Calibration of the Mg/Ca of Globorotalia truncatulinoides (R) for the reconstruction of marine temperature gradients



[1] Recent work has provided useful Mg/Ca to water temperature calibrations for shallow-dwelling planktonic foraminifer species. Globorotalia truncatulinoides (right coiling (R)) is a deep-dwelling species that can serve as a source of information about the temporal variability in the water characteristics of the thermocline. We present a temperature calibration for the Mg/Ca in the shell of G. truncatulinoides (R) and examine some of the practical issues associated with evaluating the usefulness of the technique. The Mg/Ca in the primary and the secondary calcite of individual G. truncatulinoides (R) correlates exponentially with water column temperatures, showing a change of ∼10% in the Mg/Ca per 1°C (R2 = 0.92). A limited comparison with plankton tow samples demonstrates that the average Mg/Ca temperature was offset +1°C from the average temperature calculated using the δ18O calibration of O'Neil et al. [1969], and the Mg/Ca temperatures have a range similar to the δ18O temperatures. Comparisons of the [Mg] in the core top samples to water depth of deposition indicates that dissolution does not alter the measured value of Mg in the primary calcite.

1. Introduction

[2] The determination of sea surface temperature is an important element in studies of past climate because this knowledge helps quantify the extent of global temperature change and provides information on both surface ocean and atmospheric circulations. Temperature changes within the thermocline can indicate changes in vertical mixing and the circulation between subtropical gyres and equatorial regions. While paleoceanographic studies commonly calculate open ocean temperatures from measurements of δ18O in planktonic foraminifers, the calculation relies on important assumptions because δ18Ocalcite is influenced by both ocean salinity and ice volume. In order to simplify assumptions and gain more information, efforts to quantify temperature through other methods such as faunal abundances [Imbrie and Kipp, 1971; Prell, 1985], alkenone studies [Rostek et al., 1993; Sikes and Keigwin, 1994] and Sr/Ca analysis of tropical corals [Guilderson et al., 1994] have produced inconsistent results. The coordinated use of this suite of marine proxies is complex because the organisms recording the various proxies live at different ocean depths and vary in their responses to the seasonal cycle.

[3] The ability to generate multiple proxies from the same sample material allows for a clearer definition of the influences of the environmental variables. The possibility that [Mg] in marine calcite is dependent on water temperature was initially recognized by Chave [1954]. Further studies investigated the utility and sensitivity of the possible temperature dependence of [Mg] with inconclusive results [Blackmon and Todd, 1959; Lipps and Ribbe, 1967; Bender et al., 1975; Savin and Douglas, 1973; Duckworth, 1977]. More recent studies have demonstrated a convincing link between water temperature and [Mg]. Using data generated from laboratory studies, Lea et al. [1999, 2000], Mashiotta et al. [1999], and Nürnberg et al. [1996a, 1996b] published calibrations for water temperature based on Mg/Ca for Globigerinoides ruber, Globigerina bulloides, Orbulina universa, and G. sacculifer, respectively. Paleoceanographic studies have interpreted the down core variation of Mg/Ca as a response to changing ocean temperature [Hastings et al., 1998; Mashiotta et al., 1999; Nürnberg et al., 2000].

[4] Globorotalia truncatulinoides (d'Orbigny), right coiling (R), is a planktonic foraminifer that is most abundant in the subtropics and has an annual life cycle with a large vertical migration during ontogeny [, 1977; Hemleben et al., 1989]. The species is seasonally common in the upper thermocline and occasionally in the mixed layer, where it grows from the juvenile to the adult stage [Bé and Ericson, 1963; Hemleben and Spindler, 1983]. Results from deep plankton tows and oxygen isotope studies suggest that mature individuals descend through the water column to depths of up to 800 m [Bé and Ericson, 1963; Erez and Honjo, 1981; Hemleben and Spindler, 1983] where they add a secondary crust that covers much of the surface area and approximately doubles the weight of the shell [Bé and Ericson, 1963; Bé and Lott, 1964; Lohmann, 1993; McKenna, 2000]. These two rather discrete levels in the water column generally correspond to the top and bottom of the permanent thermocline. Therefore shell characteristics are correlated to the hydrographic environments across the permanent thermocline.

[5] In this work, we use samples of G. truncatulinoides (R) from core top sediments to develop and assess a calibration between Mg/Ca and water temperature during calcification. We also examine the distribution of Mg within the shell and the potential influence of dissolution. Our inferences, drawn from the sediment data, are applied to data derived from a limited number of plankton tow samples.

2. Experiment Design

2.1. Oceanographic Setting

[6] The Indian Ocean is a unique area with the characteristics of near surface waters (the top 1000 m) determined by the atmosphere/ocean interaction of the seasonally reversing monsoon winds and the pattern of circulation within the permanent thermocline. A hydrochemical front at 10°S and over a depth range of 100 to 800 m divides the upper Indian Ocean into two distinct provinces that are characterized by the distributions of oxygen and nutrient concentrations. North of the front, the water has a low oxygen concentration, high nutrient concentrations, and high salinity. The water of the subtropical gyre, south of the front, is better oxygenated and has lower nutrient concentrations and moderate salinity [Wyrtki, 1973]. Because of the relationship between the growth of G. truncatulinoides (R) individuals and the annual cycle of water column stability, we are particularly interested in the patterns of stability found on either side of this front.

[7] We calculate the simple stability (Δσt/Δz) at 18 levels in the upper 1000 m of the water column above our core top locations using the data from Levitus and Boyer [1994] and Levitus et al. [1994] and divide the sample locations into two regional patterns based on the annual cycle of the stability gradient across our chosen depth range [McKenna, 2000]. In pattern I, across the southern Indian Ocean, the water column stability is weakest between July and December. Winter cooling at the surface leads to mixing in the upper water column as the more dense surface water sinks to a new level consistent with its potential density. The stability of the water column becomes weak enough to allow easy mixing into the very top layers of the permanent thermocline. In pattern II, north of 10°S and in the Arabian Sea, the stability at 100 m also weakens during the same time period, but the mechanism is different. Wind stress on the surface generated by the strong monsoon winds causes a deepening of the surface mixed layer. Cooling during the beginning of the northeast monsoon increases the wind's effect. This also results in a weakening of the stability down past 450 m, although a secondary layer of higher stability continues to cut off easy mixing into the top 100 m of the water column. Consequently, the annual variation of the stability gradient controls the timing and the depth of G. truncatulinoides (R) ascent to the upper water column and, indirectly, the environmental conditions during G. truncatulinoides (R) growth.

2.2. Sampling Criteria

[8] We selected G. truncatulinoides (R) from eight Holocene core top populations that span a geographic region in the western Indian Ocean from latitudes 10°N to 40°S (Table 1 and Figure 1). Across this region, the annually averaged surface temperature (Tw) ranges from 28°C to 13.6°C, and at the depth of 1000 m Tw ranges from 8.3°C to 4.9°C. At any one location, the contrast between surface and deep temperatures ranges from 15°C to 4°C [Levitus and Boyer, 1994] (Table 2). The selection of sampling locations, with deliberate duplication at some latitudes, was designed as a measure for the robustness of any regression calculated between Tw and [Mg].

Figure 1.

Map of core top locations. The eight core tops that supplied Holocene sediment for this study cover a latitude range of almost 50° and represent regions of varying G. truncatulinoides (R) relative abundance (dashed lines). The core tops are paired by latitude where possible as a check on the robustness of the results. For the majority of core tops, magnesium concentrations are measured in two individuals.

Table 1. Core Top Locations
Core IdentificationLatitudeLongitudeWater Depth, mStability PatternCore Top Documentation
V34-8310.40°57.95°1929pattern IIPrell et al. [1999]
RC12-328−3.95°60.60°3087pattern IICurry and Matthews [1981]
RC17-98−13.22°65.62°3409pattern ICurry and Matthews [1981]
V18-200−20.58°63.00°3305pattern IPrell et al. [1999]
V20-170−21.80°69.23°2479pattern IPrell et al. [1999]
V14-81−28.43°43.78°3634pattern IPrell et al. [1999]
V16-96−31.53°99.25°2417pattern IPrell et al. [1999]
RC11-121−39.72°82.25°3426pattern IPrell et al. [1999]
Table 2. Water Column Environments
Core IdentificationSeasonal Upper EnvironmentDeep Environment
MonthDepth, mσtWater Temperature, °CDepth, mWater Temperature at σt = 27, °C
V34-83January16625.917.1-no crust
RC11-121January4526.015.0-no crust

[9] We also analyzed a limited number of G. truncatulinoides (R) individuals that were collected by plankton tow from Meteor cruise M35/2, Station 160, located at 23°0.023′N, 63°29.74′W in the Atlantic Ocean. Tow MSN 1085 was collected on 5/23/96, across a depth range of 0 to 700 m, and the surface Tw was measured at 25.3°C.

2.3. Analytic Method

[10] Individual shells were cleaned ultrasonically and then imbedded in epoxy, ground to the desired level, polished, and carbon coated. We determined the elemental concentration in the foraminifer shells by using a Cameca Camebax™ electron-probe, counting for 100 s with 5 to 25 counts per second for magnesium measurements, and counting for 20 s with more than 1000 counts per second for calcium. For the sediment samples, the beam diameter was 10 μm and the specimen current was 30 nA. The measurements were made in two series, one series with an accelerating voltage of 10 kV and the second with an accelerating voltage of 20 kV. Plankton tow samples were measured with a beam diameter of 5 μm and an accelerating voltage of 20 kV. We used Dolomite USMN 10057 Oberdorf, Austria, as documented by Jarosewich and MacIntyre [1983] for our analysis standard. ZAF corrections were performed as documented by Pouchou and Pichoir [1991]. The counting uncertainty is calculated by the control protocol and is dependent on the number of counts per second above the average background value. Average counting uncertainty for measurements of Mg (wt %) is 0.004.

[11] A typical set of measurements in one individual consists of one or two measurements in each of the septa associated with the last four chambers, four to six measurements located in the secondary crust, and a similar number in the outer wall (Figure 2a). Initial results are expressed in element wt %, but calculation of Mg/Ca (mmol/mol) provides an approach that reduces variability in the results due to variable absorption of the incident energy. Our calibration between water temperature and Mg/Ca uses average values for each shell. The error associated with each average is calculated using the standard deviation of the mean for all analyses from each class of measurement location. See McKenna [2000] for a more detailed discussion of the accuracy and reproducibility of these results. The data set of individual [Mg] measurements is available from the National Geophysical Data Center (NGDC) at

Figure 2.

Importance of measurement location. (a) This cross section of a G. truncatulinoides shell shows the three different types of measurement locations within the shell. The symbols correlate to the graphs below. (b) M-727 is a shell with a large amount of secondary crust. The [Mg] measured in the outer wall would fall along a mixing line drawn between the concentrations measured in the septa and in the secondary crust. (c) When the shell has no secondary crust, M-352, the variability of the measurements made in the outer wall has the same range as the measurements made in the septa.

3. Results and Discussion

3.1. Distribution of Mg

[12] Transects across individual G. truncatulinoides (R) show that the primary calcite has a higher [Mg] than the secondary crust (Table 3 and Figures 2b and 2c), consistent with the formation of primary, higher [Mg] calcite in the warmer upper water and with the addition of a lower [Mg] secondary crust at depth in cooler water. This trend is in agreement with the previous work of Duckworth [1977] and Brown and Elderfield [1996], but in direct contrast to that found by Nürnberg et al. [1996a, 1996b] for G. sacculifer. Using shells cultured in the laboratory, electron microprobe analysis demonstrated that the gametogenic crust had [Mg] enriched by 230% above the concentration of the primary calcite. For comparison, this enrichment was larger than the increased [Mg] due either to a 10°C increase in temperature or a 10‰ change in salinity. However, Rosenthal et al. [2000] find a decreasing [Mg] in G. sacculifer with both addition of a gametogenic crust and dissolution. They suggest that the crust does have a lower [Mg] as expected with calcification at cooler temperatures, and that the higher values found by Nürnberg et al. [1996a, 1996b] result from an inorganic precipitate that is preferentially removed in contrast to the Mg incorporated into the crystal lattice.

Table 3. Magnesium Concentrations, Mean Values for Individual Shells
Core/Sample IdentificationType of Analysis SiteCa, wt %Mg, wt %Mg/Ca, mmol/moln
Sediment Samples, Indian Ocean
M-358outer wall39.2330.1180.0620.0052.60.26
secondary crustNANANANANANA 
M-352outer wallNANANANANANA 
secondary crustNANANANANANA 
M-334outer wallNANANANANANA 
secondary crust39.9400.3380.0160.0010.70.15
M-331outer wallNANANANANA  
secondary crust40.0220.4260.0180.0030.70.13
M-282outer wall35.5670.5980.0420.0031.90.16
secondary crust39.3370.1730.0170.0010.70.046
M-727outer wall37.0400.5820.0440.0052.00.212
secondary crust40.1380.5670.0080.0010.30.0515
M-728outer wallNANANANANANA 
secondary crust38.9290.3040.0160.0010.70.045
M-170outer wall41.6470.2670.0480.0071.90.36
secondary crust41.5830.1430.0130.0040.50.26
M-192outer wallNANANANANANA 
secondary crust39.2890.4370.0170.0040.70.25
M-625outer wall38.1420.3330.0410.0051.80.26
secondary crust39.6090.1100.0190.0020.80.16
M-579outer wallNANANANANANA 
secondary crust40.0260.4500.0130.0020.50.15
M-493outer wall38.8650.1050.0420.0011.80.16
secondary crust38.9690.1450.0190.0020.80.16
M-507outer wallNANANANANA  
secondary crust40.6130.2130.0080.0020.30.15
M-409outer wall37.9850.4720.0450.0022.00.16
secondary crustNANANANANA  
Plankton Tow Samples, Atlantic Ocean

[13] The magnesium in the primary calcite is not distributed evenly throughout [Duckworth, 1977; Nürnberg et al., 1996a, 1996b], and this inhomogeneity introduces more variation to the measurements than the error associated with analytical uncertainty (Figure 3). Indeed, Nürnberg et al. [1996a, 1996b] found scatter among individual measurements so that the maximum measured value might be on the order of 2 to 3 times the minimum value. A probable explanation for the smaller scatter in our data lies in our use of a 10 μm diameter beam rather than the 2 to 4 μm beam used by Nürnberg et al. [1996a, 1996b]. Our beam size would integrate over a larger area, smoothing some of the internal inhomogeneity and reducing the scatter of the resulting measurements.

Figure 3.

Assessment of the effects of dissolution. Magnesium concentrations from the septa are presented in the context of core top water depth. Depths below the regional lysocline are indicated by the shaded area. The data from core top samples show no systematic variation of lower [Mg] wt % with increasing depth. We see neither a general trend toward lower values with depth nor a discrete drop at the depth of the regional lysocline. Individual measurements and shell average [Mg] for the plankton tow samples cover a similar range in concentrations. The pattern of the data supports a reduced concern that dissolution has altered the original magnesium concentration.

[14] Our measured values from the secondary crust are extremely low, with values not much larger than the counting uncertainty. However, we can have confidence that the values do accurately represent the [Mg] of the secondary crust because the values are more closely grouped than the counting uncertainty might predict. These data (Table 3) would indicate a uniformity of conditions for deposition of secondary crust.

[15] The location of measurement sites in the primary calcite is an important consideration. While the concentrations measured in the septa and the secondary crust are significantly separated beyond the error associated with each measurement, the measurements in the outer wall exhibit larger variability. Indeed, the value of any [Mg] measured in the outer wall can be calculated using a mass balance argument with the averages of the septal [Mg] and crustal [Mg] as end-members (Figure 2b). When the shell does not have a secondary crust, the variability measured in the outer wall is identical to the variability measured in the septal calcite (Figure 2c).

[16] The outer wall has a surface that is irregular, and the wall itself is porous. Since the secondary crust is deposited on this surface, the crust cannot have a smooth lower boundary. The secondary crust must fill in the gaps on the surface of the outer wall. Even with carefully selected measurement locations, we conclude that the large relative variability of the outer wall measurements results from inadvertent excitation of the secondary crust. Consequently, the measurements from the septa more accurately represent the [Mg] of the primary calcite.

3.2. Dissolution

[17] Previous studies have demonstrated that carbonate dissolution affects the bulk [Mg] of foraminifers [Lea et al., 2000; Rosenthal et al., 2000; Brown and Elderfield, 1996; Rosenthal and Boyle, 1993; Russell et al., 1994; Savin and Douglas, 1973]. Even above the lysocline dissolution lowers the bulk [Mg] so that calculated Tw would be lower than environmental values. However, G. truncatulinoides (R) is a species that is moderately dissolution resistant; Berger [1970] ranks it 13th out of 22 species with higher numbers corresponding to greater resistance.

[18] Dissolution is relevant to higher-resolution studies using an electron microprobe only if the shell's carbonate is susceptible to incongruent dissolution or preferential release, for then the scale of the alteration of shell chemistry is comparable to the resolution of the analysis technique. Duckworth [1977] points out that inorganic precipitation of magnesium would result in a [Mg] 2 orders of magnitude higher than that measured in the foraminifers of this or any previous study. Thus she concludes that a post depositional effect would be the deposition of magnesium rather than its removal by dissolution. Rosenthal and Boyle [1993] conclude that incongruent dissolution is not important when considering the [Mg] in foraminifers for two reasons. The decrease in [Mg] correlates with an increase in percent fragmentation data indicating that congruent dissolution is the dominant process. Also, a down-core [Mg] record shows cyclic variation rather than a monotonic decrease. The data presented by Brown et al. [1980] support the premise that electron microprobe measurements are not affected by dissolution. They analyzed the concentrations in the primary calcite and crust of the individual from the shallow core top and the crust of the individual from the deeper core top. Dissolution had removed all of the primary calcite from this second individual. If dissolution lowers the measured concentration on a scale of tens of microns, the measurements (n = 22) from the crust of the individual deposited at a deeper depth should be lower than the measurements (n = 30) from the individual from a shallower depth. Only three measurements, or 10%, from the shallower individual were higher than the measurements from the deeper individual.

[19] While no conclusive proof is possible, three lines of evidence provide confidence that the [Mg] of the samples shells have not been altered by dissolution. First, visual inspection of the samples in the study, both the outer surfaces of the intact shells and the polished surfaces of cross-sections, shows no obvious evidence of dissolution. Second, the core tops used in this study are from a wide range of depths (1929 m to 3634 m) that includes depths below the average regional lysocline. We can allow water depth of the core top to serve as a proxy for dissolution potential because carbonate dissolution is greater with colder temperatures and greater pressure. Any effects of dissolution on our shells should appear first in the primary calcite since the higher [Mg] calcite of the septa should be more susceptible to dissolution than the lower [Mg] calcite of the crust. A comparison of water depth to [Mg] of the septa (Figure 3) shows no obvious relationship between the two variables. We see neither a general trend toward lower concentrations with depth or a discrete shift toward lower concentrations at the approximate level of the lysocline. For comparison, note that the range of [Mg] in the primary calcite of the sediment samples covers the same range as the [Mg] in the primary calcite of the plankton tow samples. Finally, the individual measurements of [Mg] within single shells show no pattern that could be explained by the effects of dissolution.

3.3. Strategy for Water Temperature Selection

[20] G. truncatulinoides (R) is not successfully grown under laboratory conditions, and the calibration between [Mg] and calcification temperature must be determined with a field study using specimens from the sediment. We use the established relationships between the relative abundance of G. truncatulinoides (R) in surface waters, the results of oxygen isotope studies, and the seasonal variation in water column structure in the western North Atlantic to infer the timing, depth, and temperature of calcification.

[21] The seasonal, upper water temperatures are used to establish the connection between the [Mg] of the primary calcite and the Tw at the top of the permanent thermocline. Figure 4 displays the relationship between the annual cycle of simple stability in the upper water column at Station S, Bermuda (32.1°N, 64.25°W), and the annual pattern of the flux of G. truncatulinoides (R) into sediment traps. The annual cycle of the vertical migration is synchronized to changes in the water column stability [Lohmann, 1992, 1993; McKenna, 2000]. As the stability in the water column decreases in November, the population in the surface water increases in numbers as the individuals migrate upward. Growth occurs from November to May until the mature individuals descend in the water column. The timing of the descent is linked to and precedes the development of a layer of increased stability that would impede movement deeper into the water column. As the stability weakens in November G. truncatulinoides (R) migrates from the depths to the surface waters. This pattern is supported by the annual variation of foraminifer flux and mean shell size in the data collected from sediment traps located near Station S, Bermuda (Figure 4) [Deuser and Ross, 1989], as well as plankton tow results from the North Atlantic [Bé and Spindler, 1983; Tolderlund and Bé, 1971]. A stability equal to 12 × 10−3 kg/m2 appears to be a critical stability (Sbcr) because its establishment in the water column seems associated with the reduced flux into the sediment traps at Station S, Bermuda [Deuser and Ross, 1989].

Figure 4.

Life cycle of G. truncatulinoides (R) and its relation to water column properties. The reproductive cycle of G. truncatulinoides (R) is synchronized to the annual variation of simple stability (Δσt/Δz) in the upper water column. The annual flux of the species in sediment traps at station S, Bermuda, occurs during those months when the water column stability is weakest [Deuser and Ross, 1989].

[22] We assume that G. truncatulinoides (R) is only present in the surface waters of the southern Pattern I (Table 1) sites when the layer of maximum stability is less than Sbcr. We further assume that all growth occurs in the surface layer and that maximum size is reached right before descent in the water column. We chose the Tw from the month and at a depth that the Sbcr first develops in the spring. Depending on core top location, this is either December or January at depths in the top 100 m (Table 2). A northern core top with an annual stability pattern similar to Pattern II (Table 1) presents a more complicated problem because, here, the stability is never less that Sbcr. We assume that G. truncatulinoides (R) only rises to the depth where stability equals Sbcr. Conditions will be most appropriate for growth when this is highest in the water column. At the Pattern II site, we chose Tw from the month and depth when the stability that equaled Sbcr was at its shallowest depth for the year. Tw is chosen from January, at the depth of 166 m (Table 2).

[23] Our choice of Sbcr is informed by additional data from G. truncatulinoides (R) in four of these core tops that were included in a previous analysis for stable isotopes. The δ18Owater and Tw at the calculated depth of calcification are combined with the average δ18O of the primary calcite of the samples [McKenna, 2000] to determine the relationship between Tw and δ18Ocalcite-water (Table 4 and Figure 5). This relationship is consistent with the O'Neil et al. [1969] relationship, having a similar slope. The observed offset between our data and the O'Neil et al. [1969] relationship results from the nature of the comparison. The δ18O measurements represent an homogenized value for the entire shell and an entire season's growth. Our Mg/Ca calibration uses only the last chambers formed over a shorter portion of the seasonal cycle of water column variation.

Figure 5.

Selection of water temperature for calibration to Mg/Ca. Shells from four of the core top populations used in this study had been used for previous stable isotope analysis [McKenna, 2000]. Water column data [Levitus et al., 1994] from the times and depths of calcification chosen for our calibration are used to calculate the δ18Owater. The offset between the measured δ18Ocalcite and δ18Owater is compared to the temperature [Levitus and Boyer, 1994]. The relationship of the data among the four sample locations is consistent with the O'Neil et al. [1969] relationship (solid line). The dashed line reproduces the slope of this fractionation and is shown in a shifted position for ease of comparison to our calibration selections.

Table 4. Stable Isotope Data From Selected Core Top Locations
Core Identificationδ18Ocalciteδ18OwaterTemperature, °C
V20-170 primary calcite0.260.3720.3
V20-170 secondary crust2.20−0.137.3

[24] The deep temperatures are used to establish the connection between [Mg] of the secondary crust and the Tw near the bottom of the permanent thermocline. Selection of an appropriate temperature for the addition of a secondary crust is based on stable isotope studies of Curry and Matthews [1981], Deuser and Ross [1989], Erez and Honjo [1981], and McKenna [2000]. We select environmental data from Levitus and Boyer [1994] and Levitus et al. [1994] at a constant density surface of σt = 27 (Table 2). The timing for the addition of the secondary crust is placed one month past the time of descent in the water column as determined in the section on selection of seasonal upper water column temperatures.

3.4. Temperature Calibration

[25] Sampling of G. truncatulinoides (R) from core tops across a wide geographic range provides two temperature gradients that can be calibrated to [Mg]. Variation of [Mg] in the primary calcite of all the individuals measured should reflect the latitudinal gradient of water temperature at the top of the thermocline. The difference of [Mg] in the primary calcite and the secondary calcite of any one individual should reflect the temperature variation between the top and the bottom of the permanent thermocline overlying that core top location.

[26] The relationship between Tw and [Mg] was initially calculated using only the average Mg/Ca from the primary calcite of septa from each shell. A linear relationship described by the equation:

display math

indicates a change of ∼163 ppm in the Mg/Ca per 1°C. The extrapolation of this septa-based relationship to lower temperatures passes directly through the data from the secondary crust. Including the crustal data in the regression does not significantly change the calculated relationship:

display math

(see Figure 6).

Figure 6.

Calibration between Mg/Ca and temperature. The data from this study can be related to water temperature by either a linear or an exponential relationship. The practical difference between the two approaches exists at water temperatures above 20°C.

[27] If Mg/Ca = (a + b)Tw the error associated with each term is

display math

Since use of the combined data set produces the same answer for temperature dependence as the data set compiled from only septal calcite measurements, the mechanism for exclusion of magnesium from the secondary crust appears to respond to temperature in the same manner as the mechanism that excludes magnesium from the primary calcite.

[28] While the above calibration appears sufficient, previous work using planktonic foraminifers [Lea et al., 1999, 2000; Mashiotta et al., 1999; Nürnberg et al., 1996a, 1996b] and benthic foraminifers [Rosenthal et al., 1997] have resulted in exponential relationships. While a single calibration is not appropriate for all species of planktonic foraminifers, we assume that the shape of the relationship remains consistent across species. Using this assumption, the calibration is:

display math

(see Figure 6).

[29] If Mg/Ca = ae (bTw) the error associated with each term is

display math

A 1°C change in water temperature produces a change of ∼10% in the Mg/Ca.

3.5. Error Analysis

[30] The uncertainty associated with the relationship between [Mg] in the calcite of G. truncatulinoides (R) and water has a number of sources of error. Each measurement has a counting uncertainty that is recorded in the raw data [McKenna and Prell, 2000]. The means of [Mg] are calculated for each type of measurement site within a shell and are the basis for the regressions just calculated. The standard deviation of the mean is an appropriate measure of the analytical error associated with the magnesium concentrations within a particular shell, and these are reported in Table 3. We also have assigned appropriate error estimates to the statistical uncertainty of each equation, but our method for selecting Tw for the calibration is the source of a statistically unquantifiable error, for we cannot assume that Tw is well-known. Indeed, the error associated with Tw might well exceed the uncertainties associated with either measurements of [Mg] or regression parameters. A data set of [Mg] and δ18O for G. truncatulinoides (R) from the same sample would provide a reliable approach for initial testing of our regression equations. We apply our regression equations to a limited number of plankton tow samples. The calculations of Tw from [Mg] for these samples are reported with the error from the analytical uncertainty of the measurements as well with the highest and lowest possible temperatures based on the error associated with the calculation of the regression parameters (Table 5a).

Table 5a. Calculated Temperatures for Plankton Tow Samples From Mg/Ca Measurements
Sample IdentificationMg/Ca, mmol/molStandard Deviation of MeanTemperature Linear Calculation, °CAnalytical Error, °CMaximum Temperature, °CMinimum Temperature, °CTemperature Exponential Calculation, °CAnalytical Error, °CMaximum Temperature, °CMinimum Temperature, °C

3.6. Application to Plankton Tow Samples

[31] Individual G. truncatulinoides (R) collected by plankton tow in the western North Atlantic provide an avenue to assess the accuracy of our temperature calibration. The information that the individuals from this tow can provide is limited because the tow is integrated over 700 m and the only environmental data recorded was the surface water temperature. However the availability of the plankton tow samples provides two opportunities. First, the results of the two possible Mg/Ca equations (linear and exponential) can be compared to environmental temperatures extracted from a published data set. Second, Mg/Ca temperatures can be compared to temperatures calculated from δ18O using contemporaneous shells from one location.

[32] These shells had no secondary crust, and the range in their sizes (50,000 μm2 to 400,000 μm2) was roughly equivalent to the size range found in the populations recovered from the sediments. The [Mg] was measured in the septa of half of the shells with the electron probe (Table 3 and Figure 3). The remaining shells were individually analyzed for δ18O (Table 5b). This analysis was performed at Woods Hole Oceanographic Institution with a Finnigan MAT252 mass spectrometer with the Kiel Carbonate Device. Samples were broken apart and washed in a 3% hydrogen peroxide solution to remove organic matter. The system automatically analyzes each sample in a separate reaction vessel with 100% phosphoric acid at 70°C. A datum was rejected if the sample voltage was less than 0.8 volts and the ratio of sample voltage to standard voltage was outside the range of 0.8 to 1.2. Data are corrected so that the δ18O of NBS 19 is −2.2‰ and reported as a deviation per mil from Vienna Peedee belemnite. The analytic precision, based on 2200 National Bureau of Standards (NBS) 19 samples, is ±0.07‰ for δ18O [Ostermann and Curry, 2000]. The associated error for each temperature is calculated using the appropriate equation and the average values ± the standard deviation of the mean (Mg/Ca) or the precision of the measurement (δ18O).

Table 5b. Calculated Temperatures for Plankton Tow Samples From δ18O Measurements
Sample IdentificationOxygen (PDB)aPrecisionTemperature,b °C

[33] At the time of the plankton tow, the surface water temperature was 25.3°C. Comparison with the Levitus and Boyer [1994] data indicates that this temperature was unusually warm for May. Therefore we use the June environmental data to determine the water properties at a σt = 26.0: δ18Owater = 0.98‰ and Tw = 19.8°C. The linear equation estimates temperatures that are clearly too warm to be realistic (Table 5a and Figure 7), producing some temperatures that equal or exceed the measured surface temperature. However, the exponential equation provides temperature estimates that are compatible with the Levitus and Boyer [1994] temperature of 19.8°C at depth and the measured surface water value of 25.3°C (Table 5a5b and Figure 7).

Figure 7.

Comparison of calcification temperatures calculated using Mg/Ca and δ18O measurements. Shells from a North Atlantic plankton tow are measured for either Mg/Ca or δ18Ocalcite. Water temperatures are calculated from the resulting data using either the two Mg/Ca relationships (linear and exponential) or the δ18O fractionation of O'Neil et al. [1969]. The average temperature for each relationship is shown highlighted in the shaded box to the left of the temperature axis. The offset between the Mg/Ca average temperature (exponential) and the average δ18O temperature is 1°C, demonstrating the utility of this calibration. The average temperature calculated using the linear relationship is 2.3°C higher than the δ18O temperatures.

[34] We also see this pattern in the comparison of Mg/Ca temperatures and δ18O temperatures. The range of the temperatures calculated using our exponential Mg/Ca relationship closely matches the range of the δ18O temperatures [O'Neil et al., 1969]. The average temperatures for each relationship agree to within +1°C (exponential: 22.2°; O'Neil: 21.1°). In contrast, the average of the temperatures calculated using the linear relationship is 2.3°C higher than the average of the δ18O temperatures. While the specific offset will vary with the choice of equation for δ18O fractionation and particularly the values chosen for the upper Tw, the comparison does supports the utility of the exponential Mg/Ca calibration.

4. Conclusions

[35] G. truncatulinoides (R) is a planktonic foraminifer that has an annual life cycle during which it grows in the surface waters and adds a secondary crust at depth. These features of the species' natural history mean that an individual's shell has the potential to record the temperatures at the top and bottom of the permanent thermocline. Previous work with other species of foraminifers has established [Mg] as a proxy for water temperature during calcification. However, G. truncatulinoides (R) cannot be cultured in the laboratory, so a temperature calibration must be determined using individuals collected from either the water column or the sediments.

[36] This study has demonstrated that measurements on specimens from the sediment can be combined with information in environmental data sets to produce a useful calibration between [Mg] and Tw.

[37] 1. The [Mg] in the primary calcite of G. truncatulinoides (R) is best represented by measurements from the septa rather than from the outer wall. Useful mean values for each type of measurement site can be calculated from 4 to 6 individual measurements.

[38] 2. Dissolution does not affect the values of the [Mg] in the shell septa of G. truncatulinoides (R).

[39] 3. The Mg/Ca data can be usefully correlated with water temperature by an exponential equation that indicates a ∼10% change in Mg/Ca per 1°C temperature change.

[40] 4. Temperatures calculated from Mg/Ca are consistent with (+1°C) and have a similar range as the temperatures calculated from δ18O using the O'Neil et al. [1969] equation.

[41] The [Mg] in the primary calcite reflects Tw at the top of the permanent thermocline, and the [Mg] in the secondary crust reflects Tw at a constant density surface at a depth close to the bottom of the thermocline. In paleoceanographic studies, the [Mg] in the primary calcite and secondary crust of G. truncatulinoides (R) can be a useful proxy for temperature changes at the extremes of the permanent thermocline.


[42] We thank Joseph Devine and William Collins for technical assistance with sample preparation and use of the electron microprobe and Dorinda Osterman of WHOI for assistance with stable isotope analyses. Tim Herbert and Robley Matthews contributed valuable suggestions and discussion. We thank Eugene Jarosewich, National Museum of Natural History, Smithsonian Institution, for providing the sample dolomite USMN 10057 Oberdorf, Austria, used as a standard in our microprobe analysis. V.S.M. is currently Education Director for EAPS at Massachusetts Institute of Technology.