5.1. Mg/Ca Versus δ18O-Derived Calcification Depths/Temperatures
 In the following, calcification depths are considered to be between 20 m and 50 m for G. ruber s.s., at 50 m for G. ruber s.l., G. sacculifer, and G. bulloides, at 75 m for N. dutertrei and P. obliquiloculata, and at 100 m for G. tumida. The deeper habitat inferred for various species by applying the plankton tow and culture based equations of Mulitza et al. [2003, Figures 3c, 3g, and 4c] and Spero et al. [2003, Figure 4a] might be a result of generally higher δ18O values recorded by surface samples. Our results agree with previous studies suggesting habitat depths of G. ruber s.s., G. ruber s.l., G. sacculifer and G. bulloides within the mixed-layer, and of N. dutertrei, P. obliquiloculata, and G. tumida within the thermocline [e.g., Fairbanks and Wiebe, 1980; Fairbanks et al., 1982; Peeters, 2000; Field, 2004; Kuroyanagi and Kawahata, 2004; Farmer et al., 2007].
 Shell Mg/Ca ratios of all species show a significant, linear relationship to temperatures from the WOA 05 data at depths that correspond to the estimated calcification depth of each species (Figure 6). Since salinities in the upper water column of the study area are lower than 35 psu (Figure S2), a positive salinity effect on shell Mg/Ca ratios [Arbuszewski et al., 2010] can be neglected. The large scatter in the data (gray dots in Figure 6) might be due to the (a) relatively narrow temperature range for each species, (b) different amount of encrustation [e.g., Schmidt and Mulitza, 2002], (c) different genetic types [e.g., Darling et al., 2003], (d) different ages of the surface samples, (e) a limited coverage of local hydrography by the WOA 05 data, both spatially and vertically, and (f) interannual variability in local hydrography caused by ENSO or IOD that cannot be resolved in our core top study. Therefore, it is not surprising that average values for each basin (black dots in Figure 6) show improved correlation coefficients (R2) at or greater than 0.7 for all species by reducing the above mentioned uncertainties, although a larger R2 is also a result of fewer data points. It should be noted that for G. ruber s.s., G. bulloides, and G. tumida, correlation coefficients between Mg/Ca and WOA 05 temperatures are also significant at adjacent depths due to the fact that calcification of planktic foraminifera and hence, incorporation of Mg in their calcite shell, occur within a depth range, rather than at a fixed depth (see Figure S3 and discussion below).
Figure 6. Relationship between shell Mg/Ca ratio and the calcification temperatures of various planktic foraminiferal species in the study area. Gray: the entire data sets; black: average values for each basin. Temperatures are taken from the WOA 05 data set, at depths that correspond to the estimated calcification depth of each species (see Figures 3–5). Mg/Ca values are plotted versus annual mean temperatures at (a) 20 m and (b and c) 50 m, and versus summer temperatures at (d) 50 m, (e and f) 75 m, and (g) 100 m, respectively.
Download figure to PowerPoint
 In face of a small temperature range for each species in the study area, Mg/Ca-temperature relationship could be expressed as a linear function (Figure 6). However, previous studies showed that shell Mg/Ca in planktic foraminifera increases exponentially with temperature [e.g., Elderfield and Ganssen, 2000; Dekens et al., 2002; Anand et al., 2003; McConnell and Thunell, 2005; Cléroux et al., 2008; Regenberg et al., 2009]. The species-specific Mg/Ca-temperature exponential relationships from this study (Figure 7, black dots and lines) fit within the range of previously published equations, although they occur at the warm end of the existing calibrations (Figure 7, gray lines). The only exception is the Mg/Ca-temperature relationship of G. tumida that shows a comparable slope with the Rickaby and Halloran  equation, but a different intercept (Figure 7g), which might be due to the much shallower inferred habitat of this species in the study area, or to the different size-fractions used in the other studies (300–355μm [Rickaby and Halloran, 2005] and 355–400μm [Regenberg et al., 2009]). We therefore calculated a regional Mg/Ca-temperature calibration for G. tumida:
Figure 7. (a–g) Shell Mg/Ca ratio to δ18O-derived calcification temperature of various planktic foraminiferal species from this study (black dots and lines) compared to other species-specific temperature calibrations (gray lines). Note that the exponential fits in Figures 7a–7d (Figures 7e–7g) are extrapolated to lower (higher) temperatures than in the original publications, and that all Mg/Ca values from this study are at the warm end of the published calibrations. 1a: calibration with calculated exponential value; 1b: calibration with assumed exponential value; 8a: species-specific calibration; 8b: deep-dwelling (multispecies) calibration.
Download figure to PowerPoint
 For all the other species, the existing Mg/Ca-temperature calibrations can be used for the eastern equatorial Indian Ocean. For G. ruber s.s., G. ruber s.l., and G. sacculifer the regional calibrations fit best to the equations proposed by Anand et al.  for the same size-fractions used in this study (Mg/Ca = 0.34 exp (0.102*T) and Mg/Ca = 0.38 exp (0.09*T) for G. ruber, Figures 7a and 7b, Mg/Ca = 0.347 exp (0.09*T) for G. sacculifer without sac, Figure 7c). Regional Mg/Ca-temperature relationship of G. bulloides is best explained by the equation of Elderfield and Ganssen  (Mg/Ca = 0.81 exp (0.081*T)). For N. dutertrei, the regional calibration fits best to the equation proposed by Regenberg et al.  (Mg/Ca = 0.65 exp (0.065*T), Figure 7e). Finally, regional Mg/Ca-temperature relationship of P. obliquiloculata matches best to the equation for deep-dwelling foraminifera proposed by Cléroux et al.  (Mg/Ca = 0.78 exp (0.052*T), Figure 7f).
 As mentioned above, δ18O-disequilibrium effect on the calculation of apparent calcification depths of planktic foraminifera might also impact their Mg/Ca-temperature relationships. Correcting the measured δ18O values for a range of disequilibrium values suggested for G. sacculifer, N. dutertrei, and P. obliquiloculata [see Niebler et al., 1999; Regenberg et al., 2009, and references therein] would increase their calcification depths, but decrease significantly (0.1–0.3) the correlation coefficient of their Mg/Ca-temperature calibration. This finding, in accordance with results of Regenberg et al. , justifies the application of the disequilibrium-uncorrected calibration for these species. There is no disequilibrium effect reported for G. ruber s.l. and G. tumida. Correcting the measured δ18O values for a range of disequilibrium values suggested for G. ruber s.s. (G. bulloides) would increase (increase or decrease) its calcification depth, while the correlation coefficient of the Mg/Ca-temperature calibration remains high (Figure S3). This might be a result of a bimodal vertical distribution of these species in the water column [e.g., Peeters et al., 2002; Kuroyanagi and Kawahata, 2004] due to different hydrologic conditions in the study area (upwelling and non-upwelling areas), or different genotypes of these species in sediment samples [e.g., Darling et al., 2003]. In spite of these possible effects, the water depths with a significant Mg/Ca-temperature correlation coefficient for these species match the range of their δ18O-derived calcification depths (20–50 m for G. ruber and 20–75 m for G. bulloides, Figures 3a–3d and 4e–4h), justifying the application of the disequilibrium-uncorrected calibration for these species. Nonetheless, future field studies on both planktic foraminifera and local hydrography are imperative in order to assess the δ18O-disequilibrium effect of planktic foraminifera, and to better calculate their apparent calcification depth.
5.2. Use of Temperature Proxies for Reconstructing Upper Water Column Hydrology
 Conversion of the Mg/Ca values to temperatures using the species-specific calibrations that match the apparent calcification depths of planktic foraminifera reveals a relatively consistent picture in the study area (Figure 8). In the non-upwelling basins of SB, NB, and the NMB, Mg/Ca temperatures of the surface-dwelling species G. ruber s.s., G. ruber s.l., G. sacculifer, and G. bulloides reflect mixed-layer temperatures between 0 m and 50 m at or above 28°C, whereby the succession of these species reflects decreasing (increasing) temperatures (habitat depths). Alkenone-based temperatures appear to match Mg/Ca temperatures of G. sacculifer, also reflecting mixed-layer temperatures in these basins. Since there are no significant seasonal changes in the mixed-layer conditions of these basins, our sediment surface data suggest that these proxies can be used to reconstruct past mixed-layer conditions of the eastern equatorial Indian Ocean, from ∼4° N to ∼4° S. Mg/Ca temperatures of P. obliquiloculata and N. dutertrei reflect upper thermocline temperatures at ∼70 m (26°C–28°C) and ∼100 m (23°C–25°C), respectively. Despite the sparse data of G. tumida from these basins, it appears that this species records nearly the same temperatures as N. dutertrei, at ∼100 m water depth (23°C–24°C). Accordingly, these three species can be used to reconstruct the upper thermocline conditions in the eastern Indian Ocean, from ∼4° N to ∼4° S.
Figure 8. Temperature profiles and nitrate concentration of the upper 200 m in the different basins of the study area (WOA 05, [Garcia et al., 2006; Locarnini et al., 2006]). Solid black lines: annual mean temperatures; dashed red lines: summer temperatures; dashed blue lines: winter temperatures; solid green lines: annual mean nitrate concentrations in micromole per liter (μM L−1); dashed green lines: boreal summer nitrate concentrations; mbs: meter below surface. Color dots indicate average temperatures, calculated from shell Mg/Ca ratio of planktic foraminiferal species, and alkenone unsaturation index. Alkenone temperatures are calculated after Conte et al. . Mg/Ca-temperatures are calculated using species-specific equations of Anand et al.  for G. ruber s.s., G. ruber s.l., and G. sacculifer, Cléroux et al.  for P. obliquiloculata, Regenberg et al.  for N. dutertrei, Elderfield and Ganssen  for G. bulloides, and own equation for G. tumida. Color lines denote the range of the calculated temperatures in different samples; values in brackets refer to the number of the observed surface sediment samples. The bottom (top) three panels refer to basins that are (not) affected by seasonal upwelling.
Download figure to PowerPoint
 In the southern basins of SMB, JB, LB, and the SS, calibrated temperatures based on shell Mg/Ca- and calcification depth of planktic foraminifera suggest that G. ruber s.s., G. ruber s.l., and G. sacculifer reflect annual mean mixed-layer temperatures of 27°–28°C between 0 m and 50 m (Figure 8). Alkenone-based temperatures remain at ∼28°C, suggesting that this proxy can only be used to reconstruct annual mean SST south of 4°S. It should be noted that alkenone temperatures in the study are at the upper limit of this proxy [e.g., Conte et al., 1998] and might underestimate the growth temperature of the alkenone-producing coccolithophores, particularly during boreal winter when SST is highest. Downcore studies from the study area show a strong seasonal signal in the alkenone-based SST estimates during glacial periods when SSTs lie within the temperature response of alkenone unsaturation [Lückge et al., 2009; Mohtadi et al., 2010]. Hence, it is well possible that alkenones in surface sediments simply record the limit of their temperature response, which corresponds to modern annual mean SST in the study area but masks their seasonality.
 Shell Mg/Ca of G. bulloides seems to record summer temperatures ranging between 20 m and 75 m, with average values reflecting temperatures of 24°–26°C at the base of the mixed-layer during boreal summer, at ∼50 m. Mg/Ca temperatures of P. obliquiloculata and N. dutertrei reflect upper thermocline temperatures at ∼60 m (21–24°C) and ∼80 m (20–22°C) during boreal summer, respectively, thus recording a slightly shallower depth than in the non-upwelling basins. G. tumida appears to record temperatures between 19°C and 20°C at ∼100 m depth during boreal summer, at a similar depth as in the northern, non-upwelling basins. In general, these results are in good agreement with the inferred habitat depths from a sediment trap time series in the upwelling area off southern Java [Mohtadi et al., 2009] suggesting habitat depths of 0–30 m for G. ruber, 60–80 m for P. obliquiloculata, and 60–90 m for N. dutertrei.
 Comparison of the Mg/Ca data with the nitrate concentrations in the upper water column further supports previous findings that G. bulloides and the thermocline-dwelling species are also bound to the nutricline (Figure 8) [e.g., Fairbanks and Wiebe, 1980]. In particular, N. dutertrei and G. tumida are apparently independent of the thermocline temperature, but dwell at depths characterized by a nitrate concentration of 10–15 μM L−1. In general, it appears that temperature changes at the upper thermocline in different basins do not significantly affect the water depth that defines the shell Mg/Ca ratio. Rather, the thermocline species appear to track the top of the nutricline that mostly corresponds to the same water depths (70–100 m), regardless of the absolute temperature values at these depths (Figure 9). For instance, Mg/Ca-derived temperatures of P. obliquiloculata track the summer 75 m isotherm that represents the upper thermocline (nutricline), although temperatures at this depth vary between ∼22°C and ∼28°C in the study area (Figure 9). To this end, our data do not allow a conclusion on whether temperature or nutrient is decisive for the planktic foraminiferal calcification depths. This enterprise requires extensive regional field studies in the future.
Figure 9. Alkenones- and Shell Mg/Ca-derived temperatures averaged for each basin. Calibrations are as in Figure 8. Solid lines (dashed lines) indicate annual mean (summer) upper water column temperatures in the different basins at 0 m and 50 m (0–100 m); averaged for each depth at each basin using the WOA 05 database [Locarnini et al., 2006]. SB: Simeulue Basin; NB: Nias Basin; NMB: Northern Mentawai Basin; SMB: Southern Mentawai Basin; JB: Java Basin; LB: Lombok Basin; SS: Savu Sea. The latter four basins are characterized by monsoon-induced seasonal upwelling.
Download figure to PowerPoint
 In search of a robust proxy for reconstructing past changes in the thermocline depth and/or changes in the upper water column stratification/mixing, our data show that both the pattern and the magnitude of the temperature difference between G. ruber s.s. (or G. ruber s.l.) and P. obliquiloculata in surface sediments of the study area track the difference between temperatures at 20 m and at 75 m from the WOA 05 data (Figure 10a). Likewise, temperature difference between G. ruber s.s. and N. dutertrei effectively follows the difference between 20 m and 100 m, suggesting that these differences can be used to track past changes in the water column structure in the eastern equatorial Indian Ocean (Figure 10a). Temperature difference between G. bulloides and P. obliquiloculata (N. dutertrei) shows the same pattern as observed for the difference between summer temperatures at 50 m and at 75 m (100 m, Figure 10b). However, the calculated difference between G. bulloides and N. dutertrei temperatures lies between the WOA 05 difference of 50–75 m and 50–100 m, which might be due to the Mg/Ca temperatures of N. dutertrei that reflect temperatures between 75 m and 100 m water depth (Figure 9).
Figure 10. (a) Mg/Ca-based average temperature difference (ΔT) between G. ruber s.s. and P. obliquiloculata (black dots), G. ruber s.l. and P. obliquiloculata (gray dots), and between G. ruber s.s. and N. dutertrei (black circles) in different basins. Dashed lines represent temperature difference between 20 m and 75 m (black), and 20 m and 100 m (gray) from the WOA 05 database for boreal summer. (b) Mg/Ca-based average temperature difference (ΔT) between G. bulloides and P. obliquiloculata (black dots), and between G. bulloides and N. dutertrei (gray dots) in different basins. Dashed lines represent temperature difference between 50 m and 75 m (black), and 50 m and 100 m (gray) from the WOA 05 database for boreal summer. (c) Difference between average alkenone and G. bulloides temperatures (black dots), and between G. sacculifer and P. obliquiloculata (gray dots) temperatures in different basins. Dashed lines represent temperature difference between annual mean SST (0 m) and summer temperatures at 50 m (black), and between 20 m and 75 m during boreal summer (gray) from the WOA 05 database. Abbreviations for the different basins are as in Figure 9.
Download figure to PowerPoint
 Temperature difference between G. sacculifer and thermocline-dwelling species follows the same pattern as observed between G. bulloides (G. ruber s.s.) and thermocline species in the study area (Figure 10c). However, the temperature difference slightly overshoots the expected difference in the JB, and is less than expected in the NB. Finally, the difference between alkenone and G. bulloides temperatures appears to be an appropriate measure of the seasonality of the mixed-layer, as it depicts the difference between annual mean SST and summer temperatures at 50 m (Figure 10c). However, our results do not resolve the applicability of this proxy for reconstructing seasonality in geological past (see discussion above).
 Our study cannot resolve the interannual effect of ENSO and IOD on different proxies at different sites. Since interpretation of past changes in the thermal structure of the water column or seasonality might vary at different sites, a careful understanding of various processes contributing to variance in these proxies is essential. For this purpose, long-duration field studies (sediment trap and plankton tow) are required that record the gradient and the entire range of climate variability, and their effect on the here introduced proxies.