Advances in reference materials and measurement techniques for greenhouse gas atmospheric observations

Comparisons which follow the model where all participants send their reference materials to a central facility are most beneficial when the instrumentation used for the analysis is specific to the component, unbiased towards potential differences in the reference materials other than the amount fraction and can be operated in a way to minimise the variance of measurements during the time required to conduct the comparison. The aim is to perform analytical measurements that produce an analytical uncertainty that is negligible compared to the uncertainties on reference materials declared by participants.

Component specificity, a requirement of measurements in whole air, is normally achieved by using the same instrumentation found in air monitoring laboratories, as described in section 4. GC-FID was used during the methane and carbon dioxide key comparisons [20, 22] and GC-ECD during the nitrous oxide key comparison [23]. As GC instruments are known to present drifts on a time scale of about an hour, the pilot laboratory normally uses drift correction with a control cylinder and optimises its measurement sequence to obtain minimum repeatability uncertainty. Relative standard uncertainties of the order of 0.02% were achieved on methane measurements during CCQM-K82, and between 0.003% and 0.01% on carbon dioxide during CCQM-K120.

Another beneficial practice introduced during CCQM-K82 was to measure all submitted reference materials using two analytical techniques differing as much as possible in their principle; this increases the robustness of the comparison towards incidents, and most importantly, towards potential method-specific biases. In this way, techniques relying on absorption of infrared light by the component were also selected: CRDS during CCQM-K82, Fourier-transform infra-red (FTIR) during CCQM-K120, and TDLAS to be used in the upcoming comparison on nitrous oxide (CCQM-K68.2019). These techniques are normally characterised with better short-term precisions than GC (section 4.3), which may be mitigated by additional variances due to the time required to compare as many as sixteen reference materials together [20]. They are, however, subject to a specific bias arising from differences in the composition of the matrix (a 'component dependent pressure broadening effect' that drove tight requirements on prepared reference materials during the comparisons). In addition, spectroscopic techniques are normally specific to isotopologues, resulting in biases if the reference materials in a comparison presents different isotopic compositions. This was addressed during CCQM-K120, with a third IRIS instrument introduced to measure carbon dioxide isotope ratios in all reference materials to correct the carbon dioxide amount fractions measured by FTIR.

The property of commutability towards measurement techniques introduced in section 2.3 constitutes a strong requirement during key comparisons as well. Comparisons conducted since 2010 introduced tighter limits on the amount fraction of the main components of synthetic air (nitrogen, oxygen and argon) imposed to participants when preparing their reference materials, to ensure a negligible uncertainty due to the pressure broadening effect. During the validation work performed by the BIPM for CCQM-K82, it could be demonstrated that reference materials of methane prepared by the National Institute of Standards and Technology (NIST) in scrubbed dry air were equivalent to those prepared in synthetic air within the limits imposed by the use of CRDS as the analytical technique [24].

For comparisons of carbon dioxide in air reference materials by spectroscopic techniques such as FTIR used during CCQM-K120, the dependency of the technique to particular carbon dioxide isotopologues was considered. However participants were not asked to select a particular composition when preparing their reference materials. Instead, the pilot laboratory measured the isotopic composition of carbon dioxide in each reference material with an IRIS instrument and successfully corrected amount fraction measurements by FTIR [20].

The preferred model for comparisons organised by the CCQM-GAWG requires the preparation of gas reference materials by each participant at a defined nominal amount fraction. The measurement of all reference materials is carried out by the coordinating laboratory and considered as a set defining a calibration curve for the analytical instrument used as the comparator. The calibration curve is computed from a regression line of the measured values using uncertainties on both axes, following the procedures outlined in ISO 6143:2001 [25]. Gas reference materials from participants which contribute to the key comparison reference value (KCRV) are then selected on the basis of obtaining the largest consistent data set, as demonstrated with the goodness-of-fit parameter. Technical reasons are normally applied for removing outliers, for example a lack of stability reported in the most recent carbon dioxide comparison CCQM-K120 [20]. A KCRV is then determined for each gas reference material by applying the inverse regression equation to the value measured by the pilot laboratory. This allows degrees of equivalence D_i to be defined for each participant, as the difference between the submitted value (typically the gravimetric value) and the KCRV, with its associated uncertainty at a 95% level of confidence.

Over the last ten years, improvements in preparation of gravimetric reference materials described in section 2.2 and in the comparison methods resulted in better agreement amongst NMIs. This is most obvious for the methane and carbon dioxide comparisons. Figure 1 compares the results of CCQM-K82 with results from a comparison performed ten years earlier (CCQM-P41). The level of agreement amongst reference materials improved by a factor of ten. The standard deviation of results in 2003, around the reference value was 30 nmol mol⁻¹, and 11 nmol mol⁻¹ for a more limited set of reference materials, which improved to 1.8 nmol mol⁻¹ in 2013.

Zoom In Zoom Out Reset image size

The most recent carbon dioxide comparison CCQM-K120 also demonstrated improvements in the uncertainty of the reference materials and the comparison method, as displayed in figure 2 with its results compared to the previous exercise CCQM-K52. The improvement is most striking when only participants selected as forming the largest consistent set of reference materials are considered, as indicated in the figure. A similar improvement is expected in the future nitrous oxide comparison to be commenced in 2019 by the BIPM (CCQM-K68.2019).

Zoom In Zoom Out Reset image size

Three major sources of uncertainty for gravimetric preparation of reference materials must be characterised and controlled: mass measurement, gas purity and the relative atomic or molecular masses of the gases being weighed.

The relative atomic masses of the elements are tabulated [43] with uncertainties that, except for the mononuclidic elements, are dominated by the distribution of isotopes seen in naturally occurring elements from different environments. Reducing the uncertainties requires the measurement of isotope ratios in the sample, or by using the information available about the source(s) of the material(s) at hand [28].

Historically, the component of uncertainty that has limited accuracy has been the determination of purity of the gases being weighed, with the potential variation in isotope ratios not considered to significantly affect the amount fraction values and uncertainties achievable. Similarly, gas reference materials have been compared with techniques that were sensitive to the total amount fraction of the component of interest (e.g. GC with flame ionisation detectors (FID) or GC with thermal conductivity detectors (TCD)) and not a particular isotopocule. However, as gas analysers that are based on techniques that are sensitive to variations in isotopocule fractions have become more prevalent, further attention has been paid to the isotopic composition of the component of interest in the reference materials. Spectoscopic techniques used for high-end applications are also sensitive to differences in the composition of the matrix gas, adding a further demand to the basket. This section reviews the impact of isotope ratio variation of gas reference materials on their commutability, and how measurements of isotope ratio are being incorporated into gas reference material production and value assignment procedures to ensure the compatibility.

One of the first CCQM comparisons, in which gas reference materials were compared under repeatability conditions, CCQM-P23 [44], focused on verifying the uncertainties of gravimetrically prepared gas reference materials (carbon monoxide in nitrogen), by reducing the influence of analytical uncertainties. The comparison included gas mixtures submitted at several amount fractions to one laboratory for analysis under repeatability conditions. At these higher than ambient amount fractions, relative standard uncertainties of the order of 0.01% were reported and expected to be validated. However, the comparison results demonstrated much larger inconsistencies when the analytical method was NDIR, up to 1% in some instances. These discrepancies were largely removed if GC-TCD or GC-FID was used as the comparison method. Subsequent measurements of the absolute ¹³C/¹²C ratios of the carbon monoxide in these gas mixtures reported a subset with ratio values grouped around 0.011, close to natural abundance, but others with measured values of 0.009 and even 0.004. The sensitivity of NDIR to isotopic composition of carbon monoxide was subsequently studied and reported [45], as well as methods to screen pure carbon monoxide gas to ensure its isotopic composition was close to natural abundance if it was to be used for the manufacture of calibration reference materials. Furthermore, whilst a bias proportional to the difference in isotope ratio from natural abundance could be demonstrated for the NDIR instruments tested, the exact relationship was dependent on the particular analyser. The variation in isotope ratios found in the pure carbon monoxide gases was reported to be due to the source and treatment of the gases, with a number of the gases having been processed to remove ¹³C carbon monoxide for delivery to the ¹³C labelled chemical market, leaving the remaining carbon monoxide gas depleted of ¹³C isotopologues.

The atmospheric monitoring community has driven the requirement for gas reference materials with reduced measurement uncertainties for monitoring long-term trends of greenhouse gas levels. Global networks [30, 46] for carbon dioxide monitoring within the WMO-GAW programme aspire to keep network biases less than 0.1 µmol mol⁻¹ (0.05 µmol mol⁻¹ in the southern hemisphere).

The first comparisons of carbon dioxide in air reference materials at atmospheric background amount fractions organised by the CCQM-GAWG, CCQM-P41 [18], occurred in 2003 with a second comparison, CCQM-K52 [21] in 2006. The direct comparison of the gas reference materials (CCQM-P41 part 2) was carried out using GC-TCD. This method is insensitive to the isotopic composition of carbon dioxide in the gas reference materials, and the spread of results, which were within a range of  ±0.5 µmol mol⁻¹ of the reference value (if calculated using methods developed in later comparisons of this type—see CCQM-K82 [22]), were due to other factors. The CCQM-K52 comparison of carbon dioxide in air, at nominally 365 µmol mol⁻¹, was performed by all nineteen participants measuring reference materials that had been sent to them with reference values that had been assigned with standard uncertainties of 0.2 µmol mol⁻¹. The majority of laboratories demonstrated agreement of their measurement capabilities within  ±1 µmol mol⁻¹. Thirteen participants used GC-TCD or GC-FID analysers for comparison with their in-house reference materials and no isotope effects were observed. For the seven participants using NDIR analysers, no isotopic effects were quantified or reported and measurement uncertainties were large in comparison to the expected size of the effects, apart from two participants (who did report isotope effects or whose uncertainties were not large in comparison). One of the participants, working closely with the atmospheric monitoring community, investigated and published [47] the effect of carbon dioxide isotopic variations on NDIR measurements. A difference approaching 0.2 µmol mol⁻¹ was reported for measurements of reference materials with near ambient isotopic composition compared to those with highly depleted isotope ratios made from carbon dioxide from a combustion source.

Commercial laser-based analytical systems, such as TDLAS and CRDS, for trace gases have been developed with sufficient resolution to be able to measure the absorption of single isotopologues of carbon dioxide, which can be used for the measurement of amount fractions. Good agreement between CRDS and NDIR measurements was demonstrated [33] when a number of potential sources of bias were controlled, including the isotopic composition. The correction for isotope effects was shown to lead to the removal of a 0.15 µmol mol⁻¹ bias, resulting in the mean difference between the measurement techniques being reduced to (0.05  ±  0.09) µmol mol⁻¹.

The theoretical bias in amount fraction measurements caused by different isotope ratios in calibrant and sample gases can be readily calculated from measured isotope ratios, expressed in delta notation and traceable to conventional standards. As a first approximation, when carbon dioxide is measured as the major isotopocule ¹²C¹⁶O₂, a 1‰ difference in δ¹³C values between the calibration and sample gases leads to a difference of 4 nmol mol⁻¹ in the amount fraction of a 400 µmol mol⁻¹ mixture, and 2 nmol mol⁻¹ for the same difference in δ¹⁸O values. Blending carbon dioxide gases from different sources provides a method for matching to natural abundance isotope ratios for ¹³C, with the mixing of industrially sourced carbon dioxide with highly enriched ¹³CO₂ described [48] as well as mixing with pure carbon dioxide obtained from a gas well source [49] (δ¹³C at  −1‰ versus Vienna Peedee Belemnite (VPDB)), with δ¹⁸O values better matched to atmospheric values using the latter source.

The latest comparisons of carbon dioxide in air mixtures CCQM-K120 [20], have focused on demonstrating equivalence of reference materials at better than the 0.1 µmol mol⁻¹ level and correcting optical instrument responses for measured isotope ratios of compared carbon dioxide gases. The comparison at the BIPM consisted of forty four gas reference materials, prepared by fourteen participating laboratories. Two measurement techniques were used to compare the reference materials, to ensure no measurement method dependent bias: GC-FID and FTIR spectroscopic analysis. The latter had to be corrected for the isotopic composition of the carbon dioxide gases, involving measurements of isotope ratios in each reference material. The method used by the BIPM for measuring isotope ratios is described in a recent publication [49] and was validated with carbon dioxide in air reference materials that had been value assigned for their isotopic composition by BGC-IsoLab Jena, with traceability of the standards used to the VPDB-CO₂ scale realised by the Jena reference air set (JRAS)-06 scale. The isotope ratios of submitted gases were found to range from  −66.9‰ to  +0.2‰ versus VPDB for δ¹³C and  −34.5‰ to  −0.3‰ versus VPDB-CO₂ for δ¹⁸O. Typical standard uncertainties for the isotope ratio measurements made by laser spectroscopy were 0.2‰ and 0.5‰ for δ¹³C and δ¹⁸O respectively. Whilst these uncertainties are considerably greater than can be achieved by mass spectrometry, achieving them still requires careful consideration of the calibration strategy and traceability chain, both in terms of strategy [50] and standards to be used [49]. A comparison being coordinated jointly by the BIPM and the IAEA [51] aims to determine the compatibility of isotope ratio measurements of carbon dioxide, both in nominally pure carbon dioxide and carbon dioxide in air reference materials using a variety of analytical techniques.

The CCQM-K120 comparisons were successful in demonstrating the degrees of equivalence of carbon dioxide in air reference materials amongst participating NMIs at nominal amount fractions of 380 µmol mol⁻¹, 480 µmol mol⁻¹ and 800 µmol mol⁻¹, with a standard uncertainty of 0.05 µmol mol⁻¹ (a reduction by a factor of at least 4 relative to CCQM-K52). The methods based on spectroscopy (FTIR, with corrections for isotope ratios) and GC-FID for comparing carbon dioxide in air reference materials demonstrated excellent agreement, with the standard deviation of the difference in reference values found by the two methods being 0.07 µmol mol⁻¹ at 380 µmol mol⁻¹ and 0.05 µmol mol⁻¹ at 480 µmol mol⁻¹. It also demonstrated that providing reference materials to communities measuring atmospheric carbon dioxide with spectroscopic methods produced with carbon dioxide originating from combustion sources can lead to isotope ratios in the carbon dioxide which will produce biases in amount fractions measurements of up to 0.3 µmol mol⁻¹. These biases can be corrected by either measuring isotope ratios of the gas reference material and appropriate processing by the user, or by using carbon dioxide which has been closely matched to atmospheric isotope ratio values in the production of the reference material. Isotopic differences in the reference materials that define the WMO-CO₂-X2007 (CO₂-in-air) scale have also been measured [52] and are now accounted for in the propagation of the scale, with isotope ratios reported as characterised to  ±0.2‰ for both δ¹³C and δ¹⁸O values. Absolute ratios for ¹³C/¹²C and ¹⁸O/¹⁶O which are currently known with uncertainty at about 1‰ (see below) appear to be another source of uncertainty.

Based on these developments and the advent of optical spectroscopy for measurements of isotope ratio, some NMIs are now focusing on developing appropriate reference materials with isotope ratios traceable to reference materials produced by the IAEA.

Primary standards for methane and nitrous oxide are prepared using analytical balances, mass comparators, and gas blending manifolds as described in ISO 6142-1 [26]. In general, a known mass of the minor component is added to an evacuated cylinder and diluted with a known mass of the major component. The mass of the minor component can be determined in various ways: (a) by direct transfer of an aliquot to an evacuated cylinder followed by weighing, or (b) by adding a small amount of gas to a small transfer vessel (5–50 cm³), weighing, and then transferring the aliquot to an evacuated cylinder. Transfer from a vessel can involve expansion from the transfer vessel into the target, or expansion plus flushing [56] in which the vessel is repeatedly pressurised and the contents expanded into the cylinder. For methane, all three methods were used to create a suite of primary standards (direct addition, expansion, expansion with flushing). For nitrous oxide, only the expansion with flushing method was used [57]. An advantage of direct addition is that it ensures complete transfer because the mass of the aliquot is determined after it has been added to the cylinder. An advantage of the transfer vessel method is that relatively large dilution factors can be achieved (up to 2  ×  10⁵ in a single step). A disadvantage of the transfer vessel method is that it requires quantitative knowledge of the transfer efficiency. High transfer efficiencies can be achieved with care [58].

For both methane and nitrous oxide, primary standards were prepared gravimetrically in 5.9 l aluminium cylinders. For methane seventeen standards were prepared from 1991–1995 over the nominal range (300–2600) nmol mol⁻¹ [56] with additional standards prepared in 2013 to verify the original work and extend the range to 5900 nmol mol⁻¹ (table 2). Scrubbed whole air (i.e. whole air from which methane, carbon dioxide, and nitrous oxide is removed to produce residual levels of  <10 nmol mol⁻¹, <100 nmol mol⁻¹ and  <5 nmol mol⁻¹ respectively) was used as the dilution gas. Uncertainties were limited to some extent by difficulties in determining the amount of methane in the dilution gas (often a few nmol mol⁻¹) [24, 56].

For nitrous oxide fifteen standards over the nominal range 260–370 nmol mol⁻¹ were made gravimetrically from four parent mixtures at higher amount fractions [57, 59]. Synthetic air (oxygen and nitrogen) was used as dilution gas because scrubbed whole air often contains trace amounts of nitrous oxide, which can be difficult to quantify. This constrains the use of the current WMO-GAW primary standards as they cannot be used to calibrate spectroscopic instruments directly because they lack argon, which could be important with respect to pressure broadening [60]. Scale transfer is currently achieved using gas chromatography with electron capture detection (GC-ECD). Work is underway to prepared nitrous oxide standards in scrubbed whole air.

The sets of primary standards for nitrous oxide and methane show good internal consistency. The standard deviation of residuals from a 2nd order polynomial fit to twenty two methane standards, based on GC-FID analysis, is 3.4 nmol mol⁻¹. For nitrous oxide, the standard deviation of residuals is 0.33 nmol mol⁻¹ [57] based on GC-ECD analysis. In addition to CCQM comparisons, WMO-GAW scales have also been compared with other independent scales, such as those from the Scripps Institution of Oceanography (SIO) and Tohoku University (TU) [61]. Through comparison of more than twenty years of atmospheric measurements at multiple sites, a divergence between the SIO scale 'SIO-98' and the WMO scale 'WMO-N₂O-X2006A' has been observed. A recent update of the SIO scale (now SIO-16) has improved the comparison, but small differences remain [62]. This underscores the need for repeated comparisons.

The manometric determination of carbon dioxide in air involves loading a sample of dry whole air into a known volume ('large volume') where its temperature and pressure are determined. Carbon dioxide is then cryogenically extracted from that air sample, cryogenically purified to remove (residual) water vapour, and transferred into a much smaller volume ('small volume'). After reaching thermal equilibrium, the temperature and pressure of the purified carbon dioxide in the small volume are determined. The volumes of the large and small volumes need not be known exactly, but their ratio must be well-known. The amount fraction of carbon dioxide in dry air is derived from ratios of temperature, pressure, and volume, with appropriate corrections for non-ideality [63]. Since nitrous oxide cannot be separated from carbon dioxide cryogenically, a correction must be made for the amount of nitrous oxide and its non-ideality.

Manometry has been used to prepare standards for carbon dioxide measurements since C D Keeling began measuring carbon dioxide in background (remote) air at the South Pole in 1957, and at Mauna Loa, Hawaii in March 1958, for Scripps Institution of Oceanography (SIO) [64]. SIO data are traceable to carbon dioxide primary standards value-assigned using either a constant-volume mercury-column manometer (CMM) or an electronic constant-volume manometer (ECM) [65–68].

The CMM used at SIO had a large volume of 5014 cm³ and a small volume of 3.80 cm³ [69]. Pressure was measured by comparing the height of mercury in a column containing the sample to the height of mercury in a vacuum column. Temperatures were determined using mercury thermometers. Cathetometers were used to read the mercury column heights and temperatures through windows in a temperature-controlled housing. The volume ratio was determined by directly weighing the main volumes filled with water or mercury, and by measuring pressure and temperature after expanding gas (carbon dioxide, argon or nitrogen) from the small volume into successively larger volumes [67].

The SIO carbon dioxide amount fraction scale, as determined from CMM measurements of a series of primary standards, was used by the WMO Background Air Pollution Monitoring Network, a predecessor to the GAW programme, as the world reference for carbon dioxide from 1975–1995. Primary standards were measured using the CMM approximately every two years. Results of the CMM measurements were documented in technical reports (e.g. [70, 71]). In 1995, NOAA assumed the role of CCL and established manometric capabilities. The NOAA manometer was modelled on the Scripps CMM, but with electronic devices for pressure and temperature measurement [63, 72]. NOAA prepared a suite of 15 primary standards, which were initially value-assigned based on both the SIO CMM and the NOAA ECM from 1995–2001. In 2001, NOAA derived an independent scale based solely on NOAA measurements [72]. Similarly to SIO, NOAA analyses primary standards using their ECM approximately every two years. The current WMO-GAW scale (WMO-CO₂-X2007) is based on NOAA manometric results from 1996 through 2007. NOAA has continued to measure primary standards every two years and has presented updated manometric results at WMO-GAW meetings [73, 74].

Repeatability (estimated as the standard deviation) of CMM and ECM instruments is about 0.018% (0.07 µmol mol⁻¹ at ~400 µmol mol⁻¹) [69, 72] and has improved over time (figure 3).

Zoom In Zoom Out Reset image size

Historically, several issues complicated the long-term maintenance of carbon dioxide scales using CMM and ECM. Carbon dioxide 'drift' in cylinders is of primary concern, and can be at the limit of what can be detected over even a decade, given the typical expanded uncertainty associated with the technique (~0.05%). Changes in manometer-assigned values over time could be the result of cylinder drift, or changes in the manometer. This dilemma confounded SIO to the point that two separate scales were derived in 1999: one in which cylinders were assumed to be stable and any observed drift attributed to changes in the volume ratio (SIO scale X99A), and one in which the manometer was assumed to be stable and the cylinders thought to be drifting (SIO scale X99B) [69]. This issue was likely complicated by the fact that some of the early SIO primary standards were prepared in steel cylinders, which tend to show higher drift rates than aluminium cylinders [39]. Nevertheless, drift in the amount fraction is not easily resolved. Keeling et al [69] investigated multiple issues related to drift, and concluded that their small volume did not change between 1985 and 2016, but their thermometers drifted 0.11 °C between 1961 and 2006, and that some of the auxiliary volumes (used to determine the volume ratio) changed over time as they were annealed and 'ashed' each year to remove grease. A revised SIO scale is under development and will incorporate these latest findings.

Interactions of carbon dioxide with manometer surfaces also complicate interpretation of manometer results. SIO has documented small changes in their CMM results following cleaning to remove grease and residual mercury and NOAA is investigating the possibility that carbon dioxide adsorbs to or permeates O-rings. Despite these issues, both SIO and NOAA have maintained carbon dioxide scales that have remained remarkably stable, and in close agreement, for many years [69, 72].

BIPM has developed an all SilcoNert^ treated stainless steel ECM with metallic, instead of elastomer, O-rings in all valves to address the interaction of carbon dioxide with glass and other surfaces. Preliminary results show good agreement between this system and the all-glass NOAA system.

To address scientific needs for interpreting atmospheric observations of the major greenhouse gases, the GAW programme sets ambitious network compatibility goals, which are reviewed and, if necessary, revised during biennial meetings of the WMO-GAW community. Network compatibility goals are defined for the unpolluted troposphere and represent the maximum bias that can generally be tolerated in measurements of well- mixed air used in global models to infer fluxes. Extended network compatibility goals are provided as a guideline for many other studies in which the smallest bias tolerance is not required [30]. These network compatibility goals currently stand at  ±0.05 µmol mol⁻¹ for carbon dioxide in the Southern Hemisphere, relaxed to  ±0.1 µmol mol⁻¹ elsewhere, ±2 nmol mol⁻¹ for methane, and  ±0.1 nmol mol⁻¹ for nitrous oxide. The extended goals are set to  ±0.2 µmol mol⁻¹ for carbon dioxide, ±5 nmol mol⁻¹ for methane, and 0.3 nmol mol⁻¹ for nitrous oxide.

The WMO-GAW programme has implemented a quality management framework (QMF) to ensure consistent scales and comparability, in the recognition that 'collecting adequate information on the chemical composition of the atmosphere and on the consequences of the anthropogenic impact on different spatial and temporal scales is valuable and possible only if all the relevant measurements are expressed in the same units and on the same scale and if data from different countries and from different sites are comparable' [75]. Further details on the WMO/GAW QMF can be found in the WMO GAW Implementation Plan: 2016–2023 [75]. Briefly, the QMF requires CCLs maintain and enable access to the primary network standards and calibration scales. CCLs are encouraged to compare the WMO-GAW scales to other independent scales whenever possible, e.g. through participation in key comparisons organised by BIPM [22, 23] and other exercises as appropriate. All WMO measurements are made traceable to an appropriate scale [30, 53], which leads to high internal consistency of data within and between networks. World Calibration Centres (WCCs) organise comparison campaigns and performance audits at GAW sites using transfer standards as an independent check of the traceability to the CCL. World Data Centres (WDCs) provide archiving facilities, perform plausibility and consistency checks on submitted data, and provide feedback to the data providers when necessary. QA/QC related issues are further coordinated by Scientific Advisory Groups (SAGs) and Quality Assurance/Scientific Activity Centres (QA/SACs).

Zellweger et al [76] published results of performance audits made by the WCC for methane (2005–2014) and carbon dioxide (2010–2015) that showed newer spectroscopic techniques such as CRDS have clear advantages in stability and linearity of the response compared to traditional methods. Figure 4 shows an update of the analysis made by Zellweger et al [76] including audit data until 2018 for methane. Briefly, the bias in the centre of the relevant amount fraction of the unpolluted troposphere versus the slope of the linear regression analysis of the standard comparison is shown, which results in allowed bias/slope combinations meeting the WMO-GAW network compatibility (green area) and extended network compatibility goals (yellow area) for the relevant amount fraction. Figure 4 also shows comparisons made between the WCC and standards prepared by NMIs. The NMI standards showed an offset of about 2–3 nmol mol⁻¹ compared to the WMO-CH₄-X2004A calibration scale, which is in good agreement with the results of the BIPM key comparison CCQM-K82 [22].

Zoom In Zoom Out Reset image size

The results obtained during the WCC station performance audits are similar to those from recent WMO/IAEA Round Robin Comparison Experiment (RR6) organised and coordinated by the CCL for methane hosted by NOAA. RR6 took place in 2014/15 with 35 participating laboratories, and involved two methane standards, one containing (average 1756.2 nmol mol⁻¹) and the other containing (average 1943.1 nmol mol⁻¹) amount fraction [77]. With this data set, the analysis described above was repeated after exclusion of one laboratory, which was far outside the compatibility goal. This excluded laboratory was also audited by the WCC in 2012 and the results were excluded because the instrument was found to be unsuitable for methane analysis. The results of RR6 are also presented in figure 4 (right panel). The percentage of laboratories fulfilling the WMO network compatibility and extended network compatibility goal during WMO RR6 was similar to the results from the WCC station performance audits. The laboratories reporting data on scales other than WMO-CH₄-X2004 and WMO-CH₄-X2004A were Japanese labs that used their own gravimetric calibration scales.

Currently, QA/QC is still not systemised for measurements of stable isotopes; there is no WCC identified and there have been no performance audits performed. These shortcomings need to be addressed [30]. The work performed at the CCL aimed at the practical scale-realisation by JRAS mixtures should be independently verified by another laboratory [30]. This is crucial as the stable isotope scales are artefact-based and there is no primary method in use to verify absence of drifts and/or unrecognised biases. Round-robin inter-comparisons, the only QA tool currently in use, demonstrate lab-to-lab discrepancies in δ¹³C and δ¹⁸O on air-CO₂ which largely exceed the WMO network compatibility goals [30, 55]. These discrepancies, mainly offsets, in RR6 appear to be largely driven by the fact that most laboratories are not yet on the CCL realisation of the VPDB scale and several improvements have been proposed at GGMT-2017 for RR7 [30].

Figure 5 shows an update of the analysis made by Zellweger et al [76] including audit data until 2018 for carbon dioxide. The same analysis as above was made and also applied to the RR6 results. Again, the addition of new performance audit data confirms the advantage of the CRDS/off-axis integrated cavity output spectroscopy (OA-ICOS) techniques compared to NDIR. Figure 5 also shows comparisons with NMIs and Japanese laboratories reporting on their own calibration scales. Compared to the methane study, the proportion of laboratories complying with the WMO network compatibility goal was smaller for the audits by the WCC, which indicates that the transfer of the calibration to the actual field measurement is more challenging for carbon dioxide. However, this significantly improves if only audits of CRDS instruments are considered.

Zoom In Zoom Out Reset image size

In the 1960s the IAEA established tritium and stable isotope measurements of water in precipitation that required the development of specific stable isotope reference materials. In particular, this programme required the development of both the scale-defining reference materials and high-level reference materials aimed at scale realisation for stable isotope measurements of hydrogen, carbon, nitrogen, oxygen and sulphur. As custodian for these stable isotope scales, the IAEA maintains the stable isotope scales, monitors and introduces replacement primary reference materials, and develops new reference materials to address critical applications [15].

Currently, all stable isotope scales for δ¹³C, δ¹⁸O, δ²H and δ¹⁵N are based on artefacts, i.e. the ratio of isotopic species in the artefact used to define the scale-zero and where possible a scale span. These historical artefacts (or their later replacements) have also been used for the highest-level scale-realisation. With the exception of the δ¹⁵N_AIR scale, these are now exhausted for distribution. The scale defining materials include VSMOW, NBS19 and Air-N₂ (see table 3), and in contrast to greenhouse gas amount fraction scales, these scales are not based on an SI-traceable method, i.e. a mixture of isotopic entities (¹³C and ¹²C) with a desired isotope ratio. Nevertheless, best current measurements of the absolute ratios of scale defining primary standards exist and are the basis of calculating atomic weights of natural materials [80]. The exhausted scale-zero defining materials SMOW and PDB were first replaced by VSMOW and NBS19 (NBS19 defines ¹³C and δ¹⁸O values close to the scale zero) [84] and later have been replaced by VSMOW2 and IAEA-603 [15, 81, 82] with the value assignment based on measurements against the VSMOW and NBS19 materials. The first replacements are reflected in adding V (Vienna) to the scale abbreviations. IAEA-N-1 is the primary reference material for the N₂-Air scale [78].

The replacements, VSMOW2 and IAEA-603 were introduced to guarantee the consistency of the scale realisation. VSMOW2 was prepared at the IAEA by careful mixing of waters and is identical to VSMOW, within the uncertainties [83]. In contrast, marble carbonates of the desired isotopic composition cannot be prepared, so the isotopic composition of IAEA-603 [82] differs from that of NBS19.

All stable isotope ratio measurement instruments have been shown to suffer from some systematic effects (e.g. memory, cross-contamination, linearity, interference, sample gas preparation steps etc) that can influence the correctness of measurement results and require harmonised calibrations (see section 4). Therefore, in addition to the primary reference material a second scale 'anchor' (with its value assigned based on measurements against primary reference material under optimal conditions) is needed to synchronise the calibration scale span and to be used for data normalisation.

Standard light antarctic precipitation (SLAP), the second scale anchor accepted on the VSMOW scale (δ²H  =  −428‰ and δ¹⁸O  =  −55.50‰, values assigned without uncertainty [84]) was developed and later replaced by SLAP2 (δ²H  =  −427.5‰  ±  0.3‰ and δ¹⁸O  =  −55.50‰  ±  0.02‰ as assigned by measurements against SLAP [81]). The replacement SLAP2 was prepared at the IAEA by careful mixing of waters and is almost identical to SLAP (with 0.5‰ difference in δ²H for SLAP2) [83].

LSVEC (lithium carbonate) was introduced as the second anchor on the VPDB δ¹³C scale, with a value of  −46.6‰ accepted without uncertainty [85]. Later, LSVEC values were found to be drifting [86, 87] so IUPAC has recommended that LSVEC be discontinued as a reference material for δ¹³C data normalisation; other valid reference materials are suggested to be used for this purpose until a replacement for LSVEC is found [88]. However, values for these reference materials were assigned using LSVEC as a scale anchor so when the LSVEC replacement material is released the values of all standards that were standardised with LSVEC may need to be revised. Currently there are efforts underway to make a replacement available to the isotopic community as soon as possible.

All available stable isotope reference materials are based on and are traceable to the historical artefacts. The hierarchy of reference materials (traceability chain) is described below and as the chain is descended, the uncertainty of the reference material increases.

• Historical artefacts are used to define the scales and as a concept, each scale-defining material has its value assigned with zero uncertainty. In most cases, these artefacts are no longer accessible [78].
• Primary reference materials which either are, or have the closest link to, the historical artefacts. These primary reference materials are used to realise their respective scales, e.g. VSMOW and NBS19 and their replacements VSMOW2 and IAEA-603 (table 3). Like other reference materials, they have an uncertainty (at least due to homogeneity) and this is well characterised for VSMOW2 and IAEA-603 [81, 82].
• The scale-anchors. These materials are used to perform a two-point data normalisation for the scale. Example materials are SLAP (now replaced by SLAP2) for water and LSVEC (see below) for carbon.
• Secondary reference materials. These materials have been characterised against primary reference materials with their values normalised against the scale-anchors. They have larger uncertainty than the primary reference materials.
• Tertiary level reference materials. While there is no strict formal recognition of tertiary level reference materials there may be a valid case for recognising these in some instances. An example is carbon dioxide gas mixtures prepared and characterised against primary or secondary reference materials (carbonates), such as the JRAS air sets [89]. Other examples include matrix reference materials for GC combustion applications. In particular, this class of reference material is susceptible to stability issues and there is a need for it to be re-characterised periodically against more stable reference materials. Tertiary level materials have larger uncertainty than secondary reference materials
• Working laboratory standards. These are generally produced for specific laboratory applications and have the largest uncertainty that includes uncertainty from the reference material used to assign values and a component introduced by the two-point data normalisation.

The uncertainty introduced by the two-point data normalisation is intrinsic to all secondary and tertiary reference materials as well as working laboratory standards. Often a pair of reference materials is used for calibration purpose (addressing two-point data normalisation requirement), with a third reference material to be used for quality control.

Critical uncertainty limits for international reference materials are imposed by δ¹³C observations of atmospheric carbon dioxide. Inter laboratory comparability of such measurements should be within 0.01‰ to allow meaningful comparisons of data [30]. This, as well as the LSVEC-drift found [86, 87] and the strong need for an LSVEC-replacement, supports that IAEA focus on developing new reference materials for δ¹³C. Further improvements to the QA/QC, addressing the recently introduced ISO 17034 for reference material producers [90] and comprehensive uncertainty evaluation based on metrological principles have been recommended [15]. Using this new approach, the IAEA-603 preparation [82] included several steps as (i) a batch production of 5000 ampules flushed with argon and sealed off (material has been reserved for production of 4 such batches), (ii) 52 ampoules selected for the homogeneity study; (iii) extensive characterisation study performed at the IAEA. The assigned value δ¹³C  =  2.46‰  ±  0.01‰ (k  =  1) and δ¹⁸O  =  −2.37‰  ±  0.04‰ (k  =  1), for IAEA-603 have a dominating uncertainty component due to homogeneity (type A uncertainty) estimated over 5000 ampoules rather than characterisation measurements against the NBS19 primary reference (type B uncertainty) [82].

Three new carbonate reference materials, under development at the IAEA, have been sealed in glass ampoules in 2018 (around 11 000 in total) and material for further batches has been reserved. The development of carbon dioxide gas reference materials is planned to commence in 2019. Both types of reference material will cover a range of δ¹³C and δ¹⁸O values, have well-characterised uncertainty and be well-preserved to eliminate storage effects. These reference materials will be used for harmonised realisation of the VPDB δ¹³C scale and provide several scale-anchors on the scale. Thus, users will be able to select anchors in the most suitable form (carbonate or carbon dioxide) and with the most appropriate δ¹³C values. In particular, several scale anchors for δ¹³C will help (i) to monitor drifts in δ¹³C, if any and (ii) verify exactness of the ¹⁷O correction [91] applied in the mass-spectrometry software or in user' calculations. The value and uncertainty assignment for new reference materials shall be based on the traceability to the highest reference materials, careful validation of analytical methods in use and thoughtful assessment of all corrections involved as well as distinguishing type A and type B uncertainty components.

To realise scales independent from the historical artefacts, it is important that the absolute ratios (¹³C/¹²C, ¹⁸O/¹⁶O, ²H/¹H, ¹⁵N/¹⁴N) of the scale-artefacts are determined [15]; current best measurements of the absolute ratios as recommended by IUPAC can be found in Meija et al and references therein [79]. Several attempts to determine the ¹³C/¹²C value of the VPDB scale zero [92–95] have been made, however discrepancies among several ¹³C/¹²C determinations were observed and the absolute ratios are still not known with the uncertainty required for robust scale realisation which, for δ¹³C, must be 0.01‰ or lower. In addition, with the development of optical spectroscopic methods for stable isotope measurements, it will be necessary to develop reference materials or calibration approaches that allow both mass-spectrometry and optical spectroscopy results to be reported on the same reference scales. The realisation of isotopic scales may not be straight forward for users of optical analysers, i.e. scaling δ²H measurements of methane gas to VSMOW (primary reference is water), or δ¹³C measurements of atmospheric carbon dioxide to VPDB (primary reference is carbonate). Therefore, other practical solutions such as optical transfer standards [96] may be considered for the future. For isotopes of atmospheric carbon dioxide, the Jena reference air sets (JRAS), represent standards usable by both optical, and IRMS instruments (see section 3.3).

δ¹³C and δ¹⁸O isotopic values of atmospheric carbon dioxide are reported on the VPDB and VPDB-CO₂ scales respectively. The quantitative conversion of the carbonate to carbon dioxide using the classical method, i.e. the reaction with 100% phosphoric acid at 25 °C [102], does not lead to a fractionation of the carbon isotopic signature. However, the oxygen isotopic signature of the carbonate is not preserved as only 2 out of 3 oxygen atoms are transferred to the carbon dioxide molecule with some fractionation involved [100]. This is why δ¹⁸O values of carbon dioxide gas are reported on the VPDB-CO₂ scale, and not on the VPDB scale. Correspondingly, the oxygen isotopic composition of the produced carbon dioxide was found to be dependent on the reaction temperature, the isotopic composition of the acid, its concentration and the water content [103]. Since the results of the phosphoric acid reaction depend on numerous variables, the reaction is one of the limiting factors when considering inter-laboratory comparability of the isotopic composition of carbon dioxide gas. For example, a comparison between three laboratories has shown that carbon dioxide evolved from NBS 19 carbonate varies by no more than 0.003‰ in its δ¹³C_VPDB value, but by over 0.1‰ in its ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$ value [100]. One way to circumvent this problem is to make standard carbon dioxide gases with certified δ¹³C_VPDB and ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$ values available (NIST RM 8562, NIST RM 8563, NIST RM 8564, NARCIS I, NARCIS II, but table 4 demonstrates that more effort may still be needed).

Since GAW long term data comparability between laboratories requires that the same reference scale realisation be used, the WMO designated the Stable Isotope Lab of the Max Planck Institute for Biogeochemistry (BGC-IsoLab) as CCL for stable isotopic calibrations of measurements on atmospheric carbon dioxide. The BGC-IsoLab developed the JRAS-06 realisation of the VPDB-CO₂ scale consisting of two carbon dioxide in air mixtures [89, 104]. The carbon dioxide is derived from two carbonates, MAR-J1 (${{\delta }^{{\rm 13}}}{{{\rm C_{\rm VPDB}}}}$:  +1.958‰  ±  0.010‰; ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$:  −2.578‰  ±  0.031‰) and OMC-J1 (${{\delta }^{{\rm 13}}}{{{\rm C}}_{{\rm VPDB}}}$:  −4.373‰  ±  0.014‰; ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$:  −8.928‰  ±  0.025‰) [104]. The BGC-IsoLab has developed an automated carbon dioxide preparation system, ARAMIS (acid reaction and mixing system) that is used for the phosphoric acid reaction [89, 103] and routinely produces carbon dioxide gas from carbonates, including primary international reference materials, with a precision of 0.01‰ and 0.03‰ for δ¹³C_VPDB and ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$, respectively. The produced carbon dioxide is mixed into carbon dioxide free air achieving an amount fraction close to ambient, and these mixtures are then analysed by cryogenically extracting the carbon dioxide [89] in order to assign δ¹³C_VPDB and ${{\delta }^{{\rm 18}}}{{{\rm O}}_{{\rm VPDB{{\text -}}C}{{{\rm O}}_{{\rm 2}}}}}$ values. Since the mixing procedure and subsequent analyses introduces only small isotopic offsets between the pure carbon dioxide, and that mixed into carbon dioxide free air [100], isotopes of carbon dioxide in air are reported on the JRAS-06 scale realisation of the VPDB-CO₂ scale.

The ¹³C/¹²C and ²H/¹H isotope ratios of methane are reported on the VPDB and the VSMOW scale, respectively. The first δ¹³C measurements on atmospheric methane were made at the National Institute for Water and Atmosphere Research (NIWA, NZ) [105]. A barium carbonate standard (NZCH, later distributed as IAEA-CO-9) and a carbon dioxide standard gas (NBS-16) were used to standardise measurement results. These reference materials, re-characterised against LSVEC in 2006 [85], offered by NIST and IAEA in the past, are now exhausted.

Most laboratories have developed their own lab-standards and associated protocols to link δ¹³C and δ²H methane isotope data to the VPDB-, and VSMOW-scales, respectively [106]. The use of different calibration approaches in laboratories, different laboratory standards, inconsistent use of the ¹⁷O correction and not correcting for instrumental memory and cross-contamination have contributed to inter-laboratory measurement offset of up to 0.5‰ for δ¹³C, and up to 13‰ for δ²H measurements of atmospheric methane isotopes [106]. These inter-laboratory offsets clearly indicate that a unified effort needs to be undertaken to improve calibrations [106].

The most recent work linking isotopes of methane to primary reference materials was undertaken by Sperlich et al [107]. They selected a suite of pure methane gases that are directly characterised in the δ¹³C_VPDB and δ²H_VSMOW-SLAP values via elemental analyses (EA) and high temperature conversion (HTC) IRMS techniques, respectively. Significantly, the δ¹³C values of the methane gases were calibrated using LSVEC as a scale anchor, which means that these values may need to be revised once an LSVEC replacement is available. To address commutability requirements and have reference materials as close in composition as possible to samples, calibrated methane gas was mixed into methane free air, and these mixtures used to calibrate the working standard for methane in air measurements at the BGC-IsoLab [108]. δ²H–CH₄ scales linked to the δ²H_VSMOW-SLAP scale and based on synthetic mixtures of pure isotopologues [109] are in use at the Max-Planck-Institute for Chemistry in Mainz, Germany, and the Institute for Marine and Atmospheric research in Utrecht, Netherlands [106]. However, we note that the work by Sperlich et al [107] is the first attempt to generate a δ²H–CH₄ methane in air mixture that is directly traceable to primary reference materials.

¹⁵N/¹⁴N and ¹⁸O/¹⁶O isotope ratios of nitrous oxide are reported on the AIR-N₂ and the VSMOW scale, respectively. Being an asymmetric linear molecule the central (α) and terminal (β) nitrogen atoms of nitrous oxide provide position-specific ${{\delta }^{{\rm 15}}}{{{\rm N}}_{{\rm AIR{{\text -}}}{{{\rm N}}_{{\rm 2}}}}}$ isotopic information [110]. The difference between ${{\delta }^{{\rm 15}}}{{{\rm N}}_{{\rm AIR{{\text -}}}{{{\rm N}}_{{\rm 2}}}}}$ values of the central (δ¹⁵N^α) and terminal (δ¹⁵N^β) nitrogen position of the nitrous oxide molecule is called site preference and defined as SP  =  δ¹⁵N^α  −  δ¹⁵N^β, while the average ${{\delta }^{{\rm 15}}}{{{\rm N}}_{{\rm AIR{{\text -}}}{{{\rm N}}_{{\rm 2}}}}}$  =  (δ¹⁵N^α  +  δ¹⁵N^β)/2. The synthesis of nitrous oxide by the thermal decomposition of isotopically characterised ammonium nitrate has been suggested as an approach to link the position-dependent nitrogen isotopic composition of nitrous oxide to AIR-N₂ [110]. During decomposition, the nitrogen atom of the ${\rm NO}_{3}^{-}$ ion is converted into the central nitrogen position of N₂O, while the end nitrogen comes from the ${\rm NH}_{{\rm 4}}^{+}$ ion. However, the accuracy of this approach for the calibration of δ¹⁵N^α and δ¹⁵N^β was found to be limited by non-quantitative ammonium nitrate decomposition in combination with substantially different isotope enrichment factors for the conversion of the ${\rm NO}_{3}^{-}$ or ${\rm NH}_{{\rm 4}}^{+}$ nitrogen atom into the α or β position of the nitrous oxide molecule [111]. The yield of the ammonium nitrate decomposition reaction to nitrous oxide was found to be dependent on reaction conditions, e.g. temperature, which impedes the implementation of this technique in many laboratories [112]. Consequently, the compatibility between eleven IRMS and laser spectroscopy laboratories was found to be limited to approximately 6‰ for the nitrous oxide site preference [113]. In order to overcome such experimental artefacts when comparing data between laboratories, two new nitrous oxide reference gases, USGS51 and USGS52 have been produced [98] and are available (table 4). The delta values provided for USGS51 and USGS52 are based on a preliminary assessment by Naohiro Yoshida and Sakae Toyoda/Tokyo Institute of Technology without uncertainty estimates and offer only a small range of δ¹⁵N and δ¹⁸O values (<1‰), they are not suitable for a two-point calibration. Therefore, novel nitrous oxide reference materials are being developed within the European Metrology Programme for Innovation and Research (EMPIR) 16ENV06 project 'Metrology for Stable Isotope Reference Standards (SIRS)' [114]. Reference materials will be available as pure gas, to be used directly for traditional mass-spectrometry and/or for mixture preparation by users and as nitrous oxide diluted in whole air for laser spectroscopic measurement techniques.