Large-scale climatic forcing is impacting oceanic biogeochemical cycles and is expected to influence the water-column distribution of trace gases, including methane and nitrous oxide. Our ability as a scientific community to evaluate changes in the water-column inventories of methane and nitrous oxide depends largely on our capacity to obtain robust and accurate concentration measurements that can be validated across different laboratory groups. This study represents the first formal international intercomparison of oceanic methane and nitrous oxide measurements whereby participating laboratories received batches of seawater samples from the subtropical Pacific Ocean and the Baltic Sea. Additionally, compressed gas standards from the same calibration scale were distributed to the majority of participating laboratories to improve the analytical accuracy of the gas measurements. The computations used by each laboratory to derive the dissolved gas concentrations were also evaluated for inconsistencies (e.g., pressure and temperature corrections, solubility constants). The results from the intercomparison and intercalibration provided invaluable insights into methane and nitrous oxide measurements. It was observed that analyses of seawater samples with the lowest concentrations of methane and nitrous oxide had the lowest precisions. In comparison, while the analytical precision for samples with the highest concentrations of trace gases was better, the variability between the different laboratories was higher: 36 % for methane and 27 % for nitrous oxide. In addition, the comparison of different batches of seawater samples with methane and nitrous oxide concentrations that ranged over an order of magnitude revealed the ramifications of different calibration procedures for each trace gas. Finally, this study builds upon the intercomparison results to develop recommendations for improving oceanic methane and nitrous oxide measurements, with the aim of precluding future analytical discrepancies between laboratories.
The increasing mole fractions of greenhouse gases in the Earth's atmosphere are causing long-term climate change with unknown future consequences. Two greenhouse gases, methane and nitrous oxide, together contribute approximately 23 % of total radiative forcing attributed to well-mixed greenhouse gases (Myhre et al., 2013). It is imperative that the monitoring of methane and nitrous oxide in the Earth's atmosphere is accompanied by measurements at the Earth's surface to better inform the sources and sinks of these climatically important trace gases. This includes measurements of dissolved methane and nitrous oxide in the marine environment, which is an overall source of both gases to the overlying atmosphere (Nevison et al., 1995; Anderson et al., 2010; Naqvi et al., 2010; Freing et al., 2012; Ciais et al., 2014).
Oceanic measurements of methane and nitrous oxide are conducted as part of established time series locations, along hydrographic survey lines, and during disparate oceanographic expeditions. Within low-latitude to midlatitude regions of the open ocean, the surface waters are frequently slightly supersaturated with respect to atmospheric equilibrium for both methane and nitrous oxide. There is typically an order of magnitude range in concentration along a vertical water-column profile at any particular open ocean location (e.g., Wilson et al., 2017). In contrast to the open ocean, nearshore environments that are subject to river inputs, coastal upwelling, benthic exchange, and other processes have higher concentrations and greater spatial and temporal heterogeneity (e.g., Schmale et al., 2010; Upstill-Goddard and Barnes, 2016).
List of laboratories that participated in the intercomparison. All laboratories measured both methane and nitrous oxide except the U.S. Geological Survey (methane only), UC Santa Barbara (nitrous oxide only), and NOAA PMEL (nitrous oxide from the Pacific Ocean). Also indicated are the 12 laboratories that received the SCOR gas standards of methane and nitrous oxide.
Methods for quantifying dissolved methane and nitrous oxide have evolved and somewhat diverged since the first measurements were made in the 1960s (Craig and Gordon, 1963; Atkinson and Richards, 1967). Some laboratories employ purge-and-trap methods for extracting and concentrating the gases prior to their analysis (e.g., Zhang et al., 2004; Bullister and Wisegarver, 2008; Capelle et al., 2015; Wilson et al., 2017). Others equilibrate a seawater sample with an overlying headspace gas and inject a fixed volume of the gaseous phase into a gas analyzer (e.g., Upstill-Goddard et al., 1996; Walter et al., 2005; Farías et al., 2009). The purge-and-trap technique is typically more sensitive by 1–2 orders of magnitude over headspace equilibrium (Magen et al., 2014; Wilson et al., 2017). However, the purge-and-trap technique requires more time for sample analysis and it is more difficult to automate the injection of samples into the gas analyzer. Headspace equilibrium sampling is most suited for volatile compounds that can be efficiently partitioned into the headspace gas volume from the seawater sample. To compensate for its limited sensitivity, a large volume of seawater can be equilibrated (e.g., Upstill-Goddard et al., 1996). Additional developments for continuous underway surface seawater measurements use equilibrator systems of various designs coupled to a variety of detectors (e.g., Weiss et al., 1992; Butler et al., 1989; Gülzow et al., 2011; Arévalo-Martínez et al., 2013). Determining the level of analytical comparability between different laboratories for discrete samples of methane and nitrous oxide is an important step towards improved comprehensive global assessments. Such intercomparison exercises are critical to determining the spatial and temporal variability of methane and nitrous oxide across the world oceans with confidence, since no single laboratory can single-handedly provide all the required measurements at sufficient resolution. Previous comparative exercises have been conducted for other trace gases, e.g., carbon dioxide, dimethylsulfide, and sulfur hexafluoride (Dickson et al., 2007; Bullister and Tanhua, 2010; Swan et al., 2014), and for trace elements (Cutter, 2013). These exercises confirm the value of the intercomparison concept.
To instigate this process for methane and nitrous oxide, a series of
international intercomparison exercises were conducted between 2013 and 2017,
under the auspices of Working Group no. 143 of the Scientific Committee on
Oceanic Research (SCOR). Discrete seawater samples
collected from the subtropical Pacific Ocean and the Baltic Sea were
distributed to the participating laboratories (Table 1). The samples were
selected to cover a representative range of concentrations across marine
locations, from the oligotrophic open ocean to highly productive waters, and
in some instances sub-oxic coastal waters. An integral component of the
intercomparison exercise was the production and distribution of methane and
nitrous oxide gas standards to members of the SCOR Working Group. The
intercomparison exercise was conceived and evaluated with the following four
questions in mind.
What is the agreement between the SCOR gas standards and the
“in-house” gas standards used by each laboratory? How do measured values of dissolved methane and nitrous oxide compare
across laboratories? Despite the use of different analytical systems, are there general
recommendations to reduce uncertainty in the accuracy and precision of
methane and nitrous oxide measurements? What are the implications of interlaboratory differences for
determining the spatial and temporal variability of methane and nitrous oxide
in the oceans?
Laboratory-based measurements of oceanic methane and nitrous oxide require separation of the dissolved gas from the aqueous phase, with the analysis conducted on the gaseous phase. Calibration of the analytical instrumentation used to quantify the concentration of methane and nitrous oxide is nearly always conducted using compressed gas standards, the specifics of which vary between laboratories. Therefore, the reporting of methane and nitrous oxide datasets ought to be accompanied by a description of the standards used, including their methane and nitrous oxide mole fractions, the declared accuracies, and the composition of their balance or “makeup” gas. For both gases, the highest-accuracy commercially available standards have mole fractions close to current-day atmospheric values. These standards can be obtained from national agencies including the National Oceanic and Atmospheric Administration Global Monitoring Division (NOAA GMD), the National Institute of Metrology China, and the Central Analytical Laboratories of the European Integrated Carbon Observation System Research Infrastructure (ICOS-RI). By comparison, it is more difficult to obtain highly accurate methane and nitrous oxide gas standards with mole fractions exceeding modern-day atmospheric values. This is particularly problematic for nitrous oxide due to the nonlinearity of the widely used electron capture detector (ECD) (Butler and Elkins, 1991).
The absence of a widely available high mole fraction, high-accuracy nitrous
oxide gas standard was noted as a primary concern at the outset of the
intercomparison exercise. Therefore, a set of high-pressure primary gas
standards was prepared for the SCOR Working Group by John Bullister and David
Wisegarver at NOAA Pacific Marine and Environmental Laboratory (PMEL). One
batch, referred to as the air ratio standard (ARS), had methane and nitrous oxide
mole fractions similar to modern air, and the other batch, referred to as
the water ratio standard (WRS), had higher methane and nitrous oxide mole
fractions for the calibration of high-concentration water samples. These SCOR
primary standards were checked for stability over a 12-month period and
assigned mole fractions on the same calibration scale, known as
“SCOR-2016”. A comparison was conducted with NOAA standards prepared on the
SIO98 calibration scale for nitrous oxide and the NOAA04 calibration scale
for methane. Based on the comparison with NOAA standards, the uncertainty of
the methane and nitrous oxide mole fractions in the ARS and the uncertainty
of the methane mole fraction in the WRS were all estimated at
Pertinent information for each batch of methane and nitrous oxide samples. This includes contextual hydrographic information, median and mean concentrations of methane and nitrous oxide, range, number of outliers, and the overall average coefficient of variation (%).
Dissolved methane and nitrous oxide samples for the intercomparison exercise
were collected from the subtropical Pacific Ocean and the Baltic Sea. Pacific
samples were obtained on 28 November 2013 and 24 February 2017 from the
Hawai'i Ocean Time-series (HOT) long-term monitoring site, station ALOHA,
located at 22.75
Samples from the western Baltic Sea were collected during
15–21 October 2016 onboard the R/V
Each laboratory measured dissolved methane and nitrous oxide slightly differently. A full description of each laboratory's method can be found in Tables S6 and S7 in the Supplement for methane and nitrous oxide, respectively.
The majority of laboratories measured methane and nitrous oxide by equilibrating the seawater sample with an overlying headspace and subsequently injecting a portion of the gaseous phase into the gas analyzer. This method has been conducted since the 1960s when gas chromatography was first used to quantify dissolved hydrocarbons (McAuliffe, 1963). The headspace was created using helium, nitrogen, or high-purity air to displace a portion of the seawater sample within the sample bottle. Alternatively, a subsample of the seawater was transferred to a gastight syringe and the headspace gas subsequently added. The volume of the vessel used to conduct the headspace equilibration ranged from 20 mL borosilicate glass vials to 1 L glass vials and syringes used by Newcastle University and the U.S. Geological Survey, respectively. The dissolved gases equilibrated with the overlying headspace at a controlled temperature for a set period of time that ranged from 20 min to 24 h for the different laboratories. The longer equilibration times are due to overnight equilibrations in water baths. The majority of laboratories enhanced the equilibration process by some initial period of physical agitation. After equilibration, an aliquot of the headspace was transferred into the gas analyzer (GA) by either physical injection, displacement using a brine solution, or injection using a switching valve. Some laboratories incorporated a drying agent and a carbon dioxide scrubber prior to analysis. The gas sample passed through a multi-port injection valve containing a sample loop of known volume, which transferred the gas sample directly onto the analytical column within the oven of the GA. Calibration of the instrument was achieved by passing the gas standards through the injection valve.
The final gas concentrations using the headspace equilibration method were
calculated by
In contrast to the headspace equilibrium method, five laboratories used a
purge-and-trap system for methane and/or nitrous oxide analysis (Tables S6
and S7 in the Supplement). These systems were directly coupled to a flame
ionization detector (FID) or ECD, with the exception of the University of British
Columbia, where a quadrupole mass spectrometer with an electron impact ion
source and Faraday cup detector were used (Capelle et al., 2015). The
purge-and-trap systems were broadly similar, each transferring the seawater
sample to a sparging chamber. Sparging times typically ranged from 5–10 min
and the sparge gas was either high-purity helium or high-purity nitrogen. In
addition to commercially available gas scrubbers, purification of the sparge
gas was achieved by passing it through stainless steel tubing packed with
Poropak Q and immersed in liquid nitrogen. This is a recommended precaution
to consistently achieve a low blank signal of methane. The elutant gas was
dried using Nafion or Drierite and subsequently cryotrapped on a sample loop
packed with Porapak Q to aid the retention of methane and nitrous oxide.
Cryotrapping was achieved for methane using liquid nitrogen
(
The final concentrations of methane and nitrous oxide are reported in
nmol kg
Six laboratories compared their existing “in-house” standards of methane with the SCOR standards. This was done by calibrating in-house standards and deriving a mixing ratio for the SCOR standards, which were treated as unknowns. Four laboratories reported methane values for either the ARS or WRS within 3 % of their absolute concentration, whereas two laboratories reported an offset of 6 % and 10 % between their in-house standards and the SCOR standards (Table S6 in the Supplement). For those laboratories who measured the SCOR standards to within 3 % or better accuracy, observed offsets in methane concentrations from the overall median cannot be due to the calibration gas.
Concentrations of methane measured in nine separate seawater samples
collected from the Pacific Ocean
Seven laboratories compared their own in-house standards of nitrous oxide
with the prepared SCOR standards. Six laboratories reported values of nitrous
oxide for the ARS that were within 3 % of the absolute concentration,
with the remaining laboratory reporting an offset of 10 % (Table S7 in
the Supplement). The majority of these laboratories (five out of six groups)
compared the SCOR ARS with NOAA GMD standards, which have a balance gas of
air instead of nitrogen. Some laboratories with analytical systems that
incorporated fixed sample loops (e.g., 1 or 2 mL loops housed in a 6-port or
10-port injection valve) had difficulty analyzing the WRS, as the peak areas
created by the high mole fraction of the standard exceeded the signal
typically measured from in-house standards or acquired by sample analysis by
an order of magnitude. The high mole fraction of the WRS was not an issue
when multiple sample loops of varying sizes were incorporated into the
analytical system, which was the case for purge-and-trap-based designs. For
the two laboratories with an in-house standard of comparable mole fraction to
the WRS, an offset of 3 % and a
Overall, median methane concentrations in seawater samples collected from the
Pacific Ocean and the Baltic Sea ranged from 0.9 to 60.3 nmol kg
The two Pacific Ocean sampling sites had the lowest water-column
concentrations of methane (Fig. 1a and b). The PAC1 samples collected from
within the mesopelagic zone, where methane concentrations have been reported
to be less than 1 nmol kg
Deviation from the median methane concentration (reported as
absolute values in nmol kg
FID response to methane fitted with a linear regression
calibration. The inclusion
Three Baltic Sea sampling sites (BAL1, BAL3, and BAL6) had median methane
concentrations that ranged from 4.1 to 5.7 nmol kg
Further analysis of the data was conducted to better comprehend the factors that caused the observed interlaboratory variability in methane measurements. The deviation from median values was calculated for each sample collected from the Baltic Sea (Fig. 2). The Pacific Ocean samples (PAC1 and PAC2) were not included in this analysis due to the skewed distribution of data. There were also some instances in the Baltic Sea samples for which the median concentration might not have realistically represented the absolute in situ methane concentration. This was most likely to have occurred at low concentrations due to the skewed distribution of reported concentrations (e.g., BAL1) or at high concentrations for which there was a large range in reported values (e.g., BAL2). The results revealed that a few laboratories (Datasets D, F, and G) were consistently within or close to 5 % of the median value for all batches of seawater samples (Fig. 2). Some laboratories (e.g., Datasets B, C, and H) had a higher deviation from the median value at higher methane concentrations. Two laboratories (Datasets J and K) had a higher deviation from the median value at lower methane concentrations. Finally, in some cases it was not possible to determine a trend (Datasets A and E) due to the variability.
The reasons behind the trends for each dataset became more apparent when
considering the effect of the inclusion or exclusion of low standards in the
calibration curve on the resulting derived concentrations (Fig. 3). The FID
has a linear response to methane at nanomolar values and therefore a high
level of accuracy across a relatively wide range of in situ methane
concentrations can be obtained with the correct slope and intercept. To
demonstrate this, calibration curves for methane were provided by the
University of Hawai'i. These revealed minimal variation in the slope value
when calibration points were increased from low mole fractions (Fig. 3a) to
higher mole fractions (Fig. 3b). However, the intercept value was sensitive
to the range of calibration values used, and this effect was further
exacerbated when only the higher calibration points were included (i.e.,
Fig. 3c). The relevance to final methane concentrations is demonstrated by
considering the values reported by the University of Hawai'i for PAC2 samples
(Fig. 1b). An almost 30 % increase in final methane concentration occurs
from the use of the calibration equation in Fig. 3c compared to Fig. 3a.
This derives from a measured peak area for methane of 62 for a sample with a
volume of 0.076 L and a seawater density of 1024 kg m
Overall, median nitrous oxide concentrations in seawater samples collected
from the Pacific Ocean and the Baltic Sea ranged from 3.4 to
42.4 nmol kg
For six sets of seawater samples, BAL1, BAL2, BAL3, BAL6, BAL7, and PAC2, the
concentrations of nitrous oxide were close to atmospheric equilibrium. The
reported values ranged from 7.7 to 12.7 nmol kg
Concentrations of nitrous oxide measured in nine separate samples
from the Baltic Sea and the Pacific Ocean. The dashed grey line represents
the value of nitrous oxide at atmospheric equilibrium
For the three other sets of samples (BAL4, BAL5, and PAC1), the nitrous oxide
concentrations deviated significantly from atmospheric equilibrium (Fig. 4c,
d, and e). At one sampling site, BAL4 (Fig. 4c), nitrous oxide was
undersaturated with respect to atmospheric equilibrium and reported
concentrations ranged from 2.1–5.5 nmol kg
The deviation of individual nitrous oxide concentrations from the median value provides insight into the variability associated with their measurements (Fig. 5). The BAL1 dataset was not included in this analysis due to its skewed data distribution, and the high interlaboratory variability for BAL5 indicated that the median value may differ from the absolute nitrous oxide concentration for this sample. For the low-nitrous-oxide Baltic Sea and Pacific Ocean samples (Fig. 5a), the majority of data points were within 5 % of the median values. Furthermore, for the majority of laboratories, the data points for separate seawater samples clustered together, indicating some consistency to the extent they varied from the overall median value. Exceptions to this observation include Datasets E, C, L, and K (Fig. 5a), which demonstrated varying precision and accuracy. At high nitrous oxide concentrations (Fig. 5b), there are fewer data points within 5 % of the median value compared to low nitrous oxide concentrations (Fig. 5a). Therefore, for PAC1 and BAL5 samples, six and seven data points fall within 5 % of the median value, respectively. Furthermore, only three laboratories (Datasets F, G, and K) had data for both Pacific Ocean and Baltic Sea samples within 5 % of the median value. This could have been caused by inconsistent analysis between different batches of samples or by variable sample collection and transportation.
Deviation from the median value (reported in absolute units) for
nitrous oxide datasets. The batches of samples include BAL1, 2, 3, 6,
and 7
The likely factors that caused these offsets in nitrous oxide concentrations among laboratories include sample analysis and calibration of the gas analyzers. Calibration of the ECD is nontrivial and at least two prior publications have discussed nitrous oxide calibration issues (Butler and Elkins, 1991; Bange et al., 2001). The laboratories participating in the nitrous oxide intercomparison employed different calibration procedures (Fig. 6). Some used a linear fit and kept their analytical peak areas within a narrow range (Fig. 6a), while others used a stepwise linear fit and therefore used different slopes for low and high nitrous oxide mole fractions (Fig. 6b). Finally, some applied a polynomial curve (Fig. 6c) and sometimes two different polynomial fits for low and high concentrations. The difficulty in calibrating the ECD was evidenced by the deviation from median values as multiple datasets show good precision but consistent offsets at the lowest (Fig. 5a) and highest (Fig. 5b) final concentrations of nitrous oxide.
Three calibration curves for nitrous oxide measurements using an ECD
including linear
Because the prolonged storage of samples can influence dissolved gas
concentrations, including methane and nitrous oxide, the intercomparison
dataset was analyzed for sample storage effects (Table S5 in the Supplement).
It should, however, be noted that assessing the effect of storage time on
sample integrity was not a formal goal of the intercomparison exercise and
replicate samples were not analyzed at repeated intervals by independent
laboratories, as would normally be required for a thorough analysis.
Nonetheless our results did provide some insights into potential
storage-related problems. Most notably, there were indications that an
increase in storage time caused increased concentrations and increased
variability for methane samples with low concentrations, i.e., PAC1 and PAC2
samples, which had median methane concentrations of 0.9 and
2.3 nmol kg
Comparison of sample storage times with measured concentrations of
methane
Another variable that differed between laboratories for the intercomparison exercise was the size of sample bottles, which ranged from 25 mL to 1 L for the different laboratories. There was no observed difference between the methane and nitrous oxide values obtained from the various sampling bottles and it was concluded that sampling bottles were not a controlling factor for the observed differences between laboratories. We note, however, the potential for greater air bubble contamination in smaller bottles.
The marine methane and nitrous oxide analytical community is growing. This is reflected in the increasing number of corresponding scientific publications and the resulting development of a global database for methane and nitrous oxide (Bange et al., 2009). Like all Earth observation measurements, there is a need for intercomparison exercises of the type reported here for data quality assurance and for appropriate reporting practices (National Research Council, 1993). To the best of our knowledge, the work presented here is the first formal intercomparison of dissolved methane and nitrous oxide measurements. Based on our results, we discuss the lessons learned and our recommendations moving forward by addressing the four questions that were posed in the Introduction.
It is typical for laboratories to source some, or all, of their compressed
gas standards from commercial suppliers. National agencies, such as NOAA GMD
or the National Institute of Metrology China, also provide standards to the
scientific community. The national agencies typically offer a lower range in
concentrations than commercial suppliers, but their standards tend to have a
higher level of accuracy. Of the 12 laboratories participating in the
intercomparison, 8 reported using national agency standards, with 7
of them using gases sourced from NOAA GMD. Since the methane and nitrous
oxide mole fractions of these national agency standards are equivalent to
modern-day atmospheric mixing ratios, they are similar to the SCOR ARS
distributed to the majority of laboratories in this study. Laboratories in
receipt of the SCOR standards were asked to predict their mole fractions
based on those of their own in-house standards. For the majority that
conducted this exercise, there was good agreement (
The methane intercomparison
highlighted the variability that exists between measurements conducted by
independent laboratories. At low methane concentrations, a skewed
distribution of methane data was observed, which was particularly evident in
PAC1 (Fig. 1a). Potential causes include calibration procedures (Sect. 3.2)
and/or sample contamination, which is more prevalent at low concentrations
(Sect. 3.4). For some laboratories, the low methane concentrations are close
to their detection limit, which is determined by the relatively low
sensitivity of the FID and the small number of moles of methane in an
introduced headspace equilibrated with seawater. An approximate working
detection limit for methane analysis via headspace equilibration is
1 nmol kg
There was an improvement in the overall agreement between the laboratories for samples with higher methane concentrations. However, some of the highest variability between the laboratories was observed at the highest concentrations of methane analyzed (BAL2; Fig. 1e). This high degree of variability resulted in significant uncertainty in the absolute in situ concentration. Methane concentrations of this magnitude and higher are found in coastal environments (Zhang et al., 2004; Jakobs et al., 2014; Borges et al., 2018) and in the water-column associated with seafloor emissions (e.g., Pohlman et al., 2011). These environments are considered vulnerable to climate-induced changes and eutrophication, and therefore it is necessary that independent measurements are conducted to the highest possible accuracy to allow for interlaboratory and inter-habitat comparisons. To address this, we recommend that reference material be produced and distributed between laboratories.
Some of the trends discussed for methane were also evident in
the nitrous oxide data. For the samples with the lowest nitrous oxide
concentrations a skewed data distribution was observed, as found for methane
(Fig. 4c). Such low nitrous oxide concentrations are typical of low-oxygen
water-column environments (
The majority of seawater samples analyzed had nitrous oxide concentrations
ranging from 7–11 nmol kg
There are several analytical recommendations resulting from this study. The
use of highly accurate standards and the appropriate calibration fit is an
essential requirement for both headspace equilibration and the purge-and-trap
technique. It was shown that both analytical approaches can yield comparable
values for methane and nitrous oxide, with the main differences observed at
low methane concentrations. At sub-nanomolar methane concentrations, four out
of the six laboratories that reported methane concentrations
This study also revealed that sample storage time can be an important factor. Specifically, the results from this study corroborate the findings of Magen et al. (2014), who showed that samples with low concentrations of methane are more susceptible to increased values as a result of contamination. The contamination was most likely due to the release of methane and other hydrocarbons from the septa (Niemann et al., 2015). Since the release of hydrocarbons occurs over a period of time, it is recommended to keep storage time to a minimum and to store samples in the dark. It should be noted that sample integrity can also be compromised due to other factors including inadequate preservation, outgassing, and adsorption of gases onto septa. For all these reasons, it is recommended to conduct an evaluation of sample storage time for the environment that is being sampled.
One useful item that was not included as part of the intercomparison exercise
but can help decrease uncertainty in the accuracy and precision of methane
and nitrous oxide measurements is internal control measurement. Internal
controls represent a self-assessment quality control check to validate the
analytical method and quantify the magnitude of uncertainty. Appropriate
internal controls for methane and nitrous oxide consist of air-equilibrated
seawater samples. Their purpose is to provide checks for methane
concentrations ranging from 2–3 nmol kg
In addition to the self-assessments provided by the analysis of air-equilibrated seawater, this study revealed the need for reference seawater to help assess the accuracy of high-concentration methane and nitrous oxide measurements. Reference seawater in this instance refers to batches of dissolved methane and nitrous oxide samples prepared in the laboratory using an equilibrator setup, as used for dissolved inorganic carbon (Dickson et al., 2007). In the absence of plans for additional intercomparison exercises, the provision of reference seawater will allow laboratories to continue evaluating their own measurements. Finally, the lessons learned during the intercomparison exercises will be the basis for a forthcoming good practice guide for dissolved methane and nitrous oxide.
The key outcome of this study was the identification of differences in methane and nitrous oxide concentrations for the same batch of seawater samples measured by several independent laboratories. Emergent from this is the distinct possibility that any given laboratory will incorrectly report data, thereby increasing uncertainty over the saturation states of both gases. The tendency to overestimate methane concentrations close to atmospheric equilibrium means that marine emissions of methane to the overlying atmosphere will also be overestimated (Bange et al., 1994; Upstill-Goddard and Barnes, 2016). In contrast, for nitrous oxide there does not appear to be either an underestimation or overestimation of concentrations. Consequently, there is generally a lower inherent uncertainty in its surface ocean saturation state, as previously proposed (Law and Ling, 2001; Forster et al., 2009).
The interlaboratory differences highlighted by this study should be viewed in the context of numerous individual efforts to assess temporal and/or spatial trends in methane and nitrous oxide by way of time series observations (Bange et al., 2010; Farías et al., 2015; Wilson et al., 2017; Fenwick and Tortell, 2018), repeat hydrographic survey lines (de la Paz et al., 2017), and single expeditions. While the value of these in integrating the behavior of methane and nitrous oxide into the hydrography and biogeochemistry of local–regional ecosystems is beyond question, their value would be enhanced by the rigorous cross-validation of analytical protocols. Without this, perceived small temporal and/or spatial changes in water-column concentrations in any given region are difficult to verify unless the data all originate from a single laboratory. In addition, the value of a global methane and nitrous oxide database (e.g., Bange et al., 2009) would to some extent be compromised by the uncertainty. Taking due account of the analytical variability between laboratories will clearly be vital to any future assessment of the changing methane and nitrous oxide budgets of the oceans.
Overall, the intercomparison exercise was invaluable to the growing community of ocean scientists interested in understanding the dynamics of dissolved methane and nitrous oxide in the water column. The level of agreement between independent measurements of dissolved concentrations was evaluated in the context of several contributing factors, including sample analysis, standards, calibration procedures, and sample storage time. Importantly, the intercomparison represents a concerted effort from the scientists involved to critically assess the quality of their data and to initiate the steps required for further improvements. Recommendations arising from the intercomparison include routine cross-calibration of working gas standards against primary standards, minimizing sample storage time, incorporating internal controls (air-equilibrated seawater) alongside routine sample analysis, and the future production of reference seawater for methane and nitrous oxide measurements. These efforts will help resolve temporal and spatial variability, which is necessary for constraining methane and nitrous oxide emissions from aquatic ecosystems and for evaluating the processes that govern their production and consumption in the water column.
Data are available in the Supplement.
The supplement related to this article is available online at:
The authors declare that they have no conflict of interest.
Any use of trade names is for descriptive purposes and does not imply endorsement by the U.S. government.
During the final stages of this work, our coauthor John L. Bullister passed away. The intercomparison exercise greatly benefited from John's scientific expertise on dissolved gases. He will be deeply missed by the oceanographic community.
The methane and nitrous oxide intercomparison exercise was conducted as a
Scientific Committee on Ocean Research (SCOR) Working Group, which receives
funding from the U.S. National Science Foundation (OCE-1546580). Pacific
Ocean seawater samples were collected on HOT cruises, which are supported by
the NSF (including the most recent OCE-1260164 to DMK). Baltic Sea seawater
samples were collected during cruise no. 142 of the R/V