Laser ablation inductively coupled plasma mass spectrometry was used to analyze the individual chambers from tests of foraminiferal fossil and plankton tow Globigerinoides ruber from the southwest Pacific Ocean, from latitudes 3°S to 42°S. The variability of Mg/Ca between chambers of an individual (intraindividual) and individuals of the same population (interindividual), is such that when converted to temperature, the extent of intra-individual and interindividual variability appears to exceed that attributable to either calcification or seasonal temperature variability. The pooled mean chamber Mg/Ca from each core top and plankton tow site demonstrates a significant (p < 0.05) positive correlation with temperature. We derive chamber-specific calibrations where Mg/CaCh_F-2 = 0.798 exp0.070 T, Mg/CaCh_F-1 = 0.891 exp0.067 T and Mg/CaCh_F = 0.590 exp0.072 T. We do not observe any bias between the two morphotypes Gs. ruber ruber and Gs. ruber pyramidalis. The chamber-specific calibrations potentially offset Mg/Ca-based temperature reconstructions if used on bulk (whole) test Mg/Ca or applied to misidentified chambers. Nevertheless, these calibrations can be used to reliably estimate sea surface temperature. Although there is a general overriding temperature control on Mg/Ca, we show that removal of the effect of temperature at each site reveals a lognormal Mg/Ca distribution. This suggests that Mg/Ca variability at each site is also affected by biological mechanism(s) that may control the distribution of interindividual Mg/Ca. In addition, other TE/Ca data (Al/Ca and Mn/Ca) from laser ablation trace element depth profiles can be used to identify detrital or diagenetic phases that may bias the trace element/Ca signal.
 The ratio of Mg/Ca in planktonic foraminiferal calcite is an important proxy for past sea surface temperatures (SST) [Anand and Elderfield, 2005; Barker et al., 2005; Dekens et al., 2002]. Although there is a strong empirical correlation between calcification temperature and the Mg/Ca ratio as bulk samples (comprising ∼5–50 individual tests), planktonic foraminifera exert considerable biological control over the incorporation of Mg (and other trace elements (TE)) into their tests [Eggins et al., 2004; Elderfield et al., 1996; Erez, 2003; Zeebe and Sanyal, 2002]. Furthermore, the TE/Ca ratio measured in foraminiferal tests recovered from deep sea sediments can be complicated by postdepositional digenetic alteration [Dekens et al., 2002; Tachikawa et al., 2008]. Therefore, uncertainties remain over the relative importance of environmental versus biological versus diagenetic factors in determining what controls the Mg/Ca ratio of foraminiferal tests.
 The development of microprobe and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) microanalytical techniques now allow researchers to quantify TE/Ca ratios in individual foraminifer chambers and along profiles through test walls [Eggins et al., 2003; Elderfield, 2002; Hathorne et al., 2003; Reichart et al., 2003; Sadekov et al., 2005]. For example, in addition to more traditional estimates of sea surface temperature (SST) [e.g., Anand et al., 2003], fine-scale structural differences such as compositional differences between gametogenetic versus ontogenetic calcite can be inferred [e.g., Kozdon et al., 2009]. An additional benefit to using this technique is the minimal destruction of the test, potentially enabling further paired analyses.
 In this study an open question is posed, to what extent does interindividual and intraindividual chamber Mg/Ca variability reflect environmental conditions relative to biological vital effects, and what can this variability tell us about the limitations of Mg/Ca for paleoocean temperature reconstruction more generally? To investigate this and other questions, we examined a species of foraminifera commonly used for paleo-SST reconstruction, Globigerinoides ruber [D'Orbigny, 1839] white variety, Gs. ruber (w). Using LA-ICP-MS we present trace element depth profiles from individual chambers of Gs. ruber from core top and plankton tows from the South Pacific Ocean spanning a mean annual SST of 14.7 to 28.8°C. We present our data in terms of a LA-ICP-MS derived SST versus Mg/Ca calibration for Gs. ruber for the southwest Pacific Ocean. Following this, we consider our results in terms of potential environmental controls over Mg/Ca and argue that individual test and chamber measurements point strongly to biological control over Mg/Ca on an intraindividual and interindividual test basis, independent of temperature, but the mean Mg/Ca of a foraminiferal population is sensitive to temperature in a manner described by solution-based studies.
Gs. ruber is a symbiont-bearing planktonic foraminifera that is abundant in tropical and subtropical waters occupying a temperature and salinity range between 14 and 32°C and 22–49 psu, respectively [Bé et al., 1977; Bijma et al., 1990a]. The presence of photosynthetic symbionts [Bé et al., 1977] mean it is restricted to the photic zone, that is, the upper mixed layer of the ocean [Anand and Elderfield, 2005; Dekens et al., 2002; Huang et al., 2008; Regenberg et al., 2009]. Consequently, SST data obtained from Gs. ruber are thought to largely reflect conditions prevailing in the depth range ∼0–50 m, minimizing complications from significant temperature changes associated with migration within the water column during the foraminiferal lifecycle. Gs. ruber has been classified into two end-member morphotypes [Steinke et al., 2005] which previous workers have suggested occupy specific ecological niches based on their significantly different Mg/Ca and oxygen isotope (δ18Oc) values [Kawahata, 2005; Kuroyanagi and Kawahata, 2004; Löwemark et al., 2005; Steinke et al., 2005]. However, a separate study by Mohtadi et al.  failed to find geochemical differences between morphotypes. Kawahata  explained the apparent difference as the result of productivity occurring at different times of the year for different morphotypes. Nonetheless, because of its restricted ecological niche in a climatically important part of the water column, it is one of the most widely used planktonic foraminifera for paleoocean thermometry.
2. Materials and Methods
2.1. Regional Setting and Core Top and Plankton Tow Locations
 The study region is situated in the southwest Pacific Ocean, and includes 10 core top and 3 plankton tow samples spanning 3.4°S to 41.9°S latitude. The tropical sites lie within either the Coral Sea or Timor Sea. The Timor Sea moves warm, low-salinity, nutrient poor water between the Indonesian archipelago and Australia into the Indian Ocean [Gordon et al., 1997]. The site AIMS1631 lies seaward of the north flowing Sepik River that drains largely volcanic and igneous terrain into the Coral Sea [Brunskill, 2004]. Currents from the Coral Sea bring warm nutrient-poor waters down the east coast of Australia to the cool waters of the Tasman Sea. This flow forms a counterclockwise gyro which includes the East Australian Current (EAC), the largest ocean current off the Australian coast [Ridgway and Godfrey, 1997; Ridgway and Hill, 2009]. The midlatitude sites are located in the subtropical higher salinity waters, which flow south along the Tasman Front (TF) separated at Chatham Rise (CR). The Subtropical Front forms the boundary between this subtropical water (STW) and cold, low-salinity and macronutrient-rich Sub-Antarctic Surface Water (SAW) from the within the Sub-Antarctic Front of the Antarctic Circumpolar Current [McCave et al., 2008; Morris et al., 2001; Neil et al., 2004; Uddstrom and Oien, 1999] (Figure 1).
 The core top samples were selected from archives managed by the National Institute of Water and Atmospheric Research (NIWA), New Zealand and the Australian Institute of Marine Sciences (AIMS). These sites ranged in depth from 3290 to 430 m below sea level, and the mean and seasonal range of SSTs and sea surface salinity (SSS) for each site were taken from the World Ocean Atlas 2005 (WOA05) [Antonov et al., 2006; Locarnini et al., 2006] (Table 1). In addition, foraminifera from three plankton tow samples were also analyzed. They were collected from water depths of 10–50 m (U2322 net 7), 50–100 m (U2322 net 6) and 50–100 m (U2315) using a Multiple Opening-Closing Net and Environmental Sensing System (MOCNESS) [Wiebe et al., 1985] during March and April 2001 (voyage TAN0103, NIWA). During deployment of the MOCNESS, separate downcast measurements of temperature and salinity were taken with a conductivity/temperature/depth (CTD) probe. The temperature and salinity data from the CTD agree well with estimates obtained from WOA05.
Table 1. Core Top and MOCNESS Tow Sample Locations, Seafloor Depths, SST (Annual and Range) and Calibrated 14C Ages of Core Tops
Annual sea surface temperatures (SST), seasonality, and standard deviations (SD) were derived from the World Ocean Atlas 2005 [Locarnini et al., 2006] except for plankton tow samples that used CTD data to infer actual SST and SSS.
Annual sea surface salinities (SSS), were derived from the World Ocean Atlas 2005 [Antonov et al., 2006].
Ages were calculated using either radiometric dating within the same core or in cores located close by. Errors are in radiocarbon years.
Plankton tow temperature data derived from CTD.
Assumes similar sedimentation rate to site P71.
<100 years (G. Brunskill, personal communication, 2009)
 An important aspect of the suite of samples used in this study is that almost all of the sites are above the present-day, regional foraminiferal lysocline (3600–4000 m) with the exception of ODP Site 1123, which is the deepest site at 3290 m [Berger et al., 1976; Martínez, 1994]. Other data suggest the modern planktonic foraminifera lysocline in this region is ∼3250 m [Feely et al., 2004]. However, previous studies assessing ratios of benthic to planktonic foraminifera, fragmentation indices, test ultrastructure [Crundwell et al., 2008] and the preservation of associated calcareous nannofossils [Fenner and Di Stefano, 2004] suggest this site has not been adversely affected by dissolution. In addition, a study measuring Mg/Ca ratios in the benthic foraminifera Uvigerina spp. at ODP Site 1123, compared the Mg/Ca ratio of that species with the shell weights of the planktonic species Globigerina bulloides (G. bulloides) from the same record [Elderfield et al., 2010]. During interglacials, and vice versa during glacials, they found where Mg/Ca was high in the Uvigerina spp. record, the corresponding shell weights of G. bulloides were low. They also did not find any discernible relationship between Mg/Ca and either Al/Ca or Mn/Ca that may signal a diagenetic overprint. Using these two approaches they concluded that dissolution was unlikely to influence Mg/Ca measured in their study.
2.2. Sample Preparation
 Foraminiferal samples were dry sieved so that the size fraction 250–355 μm remained. Previous experiments on Gs. ruber have shown that an increase in size corresponds to an increase in Mg/Ca ratios [Elderfield et al., 2002] although not all studies are in agreement [Ni et al., 2007]. In the plankton tow material there were few specimens >250 μm in size, therefore individuals were selected from the >150 μm size fraction. At ODP Site 1123, the morphology of most of the specimens was smaller and more compact than archetypal subtropical specimens that have large tests and large well developed supplementary apertures. In these specimens, the majority had very small supplementary apertures consistent with specimens living at the cold limit of their biogeographic range (M. Crundwell, personal communication, 2009).
 Following sieving, each individual foraminifer was carefully picked under a binocular microscope resulting in a random sample of 12–30 individuals from each sample set. In order to test for morphotype differences, where possible, the two major morphotypes Gs. ruber ruber and Gs. ruber pyramidalis were identified prior to LA-ICP-MS analysis using the descriptions from Steinke et al. .
 Arbitrary values of 0.1 mmol/mol have been used by previous workers to reject planktic foraminiferal TE analyses with suspected high Al/Ca, Fe/Ca and Mn/Ca ratios from samples analyzed in solution [e.g., Boyle, 1983]. This is based on the assumption that high Al/Ca and Mn/Ca are indicators of contaminants such as clay minerals and surface coatings that also contain Mg, consequently contributing to the Mg/Ca that would otherwise be attributed to primary foraminiferal calcite. To ensure detrital contamination was minimized [Rollion-Bard, 2005], three methodologies were employed to identify such specimens.
 1. Specimens were visually inspected and those with obvious adhering sediment were either recleaned or discarded (e.g., red/brown colored tests were avoided);
 2. All data were analyzed for the degree of covariance of Mg/Ca with Al/Ca and Mn/Ca. A few outliers were discarded. The remaining data did not show a statistically significant correlation (r2 < 0.5; p > 0.05).
 3. Terrigenous clay mineral assemblage data was examined for the southwest Pacific Ocean and did not suggest major contamination of Mg from the sediments [Glasby, 1979]. For example, a foraminiferal analysis yielding an Al/Ca of 0.2 mmol/mol from a mixture of 25% kaolinite (Al2Si2O5(OH)4), 15% illite (K0.40 Na0.40 Fe0.24 Mg0.34 Al1.50 (Al0.57 Si3.43 O10) (OH)2) and 60% smectite (Na0.165Ca0.165Al1.2Mg0.8Si4O10 (OH)2·nH2O) would add 0.06 mmol/mol Mg/Ca to a foraminiferal Mg/Ca ratio of 1.35 mmol/mol.
 In addition, the laser ablation technique itself yields data that can resolve the contaminants by enabling the analyst to exclude data from certain regions of the test [e.g., Creech et al., 2010] and therefore avoiding the contributions from those deposits: these were the main source of concern for those analyzing foraminifera by solution in the work of Boyle [1981, 1983] and from other techniques with secondary high Mg carbonate overgrowths [Hoogakker et al., 2009]. The use of the above methodologies allows some confidence that detrital contamination was minimized by the elimination of individual samples. However, samples which were eliminated were done so without using an arbitrary “contaminant” cutoff level.
 Following selection of clean foraminifera, each sample was transferred to a clean glass vial. Each core top sample was initially given 3 rinses in ultraclean water (>18 MΩ), followed by 2 rinses in analytical grade methanol, where the first rinse was additionally treated by gentle ultrasonication (at 10% power) for <5 s. Clearly, the better preserved the specimens or the thicker the tests, the more they will stand up to the rigors of ultrasonication. Ultrasonication for longer than 5 s often resulted in partial or complete destruction of tests, particularly those obtained from the plankton tows. Broken tests were discarded from further analyses.
 Finally, each sample was rinsed three times with ultraclean water. Following cleaning, each sample was transferred to an oven and allowed to dry for 24 h at 40°C. These preparation procedures prior to laser ablation are similar to those used in other studies [Bergami et al., 2008; Eggins et al., 2003; Hathorne et al., 2009; Reichart et al., 2003; Sadekov et al., 2008]. Cleaned individual foraminifera were also weighed using a Mettler-Toledo UMX2 microbalance, mounted and photographed prior to LA-ICP-MS analysis.
 Organic material was removed from the plankton tow samples using a low-temperature oxygen plasma asher. The samples were cleaned using the same procedure as the core top samples but excluding the ultrasonic cleaning step. Other studies have used oxidative cleaning to remove organic material (e.g., buffered hydrogen peroxide or sodium hypochlorite). However, sample loss can be minimized by avoiding the use of these chemicals [Barker et al., 2003; Boyle and Keigwin, 1985; Rosenthal et al., 2004; Yu et al., 2005].
 Following cleaning, a selected number of samples were also examined with a scanning electron microscope (SEM) for any signs of dissolution and/or contaminants prior to and after LA-ICP-MS. These individuals were placed onto sticky carbon tape, gold coated, and imaged using a JEOL JSM-5300LV SEM.
2.3. Trace Element Analysis of Gs. ruber in the SW Pacific
 Prior to laser ablation analyses, each foraminifera was carefully mounted onto a NIST610 glass standard using very weak adhesive tape and a clean paint brush. The outside to the inside [cf. Sadekov et al., 2008] of the three chambers in the final whorl per individual were analyzed for the isotopes 24Mg, 27Al, 55Mn, 66Zn, 88Sr and 138Ba relative to 43Ca, using a New Wave deep UV (193 nm) solid state laser ablation system coupled to an Agilent 7500CS ICP-MS. Gs. ruber chamber formation can be either sinistral or dextral and coiling direction can be identified by examining the dorsal view of the test. Here chamber “F-2” represents the antepenultimate chamber; “F-1” is the penultimate chamber prior to the ultimate (final) chamber “F.” Measurements were standardized using known elemental compositions of the NIST610 standard [Pearce et al., 1997]. The resulting data were processed using a MATLAB script which allowed for initial screening of outliers, background correction, and Ca-corrected internal standardization.
 Background and NIST610 measurements were made for 60 s at the start and end of each run. For the NIST610 a laser spot size of 35 μm and repetition rate of 5 Hz was used. For foraminiferal measurements a laser spot size of 25 μm and repetition rate of 2 or 5 Hz was used. The washout time between standards and samples was 100 s. For individual chambers the measurement profile took a maximum of 120 s to penetrate from the outer to inner chamber wall. The LA-ICP-MS depth profiles were used to distinguish between contaminant phases and only include the ontogenetic calcite for calculation of TE/Ca ratios (e.g., Figure 2). The end of each profile was identified either visually when the laser had ablated through the test well, or from the sharp reduction of raw Ca counts from the laser ablation profile which decline rapidly once the laser has ablated through each chamber. The complete ablation of a chamber wall is sometimes observed in conjunction with elevated Al/Ca and Mn/Ca, suggesting the presence of phases containing these metals on the inside of the tests. Using the laser ablation profiles, it is possible to avoid both surface veneers and internal trapped sediment by selecting the middle portion of the test, thus, removing the influence of these contaminants.
 Duplicate measurements on subsamples of individual tests were taken to ensure reproducibility of individual chamber analyses. From these measurements the two standard deviation (2 SD) and the median of the 2 SD were 0.72 and 0.25 mmol/mol for Mg/Ca; 0.79 and 0.04 mmol/mol for Al/Ca; 0.03 and 0.01 mmol/mol for Mn/Ca; 0.03 and 0.01 for Sr/Ca; 0.09 and 0.03 mmol/mol for Zn/Ca; and 4.0 and 0.83 μmol/mol for Ba/Ca, respectively.
3.1. Elemental Profiles Through Tests of Gs. ruber
 The wall structure of a typical Gs. ruber is illustrated by SEM (Figure 3) which shows an outer calcite layer representing ∼90% of the total wall thickness separated by an inner laminar calcite layer representing the remaining ∼10% of the wall thickness by the primary organic membrane (POM). These images are consistent with previous work showing that Gs. ruber does not produce a typical gametogenic calcite crust, i.e., thickening and change in crystal structure affecting the mass, thickness and dissolution rate of the test [Hamilton et al., 2008; Williams, 2008]. Furthermore, thin sections of Gs. ruber show chamber-specific outer calcite layers that generally follow the model of lamellar wall construction as described by Reiss . However, these distinct layers do not follow the general globular shape of the chamber but instead follow the shape of the interpore ridges (Figure 3).
 The SEM images also show pores that are ∼3–6 μm in diameter at the inner wall, expanding in a funnel-like form to ∼10 μm in diameter at the outer surface. Pores are typically spaced ∼10–20 μm apart, regardless of chamber. Consequently, as our ablation profiles were measured with a spot size of 25 μm, each one will typically encompass one or two pores, providing a pathway by which surface material can “appear” in the middle of the profile. It is expected that, because of these funnel shaped pores, the TE/Ca ablation profile from the outer to the inner test will reflect, in part, the ratio between the calcite and potentially any adhering contaminants.
 With the exception of Sr/Ca, laser ablation profiles from the outer to inner test walls show a thin (<1–2 μm) TE/Ca-enriched (by several orders of magnitude) outer surface veneer compared to the innermost (ontogenetic) calcite. This enriched layer was present in every chamber analyzed, regardless of location, (i.e., independent of temperature and depth) and was also present in plankton tow samples (Figure 2e) although it is not distinguishable from our SEM images. Apart from the enriched outer veneer, there appears to be relatively uniform Mg/Ca and Sr/Ca along the inner profile; that is, there is no compositional difference between the inner and outer calcite layers identified in SEM images. Individuals within a core top or plankton tow population with high Mg/Ca values maintain such values throughout the thickness of their tests and vice versa for individuals where Mg/Ca ratios were lower. In a number of individuals, changes were observed between low and high Mg/Ca ratios in chambers F-2 and F-1 that seemed to have analogs in the SEM images (Figure 3). However, the relatively “smoothed” nature of our profiles (caused in part by ablation of an uneven surface, resulting in material being incorporated from various different depths in the profile for each ablation event) meant that the presence of submicron thick Mg-rich bands inside the tests seen were reduced or difficult to resolve. Improvements to this resolution may involve ablating the test from the smooth inside to the rougher outside [Sadekov et al., 2009], but this greatly increases preparation time and introduces difficulties with identification of specific chambers as they must be individually dismembered from the test. These high and low Mg/Ca bands are not apparent with Sr/Ca; that is, there is no significant difference in composition between the outer calcite layers and the lamellar layers.
 In contrast, for samples that were not obviously contaminated, and where there is no strong covariation between Mg/Ca and Al/Ca (e.g., Figure 2e) the Al/Ca and Mn/Ca ratios are more variable and low, with a minima occurring about 30% of the way through the test wall and elevated (by a factor of 1 to 1.5) along the inner 50% of the profile. Note the logarithmic scale enhances that apparent variability for elements occurring at very low concentrations.
 Based on this pattern of variability observed in the TE depth profiles, we have selected a midportion of each to determine a Mg/Ca value for each chamber as shown in Figure 2, specifically excluding enhanced values in the outer veneer and inner chamber wall which are particularly vulnerable to surface contamination and/or dissolution [e.g., Pena et al., 2008]. A t test rejects the null hypothesis (at a confidence interval of 95%) that there is a significant difference in Mg/Ca between the two morphotypes of Gs. ruber where the same chambers were compared. Therefore we report the following data based on all Gs. ruber.
3.2. The Distribution of Mg/Ca Within a Population
 Having obtained Mg/Ca values for each chamber of each individual and calculated pooled chamber means (i.e., the pooled chamber of mean all F-2, all F-1 or all F per site) the results are compared against WOA05-derived SST for each sample location (Figure 4 and Table 2).
Table 2. Summary of Gs. ruber Mg/Ca Data for Each Chamber at All Study Sitesa
F-2 Mg/Ca (mmol/mol)
F-1 Mg/Ca (mmol/mol)
F Mg/Ca (mmol/mol)
Samples are ordered from highest SSTs (top) to lowest (bottom). PT, plankton tow; n refers to the number of individuals analyzed.
AIMS1361 (n = 7)
AIMS3703 (n = 13)
AIMS2078 (n = 10)
AIMS1631 (n = 4)
FR1/97/GC12 (n = 12)
TAN0706 C4 (n = 26)
U2322net7 (PT) (n = 8)
U2322net6 (PT) (n = 7)
P71 (n = 7)
U2315 (PT) (n = 13)
Z7003 (n = 16)
P81 (n = 18)
ODP1123 (n = 13)
 For each site, the pooled chamber mean Mg/Ca ratio from each site is positively correlated with temperature. The chambers F-2 and F-1 show a statistically significant difference in the Mg/Ca value when compared to the ultimate chamber F (paired t test (2-tailed), p ≤ 0.05), with the exception of site FR1/97/GC12 where p = 0.058 (F-2 versus F) and p = 0.110 (F-1 versus F).
 Having considered these differences, we derive chamber-specific equations, fitted to an exponential relationship, as shown in Figure 4. In considering each equation it is observed that the exponential coefficient “A” for each chamber is similar (0.067 to 0.072) and within the confidence interval of each chamber (i.e., standard error of ±0.005 to 0.007). However, the preexponential coefficient “B” shows a considerable difference between the two earlier chambers F-2 and F-1 (0.798 and 0.891) with the ultimate chamber F (0.590). The standard error of the preexponential coefficients for chambers F-2 and F-1 (±0.091 and ±0.133) are outside the confidence interval for chamber F (±0.091). Therefore, although the gradient for the ultimate chamber F is comparable to the chamber F-2 and F-1, it is offset to lower Mg/Ca between 0.4 and 2.0 mmol/mol (Figure 4). This would potentially indicate that the ultimate chamber is calcifying in colder temperatures compared to chambers F-2 and F-1.
 For comparison, the Gs. ruber Mg/Ca versus SST calibration derived by Sadekov et al.  for sites in the Indian Ocean is shown on Figure 4. Our “F” calibration (i.e., Mg/CaCh-F = 0.590 [±0.091] exp (0.072 [±0.007] * T) is comparable to that of Sadekov et al.  (i.e., Mg/Ca = 0.520 [±0.08] exp (0.076 [±0.002] * T) which was derived using a similar LA-ICP-MS technique to measure Mg/Ca on the same ultimate chamber (Table 3). In addition, Anand et al.  summarize a number of Mg/Ca versus temperature calibrations for planktonic foraminifers determined using solution-based analysis, including several specifically for Gs. ruber (w). Their study yielded a widely used generic (combining data from 10 planktonic species recovered from N. Atlantic sediment traps) Mg/Ca versus temperature relationship of Mg/Ca = 0.38 [±0.02] exp (0.09[±0.003] * SST), identical to the Gs. ruber (w) (250–350 μm) core top calibration of Dekens et al. . A similar relationship in Gs. ruber (w) (250–350 μm) from sediment traps, revealed a relationship of Mg/Ca = 0.34 [±0.08] exp (0.102 [±0.01] * T). Therefore, published solution-based Mg/Ca versus temperature relationships for Gs. ruber (w) for tests <350 μm in size have yielded a smaller preexponential coefficient and a larger exponential although the significance of this, if any, remains to be explored.
Table 3. Summary of this Study's Mg/Ca Chamber-Specific Calibrations and Selected Existing Calibrations for Gs. ruber (white)
3.3. A Lognormal Model for the Distribution of Mg/Ca Measurements
 There are currently a limited number of microanalytical studies that have analyzed Mg/Ca from individual chambers. Sadekov et al.  produced element maps that show high and low Mg/Ca bands in individual chambers from several planktonic foraminiferal species, describing individual measurements across chambers by comparing the arithmetic and harmonic mean. These authors found that the few high Mg/Ca ratios skew the data away from the more uniform low Mg/Ca. Similarly, Anand and Elderfield  fitted individual chamber measurements that gave Gaussian averages smaller than the arithmetic means. This skew was also interpreted to reflect intraindividual Mg/Ca variation, i.e., the presence of occasional high Mg/Ca ratios from within individual chambers. This study contributes to these studies by examining the distribution of Mg/Ca measurements at any site (i.e., regardless of temperature) that show a number of large Mg/Ca values that exceed those expected from a Gaussian distribution; that is, the distribution is positively skewed.
 From these observations, it is hypothesized that the Mg/Ca distribution at any of our sites is lognormal, i.e., the distribution of log(Mg/Ca) values at any temperature is Gaussian. These lognormal distributions are found regardless of which chamber is examined. To demonstrate this, first, the distributions of Mg/Ca for each chamber at different temperatures, were normalized by dividing them by their mean temperatures (the latter being obtained by simple regression, chamber by chamber, of Mg/Ca versus temperature). This method of normalizing was used so that log values normalized to 1 can be converted to zero, allowing the data for each chamber to be pooled. Further, maximum likelihood estimates of the parameters of the best fitting lognormal distribution for each of the three chambers were also obtained. Specifically, these are the mean and standard deviations of the logarithms of the pooled data. The results are illustrated in Figure 5 and also show that the lognormal model fits the data distinctly better than a Gaussian model.
 Second, the temperatures were put into seven bins, where the temperatures are within a range of 0.5°C. For each bin, the normalized Mg/Ca data for the three chambers were pooled and their distribution plotted, along with the best fitting lognormal (Figure 6). We infer that the lognormal distribution models the data better than the normal distribution at any temperature.
 However, the seven SD of the logs of pooled data in each temperature bin are dissimilar, ranging from 0.19 to 0.32 (rms 0.27). The hypothesis that the seven temperature sets have a common variance of their logs is rejected by a Levene or Bartlett test of homogeneity of variance. But their variability is apparently not due to temperature: the slope of a regression of the SD on temperature is not significantly different from zero (95% C.I).
4.1. Impacts of Dissolution
 All of the sites except ODP Site 1123 are located well above the regional lysocline (∼3250 m) [Feely et al., 2004]. We have observed smaller test sizes and kummerform chambers at ODP Site 1123, indicators of environmental stress [Schmidt et al., 2006]. In addition, this site has highly variable frontal systems that have been demonstrated to result in decreased foraminiferal test size [Hecht, 1974]. Gs. ruber at this latitude are at the cold limit of their temperature tolerance [Bé, 1977; Bijma et al., 1990a] therefore environmental stress is likely to result in decreased growth rates [Bijma et al., 1990a; Caron et al., 1990] and may influence Mg/Ca incorporation [Elderfield et al., 1996]. However, the supporting evidence from SEM images [Williams et al., 2007], and the presence of the typical enriched surface veneer found in samples from other locations suggest that at least the outer surface of Mg-rich calcite had not been subject to dissolution.
4.2. LA-ICP-MS Trace Element Depth Profiles
 The test profiles reveal an elevated TE-surface veneer (except Sr/Ca) on all the profiles that were measured. Such TE-rich veneers were also reported in previous LA-ICP-MS studies [Sadekov et al., 2008] although their origin is not clear. Eggins et al.  suggested these are of biogenic origin, formed during the lifetime of the foraminifera, while Hathorne et al.  proposed these are of postdepositional diagenetic origin. An open question is whether this high Mg/Ca veneer is different to that substituted into the calcite lattice. For example, TE located at grain boundaries and/or dislocations may be higher than the TE in the calcite lattice [Weiner and Dove, 2003]. Our data support a biogenic origin, based on the presence of such a veneer on individuals recovered from plankton tow samples where there is no possibility for diagenetic alteration or the additional of surficial contaminant phases from the seafloor, as is the case for core top samples. The exclusion of these elevated surface veneers offers one explanation for the differences between our chamber-specific calibrations and those from bulk solution-derived Mg/Ca methods (see section 4.4).
4.3. An LA-ICP-MS-Derived Mg/Ca Versus SST Calibration for Gs. ruber in the Southwest Pacific Ocean
 Our data support the notion that LA-ICP-MS is a robust method for determining Mg/Ca ratios in Gs. ruber for paleoceanographic studies with the previously noted benefits of rapidity, test preservation (measurements are minimally destructive) and the ability to measure other elements that are useful quality checks for measuring primary foraminiferal Mg/Ca ratios.
 Regardless of the wide scatter of Mg/Ca within individual populations, the mean Mg/Ca value for each population shows a positive correlation with SST. On this basis we have generated three empirical calibrations between Mg/Ca (mmol/mol) and SST (°C) for the three final chambers of Gs. ruber. The Mg/Ca versus SST relationship for chambers F-2 and F-1 is essentially identical, however, although chamber F has a similar gradient (exponential coefficient “A”), the mean Mg/Ca in chamber F is 4 to 35% depleted Mg/Ca, thereby resulting in apparently higher Mg/Ca derived temperatures (preexponential “B”) (Table 3). Examination of Mg/Ca from the final chamber of Gs. ruber in the Indian Ocean measured using a similar LA-ICP-MS technique shows very comparable A and B values, and therefore similar Mg/Ca versus SST calibrations [Sadekov et al., 2008]. Interestingly, their study did not find a systematic relationship between a chamber's mean Mg/Ca value and its position within the test whorl; that is, the Mg/Ca value for F was not measurably less than that derived for preceding chamber F-1 or F-2. In contrast to the Sadekov et al.  study, Anand and Elderfield  used Gs. ruber from sediment traps to derive their calibration which lies in between the F-2/F-1 and F chamber-specific calibrations from our work (Figure 4). This suggests in general terms that Mg/Ca from our laser ablation depth profiles are characteristic of the bulk test composition, although other variables such as cleaning techniques [Barker et al., 2003; Martin and Lea, 2002], preservation state [Dekens et al., 2002] and regional and/or site-specific oceanography [Barker et al., 2005] may also contribute to these differences. Constraining the relative importance of analytical versus environmental influences on foraminiferal Mg/Ca can perhaps best be resolved with paired laser ablation and solution Mg/Ca measurements.
 For the purposes of determining mean Mg/Ca in a sample, it is sufficient to measure one chamber per individual. While the reason(s) for the offset between chambers F-2 and F-1 with F remain unclear (and are discussed in section 4.4), it would appear either calibration is valid for paleoceanographic purposes. However, it is important that each chamber is identified prior to analysis and that the same chamber is consistently ablated, otherwise, the single chamber calibrations will not yield meaningful paleotemperature estimates. Regardless, for Gs. ruber, calibrations are not interchangeable between LA-ICP-MS and solution methods of measuring Mg/Ca.
4.4. Intraindividual Mean Mg/Ca Variability
 Intraindividual mean Mg/Ca compositional measurements show the ultimate chamber “F” is consistently depleted in Mg relative to the mean value in chambers “F-2 and F-1,” the latter two chambers having no statistical compositional difference. If we consider the construction of lamellar walls [Bé and Hemleben, 1970; Bé and Lott, 1964; Hansen, 1999], found in this species, then the lower Mg/Ca associated with the formation of chamber F should also be visible as a low Mg/Ca zone in our LA profiles through F-2 and F-1 chambers. However, there is no evidence of such a zone, implying calcification associated with chamber F is limited to that chamber. In addition, Gs. ruber does not precipitate a gametogenic calcite crust of any significance [Caron et al., 1990], if it produces one at all [Hamilton et al., 2008]. Although these crusts are also considerably lower in Mg/Ca compared to ontogenetic calcite in other planktonic species [Anand and Elderfield, 2005; Klinkhammer et al., 2004; Lombard et al., 2010; Sadekov et al., 2005], it is unlikely that the lower Mg/Ca in chamber F is what is considered to be a “typical” gametogenic crust.
 Therefore, potential reasons for this offset include (1) the mean Mg/Ca in the chamber F represents a systematically lower calcification temperature than for previous chambers, by an amount equivalent to ∼2.3 to 4.4°C, (2) lower Mg/Ca is the product of a biologically mediated “vital” effect that does not directly relate to lower calcification temperature, or (3) some combination of the two.
 With respect to reason 1, the variability between chambers where Mg/Ca ratios between F-2 and F-1 are similar could be interpreted as both chambers calcifying at the same water temperature and hence water depth. By contrast, the relatively lower Mg/Ca from chamber F may represent calcification in cooler, deeper water during the terminal stage of its ontogeny. The interpretation that the F chamber is formed deeper in the water column than F-1 and F-2 is consistent with the known life cycle of various spinose taxa, as discussed by Bijma et al. [1990b, and references therein]. However, the occurrence of a relatively Mg depleted “F” in our shallow plankton tows would tend to count against this as the primary reason for the offset.
 With respect to reason 2, internal or genetic controls that are not strongly correlated with the surrounding environment are known to produce variations in skeletal Mg content, the physical basis of which is not well understood [Wang et al., 2009; Williams, 2008]. These “vital” effects could in turn affect the biologically induced mineralization (biomineralization) of Mg between different chambers. Anand and Elderfield  attributed lower Mg/Ca in the ultimate chamber of G. bulloides from North Atlantic core tops to such a biomineralization effect. Similarly, culture experiments on tests of Globerinoides sacculifer kept at a constant temperature and salinity in the laboratory also show a statistically significant depletion in Mg in the final gametogenic chamber clearly unrelated to environmental conditions [Dueñas-Bohórquez et al., 2011].
 Considering these, in reason 3 the symbiotic algae of Gs. ruber may, indirectly modulate the calcification process as suggested in other symbiont-bearing species where lower Mg/Ca of the ultimate chamber may be due to loss of symbiotic activity [Eggins et al., 2003]. Furthermore, symbionts may also alter the calcification rate toward the end of this species' life cycle, in turn influencing the elemental composition of this ultimate chamber [Lohmann, 1995; Nürnberg et al., 1996]. For example in selected globigerinid planktonic foraminifer, Brummer et al.  described the terminal stage of ontogeny with the disappearance (lysis and digestion) of symbionts and regulation of the calcium budget (resorption, calcification) prior to gamete release. However, in Globerinoides sacculifer the precipitation of the ultimate chamber occurs prior to the expulsion or consumption of algal symbionts [Hemleben et al., 1989]. Moreover, the nonsymbiont bearing species G. bulloides was recently demonstrated to have reduced Mg/Ca in the final chamber [Anand and Elderfield, 2005; Marr et al., 2011] suggesting that symbionts do not regulate the calcification process, but instead both Gs. ruber and G. bulloides share a similar biological mechanism. Thus, these studies suggest that biological internal regulation plays a more prominent role in the Mg variability of tests [Bentov and Erez, 2006] and the mechanism(s) which account for low Mg/Ca in the final chamber of G. bulloides [Anand and Elderfield, 2005; Marr et al., 2011] and Gs. sacculifer [Dueñas-Bohórquez et al., 2011; Sadekov et al., 2008] may also be applicable to Gs ruber.
4.5. Interindividual Mg/Ca variability
 Our results show a wide range of Mg/Ca values between individual foraminifers from each sample site that prima facie can be described by a lognormal distribution. An open question is whether or not this is the result of seasonal temperature and productivity variability at the site, or the product of biologically mediated “vital” effects unrelated to temperature. Several lines of evidence point toward the latter. First, if we assume Mg/Ca is temperature related, there is a much wider range of values than expected from environmental conditions, particularly if productivity is largely confined to one season (hence narrower range of temperature). Using the chamber-specific equations to convert each Mg/Ca from chambers F-2 and F to temperature (i.e., Figure 4), shows the range of values estimated for each site is much greater than can be explained by the modern seasonal range of SST (Figure 7). Second, although some of this variability may be explained by the age of the core top samples; that is, the mean and seasonal SSTs may have differed (slightly) from today, the spread of Mg/Ca in plankton tow samples follows a similarly wide variability (see Figure 7). One would expect that these samples would reflect a narrower range of Mg/Ca as the calcification temperature is more or less known for each site (assuming that each chamber calcifies within a few hours) [Bé et al., 1979], however this is not the case in our samples. Similar observations from individually pooled chambers from cultured foraminifera kept at constant temperatures also suggest that there is a wide variability in Mg/Ca (∼2.3 mmol/mol at 25°C in G. bulloides) [Lea et al., 1999]. Third, the observation that the spread between individuals can be described by a common distribution irrespective of site points to a vital effect. Thus, our data suggest that interindividual variability is most likely caused by a biological vital effect, placing a limit on detailed Mg/Ca SST reconstructions, especially where only a small number of individuals are available from a sample.
5. Summary and Conclusions
 In this study we have used LA-ICP-MS to examine the intrachamber and interchamber distribution of Mg/Ca within individual tests of Gs. ruber. We have shown that LA-ICP-MS profiles can be used to identify areas of secondary calcite crusts, and/or surface contamination and thus, isolate them from ontogenetic calcite by careful screening of profiles. From these profiles we show a lack of any surface layer on either of the chambers F-2 or F-1 that may typify gametogenic calcite crusts (i.e., low Mg/Ca). We therefore infer that Gs. ruber does not produce a gametogenic layer over previously precipitated chambers that may bias the Mg/Ca signal.
 However, the ultimate chamber F of Gs. ruber is systematically (4–35%) lower in Mg/Ca compared to chambers F-1 and F-2. The variability of Mg/Ca ratios found between these chambers could, in part, be explained by the ecology of the species. On account of these systematic changes we devised chamber-specific empirical calibrations that reliably estimate SST. These calibrations are comparable to extant calibrations based on LA-ICP-MS and demonstrate that chamber-specific calibrations are an alternative to bulk Mg/Ca analyses of foraminifera. These single-chamber calibrations provide a time efficient method, requiring minimal cleaning, fast analysis time and leave material for use in other analyses.
 The interindividual variability of TE/Ca of Gs. ruber within populations suggest the temperatures (Mg/Ca) recorded from individual foraminifera, including plankton tow samples, are often outside the seasonal ranges of the sample sites. Removal of the temperature effect from each site, by dividing by the mean temperature value of each site and normalizing the data to a single value reveals a log normal distribution. The lognormal distribution describes well the variability of Mg/Ca values but the spread parameter of this distribution, being the standard deviation of log(Mg/Ca) values, is not universal. This variation with temperature is at most weak, suggesting that it is due to some other, unknown, possible vital effect. Microanalytical techniques in other foraminifera may shed more light on this observation.
 We thank Gregg Brunskill and Irena Zagorskis (Australian Institute of Marine Sciences) for provision of sample material. We also thank Martin Crundwell and George Scott (Geological and Nuclear Sciences) and Helen Bostock and Lisa Northcote (National Institute of Water and Atmospheric Research) for provision of material, comments and discussion of results and analyses. We thank Richard Tilley and David Flynn for the use of the JEOL Scanning Electron Microscope and John Patterson for his assistance with the plasma asher at the Victoria University of Wellington. Finally, we are grateful to Graham Mortyn and an anonymous reviewer for their useful and pleasant reviews of this manuscript.