An interlaboratory study of TEX86 and BIT analysis of sediments, extracts, and standard mixtures
Abstract
[2] Two commonly used proxies based on the distribution of glycerol dialkyl glycerol tetraethers (GDGTs) are the TEX86 (TetraEther indeX of 86 carbon atoms) paleothermometer for sea surface temperature reconstructions and the BIT (Branched Isoprenoid Tetraether) index for reconstructing soil organic matter input to the ocean. An initial round-robin study of two sediment extracts, in which 15 laboratories participated, showed relatively consistent TEX86 values (reproducibility ±3–4°C when translated to temperature) but a large spread in BIT measurements (reproducibility ±0.41 on a scale of 0–1). Here we report results of a second round-robin study with 35 laboratories in which three sediments, one sediment extract, and two mixtures of pure, isolated GDGTs were analyzed. The results for TEX86 and BIT index showed improvement compared to the previous round-robin study. The reproducibility, indicating interlaboratory variation, of TEX86 values ranged from 1.3 to 3.0°C when translated to temperature. These results are similar to those of other temperature proxies used in paleoceanography. Comparison of the results obtained from one of the three sediments showed that TEX86 and BIT indices are not significantly affected by interlaboratory differences in sediment extraction techniques. BIT values of the sediments and extracts were at the extremes of the index with values close to 0 or 1, and showed good reproducibility (ranging from 0.013 to 0.042). However, the measured BIT values for the two GDGT mixtures, with known molar ratios of crenarchaeol and branched GDGTs, had intermediate BIT values and showed poor reproducibility and a large overestimation of the “true” (i.e., molar-based) BIT index. The latter is likely due to, among other factors, the higher mass spectrometric response of branched GDGTs compared to crenarchaeol, which also varies among mass spectrometers. Correction for this different mass spectrometric response showed a considerable improvement in the reproducibility of BIT index measurements among laboratories, as well as a substantially improved estimation of molar-based BIT values. This suggests that standard mixtures should be used in order to obtain consistent, and molar-based, BIT values.
Key Points
- Round robin study of TEX86 and BIT
- Interlaboratory consistency of TEX86 substantially improved
- BIT analysis requires standard mixtures
1. Introduction
ratio based on long-chain C37 alkenones derived from haptophyte algae [Brassell et al., 1986; Prahl and Wakeham, 1987]. Another organic SST proxy is based on archaeal glycerol dialkyl glycerol tetraether (GDGT) lipids, the TEX86 index [Schouten et al., 2002]. These lipids are biosynthesized by marine archaea that synthesize GDGTs containing 0–3 cyclopentyl moieties (GDGT-0–GDGT-3; see structures in Figure 1). Members of the phylum Thaumarchaeota also synthesize crenarchaeol, a compound with a cyclohexyl moiety in addition to four cyclopentyl moieties (Figure 1), and smaller quantities of a crenarchaeol regioisomer (Cren′). The TEX86 ratio was proposed as a mean to quantify the relative distribution of GDGTs [Schouten et al., 2002]:
(1)
(2)
(3)[5] The TEXH86 is suggested to be applicable in warm, tropical regions, whereas TEXL86 is calibrated for cold, polar regions [Kim et al., 2010a]. Besides SST reconstruction, the TEX86 proxy is also used in some lakes to infer (paleo) surface water temperatures using a lake core top calibration [Powers et al., 2004, 2010].
(4)[7] Subsequent studies have shown that this proxy can be applied to trace soil OM in coastal marine environments [e.g., Huguet et al., 2007; Walsh et al., 2008; Smith et al., 2010] as long as the export fluxes of crenarchaeol to sediments are comparable among sites [Fietz et al., 2011; Smith et al., 2012] and the relative proportions of crenarchaeol versus marine OM and branched GDGTs versus soil OM remain constant. Furthermore, Weijers et al. [2006] found that a high input of soil OM in marine sediments can potentially bias TEX86 values as soils can also contribute isoprenoidal GDGTs. They recommended simultaneous reporting of BIT indices in order to monitor for this effect. BIT values can range from 0.01 in open marine sediments to 1 in some soils [Schouten et al., 2013, and references cited therein].
[8] A prerequisite for the wider application of geochemical proxies is the robustness of analytical reproducibility. GDGTs are analyzed by high-performance liquid chromatography (HPLC) coupled to mass spectrometry (MS) [Hopmans et al., 2000; Schouten et al., 2007; Escala et al., 2007]. A common procedure to validate analytical methodology and help laboratories detect and remediate inaccuracies in their results is to conduct a round-robin test (also known as proficiency test or interlaboratory comparison) [Thompson et al., 2006], as has been done for the
ratio of long-chain C37 alkenones [Rosell-Melé et al., 2001] and for the Mg/Ca ratio of (foraminiferal) carbonates [Rosenthal et al., 2004; Greaves et al., 2008]. An initial round-robin study for TEX86 and BIT analyses was performed in 2008 [Schouten et al., 2009] on filtered polar fractions obtained from extracts of two sediments. For TEX86 the repeatability (i.e., intralaboratory variation) was 0.028 and 0.017, respectively, for the two sediment extracts. This translates to ±1–2°C of calculated temperature variation using contemporary TEX86-SST calibrations. The reproducibility, indicating interlaboratory variation, of TEX86 measurements was substantially higher: 0.050 and 0.067, respectively, or ±3–4°C when translated to temperature. These temperature uncertainties were higher than those obtained in round-robin studies of the Mg/Ca (2–3°C) and
(1–2°C) paleothermometers and suggest relatively large variations between laboratories in TEX86 measurements. Repeatability of BIT measurements for the sediment extract with substantial amounts of soil OM input was relatively small, 0.029, but reproducibility between laboratories was large, 0.410 on a scale of 0–1. This large dispersion was attributed to the large structural differences between the GDGTs used in the BIT index (e.g., the higher-molecular-weight crenarchaeol versus the lower-molecular-weight branched GDGTs), which may give rise to variable responses in the different mass spectrometers used.
[9] Here we describe the results of an anonymized second round-robin study involving 35 laboratories. The six samples distributed consisted of three different homogenized sediments, one sediment extract, and two mixtures of isolated branched GDGTs and crenarchaeol mixed in known, preweighed quantities. The results shed light on the effect of extraction and separation techniques on the analyses of TEX86 and BIT indices of sediments as well as on the necessity of using standard mixtures to calibrate the BIT index.
2. Materials and Methods
[10] A general invitation was sent to a large number of laboratories to participate in an anonymous round-robin study, to which 36 laboratories responded positively. These laboratories received three 60 mL vials containing homogenized sediment (labeled “Sediments A, B, or C”) prepared at Harvard University and at the Royal Netherlands Institute for Sea Research (NIOZ) and three 3 mL vials containing mixtures of organic compounds or extracts (labeled “organic fractions D, E, and F”) prepared at the NIOZ. Laboratories were requested to analyze the samples when their HPLC-MS setup was performing well according to their criteria and to analyze sediments at least in triplicate and the organic fractions at least fivefold. The vials were distributed at the start of November 2011, and results reported here are those of the 35 laboratories, which reported their results before 1 April 2012 (see supporting information).1 Results reported after the deadline were not considered in this study. The study was performed “doubly blind,” i.e., statistical treatment of the results was performed by individuals (A.R.M. and S.S.) unaware of the contents of the round-robin samples (prepared by A.P. and E.C.H.), and identities of the laboratories were anonymized by Lloyd Snowdon (Department of Geoscience, University of Calgary, Canada), who was not involved in analysis or statistical treatment of the data.
2.1. Sediment Origin
[11] Sediment A was obtained from Salt Pond, Falmouth, Massachusetts, USA (41°32′N, 70°37′W; water depth 3 m). In July of 2011, 33 kg of wet sediment was homogenized, dried, pulverized, and rehomogenized. This yielded 4 kg dry sediment, of which ca. 500 g was aliquoted in fifty 60 mL amber glass vials in 10 g portions.
[12] Sediment B was obtained from a 46 kg box core from the Carolina Margin (35°50′N, 74°50′W; water depth ca. 600 m) in July of 1996 and stored frozen at −20°C. In August of 2011 it was thawed and processed analogously to the Salt Pond sample, yielding 18.2 kg dry sediment, of which ca. 1 kg was aliquoted in fifty 60 mL amber glass vials in 20 g portions.
[13] Sediment C was derived from the upper part of a piston core (TY92-310G; 16°03′N, 52°71′E; 880 m water depth; 0 to 42 cm depth) taken in the Arabian Sea. The sediment was freeze-dried and ground using mortar and pestle to obtain 1.5 kg of dry sediment. Approximately 500 g of the sediment was extracted to obtain “organic fraction F” (see below), while 1 kg of sediment was aliquoted in fifty 60 mL amber glass vials in 20 g portions.
2.2. Preparation of Extract and Mixtures
[14] To obtain organic fraction F, Sediment C was divided in several aliquots and extracted using an Automated Solvent Extractor (ASE) 200, DIONEX, 100°C, and 7.6 × 106 Pa with a mixture of dichloromethane (DCM):methanol (MeOH) (9:1, vol/vol). Total lipid extracts were separated over a column filled with aluminum oxide into apolar and polar fractions using hexane:DCM (9:1, vol/vol) and DCM:MeOH (1:1, vol/vol), respectively. Polar fractions were combined, condensed by rotary evaporation, dried under a stream of nitrogen, and weighed and dissolved in hexane/isopropanol (99:1, vol/vol) in a concentration of 2 mg mL−1. Aliquots of 0.5 mg of polar fraction (labeled organic fraction F) were filtered using a PTFE (Polytetrafluoroethylene) 0.4 µm filter, dried under a stream of nitrogen, and placed in fifty 3 mL vials.
[15] Organic fractions D and E contained mixtures of three isolated GDGT standards: crenarchaeol, GDGT-I, and GDGT-II. The branched GDGTs were isolated from a large extract of sediment derived from a piston core taken in the Drammensfjord, Norway (D2-H; 59°40.11′N, 10° 23.76′E; water depth 113 m; sediment depth 746–797 cm), while crenarchaeol was isolated from the remainder of Sediment C. The sediments were Soxhlet-extracted (24 h) using a mixture of DCM and MeOH (7:1, vol/vol). The combined extracts were separated over a column filled with aluminum oxide into an apolar and polar fraction using hexane:DCM (9:1, vol/vol) and DCM: MeOH (1:1, vol/vol), respectively. GDGTs were first isolated in two stages using normal phase HPLC followed by flow injection analysis according to Smittenberg et al. [2002]. Columns used were a semipreparative and an analytical Alltech Prevail Cyano column (250 mm × 10 mm, 5 µm, and flow rate 3 mL min−1 and 250 mm × 4.6 mm, 5 µm, and flow rate 1 mL min−1, respectively). The isolated GDGTs were further cleaned using reversed phase chromatography modified from Ingalls et al. [2006]. Briefly, GDGTs were dissolved in ethyl acetate and injected onto a Zorbax Eclipse XDB C-8 column (4.6 mm × 150 mm; 5 µm; Agilent Technologies). GDGTs were eluted with the following program with acetonitrile (A) and ethyl acetate (B) as mobile phase: 0–10% B in 4 min, 10–35% B in 10 min, 35–69% B in 6 min, 69–100% B in 7 min, with a flow rate of 1 mL min−1. This yielded 5.0 mg of crenarchaeol, 3.8 mg of GDGT-I, and 3.7 mg of GDGT-II. The purity of the GDGTs was first assessed by full scan (m/z 300–2000) HPLC-atmospheric pressure chemical ionization (APCI)/MS, which did not reveal major fragments other than those from the GDGTs [cf. Hopmans et al., 2000]. In addition, we performed 1H and 13C NMR analyses on the purified compounds. All the major signals of the GDGTs could be assigned [see Schouten et al., 2013, Figure 7]. Only in the case of GDGT-II some minor signals were found that could not be attributed to GDGTs. Although the exact purity cannot be assessed, the MS and NMR data suggest that GDGTs are likely to be more than ∼90% pure. Isolated GDGTs were weighed on a microbalance (accuracy 0.1 µg), and two mixtures were prepared, Mixtures D and E. Each vial of Mixture D contained 500 µL, pipetted using a positive displacement micropipette with precision of 0.5%, of a 25 mL solution containing 200 µg of crenarchaeol, 50 µg of branched GDGT-II, and 75 µg of branched GDGT-I. The final ratio of GDGTs was 200:50:75 (wt/wt/wt) leading to a mass-based “BIT” value (BITmass) of 0.385 and a molar-based “BIT” value (BITmol) of 0.440. Each vial of Mixture E contained 500 µL of a 25 mL solution containing 1000 µg of isolated crenarchaeol, 100 µg of isolated branched GDGT-II, and 50 µg of isolated branched GDGT-I. The final ratio of GDGTs was 100:10:5 (wt/wt/wt) leading to a BITmass of 0.130 and a BITmol value of 0.158. Uncertainties in the weighing of compounds, pipetting of mixtures, and compound purity lead to an uncertainty of ∼0.01 in BIT values.
2.3. TEX86 and BIT Analyses
[16] The sediments were extracted and fractionated according to the protocols used by each individual laboratory (Table 1). GDGT Mixtures D and E and Extract F were analyzed as provided. All laboratories used HPLC/APCI/MS to analyze GDGTs (Table 2).
| Laboratory Number | Extraction | Solvent | Treatment | Column Fraction |
|---|---|---|---|---|
| 1 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 2 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 3 | Ultrasonic | DCM:MeOH, 1:1 | None | |
| 4 | Ultrasonic | H2O:MeOH:DCM, 4:10:5 | Al2O3 column | DCM:MeOH, 1:1 |
| 5 | Microwave | DCM:MeOH, 3:1 | SiO2 column | MeOH |
| 6 | Ultrasonic | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 7 | Soxhlet | MeOH:DCM, 9:1 | SiO2 column | DCM:MeOH, 1:1 |
| 8 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 9 | Ultrasonic | DCM:MeOH, 3:1 | Hydrolysis, SiO2 column | DCM:MeOH, 1:1 |
| 10 | Microwave | DCM:MeOH, 9:1 | SiO2 column | DCM:MeOH, 1:1 |
| 11 | ASE | DCM:MeOH, 9:1 | SiO2 column | MeOH |
| 12 | Ultrasonic | DCM:MeOH, 2:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 13 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 14 | n.r. | |||
| 15 | Ultrasonic | MeOH, MeOH:DCM (1:1), DCM | SiO2 column | MeOH |
| 16 | Ultrasonic | CHCl3:MeOH:ammonium acetate buffer, 2:1:0.8 | ||
| 17 | Ultrasonic | MeOH, MeOH:DCM (1:1), DCM | Hydrolysis, Al2O3 column | DCM:MeOH, 1:1 |
| 18 | ASE | DCM:MeOH, 6:4 | SiO2 column | Toluene:MeOH, 3:1 |
| 19 | ASE | DCM:MeOH, 9:1 | SiO2 column | DCM:MeOH, 1:1 |
| 20 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 21 | ASE | DCM:MeOH, 93:7 | Al2O3 column | DCM:MeOH, 1:1 |
| 22 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 23 | ASE | DCM:MeOH, 9:1 | None | |
| 24 | Ultrasonic | MeOH, MeOH:DCM (1:1), DCM | SiO2 column | DCM:MeOH, 1:1 |
| 25 | Microwave | DCM:MeOH, 3:1 | Hydrolysis | None |
| 26 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 27 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 28 | ASE/Soxhlet | DCM:MeOH, 2:1 | SiO2/Al2O3 column | DCM:MeOH, 1:1 |
| 29 | Ultrasonic | MeOH, MeOH:DCM (1:1), DCM | H2O extraction of Hex:IPa | |
| 30 | Ultrasonic | CHCl3:MeOH:phosphate buffer, trichloroacetic acid | None | |
| 31 | ASE | DCM:MeOH, 9:1 | Al2O3 column | DCM:MeOH, 1:1 |
| 32 | n.r. | |||
| 33 | Microwave | DCM:MeOH, 9:1 | Hydrolysis, NH2-SiO2 column | DCM:acetone, 9:1 |
| 34 | Ultrasonic | CHCl3:MeOH:phosphate buffer, 2:1:0.8 | None | |
| 35 | Soxhlet | DCM:MeOH, 2:1 | Al2O3 column | DCM:MeOH, 1:2 |
- a ASE, accelerated solvent extraction; n.r., not reported.
| Laboratory Number | HPLC Column | HPLC Gradient | MS | MS Type | MS Method | Integration |
|---|---|---|---|---|---|---|
| 1 | Prevail Cyano | Hex:IPA | Shimadzu 2010A | Single quad | SIM | [M+H]+ ions |
| 2 | Prevail Cyano | Hex:IPA | PE Sciex API 300 | Single quad | Mass scanning | [M+H]+ ions |
| 3 | Prevail Cyano | Hex:IPA | Bruker Esquire 3000+ | Ion trap | Mass scanning | [M+H]+ ions |
| 4 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 5 | Tracer Excel CN | Hex:IPA | Thermo TSQ Quantum | Triple quad | SIM | [M+H]+ ions |
| 6 | Prevail Cyano | Hex:IPA | Agilent 6460A | Triple quad | SIM | [M+H]+ ions |
| 7 | Prevail Cyano | Hex:IPA | Agilent XCT (Bruker) | Ion trap | Mass scanning | n.r. |
| 8 | Prevail Cyano | Hex:IPA | Bruker HCTUltra ETD II | Ion trap | Mass scanning | [M+H]+ ions |
| 9 | Prevail Cyano | Hex:IPA | Micromass Quattro Ultima | Triple quad | SIM | [M+H]+ ions |
| 10 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 11 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 12 | Prevail Cyano | Hex:IPA | Agilent 6120 | Single quad | SIM | [M+H]+ ions |
| 13 | Prevail Cyano | Hex:IPA | Thermo LCQ Deca XP | Ion trap | “SIM” | [M+H]+ ions |
| 14 | n.r. | |||||
| 15 | Prevail Cyano | Hex/DCM:IPA | Agilent 6120 | Single quad | SIM | [M+H]+ ions |
| 16 | Prevail Cyano | Hex:IPA | n.r. | Ion trap | “SIM” | [M+H]+, +1 ions |
| 17 | Prevail Cyano | Hex:IPA | Agilent 6410 | Triple quad | SIM | [M+H]+ ions |
| 18 | Prevail Cyano | Hex:IPA | Bruker Daltonics µTOF | TOF-MS | Mass scanning | [M+H]+, +1 ions |
| 19 | Prevail Cyano | Hex:IPA | Thermo LSQ Fleet | Ion trap | n.r. | n.r. |
| 20 | Prevail Cyano | Hex:IPA | Agilent 6460 | Triple quad | SIM | [M+H]+ ions |
| 21 | Prevail Cyano | Hex:IPA | Micromass Quattro Ultima | Triple quad | SIM | [M+H]+ ions |
| 22 | Prevail Cyano | Hex:IPA | Thermo LTQ Orbitrap XL | Orbitrap | Mass scanning | [M+H]+ ions |
| 23 | Prevail Cyano | Hex:IPA | Thermo TSQ Quantum | Triple quad | SIM | [M+H]+, +1 ions |
| 24 | Prevail Cyano | Hex:IPA | Agilent 6460 | Triple quad | SIM | [M+H]+ ions |
| 25 | Prevail Cyano | Hex:IPA | Thermo LSQ | Ion trap | Mass scanning | [M+H]+, +1 ions |
| 26 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 27 | Prevail Cyano | Hex:IPA | Agilent 1100 | Single quad | SIM | [M+H]+ ions |
| 28 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 29 | Prevail Cyano | Hex:IPA | Agilent 1100 | Single quad | SIM | [M+H]+ ions |
| 30 | Prevail Cyano | Hex:IPA | Agilent 6130 | Single quad | SIM | [M+H]+ ions |
| 31 | Prevail Cyano | Hex:IPA | Agilent 1100 SL | Single quad | SIM | [M+H]+ ions |
| 32 | Prevail Cyano | Hex:IPA | Agilent 1200 SL | Single quad | SIM | [M+H]+ ions |
| 33 | Prevail Cyano | Hex:IPA | Agilent 1200 | Single quad | SIM | [M+H]+ ions |
| 34 | Prevail Cyano | Hex:IPA | Agilent 1200 SL | Single quad | SIM | [M+H]+ ions |
| 35 | Prevail Cyano | Hex:IPA | Thermo TSQ Quantum | Triple quad | SIM | [M+H]+ ions |
- a Hex, hexane; IPA, isopropanol; DCM, dichloromethane; SIM, selected ion monitoring; TOF, time of flight; n.r., not reported.
2.4. Statistical Analysis
[17] The reporting and analysis of the data were performed following some of the recommendations of the IUPAC (International Union of Pure and Applied Chemistry, Basel, Switzerland) for the proficiency testing of analytical chemistry laboratories [Thompson et al., 2006]. The results, i.e., the laboratory means, were assessed using histograms and Tukey box plots. The latter were used as the only means employed to identify outliers, which are those data that fall beyond the whiskers (±1.5 times the difference between the 3rd and 1st quartiles or the interquartile range which includes 25% of all data higher and 25% of all the data lower than the median). Performance of each lab was assessed using the Z-score, which is a measure of the distance between their data and the community mean (see Table S1). In addition, a one-way analysis of variance test was used to analyze the means and calculate the interlaboratory and the intralaboratory variance. Outliers were removed before these calculations. Values of variance have also been expressed in the form of relative standard deviation (equivalent to coefficient of variation, in percentage units). The reproducibility (sR) is the interlaboratory precision, and the repeatability (sr) is an estimate of the reliability of a method for a particular laboratory [Nilsson et al., 1997] that reflects the precision of the analysis of replicate test samples. The repeatability and reproducibility values have also been expressed using confidence intervals as recommended by the ISO 5725 guidelines [International Organization for Standardization, 1986].
3. Results and Discussion
[18] The results of the TEX86 and BIT analyses of the different laboratories are listed in Tables 3 and 4, respectively, and plotted in Figures 2 and 3, respectively, while the extraction methods and HPLC-MS conditions used are summarized in Tables 1 and 2, respectively. The most common method involved extractions using a DCM/MeOH mixture, typically followed by some form of column chromatography (Table 1). The HPLC methods used by the different laboratories are listed in Table 2 and nearly all were similar to that of Schouten et al. [2007] (i.e., a cyano column eluted with a gradient of isopropanol in hexane). However, a variety of mass spectrometry techniques were used: 24 laboratories used quadrupole-MS (15 used a single quadrupole and 9 used a triple quadrupole in single quadrupole mode), 7 laboratories used ion trap-MS, 1 laboratory used a time-of-flight-MS, and 1 laboratory used an Orbitrap MS. Note that most laboratories analyzed the samples within 1–2 days, and thus the reported standard deviations do not represent the long-term laboratory repeatability.
| Laboratory Number | TEX86 A | SD | n | TEX86 B | SD | n | TEX86 C | SD | n | TEX86 F | SD | n |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 0.426 | 0.042 | 2 | 0.510 | 0.024 | 3 | 0.681 | 0.014 | 3 | 0.681 | 0.006 | 8 |
| 2 | 0.603 | 0.006 | 3 | 0.487 | 0.006 | 3 | 0.740 | 0.010 | 3 | 0.740 | 0.007 | 5 |
| 3 | 0.562 | 0.013 | 3 | 0.520 | 0.005 | 3 | 0.686 | 0.002 | 3 | 0.697 | 0.017 | 6 |
| 4 | 0.589 | 0.004 | 3 | 0.562 | 0.008 | 3 | 0.724 | 0.005 | 3 | 0.708 | 0.004 | 5 |
| 5 | 0.574 | 0.001 | 3 | 0.602 | 0.001 | 3 | 0.676 | 0.002 | 3 | 0.677 | 0.005 | 3 |
| 6 | 0.590 | 0.004 | 3 | 0.536 | 0.002 | 3 | 0.694 | 0.003 | 3 | 0.685 | 0.002 | 5 |
| 7 | 0.540 | 0.009 | 3 | 0.553 | 0.005 | 3 | 0.716 | 0.001 | 3 | 0.708 | 0.008 | 5 |
| 8 | 0.577 | 0.013 | 3 | 0.541 | 0.013 | 3 | 0.696 | 0.002 | 3 | 0.695 | 0.003 | 5 |
| 9 | 0.414 | 0.006 | 3 | 0.505 | 0.001 | 3 | 0.662 | 0.003 | 3 | 0.682 | 0.002 | 3 |
| 10 | 0.586 | 0.003 | 3 | 0.542 | 0.005 | 3 | 0.710 | 0.003 | 3 | 0.707 | 0.006 | 3 |
| 11 | 0.490 | 0.000 | 3 | 0.583 | 0.006 | 3 | 0.740 | 0.000 | 3 | 0.720 | 0.000 | 5 |
| 12 | 0.504 | 0.009 | 3 | 0.546 | 0.012 | 3 | 0.709 | 0.010 | 3 | 0.710 | 0.004 | 7 |
| 13 | 0.497 | 0.000 | 1 | 0.476 | 0.000 | 1 | 0.637 | 0.000 | 1 | 0.637 | 0.015 | 4 |
| 14 | 0.453 | 0.000 | 1 | 0.560 | 0.000 | 1 | 0.695 | 0.000 | 1 | 0.699 | 0.014 | 3 |
| 15 | 0.552 | 0.009 | 3 | 0.527 | 0.006 | 3 | 0.680 | 0.002 | 3 | 0.679 | 0.002 | 5 |
| 16 | n.r. | 0.473 | 0.006 | 3 | 0.648 | 0.003 | 3 | 0.645 | 0.003 | 4 | ||
| 17 | 0.491 | 0.019 | 3 | 0.558 | 0.021 | 3 | 0.726 | 0.003 | 3 | 0.738 | 0.002 | 3 |
| 18 | 0.590 | 0.002 | 2 | 0.547 | 0.009 | 3 | 0.711 | 0.002 | 3 | 0.722 | 0.003 | 6 |
| 19 | 0.529 | 0.002 | 3 | 0.545 | 0.001 | 3 | 0.707 | 0.002 | 3 | 0.702 | 0.002 | 3 |
| 20 | 0.547 | 0.010 | 3 | 0.549 | 0.002 | 3 | 0.730 | 0.021 | 3 | 0.740 | 0.095 | 3 |
| 21 | 0.481 | 0.005 | 3 | 0.514 | 0.011 | 3 | 0.664 | 0.004 | 3 | 0.668 | 0.004 | 5 |
| 22 | n.r. | 0.575 | 0.000 | 1 | 0.687 | 0.000 | 1 | 0.685 | 0.003 | 5 | ||
| 23 | 0.528 | 0.043 | 3 | 0.543 | 0.017 | 3 | 0.701 | 0.002 | 3 | 0.682 | 0.007 | 3 |
| 24 | 0.573 | 0.014 | 2 | 0.547 | 0.014 | 2 | 0.710 | 0.000 | 2 | 0.716 | 0.005 | 5 |
| 25 | 0.588 | 0.011 | 3 | 0.573 | 0.019 | 3 | 0.701 | 0.009 | 3 | 0.706 | 0.002 | 5 |
| 26 | 0.553 | 0.031 | 3 | 0.568 | 0.004 | 3 | 0.728 | 0.003 | 3 | 0.729 | 0.004 | 7 |
| 27 | 0.557 | 0.037 | 3 | 0.538 | 0.004 | 3 | 0.709 | 0.001 | 3 | 0.693 | 0.005 | 5 |
| 28 | 0.579 | 0.006 | 3 | 0.544 | 0.011 | 3 | 0.703 | 0.009 | 3 | 0.704 | 0.003 | 3 |
| 29 | 0.520 | 0.030 | 3 | 0.636 | 0.011 | 3 | 0.723 | 0.009 | 3 | 0.717 | 0.015 | 3 |
| 30 | 0.464 | 0.005 | 3 | 0.535 | 0.003 | 3 | 0.702 | 0.002 | 3 | 0.701 | 0.001 | 5 |
| 31 | 0.574 | 0.005 | 3 | 0.520 | 0.000 | 3 | 0.692 | 0.002 | 3 | 0.692 | 0.004 | 5 |
| 32 | 0.646 | 0.006 | 3 | 0.563 | 0.003 | 3 | 0.724 | 0.002 | 3 | 0.717 | 0.003 | 3 |
| 33 | 0.497 | 0.004 | 3 | 0.574 | 0.002 | 3 | 0.731 | 0.001 | 3 | 0.726 | 0.006 | 6 |
| 34 | 0.610 | 0.014 | 3 | 0.569 | 0.048 | 3 | 0.729 | 0.003 | 3 | 0.750 | 0.006 | 5 |
| 35 | 0.533 | 0.009 | 3 | 0.577 | 0.003 | 3 | 0.704 | 0.002 | 3 | 0.704 | 0.007 | 5 |
- a n.r., not reported.
| Laboratory Number | BIT A | SD | n | BIT B | SD | n | BIT C | SD | n | BIT D | SD | n | BIT E | SD | n | BIT F | SD | n |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 0.975 | 0.010 | 3 | 0.285 | 0.147 | 3 | 0.074 | 0.000 | 1 | 0.827 | 0.012 | 7 | 0.670 | 0.008 | 5 | 0.085 | 0.006 | 8 |
| 2 | 0.957 | 0.006 | 3 | 0.267 | 0.006 | 3 | 0.247 | 0.006 | 3 | 0.830 | 0.000 | 5 | 0.706 | 0.009 | 5 | 0.296 | 0.009 | 5 |
| 3 | 0.897 | 0.009 | 3 | 0.021 | 0.000 | 3 | 0.004 | 0.000 | 3 | 0.390 | 0.023 | 6 | 0.136 | 0.013 | 6 | 0.004 | 0.000 | 6 |
| 4 | 0.972 | 0.004 | 3 | 0.156 | 0.022 | 3 | 0.043 | 0.006 | 3 | 0.742 | 0.004 | 5 | 0.458 | 0.004 | 5 | 0.048 | 0.004 | 5 |
| 5 | 0.966 | 0.001 | 3 | 0.124 | 0.013 | 3 | 0.033 | 0.001 | 3 | 0.711 | 0.022 | 3 | 0.438 | 0.019 | 3 | 0.066 | 0.002 | 3 |
| 6 | 0.945 | 0.002 | 3 | 0.092 | 0.005 | 3 | 0.028 | 0.001 | 3 | 0.689 | 0.002 | 5 | 0.374 | 0.003 | 5 | 0.047 | 0.001 | 5 |
| 7 | 0.904 | 0.003 | 3 | 0.028 | 0.001 | 3 | 0.007 | 0.000 | 3 | 0.406 | 0.006 | 5 | 0.155 | 0.008 | 5 | 0.003 | 0.000 | 5 |
| 8 | 0.957 | 0.006 | 3 | 0.067 | 0.001 | 3 | 0.036 | 0.001 | 3 | 0.610 | 0.036 | 5 | 0.354 | 0.002 | 5 | 0.038 | 0.000 | 5 |
| 9 | 0.940 | 0.003 | 3 | 0.075 | 0.003 | 3 | 0.023 | 0.000 | 3 | 0.612 | 0.003 | 3 | 0.330 | 0.005 | 3 | 0.021 | 0.001 | 3 |
| 10 | 0.926 | 0.022 | 3 | 0.067 | 0.006 | 3 | 0.020 | 0.000 | 3 | 0.620 | 0.000 | 3 | 0.280 | 0.000 | 3 | 0.020 | 0.000 | 3 |
| 11 | 0.943 | 0.006 | 3 | 0.070 | 0.000 | 3 | 0.020 | 0.000 | 3 | 0.602 | 0.008 | 5 | 0.322 | 0.004 | 5 | 0.026 | 0.005 | 5 |
| 12 | 0.963 | 0.001 | 3 | 0.110 | 0.002 | 3 | 0.034 | 0.005 | 3 | 0.754 | 0.005 | 5 | 0.441 | 0.019 | 5 | 0.021 | 0.002 | 7 |
| 13 | 0.951 | 0.000 | 1 | n.r. | n.r. | n.r. | n.r. | n.r. | ||||||||||
| 14 | 0.967 | 0.003 | 3 | 0.097 | 0.000 | 1 | 0.036 | 0.000 | 1 | 0.695 | 0.005 | 4 | 0.429 | 0.009 | 3 | 0.034 | 0.002 | 3 |
| 15 | 0.974 | 0.003 | 3 | 0.126 | 0.005 | 3 | 0.055 | 0.002 | 3 | 0.765 | 0.013 | 3 | 0.556 | 0.035 | 3 | 0.061 | 0.002 | 5 |
| 16 | n.r. | 0.157 | 0.002 | 3 | 0.079 | 0.004 | 3 | 0.785 | 0.002 | 3 | 0.582 | 0.006 | 3 | 0.104 | 0.001 | 4 | ||
| 17 | 0.920 | 0.008 | 2 | 0.149 | 0.016 | 3 | 0.043 | 0.002 | 3 | 0.757 | 0.001 | 3 | 0.428 | 0.003 | 3 | 0.040 | 0.001 | 3 |
| 18 | 0.971 | 0.001 | 3 | 0.033 | 0.003 | 3 | 0.008 | 0.000 | 3 | 0.461 | 0.004 | 6 | 0.200 | 0.005 | 6 | 0.021 | 0.005 | 6 |
| 19 | 0.960 | 0.007 | 3 | 0.122 | 0.002 | 3 | 0.026 | 0.002 | 3 | 0.650 | 0.003 | 3 | 0.329 | 0.002 | 3 | 0.024 | 0.001 | 3 |
| 20 | 0.970 | 0.000 | 3 | 0.111 | 0.019 | 3 | 0.030 | 0.000 | 3 | 0.740 | 0.104 | 3 | 0.370 | 0.017 | 3 | 0.037 | 0.006 | 3 |
| 21 | 0.961 | 0.000 | 1 | 0.199 | 0.002 | 3 | 0.081 | 0.002 | 3 | 0.814 | 0.011 | 5 | 0.540 | 0.007 | 5 | 0.080 | 0.000 | 5 |
| 22 | n.r. | 0.151 | 0.000 | 1 | 0.060 | 0.000 | 1 | 0.698 | 0.008 | 6 | 0.433 | 0.007 | 5 | 0.058 | 0.008 | 5 | ||
| 23 | 0.942 | 0.019 | 3 | 0.114 | 0.002 | 3 | 0.033 | 0.001 | 3 | 0.719 | 0.006 | 3 | 0.425 | 0.005 | 3 | 0.058 | 0.003 | 3 |
| 24 | 0.972 | 0.012 | 2 | 0.093 | 0.005 | 2 | 0.028 | 0.002 | 2 | 0.758 | 0.004 | 5 | 0.542 | 0.004 | 5 | 0.020 | 0.000 | 5 |
| 25 | 0.917 | 0.002 | 3 | 0.049 | 0.001 | 3 | 0.011 | 0.001 | 3 | 0.504 | 0.005 | 5 | 0.216 | 0.001 | 5 | 0.014 | 0.000 | 5 |
| 26 | 0.963 | 0.003 | 3 | 0.116 | 0.001 | 3 | 0.031 | 0.002 | 3 | 0.738 | 0.008 | 5 | 0.420 | 0.000 | 5 | 0.040 | 0.010 | 7 |
| 27 | 0.930 | 0.006 | 3 | 0.054 | 0.002 | 3 | 0.014 | 0.001 | 3 | 0.539 | 0.005 | 5 | 0.247 | 0.003 | 5 | 0.019 | 0.001 | 5 |
| 28 | 0.941 | 0.007 | 3 | 0.070 | 0.026 | 3 | 0.017 | 0.002 | 3 | 0.609 | 0.001 | 3 | 0.372 | 0.001 | 3 | 0.022 | 0.007 | 3 |
| 29 | 0.990 | 0.000 | 3 | 0.284 | 0.036 | 3 | 0.181 | 0.052 | 3 | 0.823 | 0.006 | 3 | 0.637 | 0.015 | 3 | 0.183 | 0.006 | 3 |
| 30 | 0.957 | 0.002 | 3 | 0.126 | 0.034 | 3 | 0.037 | 0.002 | 3 | 0.708 | 0.001 | 5 | 0.393 | 0.006 | 5 | 0.040 | 0.001 | 5 |
| 31 | 0.940 | 0.000 | 3 | 0.060 | 0.000 | 3 | 0.010 | 0.000 | 3 | 0.630 | 0.000 | 5 | 0.334 | 0.005 | 5 | 0.010 | 0.000 | 5 |
| 32 | 0.959 | 0.004 | 3 | 0.110 | 0.004 | 3 | 0.040 | 0.001 | 3 | 0.719 | 0.001 | 3 | 0.393 | 0.001 | 3 | 0.034 | 0.000 | 3 |
| 33 | 0.957 | 0.001 | 3 | 0.093 | 0.001 | 3 | 0.024 | 0.001 | 3 | 0.718 | 0.003 | 5 | 0.542 | 0.008 | 5 | 0.024 | 0.002 | 6 |
| 34 | 0.977 | 0.000 | 3 | 0.114 | 0.005 | 3 | 0.027 | 0.001 | 3 | 0.769 | 0.009 | 5 | 0.361 | 0.007 | 5 | 0.025 | 0.001 | 5 |
| 35 | 0.932 | 0.002 | 3 | 0.069 | 0.001 | 3 | 0.024 | 0.001 | 3 | 0.579 | 0.001 | 5 | 0.291 | 0.009 | 5 | 0.031 | 0.004 | 5 |
3.1. TEX86 Analysis of Sediments and Extract
[19] The results of the TEX86 analysis are listed in Table 3, summarized in Table 5, and shown in Figure 2. Sediment C is from a tropical marine environment, the Arabian Sea, while Sediments A and B are from a temperate lake (Salt Pond, USA) and coastal shelf (Carolina Margin), respectively. These different environments are well reflected in the TEX86 values, which are substantially higher (mean value 0.702, median value 0.704) for the tropical sediment than for the sediments from the temperate environments (mean 0.540 and 0.546, respectively, and median 0.552 and 0.546, respectively; Table 5). The results have a reasonably Gaussian-like distribution (Figure 4), with smaller ranges in TEX86 values for Sediment C and Extract F compared to Sediments A and B. We statistically identified (see section 2.4) one outlier for Sediment B (Laboratory 29), which was removed from subsequent statistical treatment.
| Sediment A | Sediment B | Sediment C | Extract F | |
|---|---|---|---|---|
| Number of laboratories reporting results | 33 | 35 | 35 | 35 |
| Number of outliersb | 0 | 1 | 0 | 0 |
| Outliers laboratory number | 29 | |||
| Mean | 0.540 | 0.546 | 0.702 | 0.702 |
| Mean exc. outliers | 0.540 | 0.543 | 0.702 | 0.702 |
| Median | 0.552 | 0.546 | 0.704 | 0.704 |
| Repeatability standard deviation, sr | 0.016 | 0.013 | 0.006 | 0.014 |
| Repeatability relative standard deviation | 3% | 2% | 1% | 2% |
| Repeatability limit | 0.046 | 0.036 | 0.017 | 0.039 |
| Reproducibility standard deviation, sR | 0.053 | 0.030 | 0.023 | 0.027 |
| Reproducibility relative standard deviation | 10% | 5% | 3% | 4% |
| Reproducibility limit | 0.148 | 0.083 | 0.066 | 0.075 |
- a Median, repeatability, and reproducibility are calculated after outlier removal.
- b Outliers were determined from box plots of the data (i.e., data that fall beyond the whiskers).
[20] The estimated repeatability for TEX86, after removal of the outlier, ranged from 0.006 to 0.016 (Table 5). Reproducibility, however, is larger and ranged from 0.023 to 0.053. The better reproducibility and repeatability of Sediment C is probably due to the higher abundances of the minor GDGTs, GDGT-1–GDGT-3 and the crenarchaeol regioisomer, relative to GDGT-0 and crenarchaeol. This likely has enabled a more reliable quantification of these minor compounds, as amounts were not only above the limit of detection but also above the limit of quantification, which is probably an order of magnitude higher than the limit of detection [cf. Schouten et al., 2007]. Oddly enough, Extract F, the extract of Sediment C, has a somewhat worse repeatability and reproducibility than the results of the sediment, in part due to the poor repeatability of the measurement of Extract F by Laboratory 20 (Figure 2d).
[21] If we convert these TEX86 values to temperatures (based on Kim et al. [2008], rather than Kim et al. [2010a], to allow comparison with the round-robin results of Schouten et al. [2009]), then the repeatability of TEX86 analysis ranges from 0.4 to 0.9°C, while the reproducibility ranges from 1.3 to 3.0°C. This performance is better than the initial round-robin exercise of Schouten et al. [2009]. There, for the two samples (an Arabian Sea sediment extract and a Drammensfjord sediment extract) the estimated repeatability was 0.028 (1.6°C) and 0.017 (1.0°C), respectively, with reproducibility of 0.067 (3.8°C) and 0.050 (2.8°C), respectively.
[22] To investigate potential causes for differences among laboratories, we plotted TEX86 values of Sediment B against Sediment C (Figure 5a). The differences between TEX86 measurements are consistent within individual laboratories, i.e., most laboratories tend to have all of their values either lower or higher than the mean. This suggests that the differences between laboratories are not caused by heterogeneity between individual vials of the samples. Rather, it suggests that differences are caused by instrumental characteristics, as previously suggested [Schouten et al., 2009].
[23] The results obtained for TEX86 analysis thus compare well to those obtained in the previous round-robin study, particularly in light of the fact that most results are now obtained for sediments rather than extracts. Intralaboratory precision (repeatability) has improved to <1°C. This suggests that improvements have been made in internal lab consistency over the years, possibly by increased experience with HPLC-MS techniques as speculated by Schouten et al. [2009] and/or improved MS systems with higher sensitivity. There is also some improvement in consistency between labs (i.e., the reproducibility improved) which, when translated into temperature, corresponds to reducing the variability from 3–4 to 1–3°C. The TEX86 analysis now performs relatively well compared to round-robin studies of other paleothermometers. Rosell-Melé et al. [2001] found for
analyses of several sediments a repeatability of 1.6°C, slightly worse than our results, but their reproducibility of 2.1°C is similar or better than obtained in our study. Rosenthal et al. [2004] reported a repeatability of 1–2°C and a reproducibility of 2–3°C for Mg/Ca analysis of foraminifera, numbers which are similar to this study. Hence, it seems that TEX86 measurements in the different geochemical labs have become comparable in robustness and consistency to those of other temperature proxy measurements.
3.2. BIT Analysis of Sediments and Extract
[24] The results of the analysis of the BIT index are displayed in Tables 4 and 6 and Figures 3, 6, and 7b. Sediment C is an open marine setting with a small contribution of soil OM, and thus, values are nearly all below 0.1 (Figures 3c and 6c, Table 4) with a mean value, after outlier removal, of 0.027 (Table 6). Extract F, the extract of Sediment C, had a similar mean BIT value of 0.037 (Figures 3f and 6f and Tables 4 and 6). Sediment A is from a lake (Salt Pond, MA, USA) that likely contains substantial amounts of soil organic carbon. Indeed, higher BIT indices were measured for this sediment than for Sediment C (Figures 3a and 6a, Table 4) with a mean value of 0.951 (Table 6). Sediment B is from a coastal shelf (Carolina Margin), which in principle can contain a range of soil OM input [Schouten et al., 2013, and references cited therein]. The BIT values were consistently low (Figures 3b and 6c, Table 4) with a mean value, after outlier removal, of 0.096 (Table 6), suggesting relatively little input of soil OM in this region.
| Sediment A | Sediment B | Sediment C | Mixture D | Mixture E | Extract F | |
|---|---|---|---|---|---|---|
| Number of laboratories | 33 | 34 | 34 | 34 | 34 | 35 |
| Number of outliersb | 0 | 3 | 4 | 2 | 1 | 2 |
| Outliers I.D. | 1, 2, 29 | 2, 16, 21, 29 | 3, 7 | 2 | 2, 29 | |
| Mean | 0.951 | 0.114 | 0.043 | 0.676 | 0.403 | 0.047 |
| Mean exc. Outliers | 0.951 | 0.096 | 0.027 | 0.690 | 0.384 | 0.037 |
| Median | 0.957 | 0.110 | 0.031 | 0.710 | 0.393 | 0.034 |
| Repeatability standard deviation, sr | 0.007 | 0.011 | 0.002 | 0.017 | 0.009 | 0.004 |
| Repeatability relative standard deviation | 1% | 11% | 7% | 3% | 2% | 11% |
| Repeatability limit | 0.020 | 0.030 | 0.005 | 0.048 | 0.026 | 0.011 |
| Reproducibility standard deviation, sR | 0.022 | 0.042 | 0.013 | 0.107 | 0.139 | 0.024 |
| Reproducibility relative standard deviation | 2% | 43% | 49% | 15% | 36% | 66% |
| Reproducibility limit | 0.062 | 0.177 | 0.038 | 0.299 | 0.390 | 0.068 |
- a Median, repeatability, and reproducibility are calculated after outlier removal.
- b Outliers were determined from box plots of the data (i.e., data that fall beyond the whiskers).
[25] The three sediments and the extract thus have fairly extreme values, close to 0 (Sediments B and C and Extract F; Figure 7b) or close to 1 (Sediment A; Figure 7b). This has consequences for the repeatability and reproducibility, as numbers close to the extreme end of the indices will artificially have a better repeatability and reproducibility. Indeed, repeatability varied between 0.002 and 0.017 and reproducibility varied between 0.013 and 0.042, slightly better than for TEX86 analysis (Tables 5 and 6). The results are similar to the Arabian Sea extract analyzed in the previous round-robin study, which also had a low BIT index value, but much better than the repeatability and reproducibility of the Drammensfjord extract which had an intermediate BIT index value of 0.588 (mean of measurements) [Schouten et al., 2009]. Indeed, the two mixtures of isolated GDGTs, which had intermediate BIT indices, show a much larger repeatability and reproducibility, as their values are not close to the extremes (see 3.4). Nevertheless, the results do suggest that extreme values of BIT indices are fairly consistently measured between laboratories (Figure 7b), in contrast to intermediate values between ∼0.2 and ∼0.8, suggesting at least sediments with relatively “low” or “high” soil OM input can be distinguished.
3.3. Impact of Sample Work Up and Mass Spectrometry Techniques
[26] The inclusion of both sediment and its extract (Sediment C and Extract F, respectively) in the round-robin analysis allowed the impact of extraction methods to be evaluated. Several different extraction techniques were used including ultrasonic, microwave, accelerated solvent and Soxhlet extraction (Table 1). Most of the solvents used consisted of mixtures of DCM and MeOH, although in some cases a “Bligh & Dyer” type extraction, using a buffer [Bligh and Dyer, 1959], was used. The extracts were also processed in different ways including no treatment, hydrolysis, or column separations using SiO2 or Al2O3 (Table 1). A first evaluation can be made by comparing the mean and median BIT and TEX86 values of Sediment C and Extract F. This showed that both TEX86 (0.702 versus 0.702) and BIT values (0.027 versus 0.037) are nearly identical and well within the repeatability limits (Tables 5 and 6). Thus, on a general level the impact of sample processing is relatively small, although it should be noted that a substantial number of the participants used a similar workup for the Sediment C as was used for preparing Extract F. Comparison on the individual laboratory level between TEX86 values obtained from Sediment C and Extract F shows differences varying from −0.019 to 0.021 corresponding to −0.8 to 0.9°C when converted to temperature using Kim et al. [2008]. Differences in BIT values vary only between −0.013 and 0.049. These differences are all relatively minor, suggesting that the type of extraction method and extract processing do not have a large impact on GDGT distributions, in agreement with previous observations [Schouten et al., 2007; Escala et al., 2009; Lengger et al., 2012].
[27] The different types of mass spectrometers used allow us to assess potential differences between mass spectrometry techniques (Table S2). Comparison of the TEX86 and BIT measurements of triple quadrupole mass spectrometers with those of single quadrupole mass spectrometers, the most commonly used technique, shows no significant differences (Student's t test, p > 0.05). This is perhaps not surprising as the triple quadrupole mass spectrometers were used in single quadrupole mode. However, comparison of the results of ion trap mass spectrometers with those of single quadrupole mass spectrometers shows significant differences (Student's t test, p < 0.05), i.e., slightly lower values for TEX86 values of Sediment C (0.684 versus 0.713) and Extract F (0.684 versus 0.710) and substantially lower BIT values for standard Mixtures D (0.558 versus 0.711) and E (0.295 versus 0.445). This may suggest that ion trap mass spectrometers, in general, yield slightly lower TEX86 values but especially lower BIT values.
3.4. Comparison of MS-Based BIT Index and Mass-Based BIT Index
[28] Until now the BIT index, as well as the TEX86, has been an empirical ratio solely based on MS response. The last round-robin exercise demonstrated that this approach had especially large consequences for the BIT index as the extract analyzed with intermediate BIT value (mean 0.588) had a very large spread in values (from 0.340 to 0.821) due to the different MS instrument responses [cf. Escala et al., 2009]. The use of mixtures of GDGT standards in the current round-robin analysis allows, for the first time, a comparison between BIT values measured by the HPLC-MS (BITMS) with those based on mass (BITmass) or moles (BITmol) of GDGTs. Two GDGT mixtures were prepared with different ratios between crenarchaeol versus branched GDGT-I and GDGT-II (i.e., a BITmol of 0.440: Mixture D and 0.158: Mixture E), respectively. As expected, highly variable BITMS values ranging from 0.390 to 0.830 for Mixture D and 0.136 to 0.706 for Mixture E, respectively, were reported (Figures 3d and 3e and Table 4). A broad range of nonuniform distributions was found (Figures 6d and 6e), similar to the previous round-robin results, leading to poor reproducibilities for BIT measurements of the GDGT mixtures (Table 6). Comparison of the BITMS values of each laboratory showed that the trends were consistent between laboratories, i.e., laboratories producing high BITMS values of Mixture D also produced high BITMS values for Mixture E and vice versa (Figure 5). Interestingly, nearly all BITMS values reported were higher than the BITmol of the GDGT mixtures (Figure 5b). Only laboratories 3, 7, and 18 obtained BITMS values similar (i.e., within 0.05) to that of the BITmol of the GDGT mixtures. This suggests that BIT values previously reported in the literature were nearly all overestimating the “true” molar-based BIT index. The cause of the overestimation is likely due to a higher response factor of branched GDGTs compared to crenarchaeol in most of the MS systems used. The three laboratories that have BITMS values close to BITmol values all use a Bruker-manufactured MS, suggesting that certain MS systems may suffer less from this differential MS response. In any case, the results highlight the need to use GDGT standards to properly estimate BITmol values or even concentrations of branched GDGTs [cf. Huguet et al., 2006].
[29] The overestimation of BIT values may have several implications. For example, it likely means that past efforts to estimate the relative amount of soil organic carbon in marine sediments based on the BIT index [e.g., Weijers et al., 2009; Belicka and Harvey, 2009; Kim et al., 2010b] may have overestimated the contribution of soil organic carbon. However, this error is likely minor compared to the assumption that concentrations of branched GDGTs in soil organic carbon and crenarchaeol in marine organic carbon are similar and do not vary over time [cf. Weijers et al., 2009]. Nevertheless, the use of GDGT standards will likely lead to a better assessment of the relative contribution of soil organic carbon in the marine environment. Another point is that BIT values >0.3 are used to indicate potential biases in TEX86 due to input of soil-derived isoprenoid GDGTs [Weijers et al., 2006]. However, as discussed in Schouten et al. [2013], the recommended cutoff value heavily depends on the “terrestrial TEX86” value as well as the concentrations of isoprenoid versus branched GDGTs, which will likely vary on a regional scale. Thus, TEX86 might be biased at BIT values <0.3 or not biased at BIT values >0.3. In our previous round-robin study [Schouten et al., 2009] we suggested to correlate BIT values with TEX86 values and in case of significant correlations use this as a red flag. Our results here do not change this recommendation.
(5)
(6)[31] The calculations show that corrected BITmol values are in a much smaller range than the reported BITMS values and have a more unimodal distribution (Figures 8a and 8b and Table 7). Reproducibility also substantially improves from 0.107 to 0.059, and the mean and median values of the estimated BITmol values, 0.386 and 0.391, are now much closer to the actual BITmol value of 0.440 (Figure 8c). This shows that the use of GDGT standard mixtures can substantially improve interlaboratory consistency of the BIT index and leads to better estimates of BITmol values.
| Laboratory Number | BITMS Mixture E | Correction factor | BITMS Mixture D | SD | n | Estimated BITmol Mixture D |
|---|---|---|---|---|---|---|
| 1 | 0.670 | 10.8 | 0.827 | 0.012 | 7 | 0.307 |
| 2 | 0.706 | 12.8 | 0.830 | 0.000 | 5 | 0.276 |
| 3 | 0.136 | 0.84 | 0.390 | 0.023 | 6 | 0.433 |
| 4 | 0.458 | 4.50 | 0.742 | 0.004 | 5 | 0.390 |
| 5 | 0.438 | 4.15 | 0.711 | 0.022 | 3 | 0.372 |
| 6 | 0.374 | 3.18 | 0.689 | 0.002 | 5 | 0.411 |
| 7 | 0.155 | 0.98 | 0.406 | 0.006 | 5 | 0.412 |
| 8 | 0.354 | 2.92 | 0.610 | 0.036 | 5 | 0.349 |
| 9 | 0.330 | 2.62 | 0.612 | 0.003 | 3 | 0.376 |
| 10 | 0.280 | 2.07 | 0.620 | 0.000 | 3 | 0.441 |
| 11 | 0.322 | 2.53 | 0.602 | 0.008 | 5 | 0.374 |
| 12 | 0.441 | 4.20 | 0.754 | 0.005 | 5 | 0.422 |
| 13 | n.r. | n.r. | ||||
| 14 | 0.429 | 4.00 | 0.695 | 0.005 | 4 | 0.363 |
| 15 | 0.556 | 6.66 | 0.765 | 0.013 | 3 | 0.328 |
| 16 | 0.582 | 7.41 | 0.785 | 0.002 | 3 | 0.330 |
| 17 | 0.428 | 3.98 | 0.757 | 0.001 | 3 | 0.439 |
| 18 | 0.200 | 1.33 | 0.461 | 0.004 | 6 | 0.391 |
| 19 | 0.329 | 2.61 | 0.650 | 0.003 | 3 | 0.416 |
| 20 | 0.370 | 3.12 | 0.740 | 0.104 | 3 | 0.477 |
| 21 | 0.540 | 6.25 | 0.814 | 0.011 | 5 | 0.412 |
| 22 | 0.433 | 4.06 | 0.698 | 0.008 | 6 | 0.363 |
| 23 | 0.425 | 3.93 | 0.719 | 0.006 | 3 | 0.394 |
| 24 | 0.542 | 6.3 | 0.758 | 0.004 | 5 | 0.332 |
| 25 | 0.216 | 1.47 | 0.504 | 0.005 | 5 | 0.409 |
| 26 | 0.420 | 3.85 | 0.738 | 0.008 | 5 | 0.422 |
| 27 | 0.247 | 1.75 | 0.539 | 0.005 | 5 | 0.401 |
| 28 | 0.372 | 3.15 | 0.609 | 0.001 | 3 | 0.331 |
| 29 | 0.637 | 9.34 | 0.823 | 0.006 | 3 | 0.332 |
| 30 | 0.393 | 3.44 | 0.708 | 0.001 | 5 | 0.413 |
| 31 | 0.334 | 2.67 | 0.630 | 0.000 | 5 | 0.390 |
| 32 | 0.393 | 3.44 | 0.719 | 0.001 | 3 | 0.426 |
| 33 | 0.542 | 6.30 | 0.718 | 0.003 | 5 | 0.288 |
| 34 | 0.361 | 3.01 | 0.769 | 0.009 | 5 | 0.526 |
| 35 | 0.291 | 2.18 | 0.579 | 0.001 | 5 | 0.386 |
- a n.r., not reported.
4. Conclusions
[32] Our extensive round-robin study of TEX86 and BIT analyses, involving 35 laboratories and using 3 sediments, 1 sediment extract, and 2 GDGT mixtures, showed that measurements of the TEX86 and BIT index were improved compared to the previous round-robin study, i.e., an improved intralaboratory precision (repeatability) as well as improved consistency (reproducibility) between labs. Importantly, comparison of the results obtained from one sediment and its extract showed that TEX86 and BIT index are not affected substantially by sediment extraction and processing techniques. Comparison of measured BIT values with those of two GDGT mixtures with known ratios of crenarchaeol and branched GDGTs showed that measured BIT values generally overestimate the BIT index based on the molar ratios of the GDGTs. A correction for this different mass spectrometric response based on the GDGT mixture showed a considerable improvement in the reproducibility of BIT index measurements between laboratories, suggesting that standard mixtures should be used in order to obtain consistent BIT values as well as molar-based BIT values.
Acknowledgments
[33] We thank Lloyd Snowdon for anonymizing the lab results and Dr. Feakins and an anonymous reviewer for their useful comments. S.S. thanks the Netherlands Organisation for Scientific Research (NWO) for financial support through a VICI grant and Jaap van der Meer for advice and support on the statistical analysis. A.P. thanks Susan Carter for laboratory assistance and NSF-OCE for funding. A.R.M. thanks Jordi Coello and Núria Moraleda for advice and support on the statistical analysis and Spanish Ministry for research and innovation (MICIIN) for funding. V.G. thanks Xavier Philippon and Carl Johnson for technical assistance. K.G. and M.W. thank the Australian Research Council and John de Laeter Centre for funding toward the LC-MS system, and ARC Fellowship awarded to K.G. C.L.Z. thanks the State Key Laboratory of Marine Geology and the Chinese “National Thousand Talents” program for supporting the LC-MS work performed at Tongji University.
