial Pacific, North Pacific, or Indian Ocean (Kawahata et al., 2002; Kuroyanagi & Kawahata, 2004; Mohtadi et al., 2011; Peeters et al., 2002; Rippert et al., 2016; Watkins et al., 1996). In addition, precise Mg/Ca-temperature calibrations are a prerequisite to convert Mg/Ca into temperature. For the WPWP there are only two regional Mg/Ca-temperature calibrations (Lea et al., 2000; Sagawa et al., 2012). Lea et al. (2000) provide a species-specific calibration for G. ruber, and Sagawa et al. (2012) present a multispecies calibration. Both calibrations are exposed to certain limitations. While the species-specific calibration might be biased by postdepositional effects on the core top planktic foraminifera sample material (Lea et al., 2000), the multispecies equation of Sagawa et al. (2012) bases only on data from a single station. Besides, both calibrations base on late Holocene sediments and therefore might lack comparability to present hydrography. Regional species-specific calibrations for subsurface planktic foraminifera species do not exist for the WPWP. Subsequently, most proxy studies use Mg/Ca-temperature calibrations from other areas (e.g., Bolliet et al., 2011; de Garidel-Thoron et al., 2007; Regoli et al., 2015; Tachikawa et al., 2014). However, the adequacy of previously published Mg/Ca-temperature calibrations for the WPWP has not yet been tested.

Here we present paired Mg/Ca and δ18O measurements on planktic foraminifera tests from radiocarbon-dated, modern surface sediments in combination with water column data from stations offshore the Philippines and Papua New Guinea (PNG) (Figure 1a and Table 1). Together, these areas represent a major part of the WPWP. We estimate species-specific calcification depths and temperatures of various planktic foraminifera species and establish regional Mg/Ca-temperature calibrations for the WPWP. In order to estimate calcification depths, we compare shell δ18O with depth profiles of expected equilibrium δ18O of calcite (δ18OC) at the respective locations. While this is a common approach, an advantage of our study is the availability of concurrently measured salinity and seawater δ18O (δ18OSW) data, upon which the calculated depth profiles of δ18OC are based on. In this framework, we also calculate and provide regional δ18OSW-salinity regressions for surface and subsurface water masses in the WPWP. Finally, we relate Mg/Ca to calcification temperatures to find the most appropriate Mg/Ca calibration for each species and establish a regional multispecies as well as monospecific Mg/Ca-temperature calibrations. We identify the appropriate species to reconstruct past variations in mixed layer depth and thermocline structure as well as the appropriate calibrations to convert Mg/Ca into temperature.

Details are in the caption following the image
(a) Schematic map of the study area showing the stations, where water column data (stars) and surface sediments (dots) were collected during expeditions SO-228 (blue) and RR-1313 (red). Sites are labeled with station numbers. (b) Mean annual sea surface temperature and (c) sea surface salinity derived from WOA13 (Locarnini et al., 2013; Zweng et al., 2013). Temperature and salinity maps were created with the Ocean data view software (Schlitzer, 2014). Dashed and solid arrows indicate main surface and (sub)surface currents influencing the study sites. MC = Mindanao Current, NEC = North Equatorial Current, NECC = North Equatorial Counter Current, EUC = Equatorial Undercurrent, NGCC = New Guinea Coastal Current, NGCUC = New Guinea Coastal Undercurrent, SEC = South Equatorial Current, SECC = South Equatorial Countercurrent, ITF = Indonesian Throughflow.
Table 1. Shell Stable Oxygen Isotopes and Mg/Ca Ratios of Various Planktic Foraminifera Species in Surface Sediments From the Western Pacific Ocean
Core Latitude (°N) Longitude (°E) Water depth (m) G. ruber δ18O (‰ VPDB) G. ruber Mg/Ca (mmol/mol) G. elongatus δ18O (‰ VPDB) G. elongatus Mg/Ca (mmol/mol) G. sacc. δ18O (‰ VPDB) G. sacc. Mg/Ca (mmol/mol) P. obliq. δ18O (‰ VPDB) P. obliq. Mg/Ca (mmol/mol) N. dutertrei δ18O (‰ VPDB) N. dutertrei Mg/Ca (mmol/mol) G. tumida (355–425 μm) δ18O (‰ VPDB) G. tumida (355–425 μm) Mg/Ca (mmol/mol) G. tumida (>425 μm) δ18O (‰ VPDB) G. tumida (>425 μm) Mg/Ca (mmol/mol)
RR1313-24 50MC 13.57 123.73 1055



5.28 −3.24 - −2.78 3.81 −3.47 3.17 −1.84 - - - - -
RR1313-25 53MC 12.78 123.48 559 −2.95 5.30 −3.30 - - 3.95 - - - 4.23 - - - -
GeoB 17404-2 7.90 126.54 404 −3.03 5.47 −2.91 - −2.69 3.95 - - - - - - - -
GeoB 17410-3 7.87 126.59 771 −3.08



−3.79 5.46






4.04 - - - 2.68 - 3.27 - -
GeoB 17414-2 6.26 125.83 2188 −3.13 4.95 −3.01 - −2.90 3.55 −2.34








2.62 - - - -
Papua New Guinea
GeoB 17419-2 −2.81 144.50 1887








−3.05 5.37 −2.95 3.77








−1.64 2.30






−0.94 1.63
RR1313-12 30MC −3.13 142.76 994 −3.74 5.30 −3.01 - −3.39 4.26 −2.40 2.91 - 2.59 - - - -
GeoB 17421-2 −3.55 144.20 588 −3.45 4.94 −3.43 5.42 −2.64 3.66 −2.20 2.70





1.95 −0.31 2.15 - -
GeoB 17429-1 −4.10 145.20 1604 −3.02 5.13






5.21 −2.62



−2.00 2.93 −2.01



−0.07 1.41 −0.08 1.72
GeoB 17430-2 −4.22 145.03 1160






5.19 −2.98 5.14 −2.73



−2.15 2.80 −1.75 2.78 −0.01



−0.31 1.90
GeoB 17432-3 −5.34 146.20 1388 −3.24 5.30 −2.86








3.58 −2.13 2.63 −2.09



- 1.35 0.06 1.56
GeoB 17435-2 −7.27 147.34 1001 −3.29







5.68 −2.81 3.88 −2.23



−2.09 2.26 −0.15 1.68






We note that the study is subject to certain limitations. First, it is based on surface sediments. This bears the disadvantages that we do not have direct information about hydrographic parameters at periods, when the calcite shells were built. Besides, we cannot fully exclude secondary influences (e.g., dissolution) on our data and, we cannot resolve (intra)seasonal changes in hydrography. Second, our study includes water column data measured during two expeditions. These data provide only snapshots of the WPWP hydrography. Moreover, we did not measure pH or [CO32−] and therefore cannot fully exclude an effect of pH or [CO32−] on shell Mg/Ca, although it appears negligible under ambient seawater conditions (Kısakürek et al., 2008; Russell et al., 2004). Third, since the availability of modern sample material is a prerequisite for calibration studies, our sites are exclusively located in the coastal WPWP, where sedimentation rates are higher than in the open Pacific Ocean. The applicability of our calibrations for the open ocean WPWP needs to be tested in future studies. Finally, especially species-specific calibrations are restricted by the fact that the temperature range within the study area is rather small. Where necessary, we discuss the limitations of the data in more detail (see section 5). Overall, the strengths of this study outweigh the mentioned limitations. Sediment-based studies have the great advantage that they are performed on the same material used for paleostudies. In relation to this, a great advantage of our study is the availability of (radiocarbon) dated, modern surface sediments. In addition, as mentioned above, our study greatly benefits from concurrent measurements of temperature, salinity, and δ18OSW.

2 Study Area

The WPWP is characterized by exceptionally high ocean temperatures with sea surface temperatures exceeding 28°C (Figure 1b) (Locarnini et al., 2013). Sea surface salinity is about 34 psu (Figure 1c) (Zweng et al., 2013). The average mixed layer depth is about 50–100 m (Locarnini et al., 2013). The upper thermocline waters are characterized by higher salinities with maxima of around 34.5–35.0 psu off the Philippines and 35.5 psu off PNG. Salinity maxima correspond to the North and South Pacific Tropical Waters (NPTW and SPTW). The NPTW is formed within the western North Pacific Subtropical Gyre and transported along the Philippines toward the equator by the Mindanao Current (MC) (Fine et al., 1994). The SPTW originates in the South Pacific Suptropical Gyre (Tsuchiya et al., 1989). It is transported westward by the SEC and along the coast of PNG by the New Guinea Coastal Current (NGCC) and Undercurrent (NGCUC) systems. Below the NPTW and SPTW waters are characterized by lower salinities and are influenced by the Antarctic Intermediate Water (AAIW) originating in the Southern Ocean and/or the North Pacific Intermediate water (NPIW) with stronger predominance of AAIW offshore PNG and of NPIW offshore the Philippines (Fine et al., 1994; Zenk et al., 2005).

The seasonal climate variability is mainly controlled by the Austral-Asian monsoon and leads to only minor changes in the WPWP hydrography. Temperature and salinity variations are smaller than 1–2°C and 1 psu (Locarnini et al., 2013; Zweng et al., 2013). On interannual timescales, El Niño–Southern Oscillation (ENSO) affects surface ocean conditions as well as the vertical structure of the water column in the WPWP with drier (wetter) conditions and a shallower (deeper) thermocline during El Niño (La Niña) years.

Generally, the northern part of the study area is characterized by oligotrophic surface conditions and a deep chlorophyll maximum at the top of the thermocline (Radenac & Rodier, 1996). North of PNG nutrient concentrations and biological productivity are higher than elsewhere in the WPWP (Radenac et al., 2016). However, the (vertical) distribution of nutrients and chlorophyll is variable on (intra)seasonal timescales (e.g., Higgins et al., 2006; Radenac et al., 2016; Radenac & Rodier, 1996).

The (intra)seasonal distribution of planktic foraminifera is controlled by different factors, such as temperature, salinity, and the availability of light and nutrients. Sediment trap data do not reveal a clear picture of (intra)seasonal preferences of planktic foraminifera in the study area (Kawahata et al., 2002; Yamasaki et al., 2008). Flux data from the equatorial Pacific showed large (small) peaks during boreal summer (winter) under El Niño conditions and an increased shell flux during the first half of the year under La Niña conditions (Kawahata et al., 2002; Yamasaki et al., 2008). Since no clear (intra)seasonal pattern is indicated by these data, we assume that planktic foraminifera calcify perennially in the WPWP.

3 Materials and Methods

3.1 Water Column Data and δ18OSW-Salinity Regressions

For this study, we used profiles of water column data, measured at 15 stations offshore the Philippines and PNG in May–June 2013 during R/V Sonne expedition SO-228 (Mohtadi et al., 2013) and at six stations in August 2013 during R/V Revelle expedition RR-1313 (Figure 2) (Y. Rosenthal, personal communication, 2013). Temperature and salinity profiles are based on CTD (conductivity, temperature, and depth) data. CTD data were measured using Seabird SBE911 (plus) CTD profilers during both expeditions. During CTD casts, water samples for stable isotope analyses were collected from several water depths (supporting information Table S1). Sampling was performed with CTD sampling rosettes equipped with 24 Niskin bottles of 10–15 L volume. A part of the collected water was siphoned into 100 mL glass bottles for stable isotope analyses; care was taken to avoid getting bubbles in the samples (Mohtadi et al., 2013; Y. Rosenthal, personal communication, 2013). All SO-228 samples (labeled as GeoB samples in Table 1) were sealed with wax and stored at 4°C before analysis. δ18OSW was determined with a Picarro L1102-i CRDS water analyzer with vaporization module V1102-i coupled to a CTC/Leaptec PAL autosampler at the Department of Geography and Earth Sciences, University of Erlangen-Nuremberg. Calibration against Vienna standard mean ocean water (VSMOW) was achieved by calibration to laboratory water standards calibrated against International Atomic Energy Agency standards VSMOW2 and Standard Light Antarctic Precipitation 2 (SLAP2) (van Geldern & Barth, 2012). External reproducibility was 0.05‰. For the RR1313 samples, δ18Osw measurements were made at Rutgers University, New Jersey, on a FISONS OPTIMA Mass Spectrometer equipped with a MicroMass Multiprep automatic sample processing system after water sample equilibration with CO2 using standard methods (Epstein & Mayeda, 1953; Fairbanks, 1982). All samples were run in duplicate. Precision was estimated to be ±0.03‰ (1σ) as determined by multiple (n = 12) daily analyses of a laboratory standard. Replicates must measure to within 0.068‰ to be included in the final data set. Instrument linearity and accuracy were determined by comparison of the laboratory standard to National Bureau of Standards (NBS) standard water VSMOW, Greenland Ice Sheet Project (GISP), and SLAP. Accuracy was estimated to be within 0.03‰ by comparison of measurements of North Atlantic Bottom Water with VSMOW.

Details are in the caption following the image
(a) Salinity, δ18OSW, and temperature profiles and (b) temperature-salinity relations at stations offshore the Philippines (red) and Papua New Guinea (black). Water masses are labeled as follows: NPTW = North Pacific Tropical Water, SPTW = South Pacific Tropical Water, NPIW = North Pacific Intermediate Water, AAIW = Antarctic Intermediate Water.

Generally, δ18OSW is linearly related to salinity (Craig & Gordon, 1965; Fairbanks et al., 1992). However, since both parameters are controlled by a number of factors, intercept and slope of their relation are not the same for all ocean regions. Therefore, regional calibrations are required (LeGrande & Schmidt, 2006). Because the δ18OSW-salinity relation is not constant over depth, we generated separate equations for surface and subsurface (SPTW and NPTW) waters. For both, we investigated individual relations for the areas offshore the Philippines and PNG as well as combined, more general, relations representing the entire study area. We do not necessarily consider water samples from all stations and depth intervals. Rather, for each equation we include samples that best characterize the relevant water masses (supporting information, Figure S1).The SPTW and NPTW regressions include only samples that represent the core of these water masses; the WPWP wide subsurface regression is more general and covers a slightly wider depth range.

3.2 Sediment Samples

Surface sediments were collected with multicorer devices during expedition SO-228 (nine sampling sites) (Mohtadi et al., 2013) and during expedition RR-1313 (three sampling sites) (Y. Rosenthal, personal communication, 2013). The upper 1or 2 cm of each multicore were washed over 63 μm sieves and dried.

All core sites are situated well above the present lysocline depth of 3,300 m where carbonate preservation is expected to be good (see Berger, Bonneau, & Parker, 1982, and references therein). The presence of aragonitic pteropod shells in most SO-228/GeoB core tops (including the core top from our deepest site) further indicates a good carbonate preservation in our samples.

3.3 Dating

Age estimates of all GeoB sediment samples are based on accelerator mass spectrometry (AMS) 14C ages (Table 2). All 14C ages were measured on monospecific Globigerinoides sacculifer or on mixed Globigerinoides ruber, G. elongatus, and G. sacculifer samples. The measurements were carried out at the Keck Carbon Cycle Accelerator Mass Spectrometry Laboratory, University of California, Irvine. Fraction modern carbon (F14C) values above 1 indicate modern ages for all 14C dated samples (Table 2). In addition, a few GeoB samples were treated with Rose Bengal and contained stained individuals, indicating that these individuals died only very recently.

Table 2. Radiocarbon Dating of Surface Sediments From the Western Pacific Warm Pool
Core Lab-ID Depth (cm) Species F14C ± error 14C age ± error (years) Cal. Age
GeoB 17404-2 142715 1–2 G. ruber, G. elongatus, and G. sacculifer 1.042 ± 0.002 −325 ± 20 >1950 A.D.
GeoB 17410-3 158806 1–2 G. ruber, G. elongatus, and G. sacculifer 1.064 ± 0.002 −490 ± 20 >1950 A.D.
GeoB 17414-2 158804 0–1 G. ruber, G. elongatus, and G. sacculifer 1.058 ± 0.002 −445 ± 20 >1950 A.D.
GeoB 17419-2 142718 0–1 G. sacculifer 1.072 ± 0.002 −550 ± 20 >1950 A.D.
142719 0–1 G. sacculifer 1.068 ± 0.003 −515 ± 20 >1950 A.D.
GeoB 17421-2 158805 0–1 G. ruber, G. elongatus, and G. sacculifer 1.054 ± 0.002 −420 ± 15 >1950 A.D.
GeoB 17429-1 142725 0–1 G. sacculifer 1.054 ± 0.002 −415 ± 20 >1950 A.D.
GeoB 17430-2 142717 0–1 G. ruber, G. elongatus, and G. sacculifer 1.058 ± 0.002 −445 ± 20 >1950 A.D.
GeoB 17432-3 142716 0–1 G. sacculifer 1.049 ± 0.002 −375 ± 20 >1950 A.D.
GeoB 17435-2 158803 0–1 G. sacculifer 1.066 ± 0.002 −500 ± 20 >1950 A.D.
  • Note. Results are expressed as fraction modern carbon (F14C) and conventional 14C ages.

The RR-1313 samples did not contain enough foraminifera for radiocarbon dating. For age control of these samples, we measured the carbon isotopic composition (δ13C) on about 10 specimens of G. ruber from the 250–300 μm size fraction from the upper 10 cm of each multicore to check for the Suess effect (Figure S2). The Suess effect describes a rapid decrease in seawater δ13C going along with the depletion in 13C of atmospheric CO2, which has been caused by an increase in deforestation and burning of 12C-rich fossil fuels since the industrial revolution (e.g., Böhm et al., 1996; Friedli et al., 1986). The δ13C measurements were performed at Rutgers University, New Jersey, using a Micromass Optima mass spectrometer, coupled to an automatic line for carbonate preparation. The stable isotope values were calibrated against the international Vienna Peedee belemnite (VPDB) standard using an internal standard, which is calibrated against the National Bureau of Standards (NBS) 19 standard. The long-term standard deviation for δ13C was 0.06‰. Rapid drops in δ13C from our RR-1313 multicores are indicative for Suess effect and hence modern ages of these surface sediments (Figure S2). Thus, all samples used in this study reflect modern hydrographic conditions of the WPWP.

3.4 Isotope and Trace Element Analyses

Tests from the foraminiferal species G. ruber, G.elongatus, and G. sacculifer (without sac-like final chamber) (all taken from the 250–355 μm size fraction), Neogloboquadrina dutertrei and Pulleniatina obliquiloculata (355–425 μm), and Globorotalia tumida (355–425 μm and >425 μm) were picked under a binocular for δ18O and Mg/Ca analyses. For all species but G. tumida, specimens were separately picked for isotope and Mg/Ca analyses. G. tumida specimens were very rare in most of the samples. In order to ensure that fragments of several individuals were used for each measurement, 10 (where available) individuals were picked, crushed, homogenized, and then separated for isotope and trace element analyses.

Around 40–120 μg carbonate (around 3–10 specimens) were used for stable isotope analyses. The isotopic composition of all samples was measured at the MARUM-isotope laboratory, University of Bremen, Germany, using a Finnigan MAT 251 mass spectrometer, connected to an automatic line for carbonate preparation (type “Kiel III”). All isotope values were calibrated against the international VPDB standard. The internal carbonate standard is a Solnhofen Limestone, which is calibrated to the NBS 19 standard. The analytical standard deviation for δ18O is below ±0.07‰. To check the reproducibility of the data, we performed up to four (depending on the available material) replicate measurements on 11 samples (Table 1). The results indicate an average standard deviation of 0.22‰ for δ18O.

For Mg/Ca analyses we used 30 (where available) well-preserved specimens of G. ruber and G. elongatus, 25 specimens of G. sacculifer, 15 specimens of P. obliquiloculata and N. dutertrei, and between 2 and 10 individuals of G. tumida. Although the presence of pteropods indicates negligible effect of carbonate dissolution, all samples were weighed to estimate a potential influence of dissolution on the Mg/Ca records. Samples were gently crushed between two glass plates to open the chambers. The full trace metal cleaning procedure followed the protocol described by Barker et al. (2003) with an additional reductive step (Boyle & Keigwin, 1985; Rosenthal et al., 1997, 1999). Samples were dissolved in 0.0065 M HNO3, centrifuged for 10 min at 10,000 rpm, and diluted with 0.5 N HNO3. The final calcium concentration of the samples was on average 3.2 mM. Trace metal ratios were measured at Rutgers University, New Jersey, with a Thermo Fisher/Finnigan Element XR sector field inductively coupled plasma mass spectrometer (ICP-MS). Mg/Ca measurements were performed in low resolution (Δm/m = 300). Measured ratios were blank corrected. Mass drift and matrix effects and the long-term precision of the data were controlled within house standard solutions (Rosenthal et al., 1999). All Mg/Ca values are given in mmol/mol. Replicate measurements on 14 samples revealed an average standard deviation of 0.23 mmol/mol (Table 1). To monitor the cleaning efficacy, Al/Ca, Fe/Ca, and Mn/Ca were measured alongside Mg/Ca. None of these ratios showed a covariance with Mg/Ca (Figure S3). Mg/Ca ratios of individual species do not show a correlation to water depth or shell-normalized weights (Figure S4). Thus, we exclude any substantial effect of carbonate dissolution on shell Mg/Ca values.

3.5 Estimation of Calcification Depths and Temperatures

In order to estimate species-specific calcification depths, we compare shell δ18O of individual species with depth profiles of δ18OC, assuming that all species calcified in isotopic equilibrium with seawater. The water depth where shell δ18O matches the expected δ18OC is considered to reflect the calcification depth of the respective species at the site.

Depth rofiles of expected δ18OC were calculated from SO-228 and RR-1313 CTD salinity profiles as follows. First, δ18OSW was calculated from salinity using here established regional (WPWP) δ18OSW-salinity equations (equations 1 and 2 in section 4.1). The average uncertainty in the calculated δ18OSW given as the average standard deviation between measured and calculated δ18OSW is 0.05‰. It was necessary to calculate δ18OSW from CTD salinity instead of using the measured δ18OSW to obtain continuous profiles. Results were converted from VSMOW to the VPDB scale by subtracting 0.27‰ (Hut, 1987). Then, we applied a set of commonly used δ18O-temperature equations (Table 3) to predict equilibrium δ18OC using CTD temperature and the previously calculated δ18OSW. Finally, we matched shell δ18O to the δ18OC profiles. Since surface sediments were not always taken at the same positions as CTD, we matched shell δ18O to profile(s) from those station(s), which are nearest to the core sites. To determine uncertainties of the calcification depth of each species and core site, we estimated the calcification depths based on shell δ18O with added/subtracted species-specific standard deviations. Depending on the standard deviation and the shape of the profiles, average uncertainties range between ±10 and 30 m.

Table 3. General and Species-Specific δ18O Temperature Equations Used in This Study
Reference Species Linear equations T (°C) = a − b18OC − δ18OSW)
a b
Bemis et al. (1998) O. universa (HL) 14.9 −4.8
Bouvier-Soumagnac and Duplessy (1985) N. dutertrei 10.5 −6.58
Farmer et al. (2007) G. ruber 15.4 −4.78
G. sacculifer 16.2 −4.94
N. dutertrei 14.6 −5.09
P. obliquiloculata 16.8 −5.22
G. tumida 13.1 −4.95
Mulitza et al. (2003) G. ruber 14.2 −4.44
G. sacculifer 14.91 −4.35
Shackleton (1974) Uvigerina sp. 16.9 −4.0
Spero et al. (2003) G. sacculifer 12.0 −5.67
Quadratic equations T (°C) = a − b18OC − δ18OSW) + c18OC − δ18OSW)2
a b c
Kim and O'Neil (1997) Inorganic 16.1 −4.64 0.09

CTD temperatures at depths corresponding to the estimated calcification depths give an estimate of the calcification temperatures for each species and core site. For G. elongatus, P. obliquiloculata, and N. dutertrei only a few samples from the Philippines contained enough individuals to perform isotope analyses. For G. tumida none of the Philippines' samples contained enough specimens and therefore depth and temperature estimates for this species are only representative for the area offshore PNG.

4 Results

4.1 Water Column Data and δ18OSW-Salinity Regressions

The δ18OSW-salinity relations for the study area during summer 2013 reveal that a δ18OSW increase of about 0.3 to 0.4‰ corresponds to a salinity increase of 1 psu (Figure 3). Regressions for surface waters offshore the Philippines and PNG (considering only samples from the Bismarck and Solomon Seas) are very similar (Figure 3a, blue and red dots, respectively). The regression line for surface waters offshore PNG shows only a slightly steeper slope and, consequently, a smaller intercept. Taking all SO-228 and RR-1313 stations across the WPWP into account (Figure 3a, all black and colored dots), the surface δ18OSW-salinity relation can be described as

Details are in the caption following the image
Regional δ18OSW-salinity relations for (a) surface and (b) subsurface waters in the Western Pacific Warm Pool. Dots indicate individual samples included in the regressions. Red and blue colors indicate subsets of samples included in the regional regressions off Papua New Guinea (PNG) and the Philippines, respectively. Black dots indicate samples from sites that are not included in the regional regressions. Black lines indicate regression lines for the entire study area and colored lines for the subareas. Gray stippled lines show regional δ18Osw-salinity equations published by (1) Fairbanks et al. (1997), (2) Morimoto et al. (2002), (3/5) Leech et al. (2013), and (4) LeGrande and Schmidt (2011). Bars indicate laboratory standard deviations for oxygen isotope measurements.

Regression lines describing the δ18OSW-salinity relations for subsurface waters (NPTW and SPTW) show a shallower slope than for surface waters with slightly higher intercepts (Figure 3b). Due to the very small salinity ranges within the NPTW and SPTW, it is difficult to assess a robust relationship for each of these water masses. The regressions also greatly depend on the depths and number of stations included in each equation. However, although NPTW and SPTW have different characteristics (see Figure 2), their δ18OSW-salinity regressions are very similar to each other. Therefore, we also provide one equation for WPWP subsurface waters that includes all SO-228 and RR-1313 sample stations (Figure 3b, all black and colored dots).


4.2 Calcification Depth

In general, shell δ18O increases with increasing calcification depth of the species. G. ruber and G. elongatus record the lowest δ18O values, and G. tumida records the highest δ18O values. G. sacculifer, P. obliquiloculata, and N. dutertrei calcify in between. Shell δ18O and our derived mean calcification depths of the species used are very similar offshore the Philippines and PNG. Therefore, in our discussion we do not distinguish between samples from the Philippines and PNG. Figure 4 and Table 4 give δ18O-derived depths for all species as estimated by applying the δ18O-temperature equation of Bemis et al. (1998). Depth estimates for each species applying different δ18O-temperature equations in comparison are provided in the supporting information (Figures S4–S6).

Details are in the caption following the image
Shell δ18O-derived calcification depth estimates for several planktic foraminifera (colored dots) in the Western Pacific Warm Pool. Gray lines indicate depth profiles of predicted δ18OC calculated from individual CTD casts (see section 3.5 for details). The average calcification depth at each core site is determined by matching shell δ18O to δ18OC of the profile(s) nearest to the respective core site. Horizontal bars indicate species-specific standard deviations for shell δ18O, and vertical bars on black dots show exemplarily average uncertainties in calcification depth derived by projecting shell δ18O with added/subtracted standard deviations on the δ18OC profiles. The uncertainty of the δ18OC profiles arising from the average standard deviation between measured and calculated δ18OSW is 0.05‰ (not shown). Colored shading indicates calcification depth ranges for the individual species in the study area. The average calcification depth is highlighted by colored horizontal bars. Note the different scaling of the axes in different panels. Individual data points, which are shown outside the range of predicted δ18OC, do not match any profile (see text).
Table 4. Calcification Depth and Temperatures
Core G. ruber (250–355 μm) G. elongatus (250–355 μm) G. sacculifer (250–355 μm) P. obliquiloculata (355–425 μm) N. dutertrei (355–425 μm) G. tumida (355–425 μm) G. tumida (>425 μm)
Depth (m) Temperature (°C) Depth (m) Temperature (°C) Depth (m) Temperature (°C) Depth (m) Temperature (°C) Depth (m) Temerature (°C) Depth (m) Temperature (°C) Depth (m) Temperature (°C)
RR1313-24 50MC 7 29.8 0 29.6 85 28.4 0 29.8 140 24.3 - - - -
RR1313-25 53MC 77 28.3 0 28.8 - - - - - - - - - -
GeoB 17404-2 47 28.9 57 28.5 83 27.8 - - - - - - - -
GeoB 17410-3 22 29.3 0 29.9 70 28.2 - - - - - - - -
GeoB 17414-2 25 29.2 37 28.7 45 28.2 88 25.8 94 25.3 - - - -
Papua New Guinea
GeoB 17419-2 32 29.7 40 29.8 60 29.6 147 24.6 150 24.2 200 18.8 175 20.9
RR1313-12 30MC 0 29.7 - - 0 29.7 114 27.5 - - - - - -
GeoB 17421-2 0 29.7 0 29.7 77 28.3 117 26.7 157 23.1 240 17.7 - -
GeoB 17429-1 60 29.0 67 28.6 72 27.9 125 25.8 123 25.8 248 16.4 250 15.9
GeoB 17430-2 42 29.2 63 28.7 68 28.5 108 26.3 153 24.7 252 15.8 230 17.6
GeoB 17432-3 0 29.4 66 28.6 75 27.5 103 26.2 107 26.0 - - 263 15.6
GeoB 17435-2 0 28.9 40 28.9 45 26.7 90 26.9 97 26.2 258 16.6 238 17.7
  • Note. The calcification depth of each species and core site was estimated by comparing shell δ18O to depth profiles of predicted δ18O calcite (δ18OC). Predicted δ18OC was calculated from CTD salinity applying the δ18OSW-salinity equations of this study and the δ18O temperature equation of Bemis et al. (1998) using CTD temperature. We assigned a depth of 0 m to samples that did not match any profile. Calcification temperatures are CTD temperatures at depth corresponding to the estimated calcification depths.

4.2.1 Mixed Layer Species: G. ruber, G. elongatus, and G. sacculifer

Average shell δ18O values are very similar for G. ruber and G. elongatus (−3.19 and −3.11‰, Table 1). This implies very similar calcification depths for both species in the WPWP. Depending on the selected δ18O-temperature equation, shell δ18O derived mean calcification depths range from 0 to 45–105 m (Figures 4 and S4). Applying the δ18O-temperature equations of Bemis et al. (1998) and Shackleton (1974), the estimated depth range is within the upper 80 m of the water column for both G. ruber and G. elongatus. By applying the equations of Kim and O'Neil (1997) and Farmer et al. (2007), calcification depths shoal to 0–45 m and 0–65 m, respectively, while the application of the species-specific equation of Mulitza et al. (2003) results in a slightly deeper depth range (0–105 m) (Figure S5a).

Shell δ18O of G. sacculifer varies around −2.73‰ (excluding RR1313-12 30MC) (Table 1). Applying different δ18O-temperature equations, the mean calcification depth varies between 45 and 95 m. It ranges from 45 to 85 m using the equation of Bemis et al. (1998) (Figure 4c) and from 55 to 95 m applying the δ18O-temperature equations of Shackleton (1974) and Spero et al. (2003). The equations of Farmer et al. (2007) and Kim and O'Neil (1997) lead to slightly shallower calcification depths (0–70 m) and the equation of Mulitza et al. (2003) to deeper calcification depths (70–110 m) (Figure S5b). Note that sample RR1313-12 30MC shows an extremely low δ18O value (−3.39‰), which does not match any profile.

4.2.2 Upper Thermocline Species: P. obliquiloculata and N. dutertrei

Shell δ18O of P. obliquiloculata varies around −2.15‰ (excluding RR1313-24 50MC) (Table 1). For the majority of the samples this indicates a calcification depth between 90 and 125 m when applying the δ18O-temperature equation of Bemis et al. (1998) (Figure 4d). Applying the equations of Shackleton (1974) or Bouvier-Soumagnac and Duplessy (1985) leads to very similar calcification depths (85–120 and 90–130 m), and the equations of Kim and O'Neil (1997) and Farmer et al. (2007) leads to slightly shallower calcification depths (70–105 and 40–75 m) (Figure S6a). Our foraminifera tests from site GeoB 17419-2 record exceptionally high δ18O values resulting in, compared to the other core sites, relatively deeper (20–40 m) calcification depths. With a δ18O value of −3.47‰ (Table 1), sample RR1313-24 50MC does not match any δ18OC profile.

The N. dutertrei average shell δ18O of−1.88‰ is slightly higher than that of P. obliquiloculata. By using the equations of Bemis et al. (1998), Shackleton (1974), or Farmer et al. (2007), shell-derived calcification depths vary between around 90 and 160 m (Figures 4e and S5b). The application of the δ18O-temperature equation from Kim and O'Neil (1997) or Bouvier-Soumagnac and Duplessy (1985) results in slightly shallower or deeper depth ranges (80–150 or 95–155 m) (Figure S6b).

4.2.3 Lower Thermocline Species: G. tumida

Average shell δ18O of G. tumida is −0.15‰ for the 355–425 μm size fraction and −0.14‰ for the >425 μm size fraction (excluding GeoB 17419-2) (Table 1). Shell δ18O of individual samples differ by maximal 0.3‰. Hence, under the restriction of the small set of samples, our data indicate that there is no major size effect on the calcification depths of G. tumida. Based on the equation of Bemis et al. (1998), the calcification depth of G. tumida at most sites is between 230 and 265 m water depth (Figure 4f). The application of the equations from Shackleton (1974) and Kim and O'Neil (1997) results in slightly shallower calcification depths, ranging between 195 and 235 m. Based on the species-specific equation of Farmer et al. (2007), the estimated calcification depths are slightly deeper, ranging between 255 and 310 m (Figure S7). Exceptionally low δ18O in G. tumida shells from GeoB 17419-2 indicates up to 70 m shallower calcification depth at this site, independent of the size fraction used.

4.3 Shell Mg/Ca

Mg/Ca ratios recorded in the Philippines and PNG samples are very similar. G. ruber and G. elongatus record highest Mg/Ca, averaging 5.26 and 5.37 mmol/mol, respectively. Shell Mg/Ca in G. sacculifer varies around 3.85 mmol/mol. N. dutertrei and P. obliquiloculata show similar shell Mg/Ca varying around 2.60 and 2.66 mmol/mol, respectively. G. tumida shows lowest Mg/Ca ratios around 1.88 mmol/mol (355–425 μm) and 1.69 mmol/mol (>425 μm). Hence, average Mg/Ca of the different species confirms the results obtained from oxygen isotope ratios with calcification depths being shallowest for G. ruber, G. elongatus, and G. sacculifer, intermediate for P. obliquiloculata and N. dutertrei, and deepest for G. tumida. Mg/Ca ratios of P. obliquiloculata in RR1313-24 50MC, N. dutertrei in RR1313-53MC, and G. tumida in GeoB17410-3 are exceptionally high (3.17 mmol/mol, 4.23 mmol/mol, and 3.27 mmol/mol) (Table 1). The high Mg/Ca value of P. obliquiloculata in RR1313-24 50MC goes along with exceptionally low δ18O and could therefore indicate an extremely shallow calcification depth at this site. Due to the limited number of specimens of N. dutertrei in RR1313-25 53MC and G. tumida in GeoB 17410-3, we could not measure the oxygen isotope composition in these samples.

5 Discussion

5.1 Water Column Data and δ18OSW-Salinity Regressions

Generally, our surface and subsurface δ18OSW-salinity regressions are within the range of published WPWP regressions (Fairbanks et al., 1997; Leech, Lynch-Stieglitz, & Zhang, 2013; LeGrande & Schmidt, 2011; Morimoto et al., 2002). However, slope and/or intercept differ from previous regression lines. Especially, the regressions of Fairbanks et al. (1997) and LeGrande and Schmidt (2011) and the surface regression of Leech et al. (2013) show shallower slopes (around 0.3) and, accordingly, larger intercepts (between −10.47 and −9.14) than our regressions. The regression of Morimoto et al. (2002) is almost identical to our PNG surface equation (δ18OSW = 0.42 × S-14.3). Since δ18OSW-salinity relations depend on local environmental conditions, deviations are most probably due to different sampling sites, water depths, and periods.

For the first time, we generated equations for individual water masses (surface waters, NPTW, and SPTW). Our results reveal that for both, WPWP wide and regional regressions, the ratio between δ18OSW and salinity is higher in surface than in subsurface waters. Therefore, it appears reasonable to use different δ18OSW-salinity regressions for surface and subsurface water masses. The regression coefficients of our regression lines (Figure 3) indicate that the application of a more general, WPWP wide regression for combined NPTW and SPTW is more robust than individual regressions for the NPTW and SPTW. This is due to the very small salinity range within the NPTW offshore the Philippines and the SPTW offshore PNG.

5.2 Calcification Depths and Temperatures

Accurate calcification depth estimates are indispensable to deduce precise calcification temperatures for each species. The accuracy of the depth estimates depends on the precision of the δ18OSW-salinity equation applied, the choice of the δ18O-temperature equation, the availability of local water column data, the seasonal and interannual variations in local hydrography, and possible shell disequilibrium effects (see Regenberg et al., 2009; Steph et al., 2009).

A comparison between predicted δ18OC calculated using discrete δ18OSW measured in water samples and predicted δ18OC using δ18OSW calculated from CTD salinity shows that the calculated values reproduce measured δ18O precisely (Figure S8). The average deviation between calculated and measured δ18OSW is ±0.07‰, the maximal deviation is ±0.35‰, and only 6 out of 98 samples yielded δ18OSW differences larger than 0.20‰ (see Table S2). The comparison provides reliability to the accuracy of the δ18OSW-salinity equations for our sampling period and sites.

For the following analyses we use calcification depth estimates based on the δ18O-temperature equation that was generated by Bemis et al. (1998) in a culture experiment for O. universa. Although it has been obtained on a single species not used in our study, this equation is commonly used for other planktic foraminiferal species (e.g., Mohtadi et al., 2014; Spero et al., 2003; Thunell et al., 1999) and its application gives realistic results for all species used here. The equations of Shackleton (1974), Bouvier-Soumagnac and Duplessy (1985), and Spero et al. (2003) lead to similar calcification depths. The quadratic equation of Kim and O'Neil (1997) results in marginally shallower depths for surface and similar depths for subsurface dwellers. The application of the equations of Farmer et al. (2007) results in slightly shallower calcification depths. The equations of Farmer et al. (2007) are based on sample material from much greater water depths (mostly 3000–4000 m) where carbonate is affected by dissolution (Hertzberg & Schmidt, 2013) which would increase δ18O in foraminiferal tests. Consequently, the application of these equations would underestimate calcification depths. The equations proposed by Mulitza et al. (2003) result in deeper calcification depths. These equations are based on plankton tow studies and might overestimate calcification depths due to lower δ18O in shells of living foraminifera (see Regenberg et al., 2009, and references therein).

Based on the assumption that planktic foraminifera calcify perennially in the WPWP, we assume that our δ18O data represent mean annual conditions. However, we note that data from individual samples might be biased and reflect other than normal conditions (e.g., a single season or El Niño/La Niña conditions). We are well aware that our hydrographic data show a snapshot and therefore do not necessarily represent mean annual hydrography. However, seasonal temperature and salinity variations are very small in our study area and mostly restricted to the mixed layer (see section 2). Interannual variations in ocean hydrography are mainly caused by ENSO variability. Our water column data were collected during a normal year and are thus not biased to El Niño or La Niña conditions. Considering the paucity of continuous subsurface temperature, salinity, and especially δ18OSW data from the WPWP, our work provides the first simultaneous measurements on these parameters at different water depths at stations in close proximity to our core sites and thus the hitherto most suitable hydrographic estimates for the determination of calcification depths and temperatures.

Nonetheless, we also compared shell δ18O with depth profiles of δ18OC calculated by using mean annual temperature and salinity data from the World Ocean Atlas 13 (WOA13) (Locarnini et al., 2013; Zweng et al., 2013). Shell δ18O of most G. ruber and G. elongatus samples do not match the WOA13 δ18OC profiles.WOA13 derived calcification depths of G. sacculifer extent over a larger depth range than SO-228 and RR-1313 CTD (from here on referred to as CTD) calcification depths. WOA13 depth estimates of P. obliquiloculata, N. dutertrei, and G. tumida differ only slightly from CTD-derived depth estimates (maximal 35 m) with a tendency to shallower depths. Overall, for our study, the application of CTD data gives more realistic results than the application of WOA13 data.

Previous studies have shown that many species do not calcify in isotopic equilibrium with seawater (see Ravelo & Hillaire-Marcel, 2007, and references therein). Potential reasons are the photosynthetic activity of symbionts, incorporation of low δ18O metabolic CO2, species-specific calcification rates, the addition of gametogenic calcite, and possibly carbonate ion concentrations of the ambient seawater (see Ravelo & Hillaire-Marcel, 2007, and references therein). Disequilibrium effects depend on local conditions and the sample material used. For most species used in our study, negative disequilibrium effects ranging between 0.0 and 1.0‰ are presumed (see Lončarić et al., 2006; Niebler, Hubberten, & Gersonde, 1999, and references therein). No vital effects are reported for G. tumida. Correcting for negative disequilibrium effects would increase shell δ18O and subsequently result in deeper calcification depths and colder calcification temperatures (see discussion in Regenberg et al., 2009). However, the large variety of factors that could possibly influence shell δ18O makes it difficult to correct δ18O for disequilibrium effects precisely. In addition, most published δ18O-temperature equations do not take into account biological disequilibrium effects on δ18O. Therefore, we did not correct δ18O values for disequilibrium effects.

Overall, our results show that G. ruber and G. elongatus calcify within the mixed layer (0–80 m) and G. sacculifer calcifies at the bottom of the mixed layer (45–85 m). P. obliquiloculata and N. dutertrei calcify within the upper thermocline, whereby our δ18O values indicate that N. dutertrei calcifies within a larger depth range (90–160 m) than P. obliquiloculata does (90–125 m). G. tumida seems to calcify well below P. obliquiloculata and N. dutertrei within the lower thermocline at depth between 230 and 265 m. In relation to water masses, this means that G. ruber, G. elongatus, and G. sacculifer calcify in surface waters. P. obliquiloculata and N. dutertrei calcify predominantly within the NPTW offshore the Philippines and within the SPTW offshore PNG. G. tumida calcifies within the transition between NPTW/SPTW and NPIW or AAIW. Our depth estimates generally agree with results from sediment traps and plankton tows in the central equatorial Pacific, North Pacific, and Indian Oceans (Kuroyanagi & Kawahata, 2004; Mohtadi et al., 2009; Peeters et al., 2002; Rippert et al., 2016; Watkins et al., 1996). However, our data show a tendency to somewhat deeper absolute calcification depths and wider depth ranges for most species, probably owing to a generally thick mixed layer and deep thermocline in the WPWP. Recent results from the eastern WPWP indicate even deeper habitat depths of planktic foraminifera (Rippert et al., 2016). Such differences in absolute calcification depths are likely related to the specific regional hydrographic conditions in each study area. For G. ruber and G. elongatus, our data imply very similar calcification depths. This is in sharp contrast to studies from the South China Sea and North Pacific Ocean that suggest a deeper habitat depth for G. elongatus (Kawahata, 2005; Steinke et al., 2005; Wang, 2000) but in agreement with studies from the eastern Indian Ocean and the Caribbean Sea, which suggest the same habitat depth for both species (Mohtadi et al., 2009; Thirumalai et al., 2014).

5.3 Mg/Ca Versus Calcification Temperatures

A large number of studies revealed that Mg/Ca in planktic foraminiferal tests show an exponential relationship with ocean temperatures (e.g., Anand et al., 2003; Cléroux et al., 2008; Dekens et al., 2002; Elderfield & Ganssen, 2000; McConnell & Thunell, 2005; Mohtadi et al., 2009, 2011; Nürnberg et al., 1996; Regenberg et al., 2009). Some culture and core top studies also described a positive salinity effect on shell Mg/Ca of planktic foraminifera (Arbuszewski et al., 2010; Ferguson et al., 2008; Kısakürek et al., 2008; Mathien-Blard & Bassinot, 2009; Nürnberg et al., 1996). However, studies suggesting a significant salinity influence have been criticized due to substantial dissolution effects, seasonality (Arbuszewski et al., 2010), or diagenetic alteration (Ferguson et al., 2008) on the sample material (Hertzberg & Schmidt, 2013; Hönisch et al., 2013). However, in our study area the salinity range is rather small (between 33.9 and 35.6 psu). Variations between stations at water depths that correspond to the calcification depth of individual foraminifera species at the corresponding core sites are below 1.2 psu. There is no significant correlation between Mg/Ca and salinity (R values range between 0.00 and 0.35 for individual species). Besides, shell Mg/Ca in samples from offshore the Philippines and PNG are very similar, although these areas are influenced by water masses characterized by different salinities (Figure 2). Therefore, we argue that shell Mg/Ca in our samples is not biased by salinity. Some studies also indicate a negative effect of pH or [CO32−] on Mg/Ca (Evans et al., 2016; Kısakürek et al., 2008; Lea et al., 1999; Russell et al., 2004; Spero et al., 2015). To date, it is not clear whether pH or [CO32−] exert a dominant control on Mg/Ca (Allen et al., 2016; Evans et al., 2016). Moreover, the effect is neither constant over temperature (Spero et al., 2015) nor the same for different planktic foraminifera species (e.g., Allen et al., 2016) and some studies provide confidence that the effect is negligible for ambient seawater conditions (Kısakürek et al., 2008; Russell et al., 2004). Since pH or [CO32−] were not measured during both SO-228 and RR-1313 expeditions, we cannot estimate the range of these parameters over the study area. We do not apply any corrections for possible effects of carbonate chemistry on shell Mg/Ca, while we cannot fully exclude such effect on our samples. More extensive culture and/or sediment trap studies are required to quantify the effect pH or [CO32−] on Mg/Ca of the various planktic foraminifera species.

Mg/Ca to temperature calibrations are usually expressed as Mg/Ca = B exp (A*Temperature). We combined data of all species and compared shell Mg/Ca to CTD temperature at δ18O-derived calcification depths in a multispecies approach (Figure 5). A depth of 0 m was assigned to samples, where shell δ18O did not match the δ18OC. Mg/Ca and temperature show a clear exponential relationship (Figure 5). Previous studies suggest that this relationship is best described by a reduced major axis (or geometric mean) regression (RMA) of the natural Log of (Mg/Ca) against calcification temperature (Anand et al., 2003; Rosenthal & Lohmann, 2002). An advantage of the RMA is that it accounts for both uncertainties in Mg/Ca and calcification temperatures. Especially in field studies, calcification temperatures inherit an intrinsic scatter that arises from a range of different factors including uncertainties in the depth estimates or seasonality, for example. Using the RMA approach, the multispecies Mg/Ca-temperature relation is described by the following equation:

Details are in the caption following the image
Shell Mg/Ca versus calcification temperature for different planktic foraminifera species. Black bars indicate species-specific standard deviations of Mg/Ca. Black and colored dashed lines indicate regional multispecies and species-specific regressions calculated with a RMA regression (this study). Gray solid lines show published multispecies (large graph) and species-specific (small graphs) regression lines. Numbers denote published calibrations (see also Table 5). Note that a, b, and c refer to different calibrations in the same publication.

The calculation followed Isobe et al. (1990). The uncertainties of the slope and intercept are given as standard deviations assuming that the intrinsic scatter of the data dominates any errors of the measurement process.

For comparison, we also calibrated shell Mg/Ca against WOA13 mean annual temperatures. Despite the differences in estimated calcification depths, WOA13-derived calcification temperatures are very similar to CTD-derived calcification temperatures. The reason is that temperatures of the WOA13 climatology are generally lower than our CTD temperatures at the same depth levels. Due to the similarity of the calcification temperatures, the WOA13-derived regression is within the error range of the regression based on CTD-derived calcification temperatures (A = 0.101, B = 0.24).

Our multispecies Mg/Ca-temperature relation is in good agreement with previously published multispecies and species-specific temperature calibrations (Figure 5 and Table 5). However, comparing our calibration to published ones, it has to be noted that our calibration is based on samples that were treated by reductive cleaning, whereas most published calibrations are based on samples cleaned without a reductive step. Previous studies indicate that the reductive cleaning leads to a decrease in Mg/Ca (e.g., Barker et al., 2003; Xu et al., 2010). To estimate the effects on the Mg/Ca-temperature relations, we applied a correction assuming that the Mg/Ca relations were reduced by about 10% in all samples (e.g., Barker et al., 2003; Martin & Lea, 2002; Rosenthal et al., 2004). We note that some studies suggest different values for individual species (e.g., Xu et al., 2010), but since exact rates of Mg/Ca loss are unknown for most species, we presume a constant rate of 10% for all species. The resulting regression line shows the same slope as the original calibration (0.097) and an only slightly modified intercept (0.24). Thus, the modified calibration is within the error range of the original one.

Table 5. Multispecies and Species-Specific Mg/Ca Temperature in Comparison to Previously Published Relations
Reference Species Exponential relation Mg/Ca = B exp (A*Temperature)
Anand et al. (2003) (1a) Multispecies 0.09 0.38
Sagawa et al. (2012) (2) Multispecies 0.077 0.455
Elderfield and Ganssen (2000) (3) Multispecies 0.1 0.352
This Study Multispecies 0.097 0.26
Anand et al. (2003) (1b) G. ruber 0.09 0.44
McConnell and Thunell (2005) (4) G. ruber 0.068 0.69
Lea et al. (2000) (5) G. ruber 0.089 0.3
Dekens et al. (2002) (6a) G. ruber 0.09 0.38
This Study G. ruber 0.097 0.31
This Study G. elongatus 0.097 0.32
Dekens et al. (2002) (6b) G. sacculifer 0.09 0.37
Nürnberg et al. (1996) (7) G. sacculifer 0.089 0.39
Nürnberg, Müller, and Schneider (2000) (8) G. sacculifer 0.076 0.49
This Study G. sacculifer 0.097 0.24
Anand et al. (2003) (1c) P. obliquiloculata 0.09 0.328
Cléroux et al. (2008) (9b) P. obliquiloculata 0.039 1.02
This Study P. obliquiloculata 0.097 0.21
Dekens et al. (2002) (6c) N. dutertrei 0.08 0.6
This Study N. dutertrei 0.097 0.21
Cléroux et al. (2008) (9a) Deep dwelling species 0.052 0.78
Rickaby and Halloran (2005) (10) G. tumida 0.09 0.53
Regenberg et al. (2009) (11) G. tumida 0.041 1.23
Mohtadi et al. (2011) (12) G. tumida 0.068 0.41
This study G. tumida (355–425 μm) 0.097 0.33
This study G. tumida (>425 μm) 0.097 0.31
  • Note. Calibration numbers introduced in Figure 5 are given in brackets.

A multispecies equation does not seem accurate enough to describe the Mg/Ca-temperature relation of individual species in the WPWP. For example, all the G. ruber and G. elongatus samples fall above, and all the P. obliquiloculata and N. dutertrei samples fall below the regression line. That means that calcification temperatures of G. ruber and G. elongatus are overestimated, and those of P. obliquiloculata and N. dutertrei are underestimated by the multispecies regression. For that reason, species-specific regression lines are additionally required to reconstruct calcification temperatures for individual species precisely. Since the temperature range between the sites in our study area is very narrow, it is difficult to determine the temperature sensitivity for individual species unequivocally. Hence, we calculated regional species-specific regressions by assuming the same temperature sensitivity, A = 0.097, as it was calculated by the multispecies approach, for all species (Figure 5). The resulting regression lines for G. ruber and G. elongatus are very similar to the multispecies and species-specific regression published by Anand et al. (2003) or Dekens et al. (2002), respectively (Mg/Ca = 0.39 exp 0.09*T). For the temperature range relevant for this study, the regression lines of G. tumida fall next to the multispecies regression lines of Anand et al. (2003) and Sagawa et al. (2012). However, especially slope and intercept of the Sagawa et al. (2012) regression differ from those of our species-specific regression (Table 5) and it is important to note that the slope affects the amplitude of temperature variations in paleorecords. The regression lines of G. sacculifer, P. obliquiloculata, and N. dutertrei deviate from previously published correlations (Figure 5). Calcification temperatures of these species are exceptionally warm in the western tropical Pacific Ocean despite deeper calcification depths, and thus, G. sacculifer, P. obliquiloculata, and N. dutertrei require regional, species-specific Mg/Ca-temperature calibrations. Assuming the temperature sensitivity calculated by the multispecies approach (A = 0.097), our data indicate an intercept of B = 0.24 for G. sacculifer and an intercept of B = 0.21 for both P. obliquiloculata and N. dutertrei (Figure 5).

A comparison of Mg/Ca temperatures and calcification temperatures gives a measure of the uncertainty inherent in the Mg/Ca-temperature calibrations (Anand et al., 2003). We compared Mg/Ca temperatures calculated by the application of species-specific calibrations to δ18O-derived calcification temperatures (Figure 6). The average standard error between Mg/Ca and δ18O-derived calcification temperatures is 0.5°C. The consistency of Mg/Ca and δ18O-derived calcification temperatures validates our previous steps and assumptions and gives further evidence that Mg/Ca and δ18O data are not influenced by secondary effects.

Details are in the caption following the image
Consistency of temperature estimates. (a) Mg/Ca versus δ18O-derived calcification temperatures. Mg/Ca temperatures are based on species-specific regressions; δ18O temperatures were calculated with the equation of Bemis et al. (1998). The black solid line shows the correlation between Mg/Ca and δ18O temperatures. For comparison, the 1:1 relationship is shown (stippled line). (b) Comparison of Mg/Ca temperatures (filled symbols) and δ18O temperatures (open symbols). Note that Figure 6a includes only samples with paired δ18O and Mg/Ca measurements, whereas Figure 6b includes all samples where δ18O or Mg/Ca measurements are available. Samples where only δ18O or Mg/Ca measurements are available are marked with stars. Sample numbers are related to cores as defined in Table S3.

5.4 Application of Our Mg/Ca Temperature Regression

Our species-specific equations were applied to convert Mg/Ca into temperature (Figure 7). While our regional Mg/Ca to temperature relations were established on the basis of only samples for which we have δ18O measurements and thereby direct estimates of foraminiferal calcification depths at the respective core site, we applied the equations to all samples where we measured Mg/Ca ratios (see Table 1). Temperatures obtained from G. ruber and G. elongatus match mixed layer temperatures. Mg/Ca temperatures of G. sacculifer correspond to temperatures from the mixed layer bottom and uppermost thermocline. Mg/Ca temperatures calculated from P. obliquiloculata and N. dutertrei reflect upper and those calculated from G. tumida lower thermocline conditions. P. obliquiloculata and N. dutertrei are most suitable to track changes in NPTW and SPTW since they calcify at water depths corresponding to the core of these water masses.

Details are in the caption following the image
(a) Mg/Ca temperatures calculated by applying the newly established species-specific calibrations projected on CTD temperature profiles (gray lines). Horizontal bars denote temperature errors, calculated following the method described in Mohtadi et al. (2014). Colored vertical bars indicate calcification depth ranges resulting from the application of our species-specific Mg/Ca-temperature calibrations. (b) For comparison, we show shell δ18O-derived calcification depth estimates for all species in combination. Gray lines indicate depth profiles of predicted δ18OC calculated from CTD casts. Samples where only Mg/Ca or δ18O measurements are available are marked with stars.

6 Conclusions

We presented novel, paired Mg/Ca and δ18Omeasurements on multispecies planktic foraminifera tests from accurately dated modern surface sediments in combination with water column data (CTD and δ18OSW) from the WPWP. Based on CTD salinity and δ18OSW, we establish new δ18OSW-salinity regressions for surface (mixed layer) and subsurface (thermocline) waters (NPTW and SPTW) in the WPWP. Our data imply that it is reasonable to apply different δ18OSW-salinity regressions for surface and thermocline water masses. Due to very similar regression lines for the Philippines and PNG areas, individual regressions for sub(surface) waters from both areas can be combined into more robust equations representing the entire study area. These equations are δ18OSW = 0.37 × S-12.4 for surface and δ18OSW = 0.33 × S-11.0 for thermocline waters.

Shell δ18O-derived estimates reveal that the calcification depths of planktic foraminifera offshore the Philippines and offshore PNG are very similar. Our estimates match results from other areas and reveal that also in the WPWP, G. ruber and G. elongatus reflect mixed layer conditions (0–80 m), and G. sacculifer reflects bottom of mixed layer and uppermost thermocline conditions (45–85 m). P. obliquiloculata and N. dutertrei preserve upper (90–160 m) and G. tumida lower thermocline conditions (230–265 m). Hence, our data imply that these species are the best choice to reconstruct thermocline conditions in the WPWP. N. dutertrei inhabits a slightly larger depth range than P. obliquiloculata in the WPWP. Therefore, P. obliquiloculata might be more suitable to reconstruct the upper thermocline. Previous thermocline reconstructions that are based on other species need to be reconsidered in face of the new results. For example, shell δ18O does not indicate different habitat depths for G. ruber and G. elongatus. Hence, for the WPWP, the data do not support the use of G. ruber-G. elongatus records to reconstruct variations in the vertical structure of the upper water column.

Our newly established regional multispecies Mg/Ca temperature regression is within the range of published multispecies and species-specific Mg/Ca-temperature calibrations (Mg/Ca = 0.26 exp 0.097*T as calculated using reduced major axis regression). However, the Mg/Ca-temperature relation of most individual species is more accurately described by species-specific calibrations. We find that the regional regressions for G. ruber, G. elongatus, and G. tumida are similar to the species-specific or multispecies equations published by Dekens et al. (2002) and Anand et al. (2003). Calcification temperatures of G. sacculifer, P. obliquiloculata, and N. dutertrei are exceptionally warm in the western tropical Pacific and thus require regional, species-specific calibrations. The application of previously published calibrations would underestimate the calcification temperatures of these species. Using a reduced major axis regression, we calculate the species-specific calibration Mg/Ca = 0.24exp 0.097*T for G. sacculifer and Mg/Ca = 0.21exp 0.097*T for P. obliquiloculata and N. dutertrei. Nevertheless, further studies are needed to confirm the applicability of these calibrations for the open Pacific Ocean and for paleoreconstructions.


We would like to thank the captains, crews, and the scientific shipboard parties of expeditions SO-228 and RR-1313. We thank Henning Kuhnert, Birgit Meyer-Schack, and Ryan Bu for technical assistance. John Southon (UC Irvine) is acknowledged for performing radiocarbon measurements. GeoB sample material was stored, curated, and supplied by the GeoB Core Repository at the MARUM–Center for Marine Environmental Sciences, University of Bremen, Germany. RR samples were provided by the Rutgers Core Repository. We thank two anonymous reviewers for their constructive comments, which helped to improve the quality of our manuscript. The work is funded by the DFG-Research Center/Cluster of Excellence “The Ocean in the Earth System”, the BMBF project 03G0228A (EISPAC), and by the NSF project OCE1131371. The data reported in this paper will be made available on Pangaea (www.pangaea.de) and WDS (www.icsu-wds.org).