The effect of chemical pretreatment of sediment upon foraminiferal-based proxies



[1] Paleoceanographic studies routinely combine different foraminiferal proxies (i.e., weight, abundance, trace metal, and stable isotope measurements) into a cohesive narrative. The application of chemical treatment to disaggregate ocean sediments in the most efficient way to isolate the fossils of foraminifera from the other sediment components is dictated by the time available and the material used. Yet few studies have aimed to test both the physical and geochemical effects associated with such practices. In this study, we use samples with different sedimentological characteristics (i.e., varying percentages of CaCO3 and of terrigenous material) to test the impact upon these proxies of three processing methods and a control: (1) no chemicals (contol run); (2) soaking in sodium hexametaphosphate (Calgon®); (3) soaking in hydrogen peroxide; and (4) soaking in a sodium pyrophosphate. The samples were analyzed for faunal abundance, shell weight, stable isotope (δ18O, δ13C), and trace metal (Mg/Ca) geochemistry for four species of planktonic foraminifera (Globigerina bulloides, Globigerinoides ruber, Globorotalia inflata, and Globorotalia menardii). Results show that apart from the Calgon® solution, the values of faunal abundance, shell weight, Mg/Ca, and stable isotopes are similar irrespective of the cleaning treatment utilised and therefore warrant cross-comparison of results obtained with different preparation techniques. The use of Calgon® in pretreatment shows statistically different values for only foraminiferal shell weight.

1. Rationale

[2] The shells of foraminifera are routinely employed in paleoceanographic reconstructions as either biostratigraphic indicators or as the medium for faunal and geochemical proxies. These proxies allow reconstruction of seawater temperature via transfer functions [Kucera et al., 2005; Telford et al., 2013], stable isotope [Emiliani, 1954, 1955; Ganssen and Kroon, 2000; Kahn, 1979; Killingley et al., 1981; Mix, 1987] or trace metal geochemistry [Anand et al., 2003; Barker et al., 2003; Elderfield and Ganssen, 2000; Greaves et al., 2005], evaluation of environmental changes such as ecosystem shifts [de Deckker et al., 2012], or degree of dissolution [Berger, 1992; Grötsch et al., 1991], through fragmentation indices or shell weight [Broecker and Clark, 2002-2004].

[3] Isolating foraminifera from the bulk ocean sedimentary matrices consisting of terrestrial, organic, and siliceous components is a necessary preliminary step of paleoceanographic studies. A number of methodologies that involve the use of chemical agents have therefore been employed to expedite the transformation of sediment into a useable micropaleontological sample [Barker et al., 2003; de Moura et al., 1999; Lirer, 2000; Pak et al., 2004]. Aggressive chemical treatment may, however, influence the integrity of the (paleoceanographic) signal extracted [Ganssen, 1981], either through dissolution of reactive layers within test calcite [Vetter et al., 2013] or through weakening and subsequent disintegration of shells. It is critical therefore that an assessment is made of the magnitude of any potential bias upon characteristics of the foraminiferal proxies, such as the faunal assemblage, the degree of fragmentation, the foraminiferal shell weight, its stable isotope composition (δ18O and δ13C), and its Mg/Ca ratio frequently used as a proxy for ocean temperature. When collated, these proxies form the backbone of our understanding of the mechanisms, impacts and role of the oceans in climate change.

[4] To date, we found over a thousand research articles for the three cleaning methods tested within this paper: hydrogen peroxide (H2O2) [e.g. Basak et al., 2009; Killingley et al., 1981; Smart, 2008], Calgon® (sodium hexametaphosphate) [e.g., Herguera, 2000; Olson and Smart, 2004; Smart, 2008; Taricco et al., 2009], and sodium pyrophosphate (Na4P2O7) [e.g., Filipsson and Nordberg, 2004; Hald and Korsun, 1997; Thompson, 1973]. However, whilst extensively used, the consistency in the methodological application of these procedures is relatively low compared to their wide usage, thus raising concerns as to the comparability of the results. For example, the application of the hydrogen peroxide method varies in terms of: concentration used (3–30 %) [de Moura et al., 1999]; temperature at application (room temperature to 100°C); duration of exposure time (5–60 min); presence of methanol rinsing and a final drying step (45–70°C). The uncertainty surrounding whether chemical pretreatment produces any artifacts limits the confidence placed on any reconstructions proceeding from material treated in a different manner (e.g., regional surface water reconstruction from a transect of sediment cores treated in different laboratories). Contradictory results in prior studies focussing solely on the effect upon the geochemical signature [Fallet et al., 2009; Ganssen, 1981; Nagtegaal et al., 2012], and the lack of testing upon multiproxy reconstructions, has led to the genesis of this paper, and to its expansion to include the potential alteration of the physiological properties of foraminifera specimens.

2. Methodology

2.1. Core Location

[5] Core samples were selected to represent genuine material used in the majority of paleoceanographic, micropaleontological, and geochemical analysis as opposed to sediment trap or plankton tow material. Two regions with distinct sedimentary characteristics were chosen to test the application of the different chemical treatments: the low sedimentation open ocean setting of the North Atlantic and the high sedimentation closed basin of the Gulf of Aden. Samples from box cores T90-05b (3069 m); T90-10b (2162 m); and T90-12b (5058 m) collected during the JGOFS-1990 R.V. Tyro cruise in the North Atlantic [Ganssen and Kroon, 2000], these samples were selected above and below the regional foraminiferal lysocline. Core tops may be a mix of both recent and older material, potentially Holocene or glacial, due to bioturbation [Wit et al., 2013] or dissolution. For greater homogeneity, samples from the North Atlantic were sampled at a core depth of 5–8 cm, and for the Gulf of Aden (Core 178-13PC) at 138 cm (Figure 1).

Figure 1.

Location Map of core samples based upon (a) Ganssen and Kroon [2000] and (b) Heier-Nielsen et al. [1995].

2.2. Planktonic Foraminifera

[6] In order to test the influence of chemical pretreatment Globigerinoides ruber, Globigerina bulloides, Globorotalia inflata, and Globorotalia menardii were picked (Figure 2).

Figure 2.

Representative images of species utilized in this study: Globigerina bulloides; Globigerinoides ruber (sensu lato and sensu stricto morphotypes were combined); Globorotalia inflata and Globorotalia menardii in apertural, dorsal, ventral, and spiral views. Scale bar is 100 µm.

2.2.1. Globigerinoides ruber

[7] Globigerinoides ruber is an ecologically important, abundant, symbiont-bearing foraminifera constrained to the photic zone [Deuser et al., 1981; Lonarić et al., 2006; Peeters and Brummer, 2002] and therefore one of the major proxies for (summer) sea surface temperature [e.g., Dekens et al., 2002; Schmidt et al., 2004; Thunell et al., 1999]. The optimum temperature is considered to range between 18 and 25°C [Hecht et al., 1976]. There are two color variations, or chromotypes, a “pink” form found exclusively in the Atlantic [Thompson et al., 1979] during the warmer periods [Emiliani, 1971, 1974; Richey et al., 2011] and a globally distributed “white” form during colder periods [Aurahs et al., 2009] and oligotrophic conditions [, 1959; Tolderlund and Be, 1971]. Considerable plasticity of form exists within the white chromotype (Figure 2), which has driven considerable taxonomic ambivalence discussed in depth by Aurahs et al. [2011]. Wang [2000] combined the more compact and highly trochospiral species Globigerinoides elongatus [d'Orbigny, 1826], Globigerinoides pyramidalis [van den Broeck, 1876], and Globigerinoides cyclostomus [Galloway and Wissler, 1927] into G. ruber sensu lato. All others were considered to fall into the G. ruber sensu stricto. A number of studies have found that these morphotypes have different depth preferences [Kuroyanagi and Kawahata, 2004; Lin and Hsieh, 2007] and therefore, potentially, differences in stable isotope [Kawahata, 2005; Kuroyanagi et al., 2008; Lin et al., 2004; Löwemark et al., 2005; Numberger et al., 2009; Wang, 2000] and trace metal geochemical composition [Steinke et al., 2005]. Contrarily, single specimen laser ablation for trace metal (W. Feldmeijer 2011, unpublished data) and conventional oxygen isotope [Mohtadi et al., 2009] analysis suggests that there is no difference between these morphotypes. Therefore, our analysis consists of both morphotypes. Both morphotypes have a coarsely perforate wall texture with a primary aperture bordered by a smooth rim and additional “supplementary” apertures which increase the species shell's susceptibility to dissolution.

2.2.2. Globigerina bulloides

[8] In contrast to other spinose species, G. bulloides is nonsymbiotic [Hemleben and Spindler, 1983] occurring in the upper 200 m [, 1977; Deuser and Ross, 1989; Farmer et al., 2011; Hemleben and Spindler, 1983; Ottens, 1992; Schiebel et al., 1997]. The species is predominately found within the temperate to subpolar region [Bé and Tolderlund, 1971; Ganssen and Kroon, 2000], although it is also present in both polar and upwelling waters [Ganssen, 1983; Ganssen and Sarnthein, 1983; Ganssen et al., 2011], and low latitude boundary currents [, 1977]. It is generally considered a nutrient opportunistic species [Reynolds and Thunell, 1985] which potentially explains its extended biogeographic range, although this may be a consequence of the existence of genetically distinct-cryptic species with potentially distinct temperature and salinity constraints.

[9] Steineck and Fleisher [1978] defined the outer wall texture of G. bulloides as being under the spinose category, consisting of long acicular spines with circular cross sections and a flat wall with moderately large pores penetrating an unmodified surface lacking pore pits [Kennett and Srinivasan, 1983; Saito et al., 1981]. Living specimens of G. bulloides show a slight rim around pores in some instances. These pores are densely present throughout the shell, except in newly formed chambers [Cifelli, 1982], on occasion pores can become joined. In younger, newly formed chambers, the wall texture is essentially smooth with undulations, spine bases lacking the buttresses of later development. These buttresses appear to form in a terrace-like manner, Cifelli [1982] noted that pores are not the centre of this development as on occasion these structures cover the pores. Scott [1974] and Cifelli [1982] ascribe the failure of G. bulloides to construct a “neat” honeycomb wall texture of Globigerinoides spp, despite having a similar textural development, as due to the timing of the development of shell wall porosity, thus pore formation occurs at the same time as ridge development. The species is considered susceptible to dissolution [Kucera, 2007], likely a result of its wall structure and large aperture.

2.2.3. Globorotalia inflata

[10] Globorotalia (Globoconella) inflata is a nonspinose macroperforate species generally considered to be a deeper dwelling planktonic foraminifera with calcification occuring from the mixed layer down to water depths of 500–800 m [Hemleben and Spindler, 1983; Wilke et al., 2006]. Although narrower depth intervals have been proposed for the North Atlantic, 0–150 m [Ottens, 1992] and 300–400 m [Elderfield and Ganssen, 2000], and the South Atlantic, 50–300 m [Mortyn and Charles, 2003]. The wall structure of this species is considerably varied and dependent upon the maturation state of the observed individual [Groeneveld and Chiessi, 2011; van Raden et al., 2011]. In texturally immature specimens, the wall is smooth with perforations [Saito et al., 1981], following the classical wall texture delineation for globorotalids [Steineck and Fleisher, 1978]. The wall can have protusions or pustules (also referred to as “tubercle” or “pseudospine”) which are either hemispherical or subconical in shape the latter appearing to a have triradiate-spine like cap. These subconical pustules appear to form via terracing, similar to the buttresses of G. bulloides. These features appear, in some cases, to calcify over pores. A thick secondary crust, or cortex, similar to Pulleniatina spp. is apparent in texturally mature specimens, developed as a precursor to gametogenesis [Hemleben et al., 1985]. This outer crust serves to decrease the species susceptibility to dissolution, although internal features may be heavily affected [Johnstone et al., 2010].

2.2.4. Globorotalia menardii

[11] Given that the Gulf of Aden core lies outside of the main G. inflata biogeographic zone [Siccha et al., 2009] specimens of G. (Menardella) menardii was substituted in its place (Figure 3). Globorotalia menardii is a flat-disk-like nonspinose thermocline [Farmer et al., 2007; Mohtadi et al., 2009; Ravelo and Fairbanks, 1992; Regenberg et al., 2010; Schmuker and Schiebel, 2002; Schweitzer and Lohmann, 1991] to subsurface (>50 m) [Bé and Tolderlund, 1971; Bé and Hutson, 1977] dwelling foraminifera, occurring in subtropical-tropical environments (see MARGO database) [Kucera et al., 2005]. The abundance of G. menardii in Atlantic sediments has been shown to track both the glacial-interglacial changes and oxygen isotope ratios [Ericson and Wollin, 1956], potentially brought about by the opening and closing of oceanic gateways [Caley et al., 2012] or rather changes in thermocline ventilation [Sexton and Norris, 2011]. Whilst ecologically distinct, resulting in differing absolute geochemical values, from G. inflata and given the absence of the preferred species Pulleniatina spp. (see section 2.2.3) in our sample, G. menardii was selected. This is due to the similar encrustation, with earlier chambers in the final whorl covered in cuboid pustules that give it a sugary texture [Bé et al., 1966; Regenberg et al., 2010]. The similar response to dissolution that both species have also makes G. menardii an appropriate substitution for G. inflata [Parker and Berger, 1971; Thunell and Honjo, 1981].

Figure 3.

Schematic view of methodology used within this paper. Four samples (a–d) were (i) divided into four splits for different treatments. Each split was (ii) passed through a nest of sieves for subdivision into appropriate size fractions for abundance counts. The following (iii) analysis were performed on specimens of the selected species solely from the 300–355 µm size fraction, including (iv) shell weight. For (v) geochemical analysis, the specimens were crushed for a homogenized signal.

2.3. Chemical Pretreatment Procedure

[12] For an overview of the methodology, see the flow chart in Figure 3. Each sediment slice was divided into four aliquots, weighed unprocessed before wet sieving over a 63 µm mesh, using four processing methods: (1) a control with no chemical additional; (2) addition of commercially available Calgon® (sodium hexametaphosphate, (NaPO3)6) and hydrogen peroxide (H2O2) commercially available from Boom Laboratorium leverancier; (3) hydrogen peroxide (H2O2); and (4) sodium pyrophosphate (Na4P2O7) commercially available from Sigma-Aldrich® (see Table 1 for concentrations and procedures). Given the inconsistency in the use of the hydrogen peroxide method, we here use a methodology that reflects both the outline aims of our paper, and the most common application in the literature.

Table 1. Methodology of Treatments Used Within this Experiment
TreatmentTreatmentChemical TreatmentProcessing Treatment
NumberName(Step 1)(Step 2)(Step 3)(Step 4)Step 5Time (h)
1Control  Wet sieving over >63 μm meshDried overnightDry sieved over 125 µm; 250 µm; 300 µm; 355 µm mesh sizes24
2“Calgon”12.5 g Calgon diluted in 2.5 L distilled water stirred mechanically for an hour before 800 mL of this solution was added to each sample for 5 days400 mL of 30% hydrogen peroxide for 1 hWet sieving over >63 μm meshDried overnightDry sieved over 125 µm; 250 µm; 300 µm; 355 µm mesh sizes164
3“Peroxide”10% hydrogen peroxide for an hourWet sieving over >63 μm meshDried overnightDry sieved over 125 µm; 250 µm; 300 µm; 355 µm mesh sizes25
4“Sodium pyrophosphate”100 mL of 0.3% solution of Sodium pyrophosphate and brought to the boiling point Wet sieving over >63 μm meshDried overnightDry sieved over 125 µm; 250 µm; 300 µm; 355 µm mesh sizes24

[13] The initial pH of the solutions were measured with a Radiometer analytical Titralab® TIM840 titration manager in the VU Water laboratory, calibrated against solutions of known pH (4 and 7). Measurements showed that tap water had a pH of 8.28, Calgon® had a pH of 5.1 at a concentration of 0.5 %. Dependent upon the concentration hydrogen peroxide had a pH of 3.24 or 4.02 for 30 and 10% concentration by volume, respectively, and sodium pyrophosphate had a pH of 10.26 at a concentration of 0.3% (Figure 4). No apparent change in pH occurs for tap water and only a minor change is apparent in sodium pyrophosphate once the solution is brought to the boil. The largest shift in pH is seen within both concentrations of hydrogen peroxide and within the Calgon® solution. After 20 min all solutions, apart from the Calgon® solutions showed a plateau in pH values; by 60 min, the Calgon® solution value was close to that of tap water, and average seawater (pH 8.2). This value remains stable over extended periods (3–6 days). The addition of 30% hydrogen peroxide, with a pH of 3.24, as per our methodology subsequently lowered the pH of the Calgon® solution. Unlike other studies [Fallet et al., 2009; Ganssen, 1981], we choose not to buffer the hydrogen peroxide solution. Despite concerns that direct application of hydrogen peroxide to calcerous sediments would cause intense dissolution the efficiency of oxidation of organic matter by hydrogen peroxide is pH dependent [Fallet et al., 2009]. The addition of the sample material to the solution with 10% hydrogen peroxide shows the largest shift in pH value; however, this shift raises it to a neutral pH level. Furthermore, as there was no systematic weight loss seen in the >63 µm size fraction between hydrogen peroxide and other treatments this validates our choice.

Figure 4.

pH value of the different chemical treatments of sample T90-10b through time. Two concentrations of hydrogen peroxide (H2O2), 10 and 30%, were used for comparison with the concentrations used in the literature. Following our methodology, after sample addition sodium pyrophosphate was brought to the boil. After t = 60 min H2O2 was added to the Calgon® treatment as per our methodology, for all others measurements ceased.

[14] Dry sieving was used to divide the samples into the discrete size fractions: 125–250 µm; 250–300 µm; 300–355 µm; and >355 µm. Samples were split, following sieving, with an Otto microsplitter until ∼300 individuals were available for counting. No systematic difference is seen in the order (i.e., either prior to or after sieving) in which splitting occurs (F. Peeters 2012, personal communication). The division and preparation of subsamples was performed in such a way as to minimize bias; however, in order to check, sieve size fractions were weighed. The constituent size fractions between each sample have a standard deviation that ranges between 0.03 and 0.59 g. However, removing the >355 µm sample from treatment (2) in T90-10B, which includes a single large Ice Rafted Detrital (IRD) grain, lowers the standard deviation's upper value from 0.59 to 0.30 g. This shows that splitting was successful, as there is little difference between the amounts of material used between sample treatments. For ease, only G. bulloides, G. inflata, and G. ruber (Figure 2) were identified to the species level, with all others counted as “other foraminifera.” To further test the impact of chemical treatment upon each sample, a quantitative dissolution index testing the ratio between fragments and total whole foraminifera was utilized [Kucera, 2007]. Specimens of G. ruber(white) and G. inflata were picked as they are representative of both susceptible and resistant species [Kucera, 2007]. In order to dissuade any potential bias for a preferred methodology, the authors/analysts, bar the lead authors, were blinded to both the sample identity and the chemical procedure applied.

[15] For shell weight, 50 foraminifera were picked, from the 300–355 µm size fraction, and weighed in groups of five individuals on a Sartorius microbalance (1 µg precision) to quantify weight loss from chemical treatment, before geochemical analysis. Weighing in groups containing the same number of specimens greatly reduces the risk that potential contaminants (i.e., sedimentary infill) might skew the average weight.

2.4. Stable Isotope and Trace Metal Geochemical Analysis

[16] For geochemical analysis, the 50 specimens of four different species of foraminifera (G. bulloides; G. inflata; G. ruber, and G. menardii) were gently crushed between two glass slides just enough to obtain coarse fragments. This is necessary to open up the chambers for removal of contaminants such as clays. The crushed foraminiferal fragments were then subdivided into two groups for stable isotope and trace metal analysis. Stable isotope analysis was conducted on a Thermo Finnigan Delta+ mass spectrometer equipped with a GASBENCH II preparation device at the VU University Amsterdam. Samples were placed in He-filled 10 mL exetainer, before digestion in 100% orthophosphoric acid (H3PO4) at a temperature of 45°C. The effect of sample amount on the measured composition (reflected in the relative amplitude, in mV) is corrected using an internal standard, in varying amounts to the bracket the weight range of the samples. Isotopic values are reported as δ18O and δ13C in per mil (‰) relative to the Vienna PeeDee Belemnite (V-PDB) notation. The within-run reproducibility of a routinely analyzed external CaCO3 standard (UCD-SM92) is <0.10‰ (1σ) for both δ18O and δ13C.

[17] Additional cleaning is required prior to trace metal analysis, owing to the low concentration of the lattice-bound metals and the potential for contamination from clays, organic matter, Fe-Mn oxides, Mg-rich carbonate overgrowths, and sulphides from sedimentation [e.g., Barker et al., 2003; Bian and Martin, 2010; Pena et al., 2005]. To comply with the scope of our investigation, we applied the widely used procedure outlined by Barker et al. [2003], which comprises sonication, rinsing in water and methanol, oxidation in hydrogen peroxide (H2O2) oxidation, and weak acid (HNO3) leaching. Although some authors recommend the inclusion of a reductive step in the cleaning routine, for cases of especially contaminated foraminifera [e.g., Pena et al., 2005], to obtain calcite lattice that more reliably reflects paleotemperatures, we omitted this step because of the considerable calcite dissolution that it implies. The purpose of this study is to examine the difference between treatments, rather than to achieve absolute paleotemperature estimations. It is likely that the effect of the various sample treatments is masked by excessive calcite removal via reductive cleaning. The cleaned calcite samples were dissolved in 1% HNO3 and analyzed with a Varian Vista ICP-OES at the VU University Amsterdam. The reproducibility of replicate analysis on the international standard ECRM 752-1 [Greaves et al., 2005] is ±0.035 mmol/mol (<1%) for Mg/Ca.

3. Results

[18] Apart from a few samples, there is little difference, or no consistent deviation, between the values of fragmentation (Figure 5), shell weights (Figure 6), oxygen and carbon stable isotopes (Figures 7 and 8), and trace metal geochemistry (Figure 9) between chemical treatments. Minor differences are observed in the oxygen isotopes of G. inflata (cores T90-10b and T90-12b) and G. ruber (T90-12b) with treatment with hydrogen peroxide, and G. menardii with treatment with sodium pyrophosphate (178-13p). All other oxygen and carbon isotope values lie within machine error. For trace metal, geochemistry values of G. bulloides (T90-10b) and G. inflata (T90-12b) treated with hydrogen peroxide, and G. bulloides treated with Calgon® (178-13p) show a deviation from the control value.

Figure 5.

Fragmentation index values from the 125–250 µm (red), 250–300 µm (blue), and 300–355 µm (green) size fractions for (a) North Atlantic Ocean (T90-5b), (b) North Atlantic Ocean (T90-10b), (c) North Atlantic Ocean (T90-12b), and (d) Gulf of Aden (178-13p).

Figure 6.

Average shell weight of G. ruber, G. bulloides, and G. inflata/G. menardii for the cores (a) T90-5b, (b) T90-10b, (c) T90-12b, and (d) 178-13p. Note the change in y axis values between species and between the North Atlantic and Gulf of Aden samples. The values along the x axis represent the treatments (1) Control, (2) (NaPO3 + H2O2), (3) (H2O2), and (4) (N4P2O7).

Figure 7.

Oxygen isotope values of G. ruber, G. bulloides, and G. inflata/G. menardii for cores (a) T90-5b, (b) T90-10b, (c) T90-12b, and (d) 178-13p. Gray shading indicates value of control (black dots) sample, with bars representing the machine error. Note the change in y axis values between the North Atlantic and Gulf of Aden samples.

Figure 8.

Carbon isotope values of G. ruber, G. bulloides, and G. inflata/G. menardii, see Figure 7 for detailed description.

Figure 9.

Mg/Ca values of G. ruber, G. bulloides, and G. inflata/G. menardii for the four cores (a) T90-5b, (b) T90-10b, (c) T90-12b, and (d) 178-13p. Note the change in y axis scale between species and between the North Atlantic and Gulf of Aden samples.

3.1. Statistical Analysis: t Test

[19] Paired t test's were performed using PAST (vers. 2.17c) [Hammer et al., 2001] to determine if there was any statistical similarity between the means of the samples. As a t test is based upon the means (of. weight, stable isotope, and trace metal) of a sample, any deviation between treatments should become apparent when combining data for all species. Rather than statistically testing the differences individually between core and species, the data sets were grouped and tested for treatment effects. However, in order to clarify whether some treatments have a differential effect on some species (i.e., based on inherent differences in the resistance of the test structure), we also performed species-specific tests. For the combined dataset (“all values”), the t test critical value (n = 12) is ±2.201, whereas for species-specific t tests (n = 4) it is ±3.182. Using a paired t test, we can conclude that there is no statistical difference between treatments when considering all isotopic values (Table 2; Figure 10a). However, if we deconstruct this into its constituent components (individual species) and then compare treatments and species, there is a single sample (comparing treatment with Calgon® and sodium pyrophosphate) that lies outside of the critical value (Figure 10a). Likewise, for trace metal geochemistry, there is no statistical difference between treatments, apart from when we look into the individual species' response (Table 2; Figure 10b) comparing Calgon® and hydrogen peroxide. In comparison, the results of the t test for shell weight show for individual species responses that Calgon® deviates from the control and other treatments (Table 2; Figure 10c). Although when comparing the combined all species shell weight values, there is no difference between treatments.

Table 2. Stable Isotope, Trace Element and Shell Weight Paired t Test Resultsa
 All SpeciesbG. rubercG. bulloidescG. inflata/G. menardiic
TreatmentΔ18OΔ13CMg/CaShell WeightΔ18OΔ13CMg/CaShell WeightΔ18OΔ13CMg/CaShell WeightΔ18OΔ13CMg/CaShell Weight
  1. a

    Outliers are indicated in bold. At the 99% confidence interval, these critical values are ±3.106 and ±5.841, respectively.

  2. b

    At the 95% confidence interval, the critical t value for 11 of degrees of freedom (n = 12) is ±2.201.

  3. c

    At the 95% confidence interval, the critical t value for 3 degrees of freedom (n = 4) is ±3.182.

1–4Control/Sodium pyrophosphate−1.0010.9880.0340.1700.8691.0161.039−1.630−1.199−0.250−2.213−0.025−1.0591.0661.873−1.333
2–4Calgon/Sodium pyrophosphate−0.4451.083−0.3880.0001.6930.261−0.5958.5880.083−0.781−1.250−2.028−1.5523.6360.7854.402
3–4Peroxide/Sodium pyrophosphate−0.0520.4100.1720.1201.674−2.831−0.100−1.959−0.631−0.672−0.717−1.387−0.7191.6413.1540.077
Figure 10.

Results of performed paired t-test on (a) oxygen (δ18O; squares) and carbon (δ13C; circles) stable isotope values, (b) Mg/Ca ratios and (c) shell weight comparing all (open symbols) and species specific (filled symbols; blue for G. ruber, orange for G. bulloides and red for G. inflate/G. menardii) values. Critical t-values at the 95% confidence level are (+/−) 2.201 (open arrow) and (+/−) 3.182 for all values (n = 12) and species specific (n = 4) respectively. Outliers are indicated, note that symbols with similar values are offset for clarity.

3.2. Statistical Analysis: Linear Modeling

[20] Because treatments and samples are not independent of one another, due to the experimental design, we chose to corroborate these findings using a mixed modeling approach which allows for testing the significance of fixed effects (in this case, treatment type), but controls for pseudoreplication (site and species), i.e., multiple measurements of a particular species or site are not independent of one another. Analysis was performed on species within a site; therefore, the site is the top hierarchical measure of replication and then species the second level. The data structure thus requires a nested random effect—species within site (denoted by site/species). The first portion of this test controls these random effects, whilst the second concerns one-way analysis of variance (ANOVA) which looks for a significant difference between the means of the four treatment groups. The modeled treatment is compared with the control, where treatment affects are removed using ANOVA. In order to ascertain which treatments are different from one another, and therefore responsible for the overall significant difference, a posthoc test is performed as a third step. This test performs a pairwise comparison of all groups, but penalized for doing multiple tests in order to prevent the generation of false positive results (type I statistical errors). When performing such statistical analysis, it is essential that the model meets the assumption of the test. In the case of linear modeling with a normal error distribution, the residuals of the model must be normally distributed and have homogeneity of variance (homoscedastic). Apart from shell weights after treatment with Calgon®, which are significantly lower than all treatments and control, the effects of all pretreatments on Mg/Ca ratios, isotopic, and total abundance are not significant (Figure 11). While a number of responses had to be transformed in order to obtain a normal distribution, this is likely a result of the relatively low sample sizes.

Figure 11.

Statistic linear modelling results for (a) total abundance, (b) individual shell mass, (c) trace metal (Mg/Ca), and (d) stable isotope (δ18O) geochemistry.

4. Discussion

[21] Ganssen [1981] surmized the influence of chemical treatment and the storage medium (ethanol, Rose Bengal, formalin, tetrachlorethylene, and hydrogen perioxide) on the isotopic composition of two foraminiferal shells (the benthic species Haynesina germanica and planktonic Globigerinoides sacculifer) and on crushed Solnhofen limestone was due to partial dissolution. Our results suggest that the treatment of sedimentary material by chemicals does not influence the derived proxies: faunal abundance, stable isotopic, and trace metal geochemistry. In a comparison with treated, boiled for 3 h in buffered 10% hydrogen peroxide, and untreated samples, washed in hot demineralized water, Berger [1970] in a in situ calcite dissolution experiment, noted an increase in susceptibility of the treated samples to waters that were undersaturated with respect to calcium carbonate. However, whilst oxygen isotope and Mg/Ca values appear lower in the deeper samples of the North Atlantic chemical treatment does not appear to exacerbate any apparent effect of water depth, in turn a reflection of the position of sample relative to the lysocline. Whilst our research questions have been answered, we wish to draw the attention to a number of aspects concerning routine use of the proxies studied here in the following sections.

4.1. Shell Weight

[22] The shell weight of fossilized planktonic foraminifera has the potential to include contaminants (i.e., clays, coccolithophores, or detritus), which may alter or skew the record toward heavier values, whereas dissolution may lead to lighter values [Broecker and Clark, 2002]. Statistically, our results suggest that there is an effect of the Calgon® treatment on shell weight of planktonic foraminifera only. The experimental setup within this paper infers that the control has no contamination. However, given the nature of sedimentary material, chemical cleaning may dislodge sediment infill or remove potential contaminants (i.e., clays), that likely otherwise adhere to the shell, and therefore yield the mild discrepancies that bring the measured weight closer to that of a “clean” shell. Likewise, cleaning agents also have the potential to dissolve or remove outer layers of the test through chemical action or changes in the pH (Figure 4). This is dependent upon the duration of exposure to chemicals, the water used (i.e., de-ionized water, pH 7, or tap water, pH 8.2) and the storage medium [Ganssen, 1981]. With exposure time both hydrogen peroxide and Calgon® changed to a more basic level (Figure 4), potentially the result of calcium carbonate being dissolved. Whilst hydrogen peroxide, at 10   % concentration, underwent the largest shift in pH level the fact that the changes in pH became negligible over time (at 20 min) and this initial shift raised it to a neutral pH level (pH 7) likely explains the lack of difference from control. Whereas the Calgon® reaction continued to show signs of pH increase. It is likely that the sample's exposure time to Calgon® solution with the potential for substitution of calcium for sodium atoms [Thomson, 1936] and the final additional hydrogen peroxide step (pH 3.24) is the source of the apparent deviations (Figure 10c).

4.3. Stable Isotope Geochemistry

[23] Chemical cleaning in the case of oxygen and carbon isotopic analysis was commonly/historically aimed at preventing the reaction between organic matter and the orthophosphoric acid, used to liberate carbon dioxide from the skeletal carbonate. This has the potential to liberate carbon dioxide of a significantly different isotopic composition than that of the desired skeletal carbonates and thus contaminate the product [Epstein et al., 1953]. In other instances, it was thought that without it there was the potential for introduction of volatile organic impurities that may allow the formation of molecules or radicals contaminants within the ionization chamber [Wierzbowski, 2007]. Given this, a number of organic matter removal steps have been devised, and many have been subsequently abandoned. For example, the effect of roasting or oxidizing treatment prior to stable isotopic analysis has been shown to decrease δ13C and δ18O values of skeletal carbonates by typically less than ∼1 ‰, while inorganically precipitated carbonates remain unaffected (for a detailed discussion, see Wierzbowski [2007]). Even though Savin and Douglas [1973] and Kahn [1979] concluded that the affect of Chlorax® (sodium hypochlorite solution) on foraminiferal shells lowered both δ18O and δ13C, other buffering solutions such as formalin or ethyl alcohol had no isotopic deviation using field collected foraminifera. The results of Bé and Anderson [1976] and later Ganssen [1981] concluded that the storage medium may influence the composition by enrichment in the heavier (18O) isotopes in buffered formalin, and a rise in pH, through increasing [Ca2+], in poorly buffered storage medium (i.e., Formalin).

[24] Serrano et al. [2008], Grottoli et al. [2005], and Wierzbowski [2007] have shown a significant decrease in δ18O values of carbonate with hydrogen peroxide pretreatment. The reasoning for a change (Δδ) has been linked to removal of organic contaminants, alteration of mineral phases, or isotopic exchange with organics or other substances [Wierzbowski, 2007]. Whilst isotopic exchange between the hydrogen peroxide solution and carbonate has been ruled out by Boiseau and Juillet-Leclerc [1997], based upon corals, recent work has shown that this solution can remove a significant portion of shell material [Vetter et al., 2013]. Dissolution of shell material has been shown to alter δ18O and δ13C values inferred to be the result of differential dissolution between thin-walled shallow forms and thicker-walled deeper forms [Bonneau et al., 1980]. In the case of the results presented here, treatment with hydrogen peroxide amounts to a Δδ18O of less than −0.15 % for calcite. Investigation of the peroxide treatment was conducted by Waelbroeck et al. [2005] who, whilst showing a statistically insignificant impoverishment in planktonic foraminifera oxygen isotope values (−0.11 ± 0.23 %), concluded that “further systematic tests” were necessary. This impoverishment is well within the range of error associated with taking a finite sample of a virtually infinite population, in multiple shell analysis [Killingley et al., 1981; Schiffelbein and Hills, 1984] (supporting information Figure1) and is not seen within our dataset. The lack of apparent deviation in our data set is likely attributable to the fact that there is no affect of hydrogen peroxide upon the isotopic composition.

5. Conclusions

[25] We performed a systematic experiment and statistical evaluation of the resulting data set and found no evidence to indicate that three chemical treatments commonly utilized to disaggregate marine sediment significantly affect the foraminiferal based proxies: species abundance, shell fragmentation, δ18O, δ13C, and Mg/Ca. An exception is that the use of Calgon® influences the shell weight, leading to a statistical deviation from both control and other treatments, which we regard as possible due to changes in the pH and the implied exposure time. This provides a rigorous basis for direct comparison between results obtained by any such treatment.


[26] The comments by the Editor Cin-Ty Lee and an anonymous reviewer helped to improve this manuscript. The authors would like to thank Gerald Ganssen, Frank Peeters, and Geert-Jan Brummer for helpful and insightful comments, Martin Konert and the VU Sedimentology Laboratory Amsterdam for help with processing and help with chemicals. Kate Plummer (British Trust for Ornithology) is thanked for statistical support. B.M. is funded by the European Community's 7th Framework Programme, EPOCA (European Project for Ocean Acidification; FP7/211384) and the VU, W.F. is funded by the Darwin centre for Biogeosciences “Sensing Seasonality” project, K.A. and P.S. have received funding from the European Community's 7th Framework Programme FP7/2007-2013—Marie-Curie ITN, under grant agreement 238512, GATEWAYS project.