Coccolith Sr/Ca as a new indicator of coccolithophorid calcification and growth rate

Authors


Abstract

[1] Polyspecific coccolith separates from core top sediments in the eastern equatorial Pacific show variations of ∼15% in Sr/Ca ratios across the equatorial upwelling zone, with Sr/Ca highest at the equator and decreasing off-axis. These variations cannot be due to changes in the Sr/Ca of seawater, which varies by less than 2% in the surface ocean. Variations in Sr/Ca of coccolith sediments are similar to variations in primary productivity and alkenone-estimated coccolithophorid growth rates in overlying surface waters and to CaCO3 rain rates measured in sediment traps. Because of these relationships and because calcification rate exerts a strong control on Sr/Ca in abiogenic calcites, we suggest that the observed Sr/Ca variations in coccoliths may be strongly controlled by coccolithophorid growth and calcification rates, although temperature may also influence coccolith Sr/Ca to a lesser degree. Changes in dissolution intensity and coccolith assemblages appear to exert a minor influence, if any, on coccolith Sr/Ca in these sediment core tops. If further work confirms relationships between coccolith Sr/Ca and coccolithophorid productivity, Sr/Ca records of past changes in coccolithophorid productivity may be useful in reconstructing past variations in the rain ratio of organic to carbonate carbon, an important control on deep ocean pH and partitioning of CO2 between the atmosphere and ocean. In addition, coccolith Sr/Ca might provide an independent record of past changes in coccolithophorid growth rates, which in combination with data on the carbon isotopic fractionation in coccolithophorid organic matter may permit more reliable calculations of past dissolved CO2 in the surface ocean.

1. Introduction

[2] Although the minor and trace element chemistry of biogenic carbonates is widely applied to infer past oceanographic conditions, most of the work has focused on the chemistry of foraminifera and, more recently, that of corals. Ironically, the very abundant calcite produced by coccolithophorids has not been previously utilized in minor element studies. The minor element chemistry of coccoliths, external plates of calcite produced by coccolithophorids, may provide a unique perspective on past changes in climate and the carbon cycle. This is because the coccolithophorids (marine algae class Prymesiophyceae) are primary producers and play key roles in the global carbonate and carbon cycles [Westbroeck et al., 1993]. Coccoliths are a major component of calcareous sediments in the open ocean, contributing up to 80% of the CaCO3 in some locations [Fabry, 1989; Deuser and Ross, 1989; Honjo, 1978]. As coccolithophorids are the only widely distributed primary producers that preserve a fossil record, they may provide critical information on past variations in primary productivity and carbon and carbonate fluxes.

[3] Sr may be a particularly useful geochemical characteristic in coccolith carbonate. The Sr/Ca ratio of biogenic carbonate depends on the Sr/Ca ratio of the seawater and the Sr partitioning coefficient of the carbonate. Variations in carbonate Sr/Ca greater than 2% must be due to varying Sr partitioning. This is because Sr/Ca in surface water varies by less than 2% [de Villiers et al., 1994; de Villiers, 1999] in the modern ocean, and likely varied by less than 3% over the last several hundred thousand years [Stoll and Schrag, 1998; Stoll et al., 1999]. In abiogenic calcites, Sr partitioning is strongly controlled by precipitation rate [Lorens, 1981; Tesoriero and Pankow, 1996]. In coccolithophorids the rate of calcite production varies widely [e.g., Balch et al., 1996; Nimer and Merrett, 1993] and is frequently correlated with coccolithophorid growth rates [e.g., Balch et al., 1996; Nimer and Merrett, 1993]. If, as in abiogenic calcites, Sr partitioning in coccolith carbonate is strongly controlled by calcite precipitation rates, coccolith Sr/Ca may record variations in coccolithophorid productivity. If postdepositional effects and changing coccolith components do not bias coccolith Sr/Ca in the sediment record, they might be used to extract records of past variations in coccolith productivity and CaCO3 production. Such proxies are important in evaluating model scenarios proposed for changes in atmospheric CO2 over glacial cycles [e.g., Archer and Maier-Reimer, 1994; Archer et al., 2000]. In addition, if the Sr/Ca ratio of coccoliths was to track past variations in coccolithophorid growth rates, it might be possible to use coccolith Sr/Ca ratios in combination with the carbon isotopic fractionation in alkenones to estimate past dissolved CO2 concentrations in surface waters. This would improve our understanding of changes in oceanic sources and sinks of CO2 which are likely to have contributed to past CO2 and thus climatic variability of glacial/interglacial timescales.

[4] In this study, we examine Sr/Ca ratios in coccoliths from core top and downcore sediments to establish whether or not there are important variations in Sr partitioning in coccolith carbonate and how the Sr/Ca variations of coccolith carbonate relate to Sr/Ca variations in foraminiferal and bulk carbonate. We also seek to assess the evidence for relationships between coccolith Sr/Ca and coccolith productivity (growth and/or calcification rate) and other environmental parameters. Using a recently developed inductively coupled plasma–atomic emission spectroscopy (ICP–AES) technique [Schrag, 1999], we obtained high-precision (0.4% rsd) measurements of Sr/Ca in mixed coccolith assemblages from core tops along two transects from ∼10°S to 5°N at 140° and 110°W in the eastern Pacific. In this region, upwelling brings cool nutrient- and CO2-rich waters to the surface in a narrow band at the equator, leading to large latitudinal gradients in biological productivity, sea surface temperatures, and surface nutrient and dissolved CO2 concentrations. In the 140°W transect where data are available, estimated coccolithophorid growth and calcification rates also vary latitudinally. Relationships between Sr/Ca and changes in species assemblages and the relative flux of coccoliths to foraminifera across core top transects are also investigated. This approach may not provide definitive evidence on controls of Sr partitioning in coccoliths, in part because of limited modern surface-water measurements. Nevertheless, it can demonstrate systematic trends and suggest important factors meriting further investigation, perhaps through culture studies and plankton experiments where coccolith chemistry and biological and environmental factors can be more reliably linked.

2. Sediments

2.1. Location of Samples

[5] Core top sediment samples were obtained from two transects across the equator in the eastern Pacific, from 12°S to 5°N at 140°W and from 5°S to 5°N at 110°W (Figure 1). Core top depths average 4300 m for the 140°W transect and 3600 m for the 110°W transect. Sediments are high in carbonate (75–90% [Murray and Leinen, 1996]) within 5°–6° of the equator, but on the flanks of biogenic sediment pile away from the equator, the seafloor is beneath the carbonate compensation depth, and carbonate preservation decreases significantly to CaCO3 of 55% at 12°S. Biogenic opal and terrigenous detritus are the primary noncarbonate components in the 140°W transect. In the 110°W transect, hydrothermal precipitates, especially fine Mn and Fe oxides, are also present due to the proximity of the Galapagos spreading ridge [Mayer et al., 1992]. For the 140°W transect, radiocarbon dates available for core tops between 5°N and 5°S all yield late Holocene ages; no radiocarbon dates are available for the 12°S core [DeMaster et al., 1996]. Oxygen isotope stratigraphy for the 110°W transect indicates that all core top samples are Holocene in age (A. Mix, personal communication, 1999).

Figure 1.

Location of core top (solid red circles) and downcore (open red squares) samples used in this study. Core top samples from the 140°W transect are multicores from the TT013 cruise. Downcore piston core PC-72 is also from the TT013 cruise. Core top samples from the 110°W transect are from the VNTR01 cruise. Nearby ODP sites, used for additional information on sediments, are indicated by open black circles.

[6] Downcore sediment samples were obtained from two sites within the main zone of equatorial upwelling. From site TT013-PC72 (0°06′N, 139°24′W, depth 4298 m) we studied samples covering the last 150 ka, and in core W8402A-14GC at the MANOP C site (0°57.2′N, 138°57.3′W, depth 4287 m), we studied samples covering the last 250 ka. Age models for the downcore sites are based on correlation of benthic foraminiferal δ18O records with the SPECMAP “stacked” benthic foraminiferal δ18O timescale of Martinson et al. [1987]. In site TT013-PC72, age models are from Murray et al. [1995] and in MANOP C from Jasper et al. [1994].

2.2. Separation of Coccolith and Foraminiferal Fractions

[7] In all core top samples we measured Sr/Ca in bulk sediment, the foraminiferal fraction (>63 μm) and the coccolith fraction (<12 μm); for the 140°W transect we also measured Sr/Ca in monospecific samples of foraminifera Globorotalia tumida. Our method for separating coccolith calcite is based on observations of Paull et al. [1988] on the distribution of carbonate particles in the <38 μm fraction of sediments. Particle size distributions have a mode centered around 4 μm equivalent spherical diameter, corresponding to coccolith-sized particles, and a secondary mode in the range of 10–25 μm equivalent spherical diameter, corresponding to fragments of foraminifera and calcareous dinoflagellate cysts. We sought to minimize noncoccolith carbonate by separating the <12 μm size fraction from bulk sediments. Bulk carbonate was sieved to obtain the <63 μm fraction. Ethanol was used for sieving to prevent dissolution [e.g., Pingitore et al., 1993]. The 12–63 μm size fraction was then removed by settling the suspension in ethanol for 10 min. Particle size analysis with a Coulter Counter indicates that the material remaining in suspension contains particles of up to 12 μm in equivalent spherical diameter, with maximum frequency of particles around 3–5 μm equivalent spherical diameter. We extract the suspension, allow this fraction to settle overnight, and siphon off the ethanol.

[8] To obtain a carbonate fraction dominated by foraminifera, the > 63 μm size fraction was crushed and sonicated repeatedly in ethanol to remove all adhered fine carbonate and noncarbonate material. This sonication was apparently more efficiently applied to the samples from the 140°W transect than the 110°W transect, since upon dissolution some fine noncarbonate material remained in the latter samples (implying that some fine coccolith carbonate may have been present as well).

2.3. Characterization of Coccolith Assemblages in Coccolith Fraction

[9] We used scanning electron microscope (SEM) images from smear slides at a magnification of 1500 times to identify coccolith assemblages in the separated coccolith fraction. Our goal was to identify major changes in the coccolith assemblages and not to precisely determine relative abundances of all species in the samples. We recorded the number of intact coccoliths from the six most common taxa in 10 fields of view, counting on average 200 specimens per sample in the 110°W core top transect, 114 specimens per sample in the 140°W transect, and 272 specimens per sample from six depths in the MANOP C downcore record. In the case of C. leptoporus, many of the distal and proximal shields were separated, and from some views it was not possible to determine if the shields were intact or if only one shield was present, complicating estimation of the relative contribution of this species to the sample. Because examples of clearly separated shields were nearly as abundant as intact shields in ambiguous views, we assumed that 50% of all counted specimens of C. leptoporus represented single shields and weighted the counts by 0.66 when calculating the relative contribution of C. leptoporus to the sample. Florisphaera profunda is also present in these sediments, but its small size prohibited reliable counts in our smear slide preparations. Likewise, the smallest Gephyrocapsa (<2 μm) were not counted.

[10] Counts were used to estimate the relative abundance of different taxa and the relative contribution of calcite from each species. The percent calcite from each species was calculated by estimating the average weights of coccoliths from each species by measuring the average coccolith diameter in our samples and using the diameter/volume relationships for each species established by Young and Ziveri [2000]. This calculation does not include F. profunda, which is assumed to be an insignificant source of calcite in these samples. Furthermore, it assumes that the species present as fragments of coccoliths (which were not counted) are present in the same relative abundance as the intact coccoliths. However, selective breaking of more fragile species may lead to their underrepresentation in counts of intact coccoliths. In these samples, the weight contribution of the less abundant species is so minimal that their underrepresentation by a factor of 2–4 would not significantly affect the results.

3. Cleaning Techniques and Sr/Ca Analysis

3.1. Sample Cleaning and Sr/Ca Analysis

[11] Although carbonate is the dominant source of both Sr and Ca in our marine sediments, both ions may be present in contaminant phases such as Fe and Mn oxyhydroxides [Apitz, 1991]. In addition, the contribution of scavenged and adsorbed Sr and Ca may be significant, especially in the coccolith fraction where it is not possible to physically separate coccoliths from fine aluminosilicate detritus. To evaluate the influence of noncarbonate Sr and Ca on our measurements, we compared several methods for the preparation of coccolith samples for analysis. All samples were rinsed in ethanol. Some samples were shaken 30 min with ion exchange solutions, either 3 mL of 1 N NH4Cl /4 mg sediment or 6 mL of 2% NH4OH /4 mg sediment, followed by rinsing in distilled water. Others were shaken 4 hours with reducing solutions (25 g NH2OH:HCl, 200 mL concentrated NH4OH, in 300 mL distilled water; to reduce iron and manganese oxyhydroxides) followed by ion exchange and distilled water rinses. Following the precleaning procedures, ∼0.5–1.0 mg of calcite was dissolved in 5 mL of ultrapure 2% HNO3. However, splits of several samples were dissolved in 0.1 M ammonium acetate/acetic acid buffer to minimize the influence of acid strength on release of noncarbonate Sr during dissolution. Bulk carbonate and the >63 μm fraction were analyzed with no additional cleaning other than ethanol rinsing during the separation of the >63 μm fraction. Monospecific samples of the foraminifera Globorotalia tumida from the 140 transect were analyzed following reducing and ion exchange cleaning steps. Results of all cleaning experiments are reported in Table 1.

Table 1. Location and Sr/Ca Data From Different Core Top Sediment Fractions and Different Cleaning Procedures
  Coccolith Fraction, <12 μm>63 μmG. tumidaBulk
SiteLongitude, degLatitude, degDepth, mSr/CaSr/CaSr/CaSr/CaSr/CaaSr/CaSr/CaSr/Ca
  • a

    For the Sr/Ca data, a checklist of the cleaning steps (listed on left-hand margin) used in the preparation of samples is given at the top of each column. A dash indicates that the cleaning step was not used in preparing Sr/Ca results for analysis. The cross denotes that the procedure was applied. The asterisk denotes correction of Sr/Ca data for inferred changes in the calcite chemistry during cleaning, as described in the text.

Cleaning Steps
Method nameETHMNXIONXAPITZAPITZaETHMNXETH
Ethanol rinsexxxxxxx
Reducing step-x-2x-x-
Ion exchange-NH4ClNH4ClNH4OH 2x-x-
DissolutionHNO3HNO3HNO3acetateHNO3HNO3HNO3
Samples
TT013MC148−14012490010.9 
TT013MC104−140544162.182.042.041.952.041.301.231.63
TT013MC93−140244132.332.252.271.301.231.85
TT013MC88−1400.8244152.352.272.291.311.222.10
TT013MC58−1400.1143072.362.332.332.252.341.321.241.90
TT013MC23−140−243652.262.232.231.321.221.88
TT013MC41−140−542692.242.212.211.321.241.73
TT013MC4−135−1242802.582.462.492.242.331.351.292.10
 
VNTRO14−110538482.472.25 2.152.241.29 1.92
VNTRO16−110337652.162.071.381.57
VNTRO17−110137222.392.271.362.05
VNTRO18−110037912.362.292.212.301.591.97
VNTRO19−110−338602.222.131.321.85
VNTRO110−102−4.534052.202.061.331.80

[12] A potential drawback of some of these cleaning procedures is that they can partially dissolve the sample (e.g., NH4Cl) or alter the chemistry of the calcite (e.g., hydroxide solutions). Treatment of pure, abiogenic aragonites with hydroxide solutions results in replacement of CaCO3 by Ca(OH)2 [Pingiatore et al., 1993] and may decrease Sr/Ca ratios by 5% [Love and Woronow, 1991]. We discuss these complications in detail in section 3.2.

[13] Sr/Ca ratios were determined by inductively coupled plasma–emission spectroscopy (ICP–ES) on a Jobin-Yvon simultaneous instrument, model 46-P. Sr/Ca ratios were corrected for a small effect of the concentration of Ca on the measured ratio following Schrag [1999]. Resulting precision is >0.4% (1σ) based on replicate analyses of the same sample at different dilutions over the course of several days. This precision is slightly worse than that reported by Schrag [1999] because there is greater variability in Sr/Ca ratios of foraminifera and coccoliths compared to corals. Since the concentration effect depends on the Sr/Ca ratio, the correction is less precise when applied to samples with a range of Sr/Ca ratios.

3.2. Effect of Adsorbed Cations

[14] Adsorbed Sr and Ca are potentially important contributors to the Sr/Ca ratio measured in the coccolith fraction, since components with high cation exchange capacity (alumino silicates in silt and clay fraction) cannot be separated physically from the coccolith fraction. The coccolith fraction of TT013 MC148 (which contains <1% CaCO3) releases ∼5000 ppm Ca from noncarbonate sources when exposed to strong acid (2% HNO3). The Sr/Ca ratio of this exchangeable fraction is 10.9 mmol/mol, consistent with Sr/Ca ratios measured in pore fluids at the sediment/water interface in equatorial Pacific sediments [Delaney and Linn, 1993]. Simple mass balance calculations indicate that for sediments with high-CaCO3 content, the contribution of exchangeable Sr and Ca will be minor. For example, exchangeable Sr and Ca would increase the Sr/Ca ratio by only 1% for coccolith fractions comprised of 80% carbonate and 20% noncarbonate material (assuming noncarbonate material similar to that of MC148). However, exchangeable Sr and Ca may elevate the Sr/Ca ratio by 15% in coccolith fractions comprised of 20% carbonate and 80% noncarbonate material similar to that of MC148.

[15] For the 140°W transect, removal of exchangeable cations is likely responsible for much of the decreases in Sr/Ca observed with reducing (MNX) and ion exchange (IONX) cleaning procedures (Table 1 and Figure 2). There is no additional decrease in Sr/Ca with the use of reducing techniques compared to ion exchange treatments alone, comparing samples with strong acid dissolution. The variable decrease in Sr/Ca for different samples correlates with geochemical indicators of the contribution of terrigenous material and scavenging, such as ppm Al (Figure 2b) [Murray and Leinen, 1996]. However, the Sr/Ca reduction in the 12°S site is much lower than expected from the very high detrital component and low CaCO3 percent. This may indicate incomplete removal of exchangeable cations from noncarbonate phases in this sample and/or that Sr and Ca are released from other noncarbonate sites upon acidification with 2% HNO3 for carbonate dissolution. The analyzed coccolith fraction for 12°S consisted of at most 30% carbonate and 70% noncarbonate material likely similar to that of MC148. Mass balance considerations indicate that exchangeable cations should elevate the Sr/Ca ratio by at least 9%, implying coccolith carbonate Sr/Ca ratio of 3.34 mmol/mol or less.

Figure 2.

Sr/Ca ratios in the coccolith fraction with different cleaning treatments. (a) Variation in coccolith Sr/Ca for cleaning treatments ETH (ethanol rinse), IONX (ion exchange), MNX (reducing and ion exchange), and APITZ* (reducing and ion exchange with dissolution in buffered acetic acid), corrected as described in text and denoted in Table 1. (b) Difference (in mmol/mol) in sediment Sr/Ca ratios with ETH and MNX or APITZ (uncorrected) cleaning treatments, compared with bulk ppm Al [Murray and Leinen, 1996], an indicator of detrital input and scavenging. Error bars on the APITZ measurements indicate inferred changes in carbonate chemistry from hydroxide exposure during cleaning.

[16] More intense ion exchange and reducing steps (doubling the amount of reagent per mg of sediment) followed by dissolution in acetic acid/ammonium acetate buffer may minimize the contribution of noncarbonate Sr and Ca [e.g., Apitz, 1991] (procedure abbreviated “APITZ”). However, the increased exposure to hydroxide solutions may also decrease the Sr/Ca of the carbonate. With the exception of MC4, this treatment decreases the Sr/Ca ratio of coccolith fractions by 3–4% compared to MNX, maintaining a trend parallel to that of MNX treatments for 140°W samples in Figure 2b. We believe that the 3–4% reduction may be due to changes in calcite chemistry and not to removal of exchangeable cations, whereas the larger reduction in MC4 (9%) is in part due to additional removal of exchangeable cations. If this 4% effect is removed from all analyses, the final Sr/Ca ratio of MC4 is 3.33 mmol/mol, very similar to the maximum ratio calculated from mass balance considerations. It is possible that the Sr/Ca ratios of coccolith carbonate may be lower, but additional MNX and IONX cleaning procedures are likely to incur additional changes in the calcite chemistry. Consequently, extracting true Sr/Ca ratios of coccolith calcite in low-carbonate sediments may be difficult and may require further refinement of cleaning techniques. To reflect the greater uncertainty of estimating coccolith Sr/Ca in these sediments, we distinguish this measurement with a different symbol in subsequent graphs. There is little difference (<1%) in Sr/Ca between the treated and untreated fractions for the downcore coccolith samples (Table 2), consistent with the very high CaCO3 and low detrital components in these equatorial sites.

Table 2. Age Data and Sr/Ca Measurements of Downcore Samplesa
Core Depth, cmAge, kyrCoccolith Fraction
Sr/Ca; ETHSr/Ca; MNX
  • a

    Cleaning steps for each technique are delineated in Table 1 and are described in detail in the text.

Cruise TT013 Site PC72
102.292.29
842.292.29
1272.302.27
1910.52.282.27
2513.52.292.28
3116.32.24
3719.12.172.15
4422.42.202.20
5025.52.202.19
5528.42.202.20
6433.62.262.26
7037.22.242.23
80432.222.22
9048.92.222.23
10054.72.242.24
10557.32.232.22
11563.52.232.19
12268.42.19
12872.92.202.15
13275.82.202.12
13981.22.112.00
14585.82.07
15191.92.182.15
161108.12.142.12
171122.92.252.21
181129.72.232.24
195139.82.112.12
202144.32.122.15
211150.12.202.18
218155.82.162.16
2241612.212.19
230166.52.252.24
235170.12.282.28
2441762.272.27
250179.72.442.28
256183.92.332.31
2601942.392.36
2672002.372.38
275208.52.352.33
285219.62.34
290223.62.332.30
295227.62.302.29
305234.32.302.34
3212492.252.27
325252.82.282.31
Cruise TT013 Site PC72
902.282.27
195.112.312.32
299.132.342.32
3912.612.342.33
4915.82.252.26
5918.92.282.26
6922.122.282.25
7925.742.312.27
8930.322.262.24
9936.622.232.21
10943.52.272.26
11950.152.272.26
12955.162.312.33
13959.492.32.29
14963.742.32.3
15968.362.322.32
16973.882.322.34
17980.892.182.21
18989.542.212.25
19999.152.292.3
209108.32.272.29
2191162.372.38
229122.42.292.3
239128.42.252.26
249134.42.192.19
269144.42.162.16
279149.12.142.15

3.3. Effect of Sr and Ca in Mn and Fe Hydroxides in Coccolith Fraction

[17] In the 110°W transect we did not analyze samples with ion-exchange treatments alone. Consequently, it is more difficult to separate the effects of ion exchange and reduction of Fe and Mn oxides. The effects of ion exchange may be comparable to those in the 140°W transect; however, there is likely an additional effect of removing Sr present in Mn and Fe oxides, which may explain the greater average reduction in Sr/Ca with cleaning in the 110°W transect (5 versus 3%). The sites of the 110°W transect lie close to the Galapagos spreading ridge (Figure 1) and may be influenced by hydrothermal plumes. Hydrothermal Fe and Mn oxides can elevate the Sr/Ca ratio of sediments by up to 20% [Apitz, 1991]. The largest reduction in Sr/Ca occurs at the 5°N sample, where the reducing and ion exchange treatment lowers Sr/Ca ratios by 9%. This location also has the highest magnetic susceptibility (indicative of fine terrigenous or metalliferous particles) of the Ocean Drilling Program (ODP) sites coinciding with our transect and was the only ODP site where abundant Mn and Fe oxide particles were identified in smear slides [Mayer et al., 1992]. It is difficult to assess whether all Sr and Ca associated with Mn and Fe oxides were removed during the APITZ cleaning treatment. It is possible that the Sr/Ca ratios of coccolith carbonate may be lower, but as discussed previously in section 3, additional MNX and IONX cleaning procedures are likely to incur additional changes in the calcite chemistry. Extracting reliable Sr/Ca ratios of coccoliths in sediments with abundant Mn and Fe oxides may be difficult. To reflect the greater uncertainty of estimating coccolith Sr/Ca in this site, we distinguish this measurement with a different symbol in subsequent graphs.

3.4. Selective Dissolution in the Coccolith Fraction

[18] We compared the change in Sr/Ca with partial dissolution of two carbonate-rich samples and one carbonate-poor sample upon exposure to varying amounts of 0.001 N HNO3. When we dissolve <50% of the sample, Sr/Ca ratios of the remaining carbonate (dissolved in 2% HNO3) are only slightly affected and are intermediate between those of ethanol-rinsed samples and those cleaned with reducing and ion-exchange steps (Figure 3). However, extreme partial dissolution of the samples (>50% dissolved) results in increased Sr/Ca ratios, possibly due to increasing influence of noncarbonate Sr and Ca relative to carbonate Sr and Ca.

Figure 3.

Results of partial dissolution experiments. (top) Sr/Ca ratios of sediments after different extents of partial dissolution. (bottom) Fraction of carbonate remaining after treatment. Open symbols are estimates based on prior experiments; solid symbols represent measurements made in this experiment. ETH, samples rinsed in ethanol (no partial dissolution); PDIS, samples partially dissolved with different amounts of 0.001 N HNO3; IONX, samples treated with different amounts of 1 M NH4Cl; and MNX, samples treated with reducing and ion exchange solution as described in the text. After cleaning, all samples were dissolved in 2% HNO3 for analysis.

[19] Treatment of samples with 1 M NH4Cl to remove exchangeable Sr also results in partial dissolution of carbonates when larger volumes of NH4Cl are used. In this example of partial dissolution, exchangeable Sr should be partially removed. For carbonate-rich samples, extreme partial dissolution (>50%) does not lead to a large increase in Sr/Ca ratios. This suggests that the increase in Sr/Ca ratios with extreme partial dissolution in 0.001 N HNO3 was caused by exchangeable Sr. Although there is still a significant ascent of Sr/Ca ratios for the low-carbonate sample with increasing dissolution, the Sr/Ca increase is much lower for comparable levels of partial dissolution. These results suggest that while partial dissolution may affect sediment Sr/Ca by altering the proportion of cations from carbonate and noncarbonate sources, the carbonate Sr/Ca of these coccolith fractions is relatively insensitive to partial dissolution. The constancy of coccolith Sr/Ca with moderate to extreme dissolution suggests that partial dissolution is not responsible for the latitudinal trends in Sr/Ca observed in the equatorial Pacific transect.

3.5. Implications for Sr/Ca of Other Carbonate Fractions

[20] Since noncarbonate sources may contribute appreciable Sr and Ca to the carbonate fraction, these sources must also influence the Sr/Ca ratio of bulk sediment. Consequently, the Sr/Ca ratio of bulk carbonate is likely to be much lower than that of bulk sediment measured in sites MC4 (140°W, 12°S) and VNTR01 4 (110°W, 5°N) and somewhat lower than bulk sediment in MC104 (140°W, 5°N). For subsequent discussions (e.g., Figure 4), we estimate a likely range of bulk carbonate Sr/Ca for these samples, on the basis of the effect on noncarbonate Sr and Ca in the coccolith fraction. Noncarbonate Sr and Ca are likely to have much smaller effects on the Sr/Ca ratios of the foraminiferal fraction because most noncarbonate components (clay/silt and fine Fe or Mn oxides) have been physically removed from the foraminifera during picking and separation.

Figure 4.

(a) Sr/Ca ratios of coccolith carbonate (green circles) compared with Sr/Ca of bulk carbonate (blue squares) and Sr/Ca of the >63 μm foraminiferal fraction (red solid triangles) and Sr/Ca of foraminifera G. tumida (red open triangles) for 140° and 110°W transects. Open circles and open squares denote coccolith fraction and bulk carbonate sediments, respectively, with high Sr and Ca contribution from noncarbonate fractions as discussed in the text. (b) Fraction of carbonate from coccoliths, calculated as described in the text.

4. Sr/Ca in Core Top and Downcore Carbonates

4.1. Sr/Ca in Core Top Carbonates

[21] The Sr/Ca ratio of coccolith carbonate is much higher than that of bulk carbonate or foraminiferal carbonate (Figure 4a). Coccolith Sr/Ca across the core top transects ranges from 2.04 to 2.33, much larger than the Sr/Ca variation in the >63 μm foraminiferal fraction or in the monospecific G. tumida foraminiferal record. This large variation (∼15%) in coccolith Sr/Ca exceeds the <2% variation in Sr/Ca observed in modern seawater and must be due to variations in Sr partitioning in coccolith carbonate.

[22] Both coccolith and bulk carbonate show maxima in Sr/Ca at the equator in both transects. However, the variation in bulk carbonate Sr/Ca is larger than that of coccolith carbonate and must be driven both by changes in coccolith Sr/Ca and by changes in the relative proportion of coccolith versus foraminiferal carbonate across the transect, as calculated in Figure 4b. Coccoliths contribute the greatest fraction of carbonate at the equator.

4.2. Sr/Ca and Composition of Coccolith Carbonate

[23] We compare variations in the Sr/Ca ratio of coccolith carbonate across the transects with the species composition of the coccolith fraction to see if changing species composition is responsible for the Sr/Ca variations. The <12 μm fraction from the core top samples consists of intact coccoliths, fragments of coccoliths, and fragments of diatom opal. We do not observe any small foraminifera in our samples and estimate that fragments of foraminifera are not significant contributors to the calcite since many of the calcite fragments present can be identified as fragments of coccoliths. The coccolith assemblages in all samples consist of Calcidiscus leptoporus, Gephyrocapsa oceanica, Emiliania huxleyi, Umbilicosphaera sibogae, Helicosphaera spp., and Gladiolithus flabellatus (Figure 5b). C. leptoporus and G. oceanica are the most abundant taxa in all core top samples, together contributing 50–90% of the whole coccoliths (Figure 5b). There are slight variations in the relative abundance of different coccolithophorid taxa across the transects. In the 140°W transect, the relative abundance of E. huxleyi increases slightly around the equator while that of C. leptoporus decreases. In the 110°W transect, the relative abundance of E. huxleyi and G. flabellatus increases at 5°S and 3°N at the expense of C. leptoporus and G. oceanica.

Figure 5.

Sr/Ca of coccolith fraction and species composition data for core tops. (a) Sr/Ca data for coccolith carbonate. Symbols same as in Figure 4. (b) Relative abundance of species among intact coccoliths. (c) Percent of calcite contributed by each species, calculated as described in the text.

[24] According to our estimates, nearly all of the calcite in the coccolith fraction of the core top samples derives from C. leptoporus (50–80%) and G. oceanica (5–45%, Figure 5c). The contribution of calcite from C. leptoporus versus G. oceanica shows no consistent relationship with the Sr/Ca variations (Figure 5a). In general, there is no consistent correlation between the coccolith assemblage and the Sr/Ca of the coccolith fraction. This suggests that the variations in coccolith Sr/Ca across our core top transects are not primarily controlled by variations in the components of the coccolith assemblage.

4.3. Sr/Ca in Downcore Carbonates

[25] In downcore records from the equatorial Pacific, Sr/Ca in the coccolith fraction varies by nearly 20% over the last 250 ka with identical trends for the MNX cleaned and ethanol-rinsed samples (Table 2 and Figure 6b). For the 150 ky overlap between records at TT013-PC72 and MANOP C, Sr/Ca variations show similar minima at 18, 80, and 140–150 ka which correspond to maxima in the composite benthic δ18O curve [Martinson et al., 1987]. Variations in coccolith Sr/Ca are much larger than those of foraminifer G. tumida from TT013-PC72, and coccolith Sr/Ca does not covary with G. tumida Sr/Ca (Figure 6a) [Stoll et al., 1999].

Figure 6.

Sr/Ca and species composition data for downcore records. (a) Sr/Ca data of G. tumida from site PC72 in black symbols [Stoll et al., 1999]. Green curve is maximum variation in seawater Sr/Ca predicted by models of Stoll and Schrag [1998]. (b) Sr/Ca of coccolith carbonate; blue circles for ethanol rinse and red circles for reducing and ion exchange (“MNX”) cleaning. (c, d) Relative abundance of species and percent carbonate contribution by each species among intact coccoliths from selected samples in MANOP C. Legend same as in Figure 5. Stippled pattern shows glacial maxima as defined by benthic δ18O record for MANOP C [from Jasper et al., 1994].

[26] To assess whether the rapid shifts in coccolith Sr/Ca might be due to changes in the coccolith assemblage, we analyzed the species composition of six downcore samples spanning large shifts in coccolith Sr/Ca. The species composition of the six analyzed downcore samples is similar to that of core top samples in the 140°W transect (Figures 6c and 6d). C. leptoporus and G. oceanica are the most abundant taxa and contribute the majority of the calcite. No significant changes in the coccolith assemblages accompany the short-term Sr/Ca variations between 0 and 18 ka, 129 and 139 ka, and 170 and 195 ka. However, G. oceanica contributes a larger fraction of calcite in the older sediments (59%) than in the most recent sediments (29%). The older (>170 ka) sediments also exhibit the highest Sr/Ca ratios of the time series. Consequently, changes in species composition may contribute to longer-term changes in coccolith Sr/Ca. Alternatively, higher productivity may favor higher abundances of G. oceanica.

5. Interpretation of Variations in Coccolith Sr/Ca

5.1. Relation to Variations in Productivity

[27] Coccolith Sr/Ca in core top sediments generally covaries with modern measurements of primary productivity in the overlying surface waters in both transects (Figure 7), although higher coccolith Sr/Ca contrasts with low productivity at 12°S in the 140°W transect and 5°N in the 110°W transect. While primary productivity at these sites is dominated by picoplankton [Murray et al., 1994], growth rates for alkenone-producing haptophytes parallel primary productivity in the 140°W transect where alkenone data are available (Figure 7b) [Bidigare et al., 1997]. While not all coccolithophorids produce alkenones and other noncoccolithophorid haptophytes may contribute alkenones, the alkenone-estimated growth rates provide the best available approximation of coccolithophorid growth rates in field samples. These data suggest a possible relationship between coccolith Sr/Ca and coccolithophorid growth rate. The divergence of this relationship farther from the equator may be due to complications in the measurement of coccolith Sr/Ca in these two sediments due to contributions from noncarbonate sources, as discussed previously. Alternatively, or perhaps in addition, this divergence may indicate that coccolith Sr/Ca is additionally influenced by other factors.

Figure 7.

(a) Coccolith Sr/Ca from core top sediments (green squares) and long-term average productivity (blue circles; data from Chavez et al. [1990], Barber et al. [1991], and Chavez et al. [1998]) versus latitude. (b) Coccolith Sr/Ca (green squares) and alkenone-estimated coccolithophorid growth rates (red diamonds) versus latitude [Bidigare et al., 1997]. Open symbols denote sediments with high Sr and Ca contribution from noncarbonate fractions as discussed in the text.

[28] One complication in relating the chemistry of coccoliths in sediments to overlying surface water chemistry is that the coccolith assemblages may be significantly modified during deposition. For example, the coccolith assemblages in sediment core tops consist largely of dissolution-resistant species [Roth, 1994], and the relative abundance of species is different from that observed in shallow sediment traps from the same region. Sediment traps from 5°S at 140°W (1216 m) and from 12°S at 135°W (1292 m) are dominated by G. flabellatus (≈50%) with G. oceanica and F. profunda also contributing significantly to the assemblage (greater than ≈10% each [Broerse, 2000]). However, calculations of the amount of calcite contributed by each species indicate that G. oceanica contributes 42 and 30% of calcite at 12° and 5°S, respectively, while C. leptoporus contributes 7 and 14%, respectively. Extensive dissolution of fragile G. flabellatus has likely occurred in the sediments. Although there are no assemblage data for living plankton for the 140° or 110°W transects, a transect at 155°W was characterized by higher abundances of E. huxleyi and Gephyrocapsa spp. and much lower abundances of G. flabellatus around the equator, and a sharp transition to abundant Umbellosphaera spp. around 10°S [Honjo and Okada, 1974]. It is unclear if this difference reflects spatial variations in the Pacific or selective dissolution between the photic zone and the 1200 m sediment traps.

[29] Selective dissolution of coccoliths has enriched the core top sediments in the alkenone-producer G. oceanica. However, calcite is dominantly from C. leptoporus, which does not produce alkenones (H. Kinkel, personal communication, 2000). Nonetheless, C. leptoporus is likely to follow the same growth rate maximizing r selector strategy of alkenone producers E. huxleyi and G. oceanica and is likely to show similar growth rate variations across the upwelling zone [Young, 1994]. In addition, the export production of C. leptoporus covaries with that of E. huxleyi and G. oceanica in both the Gulf of California and San Pedro upwelling systems [Ziveri et al., 1995a, b; Ziveri and Thunnell, 2000]. Consequently, the core top Sr/Ca measurements for these sediments likely register the behavior of alkenone-producing coccolithophorids whose growth rates are estimated in the modern 140°W transect.

5.2. Relation to Variations in Calcification

[30] Several estimates of calcite production rates are available for the 140°W transect as a result of the Joint Global Ocean Flux Study Equatorial Pacific Process study carried out in 1992. In equatorial Pacific surface waters, coccolithophorids dominate carbonate production; counts of foraminifera and measurements of foraminiferal carbon in CTD casts [Stoecker et al., 1996] indicate that foraminifera did not contribute appreciable calcite to the particulate inorganic carbon (PIC) measurement at any of the stations. We compare coccolith Sr/Ca from core top sediments with the CaCO3 rain rate at 4000 m sediment traps and with estimated calcification rates in surface waters.

[31] The rain rate of CaCO3 recorded in sediment traps is one indicator of the rate of CaCO3 production by coccolithophorids in the equatorial Pacific. At 4000 m, the average CaCO3 rain rate was highest at the equator and decreased away from the equator, similar to the trends in coccolith Sr/Ca (Figure 8a). This suggests a possible relationship between coccolithophorid calcite production and coccolith Sr/Ca, although, as in the previous cases, the curves diverge at 12°S.

Figure 8.

Estimated growth and calcification rates in the 140°W transect. (a) Coccolith Sr/Ca (green circles) and average 1992 CaCO3 rain rate [Honjo et al., 1995] (red squares) in 4000 m sediment traps. (b) Coccolith Sr/Ca (green circles) and estimated calcification rates from incubation experiments (blue squares) and PIC measurements (violet diamonds) based on data by Balch and Kilpatrick [1996]. Units for calcification rate are g calcite C/m2 coccosphere surface area/day. Open symbols denote sediments with high Sr and Ca contribution from noncarbonate fractions as discussed in the text.

[32] Calcification rates were measured in August–September 1992 from 12°S to 12°N at depths of 0–120 m at 140°W via in situ and simulated in situ 14C incubation experiments [Balch and Kilpatrick, 1996]. Calcification rates derived from incubation experiments have high uncertainties: the correlation coefficient between simulated in situ and in situ measurements for each station is only 0.2. PIC and coccolithophorid surface area were also measured [Balch and Kilpatrick, 1996] and in combination with alkenone-estimated coccolithophorid growth rates [Bidigare et al., 1997] provide an additional estimate of calcification rate, assuming that growth rates for all calcifying species follow those of alkenone producers. Since sediments integrate calcification rates from all depths in the photic zone, we calculate the average calcification rate for each station, weighting the calcification rate/cell surface area for each depth by the fraction of calcite produced at that depth. Estimates derived from incubation and PIC data both show high calcification rates at the equator, decreasing away from the equator (Figure 8b). These trends are broadly similar to those in coccolith Sr/Ca, although there is more local variability in the calcification estimates, and the PIC estimate indicates high calcification rates per cell surface area at 12°S. The sediment Sr/Ca record may be slightly smoothed by lateral transport of carbonate rain or sediments and by seasonal migrations of the zone of most intense upwelling and maximum calcification rates. In addition, the calcification rate estimates may be noisy as indicated by the low reproducibility of the incubation measurements.

5.3. Relation to Environmental Data

[33] The variations in primary productivity and coccolithophorid productivity observed across the equatorial Pacific result from upwelling-driven variations in availability of macronutrients and micronutrients, CO2, temperature, and other environmental factors. Consequently, it can be difficult to distinguish whether variations in coccolith Sr/Ca result directly from the environmental variations or from the biological responses of coccolithophorids (e.g., varying growth rates) to these environmental variations. Here, we compare coccolith Sr/Ca with long-term averages of sea surface temperature (SST), surface nutrient concentrations, and dissolved CO2 concentrations (Figure 9).

Figure 9.

Coccolith Sr/Ca from core tops compared with variations in chemistry of overlying waters. (a) Sr/Ca in coccoliths, symbols same as in Figure 4. (b) Annual average SST. (c) Long-term average surface NO3. (d) Long-term average fCO2 (140°W) and sea-air CO2 flux in 1990 (110°W). Temperature, nutrient, and fCO2 for 140°W from Barber and Chavez [1991]; temperature and nutrient from 110°W from Levitus et al. [1994]. Sea-air CO2 flux for 110°W from Takahashi et al. [1997].

[34] In the 140°W transect, trends in coccolith Sr/Ca are generally inversely correlated with temperature and broadly covary with surface nutrient and dissolved CO2 concentrations. In the 110°W transect, there is no clear correlation between Sr/Ca and SST, nutrient concentrations, or dissolved CO2 concentrations.

[35] Temperature is an important control on Mg/Ca and Sr/Ca ratios in other biogenic calcites [e.g., Nurnberg et al., 1996; Beck et al., 1992; Cooper et al., 1999]. However, culture experiments indicate a direct, rather than inverse, relationship between temperature and Sr partitioning in calcite of planktonic foraminifera [Lea et al., 1999] and coccolithophorids (H. Stoll, unpublished data, 1997). An inverse relationship is observed in the Sr/Ca ratios of coccoliths from core tops at 140°W: Sr/Ca ratios are high at the equator where sea surface temperatures are depressed by upwelling of cold subsurface waters. Consequently, temperature cannot be the primary factor controlling coccolith Sr/Ca in core top sediments. Temperature may exert a secondary influence on coccolith Sr/Ca in our transects, partially attenuating the observed correlation between coccolith Sr/Ca and productivity near the equator. Elevated temperatures away from the upwelling zone may contribute to the higher Sr/Ca ratios at 12°S in the 140°W transect and at 5°N in the 110°W transect. Consequently, temperature may be an important influence on coccolith Sr/Ca and needs to be further investigated in culture studies. Although variations in nutrient and dissolved CO2 concentrations covary with Sr/Ca near the equator in the 140°W transect, the lack of relation in the 110°W transect suggests that these are not important direct controls on coccolith Sr/Ca ratios.

5.4. Controls of Sr Partitioning in Coccoliths

[36] In abiogenic experiments, increasing the precipitation rate over 2 orders of magnitude results in a fivefold difference in the effective Sr partitioning coefficient (Figure 10) [Lorens, 1981; Tesoriero and Pankow, 1996] due to surface enrichment effects [Watson, 1996] or surface reaction kinetics in general [Morse and Bender, 1990]. Although coccolithophorids exert a high degree of control over the timing and location of calcification [e.g., Young et al., 1992], similar thermodynamic and kinetic effects may occur. Coccolith Sr/Ca shows coherent trends across the equatorial Pacific upwelling zone which are similar to variations in the growth rates and calcite production of coccolithophorids in overlying surface waters. Growth and calcification rates of coccolithophorids are frequently correlated [Westbroek et al., 1993]. This observation, along with the strong dependence of Sr partitioning on calcification in abiogenic calcites, suggests that coccolithophorid growth (and calcification) rates may be an important control on Sr/Ca variations in coccoliths from core top sediments. This interpretation can be tested through culture studies where calcification, growth rate, temperature, and other factors can be reliably determined and compared with coccolith Sr/Ca ratios. Culture studies also will provide opportunities to investigate the extent to which the highly regulated biomineralization processes of coccolithophorids affects Sr partitioning differently from abiogenic calcite crystallization.

Figure 10.

Partitioning coefficient of Sr in abiogenic calcite (DSr) as a function of calcite precipitation rate (nmol calcite/mg calcite seed/minute). Data from Tesoriero and Pankow [1996].

6. Possible Paleoceanographic Applications

6.1. Coccolith Sr/Ca to Track Past Variations in Coccolithophorid Productivity and CaCO3 Rain Rate

[37] The relative production rates of carbonate-producing and noncarbonate-producing (frequently siliceous) organisms are important in setting the pH of the deep ocean and consequently the ocean-atmosphere partitioning of CO2. Extensive modeling experiments indicate that changes in the rain ratio of carbonate to organic carbon remains one of the “front-runners” in explaining glacial/interglacial atmospheric PCO2 cycles [Archer et al., 2000]. However, rigorous evaluation of this hypothesis has been limited by difficulties in modeling ecological partitioning of carbonate versus silica-producing phytoplankton in the ocean and by lack of proxy data for past variations in the productivity of these groups. Because the partitioning of planktonic functional types in the ocean is poorly understood, it is difficult to model how their distributions and relative productivities may have varied in response to changing environmental conditions over glacial cycles. If a reliable proxy for past variations in coccolithophorid productivity were identified, it could contribute significantly to the resolution of this dilemma.

[38] If coccolithophorid growth and calcification rates are the predominant control on coccolith Sr/Ca ratios, downcore Sr/Ca ratios might be used to infer past changes in coccolithophorid productivity. The trends in equatorial Pacific core tops suggest a relationship between coccolith Sr/Ca and coccolithophorid productivity, but culture experiments and a wider survey of coccolith Sr/Ca in other regions are needed to confirm these relationships. In particular, more work is needed to assess the extent to which temperature may also affect Sr partitioning in coccolith calcite. In addition, to be a reliable proxy of past coccolithophorid productivity, downcore coccolith Sr/Ca records must not be significantly biased by variations in dissolution intensity over time, either through selective dissolution effects on a constant species assemblage or changes in the species assemblage. Changes in the coccolith assemblages, including changes in the dominance of different phenotypes, must be minor or the effects of these changes must be well understood. Finally, it is important to consider potential variations in the Sr/Ca ratio of seawater. Fortunately, the long residence times of Sr and Ca in the ocean greatly attenuate variations in seawater Sr/Ca over glacial/interglacial timescales. Numerous modeling experiments indicate that Sr/Ca likely varies by <3% over Quaternary glacial cycles [Stoll and Schrag, 1998; Stoll et al., 1999].

[39] If the relationships between coccolithophorid productivity and coccolith Sr/Ca in equatorial Pacific core tops are general and temperature variations do not overwhelm the signal, then our downcore Sr/Ca records may reflect past variations in coccolithophorid productivity. In these records, coccolith Sr/Ca does not covary with changes in dissolution intensity calculated as the composite preservation index at an adjacent site (WEC8803B-GC51 [LaMontagne et al., 1996]). We have shown that the rapid shifts in coccolith Sr/Ca are not due to changes in the coccolith assemblage. However, it is possible that the long-term decrease in the dominance of C. leptoporus relative to G. oceanica influences long-term trends in Sr/Ca, although there is no consistent relationship between coccolith Sr/Ca and the relative abundances of these two taxa in the core top transects. Downcore Sr/Ca variations of 20% significantly exceed predicted changes of 1–3% in the Sr/Ca ratio of seawater over the last 250,000 years. Consequently, the downcore coccolith Sr/Ca records from the equatorial Pacific may indicate times of low coccolithophorid productivity during glacial maxima and around 80 ka (Figure 6). In contrast, accumulation rates of organic carbon, opal, and barite all suggest higher overall productivity in the equatorial Pacific during glacial maxima [Lyle et al., 1988; Paytan et al., 1996], although these estimates may be variably biased by errors in accumulation rates caused by carbonate dissolution cycles and sediment focusing. Nonetheless, if the organic carbon accumulation data from the equatorial Pacific are not biased by preservation or problems in the calculation of accumulation rates, then lower coccolith Sr/Ca ratios may suggest that during the last two glacial maxima high organic carbon production combined with decreased calcification led to an increased Corg:Cinorg particle flux in this region.

[40] Relationships between coccolith Sr/Ca and coccolithophorid productivity may also be useful on longer timescales, for investigating the response of this important group of primary producers to major environmental changes in the past. In the Cretaceous, for example, Sr/Ca ratios of coccolith-dominated carbonates appear to respond to environmental changes resulting from major pulses of volcanic activity [e.g., Stoll and Schrag, 1999; H. M. Stoll and D. P. Schrag, Sr/Ca variations in Cretaceous carbonates: Relation to productivity and sea level changes, submitted to Palaeogeography Palaeoclimatology Palaeoecology, 1999]. However, for longer-term studies, the possibility of changes in seawater Sr/Ca ratios must also be considered.

6.2. Coccolith Sr/Ca to Correct for Growth Rate Contribution to εp for Estimating Paleo-pCO2 (aq)

[41] Estimation of past concentrations of dissolved CO2 in the ocean is another approach used to understand past changes in the carbon cycle. To study the role of the ocean in glacial/interglacial atmospheric PCO2 variations, workers attempt to estimate past air-sea CO2 fluxes in different regions and then surmise about the underlying processes which control changes in these fluxes over glacial cycles [e.g., Jasper et al., 1994]. On longer timescales, workers have also attempted to estimate atmospheric PCO2 from calculations of dissolved CO2 for regions of the ocean presumed to be in equilibrium with the atmosphere [e.g., Pagani et al., 1999]. Estimates of past dissolved PCO2 are made from measurements of carbon isotopic fractionation in algal organic matter, which depends on dissolved CO2 concentrations, along with algal growth rate and cell geometry [Popp et al., 1998]. Through the use of taxon-specific biomarkers, it is possible to measure fractionation in compounds derived from a single cell geometry, although the precise surface area/volume relationship may vary with cell size. However, the lack of an independent proxy for past variations in algal growth rates has limited estimation of past dissolved CO2 concentrations using this approach.

[42] If coccolith Sr/Ca varies with coccolithophorid growth rates, then coccolith Sr/Ca would be particularly suited for correcting growth rate effects on carbon isotope fractionation in alkenones because coccoliths and alkenones both derive from the same group of organisms. As discussed in section 6.1, the relationships between coccolith Sr/Ca and growth rate need to be confirmed through a broader survey of coccolith Sr/Ca in core tops and ideally through culture experiments where both coccolith Sr/Ca and ɛp (isotopic fractionation between coccolithophorid biomass and dissolved CO2) are monitored. The influence of temperature on Sr partitioning in coccoliths also must be assessed. Likewise, in downcore records the potential biases of selective dissolution and changing species assemblages must be addressed. Because there are strong evolutionary variations within coccolithophorids [e.g., Bollmann, 1997; Knappertsbusch et al., 1997] the relationship between Sr/Ca and growth rates may change through time, even within the same species. However, such changes may also affect the relationship between carbon isotopic fractionation and dissolved CO2. A further caveat is that the organic biomarkers derive from E. huxleyi and Gephyrocapsa which are not the only contributors to coccoliths in the sediments.

[43] In the case of the downcore records from the equatorial Pacific, alkenone producer G. oceanica is a major contributor of coccolith calcite, and the other important species, C. leptoporus, has a similar ecology in the modern ocean. It is difficult to evaluate the significance of changing relative proportions of these species on the coccolith Sr/Ca record. The downcore increase in G. oceanica may partly drive the slight downcore increase in Sr/Ca, or environmental conditions favoring higher coccolithophorid productivity (and higher coccolith Sr/Ca) further back in the past may also have favored a higher abundance of G. oceanica. These factors complicate the interpretation of the equatorial Pacific coccolith Sr/Ca records in terms of coccolithophorid growth rates. However, as an exercise, we consider the effect of a Sr/Ca-based growth rate correction on calculated dissolved CO2 concentrations from the equatorial Pacific for the last 250,000 years.

[44] We use the downcore Sr/Ca record from MANOP C site and the alkenone carbon isotopic fractionations measured in the same core by Jasper et al. [1994]. We roughly calibrated the relationship between coccolith Sr/Ca and coccolithophorid growth rate using two approaches. In the first approach, we compared estimated average cell-specific growth rates of modern alkenone-producing haptophytes from 5°N to 5°S in the 140°W transect (data from Bidigare et al. [1997]) with Sr/Ca of core top coccoliths from the same latitude (Figure 11, top). A second approach used to calibrate the relation between Sr/Ca and growth rate takes advantage of a strong correlation observed between PO4 concentrations and the slope of the relationship between ɛp which presumably reflects variations in coccolithophorid growth rate [Bidigare et al., 1997]. We compare PO4 data from 1992 JGOFS surveys (C. Garside, personal communication, 1999) in the 140°W transect with core top Sr/Ca from the same latitude (data from 5°N to 5°S only; Figure 11, bottom). Because there are a small number of data points, and only a single point at low growth rates and PO4 concentrations, there is considerable uncertainty to these calibrations. More stable calibrations must be extracted from other regions and culture studies for more reliable inference of past variations in coccolithophorid growth rates.

Figure 11.

(top) Correlation between coccolith Sr/Ca and alkenone-estimated growth rates [Bidigare et al., 1997] between 5°S and 5°N in 140°W transect. (bottom) Correlation between coccolith Sr/Ca and average surface PO4 [from Bidigare et al., 1997] between 5°S and 5°N in 140°W transect.

[45] For the growth rate calibration, we apply the relation:

display math

from culture experiments by Bidigare et al. [1997]. For the PO4 calibration, we use the relationships from Bidigare et al. [1997]:

display math
display math

[46] Following Jasper et al. [1994], we assumed that variations in the δ13C of DIC followed those of Neogloboquadrina dutertrei. Spero et al. [1997] have shown that carbon isotope fractionation in foraminiferal calcite depends on the carbonate ion concentration of seawater. However, we calculate that the predicted variation in δ13C of N. dutertrei due to the carbonate ion effect (δ13C shifts 0.3–0.8 permil for constant δ13C of DIC) has a negligible effect on the calculated paleo-pCO2 (aq). Likewise, the small changes in seawater Sr/Ca ratios over glacial cycles [Stoll and Schrag, 1998] also have a negligible effect on coccolith Sr/Ca ratios and on calculated paleo-pCO2 (aq). Paleo-pCO2 (aq) calculated with both growth rate and phosphate calibrations are similar, although pCO2 (aq) calculated with the growth rate calculation are consistently 50 ppm higher. These results are compared with the paleo-pCO2 (aq) calculated using the approach of Jasper et al. [1994] assuming constant algal growth rates (Figure 12).

Figure 12.

Calculated dissolved CO2 concentrations for MANOP C site for last 250 ka. (a) Calculation of Jasper et al. [1994] assuming constant algal growth rates. (b) Calculation correcting for variations in algal growth rates using Sr/Ca variations. (c) Atmospheric CO2 record from Vostok ice core [Petit et al., 1999]. Black lines indicate data smoothed at 16 ka intervals.

[47] There are small but significant differences between the paleo-pCO2 (aq) records calculated assuming constant algal growth rates and those calculated using Sr/Ca to correct for likely changes in algal growth rates (Figure 12). The correlation coefficient (Spearman) of the two records is 0.71, which is statistically different from zero at the p=0.000 level. Differences between the two records are greatest between 60 and 120 ka and 170–250 ka. During the latter interval, the pCO2 (aq) estimates exceed those of any other interval for the growth-rate-corrected record, whereas the estimates in the uncorrected record are comparable with those of other interglacial intervals. The higher calculated pCO2 (aq) in this older part of the record may be due to a shift in the Sr/Ca versus growth rate relation or the influence of temperature changes on coccolith Sr/Ca. It is also possible that algal growth rates were indeed higher but that average cell size of alkenone producers was smaller, leading to overestimation of pCO2 (aq) from the ɛp relation (H. Kinkel, personal communication, 2000). Detailed study of changes in coccolithophorid assemblages in nearby Deep Sea Drilling Program site 573 (0.5°N, 133.3°W [Pujos, 1985]) indicates that prior to 150 ka, the small Gephyrocapsa protohuxleyi and G. aperta were more abundant than the medium size G. oceanica. This change in the size of alkenone-producing cells may lead to an overestimation of pCO2 (aq) from ɛp. If growth rates were higher during this interval, as suggested by Sr/Ca, the overestimate of pCO2 (aq) from equation imagep would have been partially compensated by the assumption of constant growth rates.

[48] By comparing calculated records of pCO2 (aq) with the Vostok record of atmospheric PCO2 variations, it is theoretically possible to evaluate changes in the air-sea CO2 flux from the equatorial Pacific. However, differences in the resolution of the two types of records and uncertainties in their correlation complicate this process. Overall, both records of Pacific pCO2 (aq) show elements of the ∼100 ka cyclic variation in atmospheric PCO2, indicating some covariation between Pacific pCO2 (aq) and atmospheric PCO2. The growth-rate-corrected pCO2 (aq) calculation matches the timing of interglacial PCO2 maxima in the Vostok atmospheric PCO2 curve at with local maxima at 0, ∼130, and ∼240 ka [Petit et al., 1999], whereas the uncorrected curve places the stage 5 maximum at ∼80 ka instead of ∼130 ka. However, it is difficult to statistically discern which record is more correlated with the Vostok curve and whether the implied air-sea CO2 disequilibria are meaningful. For curves with 16 ka smoothings which minimize problems of age correlation, the correlation coefficient of the uncorrected pCO2 (aq) curve with the Vostok curve (r=0.60) is statistically indistinguishable (p≫0.1) from the correlation coefficient of the corrected curve (r=0.55).

[49] In order to fully utilize the potential of coccolith Sr/Ca to produce more reliable records of past pCO2 (aq) from ɛp, the relationships between coccolith Sr/Ca and growth rate need to be confirmed through a broader survey of coccolith Sr/Ca in core tops and ideally through culture experiments where both coccolith Sr/Ca and εp are monitored.

7. Conclusions

[50] 1. Sr/Ca ratios of polyspecific fractions of coccolith carbonate separated from core tops in the equatorial Pacific vary by 15% across the equatorial upwelling front. These variations are much larger than the very limited (<2%) variations in the Sr/Ca ratio of modern seawater and must reflect variations in Sr partitioning in coccolith calcite. Sr/Ca ratios of coccolith fractions from the equatorial Pacific are much higher and more variable than Sr/Ca ratios of planktonic foraminifera.

[51] 2. Coccolith Sr/Ca covaries with primary productivity near the equator in both the 110°W and 140°W transects studied. In the 140°W transect where data are available, variations in coccolith Sr/Ca are similar to those of alkenone-estimated growth rates of coccolithophorids in overlying surface waters and variations in the CaCO3 rain rate in deep sediment traps. These relationships suggest that coccolith Sr/Ca in these core tops may depend primarily on coccolithophorid growth and calcification rates, which are commonly linked in coccolithophorids. However, temperature may also influence Sr partitioning in coccoliths, partially attenuating the relationships between coccolith Sr/Ca and coccolithophorid productivity.

[52] 3. In this case, changing species composition was not likely to be the dominant influence on coccolith Sr/Ca. Dissolution strongly modified the composition of the species assemblage and selected for dissolution-resistant species across both transects, resulting in similar sediment assemblages. However, in shallower sediments and those with higher coccolith diversity, changes in coccolith species assemblages may exert a larger influence on coccolith Sr/Ca. In addition, in most of our samples, noncarbonate contributions of Sr and Ca were insignificant. However, samples with a high detrital or hydrothermal content may require additional investigation of cleaning procedures to yield meaningful data on the Sr/Ca of coccolith carbonate.

[53] 4. The relationships between coccolith Sr/Ca and coccolithophorid growth and calcification rates suggest that coccolith Sr/Ca may be a valuable tool for investigating past changes in coccolithophorid productivity which may provide key data on changes in the carbon and carbonate cycles in the past. One advantage of coccolith Sr/Ca records is that unlike most other paleoproductivity proxies, their application in paleoproductivity studies does not require estimation of sediment accumulation rates, advantageous because accumulation rates are difficult to measure precisely and may be biased by sediment focusing. While coccolith Sr/Ca may provide information only about past variations in coccolithophorid productivity, such information may have important applications both in the determination of past variations in the rain ratio of organic to inorganic carbon [e.g., Archer and Meier-Reimer, 1994; Archer et al., 2000] and in improving estimates of past CO2 (aq) from carbon isotope fractionation in coccolithophorid biomarkers [e.g., Jasper et al., 1994].

[54] 5. Culture studies are needed to confirm the relationships suggested by this study and optimize possible paleoceanographic applications. In cultures the relationships between coccolith Sr/Ca and coccolithophorid calcification and growth rates and the effect of temperature on Sr partitioning can be reliably measured. Sr partitioning can also be investigated in different species.

Acknowledgments

[55] This material is based upon work supported by the North Atlantic Treaty Organization under a Grant awarded in 1998 (DGE-98-04555 to H.M.S.) and by the II Plan Regional de Investigacion del Principado de Asturias (H.M.S.). Critical and insightful readings by Hanno Kinkel, Jeremy Young, and Patrizia Ziveri greatly improved this manuscript. We thank David Lea and an anonymous reviewer for numerous suggestions. We thank Ricardo Anadon and Rick Murray for useful discussions and Markus Geison and Patrizia Ziveri for assistance with taxonomy and ecology of coccolithophorids. We appreciate the skillful management of the Harvard ICP-ES provided by Ethan Goddard and acknowledge the Servicio Comun de Estadistica of the University of Oviedo for advice on statistical analyses of time series. Samples were provided by repositories at the University of Rhode Island Graduate School of Oceanography (OCE-9711464) and Oregon State University (OCE-97-12024).

Ancillary