INTRODUCTION
The radiocarbon content of dissolved organic carbon (DOC) is a powerful tool for distinguishing sources and inputs of organic matter in aquatic systems. Currently, DOC is prepared for 14C analysis by accelerator mass spectrometry (AMS) using one of three offline methods. With the first method, samples are oxidized on a vacuum line using ultraviolet (UV) light (e.g. Armstrong et al. Reference Armstrong, Williams and Strickland1966; Williams Reference Williams1968; Bauer et al. Reference Bauer, Druffel, Williams, Wolast and Griffin1998; Druffel et al. Reference Druffel, Williams, Robertson, Griffin, Jull, Donahue, Toolin and Linick1989; Beaupré et al. Reference Beaupré, Druffel and Griffin2007). UV oxidation has the advantages of extremely low blanks, the ability to analyze saline samples, and large enough volumes (~1 L) to generate sufficient CO2 even for samples with low concentrations of carbon. However, it has the disadvantage that samples are analyzed at a rate of approximately 1–2 per day. In a somewhat similar approach, potassium permanganate instead of UV oxidation has been used to convert organic matter to CO2 in large reactors (500 mL). Two DOC samples can be evaporated and reacted on one vacuum line, then the CO2 subsequently extracted, purified, and trapped on a second vacuum line (Leonard et al. Reference Leonard, Castle, Burr, Lange and Thomas2013). With the third method, samples are freeze-dried in quartz tubes and combusted to CO2 in the presence of cupric oxide, in a similar fashion to solid organic carbon samples. The CO2 generated by this closed tube combustion (CTC) is then either graphitized for analysis on an AMS or is characterized directly with a gas source AMS (Palmer et al. Reference Palmer, Hope, Billett, Dawson and Bryant2001; Neff et al. Reference Neff, Finlay, Zimov, Davydov, Carrasco, Schuur and Davydova2006; Mann et al. Reference Mann, Eglinton, McIntyre, Zimov, Davydova, Vonk, Holmes and Spencer2015). Multiple samples can be prepared simultaneously (subject to number of available ports on the vacuum line), with the time from initial freeze-drying to loading on the AMS taking approximately 3 days.
We present here a new method for the analysis of 14C content of non-saline DOC samples that is based on two recently established protocols. The δ13C analysis of DOC using wet chemical oxidation (WCO) in 12-mL gas-tight Exetainer® vials was recently developed so that samples could be loaded into an automated headspace sampler interfaced with an isotope ratio mass spectrometer (Lang et al. Reference Lang, Bernasconi and Früh-Green2012). The method has the benefit of low blanks and short preparation times, although it is not amenable to saline fluids as chloride interferes with the persulfate oxidation. This oxidation approach was subsequently applied to the compound-specific 14C analysis of the individual volatile organic acids formate and acetate (Lang et al. Reference Lang, Früh-Green, Bernasconi and Wacker2013). The compounds were isolated by high-performance liquid chromatography, collected in Exetainer vials, and chemically oxidized to CO2. The vials were then loaded into an automated headspace sampler interfaced with an AMS (Fahrni et al. Reference Fahrni, Wacker, Synal and Szidat2013). The current procedure combines these previous methods, and demonstrates that non-saline DOC samples, such as those from rivers or lakes, can be similarly analyzed. The method was verified using standards of known isotopic composition, and with freshwater environmental samples that had also been previously analyzed by either UV oxidation at the National Ocean Sciences Accelerator Mass Spectrometry Facility (NOSAMS) or by CTC at ETH-Zürich.
METHODS
Collection of Environmental Samples
Fraser River samples were collected in 2009 from Fort Langley, British Columbia, Canada (49.172°N, –122.577°E). They were filtered through an in-line capsule filter (Pall AcroPak 500 Supor membrane, 0.2 μm size with 0.8-μm prefilter; as in Voss et al. Reference Voss, Peucker-Ehrenbrink, Eglinton, Spencer, Bulygina, Galy, Lamborg, Ganguli, Montluçon, Marsh, Gillies, Fanslau, Epp and Luymes2015) and acidified in the field to pH 2 with ACS certified 85% H3PO4 into precombusted amber glass bottles with acid-washed caps and stored in the dark at room temperature. These samples were prepared for 14C analysis using both UV oxidation at NOSAMS in 2010 and wet chemical oxidation at ETH-Zürich in 2014.
Arctic water samples from the Kolyma River Basin were collected in September 2012. Water samples were collected from the main stem of the Kolyma River (“Arctic stream”) approximately 2 km upstream from Chersky, Russia, and from a small-order permafrost thaw stream (“Permafrost stream”), which drained from an exposure known as Duvanni Yar (Mann et al. Reference Mann, Eglinton, McIntyre, Zimov, Davydova, Vonk, Holmes and Spencer2015; Spencer et al. Reference Spencer, Mann, Dittmar, Eglinton, McIntyre, Holmes, Zimov and Stubbins2015). Samples were filtered through precombusted (450°C) GF/F glass fiber filters to remove particles and stored frozen in acid-washed high-density polyethylene bottles (Mann et al. Reference Mann, Eglinton, McIntyre, Zimov, Davydova, Vonk, Holmes and Spencer2015; Spencer et al. Reference Spencer, Mann, Dittmar, Eglinton, McIntyre, Holmes, Zimov and Stubbins2015). These samples were prepared for 14C analysis at ETH-Zürich using both freeze-drying/CTC (June 2013) and wet chemical oxidation (2013 and 2014, Table 2).
UV Oxidation, NOSAMS
Dissolved organic carbon was oxidized using UV light by the method of Beaupré et al. (Reference Beaupré, Druffel and Griffin2007). A 50–60 g aliquot of sample was added to pre-oxidized Milli-Q™ water to bring concentrations into the normal working range of the system. The evolved CO2 was stripped from water and cryogenically collected, then reduced into graphite with the use of a catalyst in the presence of excess hydrogen gas. The graphite was pressed into target cartridges and analyzed for 14C by AMS at NOSAMS.
Freeze-Drying, ETH-Zürich
Frozen Arctic water samples were thawed and an aliquot was transferred to precombusted (850°C for 5 hr) quartz tubes. Water was removed by freeze-drying and samples were fumigated with acid to remove carbonate. Precombusted CuO was added to the tubes, which were subsequently flame-sealed under vacuum. Organic carbon was converted to CO2 by heating the vials to 850°C for 6 hr. The evolved CO2 was cryogenically quantified, sealed into a glass tube, and loaded for 14C analysis into the MICADAS (Mini Carbon Dating System) at the Laboratory of Ion Beam Physics, ETH-Zürich (Wacker et al. Reference Wacker, Christl and Synal2010, Reference Wacker, Fahrni, Hajdas, Molnár, Synal, Szidat and Zhang2013; Molnár et al. Reference Molnár, Hajdas, Janovics, Rinyu, Synal, Veres and Wacker2013).
Wet Chemical Oxidation, ETH-Zürich
The wet chemical oxidation approach was modified from one recently developed to determine δ13C values of DOC in non-saline water samples (Lang et al. Reference Lang, Bernasconi and Früh-Green2012). The integration between organic compounds oxidized in Exetainer vials and the AMS was adapted from a method to determine the 14C content of volatile organic acids (Lang et al. Reference Lang, Früh-Green, Bernasconi and Wacker2013). In brief, samples were transferred into 12-mL Exetainer screw-capped vials with a butyl rubber septum (Labco, Buckinghamshire, UK, P/N 938W). A 1-mL aliquot of acidified sodium persulfate solution (100 mL H2O + 4.0 g Na2S2O8 + 200 µL of 85% H3PO4) was added as an oxidant and samples were sealed and purged with high-purity helium gas (Grade 5.0, 99.9999% pure, for 8 min at >100 mL/min) to eliminate inorganic CO2 from the vial. The samples were then heated to 95°C for 1 hr to convert any sample DOC to CO2. All glassware was precombusted at 500°C for 5 hr to remove organic contaminants. Further specifics on optimizing the oxidation conditions and minimizing processing blanks can be found in Lang et al. (Reference Lang, Bernasconi and Früh-Green2012).
To determine the 14C content of the evolved CO2, the samples were loaded into the carbonate handling system of the MICADAS accelerator mass spectrometer (AMS) equipped with a gas-accepting ion source (GIS) (Synal et al. Reference Synal, Stocker and Suter2007; Ruff et al. Reference Ruff, Wacker, Gäggeler, Suter, Synal and Szidat2007; Molnár et al. Reference Molnár, Hajdas, Janovics, Rinyu, Synal, Veres and Wacker2013; Wacker et al. Reference Wacker, Fahrni, Hajdas, Molnár, Synal, Szidat and Zhang2013). This gas transfer system automatically moves the CO2 in septum-sealed vials over a magnesium perchlorate water trap onto a X13 zeolite molecular sieve (sodium aluminosilicate) at room temperature. The zeolite trap is then rapidly heated to 450°C to release the CO2, which is then transferred to a gas-tight syringe. An appropriate amount of helium is added to the syringe to dilute the gas to a 5% v/v CO2 in helium, and the plunger is depressed slowly to feed the mixture into the GIS at a constant rate. The carbonate handling system was modified with the addition of a sparging needle to strip any CO2 dissolved in the water. A second, shorter needle carried the displaced sample CO2 gas from the vial headspace to the zeolite trap. Further specifics on the coupling of the Exetainer samples to the AMS can be found in Molnár et al. (Reference Molnár, Hajdas, Janovics, Rinyu, Synal, Veres and Wacker2013), Wacker et al. (Reference Wacker, Fahrni, Hajdas, Molnár, Synal, Szidat and Zhang2013), and Lang et al. (Reference Lang, Früh-Green, Bernasconi and Wacker2013).
The raw 14C data are reported as fraction modern (F14C) after Reimer et al. (Reference Reimer, Brown and Reimer2004), and after correction for instrumental background, standard normalization, and evaluation of uncertainty using the software program BATS (Wacker et al. Reference Wacker, Christl and Synal2010). An additional correction was made for contamination introduced during the isolation and oxidation procedures (the processing blank), as detailed below.
A batch of 22 samples and 10 standards can be prepared in approximately 4–6 hr. Transferring the samples and standards into clean Exetainer vials and adding the oxidant requires 1–2 hr. Two vials can be flushed with helium simultaneously, with the batch completely purged within ~2.5 hr. All vials then react on the block heater for 1 hr. The helium flushing time is the rate determining step since samples can be transferred while this is ongoing; adding additional purging stations and/or automating this step would further reduce preparation times. Typically, samples were allowed to cool to room temperature overnight before loading them onto the AMS autosampler. Once the AMS had been focused and pure gas standards had been analyzed for calibration, the batch of 32 samples plus standards could be processed within ~4 hr.
Verification Approach
Two approaches were used to verify the method. First, two powdered standards with known F14C signatures were dissolved in high-purity Milli-Q water over a range of concentrations and analyzed for F14C content. The two standards, phthalic acid (Sigma Aldrich P/N 8001-100g, ≥99.5% purity, Lot 1431342V, δ13C = –12.4‰, F14C<0.0025, ETH-47292) and sucrose (Sigma Aldrich P/N S7903-250g, ≥99.5% purity, Lot 090M02112V, δ13C = –33.6‰, F14C = 1.053 ± 0.003, ETH-47293), were chosen for their distinct isotope signatures, solubility in water and, in the case of phthalic acid, and chemical recalcitrance. The standards were prepared in 4 mL of Milli-Q water in concentration ranges from 83–833 µmol C/L, corresponding to 4–40 µg of organic carbon total (Figure 1). This range was chosen to represent the approximate concentrations of DOC in rivers and lakes and to cover the lower end of the MICADAS sample size capacity. The generated data were used to both verify the method and to determine the size and isotopic composition of the blank.
F14C versus µg C of sucrose (upper panels, F14C = 1.053 ± 0.003) and phthalic acid (lower panels, F14C < 0.002). The solid line in both panels represents the idealized mixture between the standards and a blank with characteristics determined for that particular run (see Table 1). Dashed lines represent 95% confidence intervals. Individual markers are not corrected for blanks; the y axis error bars represent the instrument error only (±1σ).
Second, riverine samples from the Fraser River and the Arctic were analyzed by persulfate oxidation and compared to the F14C values determined on the same samples by other means, either UV oxidation (NOSAMS) or freeze-drying and closed tube combustion (ETH). The raw F14C data generated from the riverine samples were corrected for the presence of a blank using the sucrose and phthalic acid standards (Figure 2).
Corrected F14C values of phthalic acid standards (upper plot, F14C < 0.002) and sucrose standards (lower plot, F14C = 1.053 ± 0.003) versus µg C. Individual analyses from sequence C130304CM1G (empty squares), C130419G (black triangles), and C140708SL1G (gray triangles) are plotted with error bars representing the propagated error of analyses.
RESULTS AND DISCUSSION
Standards with Known F14C Content
14C analysis of organic matter is highly sensitive to contributions from extraneous carbon and, since this extraneous carbon is frequently too small to analyze directly, the size and isotope composition of the processing blank is instead constrained by analyzing standards of known and distinct 14C content in a similar fashion as the samples (Pearson et al. Reference Pearson, McNichol, Schneider and von Reden1998; Santos et al. Reference Santos, Southon, Griffin, Beaupre and Druffel2007; Shah and Pearson Reference Shah and Pearson2007; Mollenhauer and Rethemeyer Reference Mollenhauer and Rethemeyer2009; Ziolkowski and Druffel Reference Ziolkowski and Druffel2009; Lang et al. Reference Lang, Früh-Green, Bernasconi and Wacker2013). As has been observed with other analyses of small amounts of organic carbon, the standards analyzed by the WCO method had F14C contents similar to that of the powdered standards at high concentrations. At lower concentrations, these values converge towards the isotope signature of the blank (Figure 1).
The data from the standards were used to calculate the size and isotope composition of the blank for each suite of samples (Table 1). Processing blanks from the three different runs ranged from 0.68 ± 0.26 to 1.05 ± 0.23 µg C with F14C values of 0.170 ± 0.051 to 0.274 ± 0.151. The size of the blanks is similar to the contribution of extraneous carbon from CTC designed for small (<25 µg C) samples. For example, Santos et al. (Reference Santos, Southon, Griffin, Beaupre and Druffel2007) determined the blank associated with closed tube combustion on their system using 14C-free coal and modern OX-1 to be 0.2–1 μg of modern and 0.1–0.5 μg of 14C-free carbon.
Composition of the WCO processing blank determined for each batch of samples, as determined using two standards (sucrose and phthalic acid).
For the environmental samples analyzed here, with concentrations of 200–1700 μM, the blank contributed ~0.6–4.6% of the total measured carbon. Analytical approaches that use larger sample volumes have a similar contribution of the blank to the amount of carbon analyzed since they are designed for samples with much lower DOC concentrations. The blanks of an improved UV oxidation method are reported to be <2 μM using a 1-L reaction vial (Beaupré et al. Reference Beaupré, Druffel and Griffin2007). For even the lowest concentrations of seawater DOC of 36 μM, this would contribute only 5% of the total measured C.
The variability in size and composition of the blank emphasizes the importance of determining the processing blank independently for each suite of analyses. This variability could be caused by the introduction of small amounts of carbon to the water, vials, reagents, user error, or instrument variability. The relatively large number of samples that can be processed simultaneously by the WCO method makes the analysis of 10–12 standards for each run feasible, and is strongly recommended.
Environmental Samples
Five freshwater samples were analyzed by the current WCO method and, after correcting for the processing blank associated with each run, had measured F14C values from 0.128 ± 0.003 to 1.082 ± 0.015 (Table 2). In both the modern and 14C-free samples, the absolute errors translate to a similar relative percent error (1.8% vs. 1.5%).
Environmental samples analyzed by WCO method.
a Percent recovery is the comparison of measured µg C to the expected µg C, based on the volume of sample that was oxidized and the DOC concentration as determined by high-temperature combustion (see Methods).
The reproducibility of the analysis was determined by analyzing two samples multiple times in the same preparation run and in different preparation runs. The older permafrost-fed stream had an average F14C value of 0.130 ± 0.002 (n = 3), with a variability similar to the propagated measurement error associated with each individual analysis. Replicates for the modern Fraser River sample GRO000019 yielded an average F14C of 1.069 ± 0.019 (n = 2), which has a variability somewhat higher than the error associated with the individual analyses. For comparison, the average F14C of replicates of the same modern sample analyzed by UV oxidation was 1.071 ± 0.011 (n = 2). The larger differences in reproducibility in the modern samples likely reflects both the lower amounts of carbon analyzed in the Fraser River sample (18.0–22.8 µg C) compared to the permafrost stream sample (31.5–117 µg C), as well as a greater influence of the processing blank (Fm ~ 0.2) on the more modern samples (Figure 3; Table 2). Larger absolute corrections must be made to the lower concentration, modern samples.
Comparison of F14Cmeas (gray circles) and F14Ccorr (empty triangles) versus μg C. Error bars are either instrument error (F14Cmeas) or propagated error (F14Ccorr). For samples with large amounts of carbon and/or F14C values similar to the processing blank (0.170–0.274), the marker points overlap. Modern samples with low amounts of carbon required the largest absolute corrections to account for the presence of extraneous carbon during processing.
For GRO000019, the blank correction leads to values that disagree more, not less, with each other. In this case, F14Cmeas on the two dates is 1.041 ± 0.011 and 1.027 ± 0.013, a difference of 0.14 that is within approximately 1 standard deviation, while F14Ccorr is 1.082 ± 0.015 and 1.056 ± 0.019, a difference of 0.026 or greater than 1 standard deviation. The small number of replicates make it difficult to state with certainty the underlying cause of this observation. One possibility is that the size of the blank has been overestimated, particularly for the samples analyzed in March 2013, when fewer processing standards were used. In the current method, the size of the blank has been determined using only pure compounds of known isotopic value. One approach to improving reproducibility between analytical runs would be to analyze an environmental sample of constant and well-known composition, similar to working standards used to correct for drift in stable isotope analyses, or the deep-ocean water provided for DOC concentration analysis. The regular use of an environmental working standard would also allow calibration of 14C of DOC values across laboratories using multiple different methods.
The yield of CO2 generated by WCO was determined by comparing the expected μg C, based on the concentration of DOC in the sample and the volume oxidized, and the measured μg C, based on the amount of gas recovered in the AMS GasTight syringe. Yields ranged from 87–101% for the Fraser River samples and 115–128% for the Arctic samples (Table 2). Lower recoveries for the Fraser River samples may be due, in part, to incomplete stripping of the CO2 from samples with larger water volumes (>4 mL). The >100% recoveries observed with the Arctic samples by WCO was similar to the values determined by closed tube combustion; the Permafrost stream sample had a recovery of 113%. These values may therefore point to an issue inherent to these particular samples, e.g. that additional carbon was added after they were analyzed for DOC concentrations but before they were analyzed for 14C content by WCO and CTC. Alternatively, since the recoveries are based on the amount of gas trapped in the GasTight syringe, the presence of an interfering gas such as SO2 could also result in the higher-than-expected values.
The F14C values of these five samples were also assessed by alternate means for comparison to the current method (Table 3). The three modern riverine samples had been previously analyzed using UV oxidation at NOSAMS. The F14C values determined by the new WCO method had F14C values that were lower than the NOSAMS values by –0.002 to –0.065 (Table 3). Two additional samples, one with a modern and one with an ancient 14C signature, were analyzed by both WCO and by CTC at ETH. Values determined by the WCO method differed by+0.012 and –0.034, respectively.
Summary comparison of environmental samples analyzed by wet chemical oxidation, UV oxidation, and quartz tube combustion. For F14C, propagated errors incorporate both the measurement error and the correction for the presence of the processing blank. In cases where a sample was analyzed multiple times, the standard deviation of multiple analyses is reported. (n.d. is not determined).
a Data determined by Shimadzu TOC-V analyzer; from Voss et al. (Reference Voss, Peucker-Ehrenbrink, Eglinton, Spencer, Bulygina, Galy, Lamborg, Ganguli, Montluçon, Marsh, Gillies, Fanslau, Epp and Luymes2015) and Mann et al. (Reference Mann, Eglinton, McIntyre, Zimov, Davydova, Vonk, Holmes and Spencer2015).
The offset between the values generated by the WCO method and the other two methods could have multiple potential sources that are difficult to constrain at this time. The largest offset of 0.065 is observed for Fraser River sample GRO000018, which has the lowest DOC concentration (199 µM) and the lowest amount of carbon analyzed by WCO (12.5 µg C). At these low amounts of carbon, the precision of the AMS measurement is somewhat decreased and could contribute to the offset. Additionally, this sample was analyzed with only six standards (3 phthalic acid, 3 sucrose); therefore, the blank was less precisely constrained than for other samples. Finally, minor differences in the sample itself may have arisen during storage. The NOSAMS analyses were performed in 2010, shortly after sample collection. The WCO of sample aliquots that were collected at the same time into different containers were analyzed ~3 yr later.
Difference in values may also arise between those analyzed by WCO, UV oxidation, and freeze-drying as a result of variable amounts of purging time. While 5-mL samples are purged for 8 min in the WCO method, the larger volume samples analyzed by UV oxidation are purged for >1 hr. In the quartz-tube combustion method, samples are freeze-dried, then subjected to vapor phase acidification. While each approach will fully remove inorganic carbon, these different methodological approaches may strip different proportions of small semi-volatile organic compounds. Several studies have demonstrated that compounds such as formate and acetate are partially, but not completely, removed in acidified samples purged with a gas (Barcelona Reference Barcelona1980; Lang et al. Reference Lang, Butterfield, Schulte, Kelley and Lilley2010). Presumably, other small organic molecules with similar attributes will behave similarly.
Finally, differences may arise due to the capability of the different oxidant approaches to convert particularly recalcitrant organic molecules to CO2. While concentrations of DOC determined by WCO are identical to those determined by high-temperature combustion (Benner and Strom Reference Benner and Strom1993; Sharp et al. Reference Sharp, Benner, Bennett, Carlson, Fitzwater, Peltzer and Tupas1995), incomplete oxidation of particularly unreactive molecules cannot be ruled out. Using a UV oxidation system, Beaupré et al. (Reference Beaupré, Druffel and Griffin2007) demonstrated that seawater DOC is converted to CO2 as a continuum, with later reacting recalcitrant components depleted in 14C relative to the bulk.
Assessment
These initial tests demonstrate the utility of a WCO approach for determination of 14C contents of DOC, although additional improvements could further expand its efficacy and applicability. One advantage of this method is that the preparation time is relatively short, allowing for higher throughput than UV oxidation or, in some cases, freeze-drying. This is particularly so when the AMS is equipped with an autosampler that can rapidly introduce the sample to a CO2-gas-accepting ion source. In addition to simply being able to process more samples in a single day, the current method also simplifies the concurrent analysis of multiple processing standards over a well-controlled concentration range. On a vacuum line, there are frequently fewer than 10 ports available for the quartz tubes used for CTC, making the preparation of a large number of standards per batch overly time consuming. Additionally, preparing standards for CTC in amounts of <25 μg C can be challenging due to the difficultly in weighing out such small amounts of a powdered standard into the quartz tubes. Instead, larger standards (1 mg) are often combusted and subsequently split into smaller aliquots of gas for analysis (Santos et al. Reference Santos, Southon, Griffin, Beaupre and Druffel2007). Because the standards for the WCO method are prepared from a concentrated liquid stock, a precise volume can be easily distributed by pipette.
A second, less obvious, advantage of this approach is that it significantly decreases the potential for cross-contamination of samples, particularly those that have inadvertent contamination from 14C tracers. While great care must still be taken to ensure that samples are not contaminated with tracer 14C, the WCO method minimizes the damage that can result from a contaminated sample. Each sample is processed and oxidized independently using single-use, disposable glassware. There is a risk of cross-contamination during the sparging step, as the same needle is used to purge each sample. However, replacing a contaminated needle is significantly less costly and time consuming than cleaning numerous components of a vacuum or graphitization line. Once oxidized, the CO2 is automatically transferred from the vial into the AMS. If the operator of the AMS notices a “hot” sample, the run can be immediately terminated, precluding subsequent contamination of later samples. Some carryover does exist on the AMS system itself, most likely related to the gas lines, water trap, and zeolite traps. Repeat injections of 14C-free CO2 and sparging with helium overnight removes this contamination without the need to disassemble the autosampler or replace the lines. It is because of these attributes that researchers at ETH-Zürich have adopted WCO as the oxidation approach as a screening tool when identifying 14C contamination (McIntyre et al. Reference McIntyre, Lechleitner, Lang, Wacker, Fahrni and Eglinton2014).
Future developments should focus on expanding the analysis to saline samples and improving precision; both improvements could potentially be accomplished by increasing sample volume. Interference of Cl– ions with the oxidation currently limits the analysis to freshwater samples, precluding the analysis of seawater. Instruments that use sodium persulfate for oxidation for the 13C analysis of DOC have overcome this challenge in part by increasing the amount of oxidant relative to sample (Osburn and St-Jean Reference Osburn and St-Jean2007), which may also provide a solution for this WCO method. The second challenge is the volume limitation imposed by using 12-mL Exetainer vials. Because the CO2 is subsampled from the headspace, the total liquid volume (sample + oxidant) is limited to approximately 7 mL maximum. Increasing the volume of the sample analyzed would allow more CO2 to be introduced to the AMS, improving counting statistics and therefore instrument precision. The additional carbon would simultaneously decrease the influence of the blank and further improve the quality of the data. Larger sample vials have been used for dissolved inorganic carbon on this AMS (Molnár et al. Reference Molnár, Hajdas, Janovics, Rinyu, Synal, Veres and Wacker2013) and could potentially be adapted for use with the WCO method.
ACKNOWLEDGMENTS
The Arctic water samples were generously donated by Drs Robert Spencer and Paul Mann with support from US National Science Foundation grants ANT-1203885 and PLR-1500169. We thank colleagues Bernhard Peucker-Ehrenbrink (WHOI) and Daniel Montluçon (ETH) for assistance with collecting the Fraser River samples and Li Xu and other NOSAMS staff for sample analysis. SQL and GF-G gratefully acknowledge support by the Swiss National Science Foundation project 200020-143891 and the Deep Carbon Observatory–Deep Energy subaward 60040915. BMV and Fraser River sampling were supported by US National Science Foundation grant OCE-0851015. The manuscript was greatly improved by suggestions of two anonymous reviewers and we thank them for their time and insights.
