Toward better paleocarbonate ion reconstructions: New insights regarding the CaCO3 size index



[1] The reconstruction of paleocarbonate ion concentrations provides an important constraint on the contribution of the CaCO3 cycle to the decrease in atmospheric CO2 content during glacial time. Such reconstructions have been challenging because each of the existing paleo-[CO32−] indices has serious limitations. In this study, we reexamine the Broecker-Clark CaCO3 size index by analyzing the <20 μm, 20 to 38 μm, and 38 to 63 μm fractions in sediments from the Ontong-Java Plateau and the Ceara Rise. Scanning electron microscope analyses demonstrate that the less than 20 μm CaCO3 is dominated by coccoliths and the greater than 20 μm CaCO3 is dominated by foraminifera. Our results clearly indicate that the coccoliths are far more resistant to dissolution than the foraminifera. Referenced to a core top sample from 2.31 km depth in a core top sample from 4.04 km depth on the Ontong-Java Plateau, ∼70% of the foraminifera CaCO3 was dissolved as opposed to only ∼7% of the coccolith CaCO3. We found that the dissolution of foraminifera shells did not produce a significant amount of fragments smaller than 63 μm in size, and thus the Broecker-Clark size index is not a measure of the extent of fragmentation. Rather, it is a measure of the extent of differential dissolution of foraminifera relative to coccoliths. On the basis of these results, we propose a new dissolution index which involves the ratio of dissolution-susceptible foraminifera CaCO3 to total CaCO3.

1. Introduction

[2] Despite much effort, the cause for the decreased atmospheric CO2 content during glacial time remains unclear. Two mechanisms stand out in the many papers that have been written on this subject. One involves an increase during glacial time in the strength of the biological pump. Marine phytoplankton utilize CO2 to build their soft tissues. When they die, some of the organic matter sinks into the sea's interior before being eaten. In this way, the biological pump reduces the amount of CO2 in surface water and hence also in the atmosphere. The other mechanism involves CaCO3 storage in marine sediments and coral reefs. The CaCO3-building marine organisms utilize both the dissolved inorganic carbon and calcium to make their tests and skeletons. The overall result is a decrease in the CO32− concentration and an increase in CO2 concentration. Hence a decrease in CaCO3 storage during glacial time would have drawn down the atmosphere's CO2 content. Proxies help to constrain the contribution of each mechanism. In the case of the biological pump, it is the difference between δ13C in the planktonic and benthic foraminifera [Charles and Fairbanks, 1990]. In the case of CaCO3 storage, it is the CO32− concentration in the deep sea. In this paper, we proposed an improved method for reconstructing the latter.

[3] The CaCO3 content of sediments has been used to define the so-called calcite compensation depth (i.e., the CCD). Criteria such as “20% CaCO3” [Van Andel, 1975] and “10% CaCO3” [Farrell and Prell, 1989] have been used to define this depth. One drawback of this method is that the relationship between CaCO3 content and the fraction of CaCO3 dissolved is highly nonlinear [Broecker, 2003]. If, as usually the case in open ocean sediments, the rain rate of CaCO3 is much higher than that of noncarbonate, a significant amount of CaCO3 must be dissolved before a noticeable reduction in the CaCO3 content occurs. In addition, no matter what criterion is selected, this approach fails because the initial CaCO3 content may have differed during glacial time [Broecker, 2008].

[4] Recently, several dissolution indices have been proposed to indicate past changes in the deep sea carbonate ion concentration. The weight of size-classed foraminifera shells was demonstrated to reflect the extent of dissolution [Lohmann, 1995; Broecker and Clark, 2001a]. However, the initial shell weight appears to be influenced by growth conditions, in particular by surface water carbonate ion concentration [Barker and Elderfield, 2002]. The Broecker-Clark CaCO3 size index [Broecker and Clark, 1999], which is based on the ratio of coarse CaCO3 (>63 μm size fraction) to the total CaCO3 is another. It is presumed to be a measure of the degree of breakup of foraminifera shells due to weakening by dissolution, and it has been shown to linearly correlate to the pressure-normalized carbonate ion concentration (CO32−*) in the tropical oceans [Broecker and Clark, 1999]. Unfortunately, the initial composition of the biogenic CaCO3 during glacial time appears to have been different from that during the Holocene [Broecker and Clark, 2001b].

[5] In addition to the above mentioned physical measurements, two chemical approaches and one isotopic approach have been explored. Marchitto et al. [2005] show that combined cadmium and zinc measurements in benthic foraminifera shells provide a measure of carbonate ion concentration and Yu and Elderfield [2007] have shown that this is also the case for the boron concentration in these shells. Sanyal et al. [1995] utilized the ratio of 11B to 10B in benthic foraminifera to constrain the paleo-pH (and hence also the paleocarbonate ion concentration) of deep ocean water.

[6] In this study, we investigated the size distribution of CaCO3 grains in the <63 μm size fraction and in this way have identified a problem with the Broecker-Clark size index. On the basis of these results we propose a revised index for the dissolution status of CaCO3 and hence a new strategy for the reconstruction of down core paleocarbonate concentrations.

2. Method

2.1. Material

[7] Sediment samples used in this study were obtained from a Ceara Rise jumbo piston core (JPC), EW9209-JPC3 (5.3°N, 44.3°W, 3.29 km) and from Ontong-Java Plateau giant gravity cores (GGC) and box cores (BC): EW91-9-GGC15 (0.0°, 158°E, 2.31 km), EW91-9-GGC55 (0.0°, 162°E, 4.04 km), and EW91-9-BC56 (0.0°, 162°E, 4.04 km). The time intervals sampled represent the late Holocene and the LGM for all these cores. In addition, one sample from the Ontong-Java Plateau core EW91-9-BC56 was selected to represent the early Holocene preservation event (∼8 ka).

2.2. Experimental Procedure

2.2.1. Sample Dissociation and Size Separation

[8] In our pilot experiment, 1 to 2 g dry sediments were weighed and then wet sieved in tap water through a series of sieves (mesh: 63 μm, 38 μm, and then 20 μm), and thus separated into four size fractions (>63 μm, 63 to 38 μm, 38 to 20 μm, and <20 μm). All the fractions were oven dried and then weighed. The finest fraction (<20 μm) was allowed to settle for days (3 ∼ 4 d) and in some cases centrifuged (for ∼3 h) to collect the suspended particles before oven drying. Note that the sieves have to be cleaned ultrasonically between samples because some sediment grains become trapped in the openings. For a 20 μm mesh sieve, 3 h ultrasonic cleaning (frequency output: 40 kHz) was proven to be optimal. The effectiveness of ultrasonic cleaning is confirmed by direct inspection of the sieve grids using an optical microscope.

2.2.2. Scanning Electron Microscope Observation

[9] An aliquot of each size fraction was then mixed with distilled water in a small glass vial. The concentration was then adjusted to achieve an optimum dispersion on a glass slide within a circumscribed circular area equivalent to that of a “stub” (specimen mount for scanning electron microscopic observation). The glass slide was observed using a Nikon Diaphot inverted microscope with phase contrast optics. A new aliquot of the optimum dispersion was filtered through a 13 mm diameter filter membrane (Millipore white GSWP01300, 0.22 μm). The membrane was mounted on a stub, coated with gold using a Denton Desktop sputter coater, and then imaged using a scanning electron microscope (SEM) (LEO/ZEISS 1455 SEM) to view the major components of each size fraction. For particles smaller than 20 μm, we observed and imaged 15 areas of about 130 μm × 180 μm on each SEM stub. For particles in the 20 to 38 μm fraction, 15 areas of about 270 μm × 350 μm were observed on each stub. For particles in the 38 to 63 μm fraction, 8 to 15 areas of about 1,100 μm × 1,500 μm were observed on each stub. For the >63 μm fraction the particles were large enough to be viewed with a stereomicroscope.

2.2.3. CaCO3 Content Measurement

[10] Aliquots (5 to 6 mg) of each fraction were analyzed for CaCO3 content by coulometry (UIC CM 5130 Acidification Module). The >63 μm fraction was crushed and homogenized before analysis.

3. Results

[11] On the basis of SEM imaging the dominant CaCO3 component of the <20 μm fraction in the sediments are coccoliths (Figures 1a1b1c1d). Siliceous biogenic fragments, dinoflagellates, and lithogenic grains are also present in this size fraction, but the amount of foraminifera CaCO3 is negligible. The distinction between foraminifera fragments and other components was based on the surface structure observation and comparison to fragments in larger size fractions. In the 20 to 38 μm fraction, fragments of foraminifera, diatoms and radiolaria are the major components. In the 38 to 63 μm fraction, juvenile foraminifera and some broken foraminifera shells are found. Thus, the finest mesh available (i.e., the 20 μm wire cloth test sieve; Newark ASTM E-11) proves to be a suitable cutoff between foraminifera and coccoliths. This is in accord with Frenz et al. [2005], who concluded that a sieving at 20 μm level was an effective way to separate coccoliths and foraminifera.

Figure 1a.

Example of SEM image of the <20 μm sediments in this study: MW91-9-GGC15 (10 to 11 cm), late Holocene. Coccoliths are the dominant CaCO3 component in the <20 μm fraction.

Figure 1b.

Example of SEM image of the <20 μm sediments in this study: MW91-9-GGC15 (42 to 43 cm), LGM. Coccoliths are the dominant CaCO3 component in the <20 μm fraction.

Figure 1c.

Example of SEM image of the <20 μm sediments in this study: MW91-9-BC56 (3 to 4 cm), late Holocene. Coccoliths are the dominant CaCO3 component in the <20 μm fraction.

Figure 1d.

Example of SEM image of the <20 μm sediments in this study: MW91-9-GGC55 (22 to 23 cm), LGM. Coccoliths are the dominant CaCO3 component in the <20 μm fraction. Note that although the sample from the LGM at water depth 4.04 km has experienced severe dissolution, the coccoliths appear to be fairly well-preserved.

[12] On the basis of the qualitative observation of the surface structure and morphology under SEM, it appears that coccoliths are better preserved than foraminifera. This is particularly obvious in cores where dissolution is severe.

[13] The dry weight and the measured CaCO3 content of each size fraction for all the samples are listed in Table 1. We measured the dry weights of the material both before and after wet sieving. The sample recovery was always greater than 94% (see Table 1).

Table 1. Dry weight, CaCO3 Content, and Related Data for Each Size Fraction of the Sediment Samples From the Ontong-Java Plateau and Ceara Risea
 Ontong-Java PlateauCeara Rise
Water Depth 2.31 kmWater Depth 4.04 kmWater Depth 3.29 km
MW91-9-GGC15 (0.0°, 158°E), Late HolocenebMW91-9-GGC15 (0.0°, 158°E), LGMcMW91-9-BC56 (0.0°, 162°E), Late HolocenedMW91-9-BC56 (0.0°, 162°E), Early HoloceneeMW91-9-GGC55 (0.0°, 162°E), LGMfEW9209-JPC3 (5.3°N, 44.3°W), Late HolocenegEW9209-JPC3 (5.3°N, 44.3°W), LGMh
  • a

    Depths in cores were chosen to reflect time intervals around the last glacial maximum (LGM), late Holocene, and early Holocene.

  • b

    Depth in core is 10–11 cm.

  • c

    Depth in core is 42–43 cm.

  • d

    Depth in core is 3–4 cm.

  • e

    Depth in core is 13–14 cm.

  • f

    Depth in core is 22–23 cm.

  • g

    Depth in core is 8–9 cm.

  • h

    Depth in core is 70–71 cm.

>63 μm Fraction
Dry weight, g0.7890.6510.3130.5200.2020.5640.138
Contribution to total weight, %4537223011338
CaCO3 content, %95949093948966
Contribution to total CaCO3, %49422634144326
38 to 63 μm Fraction
Dry weight, g0.1490.1340.0630.0870.0610.1220.044
Contribution to total weight, %9845373
CaCO3 content, %90918189898751
Contribution to total CaCO3, %9855497
20 to 38 μm Fraction
Dry weight, g0.1040.0920.0450.0570.0570.0980.038
Contribution to total weight, %6533362
CaCO3 content, %89878090848031
Contribution to total CaCO3, %6534473
<20 μm Fraction
Dry weight, g0.7080.8790.9911.0641.4370.9391.511
Contribution to total weight, %40507062825487
CaCO3 content, %77747376745315
Contribution to total CaCO3, %36446657794264
Bulk Sediments
Before sieving, g1.8041.8031.5041.8331.8381.7861.805
After sieving, g1.7501.7561.4131.7281.7571.7231.731
Sample recovery, %97979494969696
CaCO3 content (calculated), %87837882786920
Broecker-Clark CaCO3 size index0.490.420.260.340.140.430.26
Index of Foram Fragmentation
(Juveniles + fragments) ÷ (whole shells)0.300.330.310.270.540.370.38
Revised Size Index
F ÷ (F + C)0.640.560.340.430.210.580.36

3.1. Rain Ratio of Foraminifera to Coccolith CaCO3

[14] As summarized in row 16 in Table 1, at 2.31 km on the Pacific's Ontong-Java Plateau, while coccoliths calcite constituted 36% of the total CaCO3 in the Holocene sample, it contributed 44% to the LGM sample. On the Atlantic's Ceara Rise at 3.29 km, while coccolith calcite constituted 42% of the CaCO3 in the Holocene sample, it contributed 64% to the LGM sample. As both sites are shallow enough to have experienced only minor dissolution, this confirms the prediction by Broecker and Clark [2001b] that the coccolith to foraminifera rain rate ratio was higher during the LGM than during the Holocene.

3.2. Comparison Between Shallow and Deeper Cores on the Ontong-Java Plateau

[15] In this study, we sampled three sediment cores from the Ontong-Java Plateau: MW91-9-GGC15 was retrieved from 2.31 km water depth; MW91-9-GGC55 and BC56 (companion cores from the same location) were retrieved from 4.04 km water depth. In order to understand how dissolution affects the sedimentary CaCO3 record, we compare results from these two water depths assuming that the CaCO3 rain rates are the same because of their proximate locations. We use the core at 2.31 km as a reference for the estimation of the excess dissolution at 4.04 km.

[16] As summarized in row 4 in Table 1, while the late Holocene sample from 2.31 km depth (column 1 in Table 1) has 49% of its total CaCO3 in the >63 μm fraction, the late Holocene sample from a depth of 4.04 km (column 3 in Table 1) has only 26% of the total CaCO3. The opposite is the case for the <20 μm fraction (row 16 in Table 1). The 2.31 km sample has 36% of the total CaCO3 and the 4.04 km sample has 66%. Hence the ratio of >63 μm to <20 μm CaCO3 drops from 1.36 at 2.31 km to 0.39 at 4.04 km. The situation for the 20 to 38 μm and 38 to 63 size fractions is similar to that for the >63 μm fraction. These differences point clearly to a far greater extent of dissolution for foraminifera than for coccoliths, consistent with the Frenz et al. [2005] conclusion based on grain-size analyses of foraminifera CaCO3 at various water depths.

[17] In order to quantify the magnitude of excess dissolution of foraminifera and of coccolith CaCO3 at 4.04 km relative to 2.31 km, we first need to estimate the difference in accumulation rate. At 4.04 km, an accumulation rate of 1.5 cm/ka was derived from the 9295 years 14C age at 14 cm depth (W. S. Broecker and E. Clark, Broecker and Clark planktonic foraminifera weight database, available at At 2.31 km an accumulation rate of 3.1 cm/ka was derived from the 9650 years 14C age at 30 cm depth (Broecker and Clark planktonic foraminifera weight database, available at We further assume that the two cores have the same bulk density (i.e., ∼1 g/cm3). Using the CaCO3 content of the bulk sediments and contribution to total CaCO3 of each size fraction (see Table 1) along with the estimated accumulation rates, we calculated that about 70% of >20 μm CaCO3 (i.e., the foraminifera calcite) but only about 7% of <20 μm CaCO3 (i.e., the coccolith calcite) present in the 2.31 km depth core has been dissolved in the 4.04 km core. This confirms our earlier qualitative SEM observation that coccoliths are far more resistant to dissolution than foraminifera.

[18] Consistent with the finding by Broecker et al. [1999], for the 8 ka preservation maximum at 4.04 km, the ratio of foraminifera calcite to coccolith calcite was higher than that for the late Holocene and that for the LGM (see Table 1).

[19] In contrast, during the LGM dissolution at 4.04 km is even more pronounced. At 2.31 km, the >63 μm size fraction housed 42% of the total CaCO3 while at 4.04 km, it housed only 14% (row 4 in Table 1). The <20 μm sample at 2.31 km contained 44% of the total CaCO3 while that at 4.04 km it contained 79% (row 16 in Table 1).

3.3. A New Carbonate Dissolution Index

[20] We propose a revised size index:

display math

[21] It represents the ratio of dissolution-susceptible CaCO3 (foraminifera) to total CaCO3 in the sediments. Thus, as in the case for the original size index, lower values for the revised index are expected for more dissolved samples. Of course, the revised index has a flaw similar to that which plagued the original size index, namely that the initial composition of the biogenic CaCO3 changes with climate. This difficulty, however, can be partially overcome by referencing the results to those for a core at a water depth shallow enough to be subject to minimal dissolution.

[22] As in most cases the amounts of CaCO3 in the 20 to 38 μm and 38 to 63 μm fractions are modest, the new index has numerical values only 30% or so higher than that of the Broecker-Clark size index. The important point is that the revised index emphasizes the new finding that changes in the index have little to do with fragmentation. Rather, they reflect the far higher dissolution tendency for foraminifera calcite relative to coccolith calcite.

4. Conclusions

[23] On the basis of the results of sediment samples from the Ontong-Java Plateau and the Ceara Rise, we conclude that (1) the dominant CaCO3 component of the <20 μm fraction in the sediments are coccoliths, (2) coccoliths dissolve far more slowly than foraminifera do, (3) the rain rate of coccoliths relative to that of foraminifera was higher during the LGM than the Holocene, and (4) fragments and foraminifera whole shells dissolve at comparable rates.


[24] D. McCorkle and W. Curry (WHOI) provided Ontong-Java Plateau samples (MW91-9-GGC15, MW91-9-GGC55, and MW91-9-BC56) and Ceara Rise samples (EW9209-JPC3), respectively. The 14C ages were determined at the ETH/PSI AMS facility in Zurich by I. Hajdas. E. Clark (LDEO) assisted with experimental setup. O. R. Anderson (LDEO) and R. Harniman (Columbia University) generously provided laboratory facilities and assistance on SEM preparation and imaging. O. R. Anderson (LDEO) and T.-N. Yang (Institute of Earth Sciences, Academia Sinica, Taiwan) gave helpful suggestions on handling and identification of nannofossils. P. Malone (LDEO) assisted with coulometry measurements. S. Baker (Cardiff University, U. K.) commented on data interpretation. This work was supported by an Abrupt Climate Change Fellowship, Comer Science and Education Foundation. This is Lamont-Doherty Earth Observatory contribution number 7136 and Institute of Earth Sciences, Academia Sinica contribution number IESAS 1249.