Introduction

Carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O) are all greenhouse gases (GHGs) that make important contributions to climate change. A recent analysis from the Global Atmosphere Watch (GAW) Observations Program of the World Meteorological Organization (WMO) shows that the globally averaged mole fractions of GHGs reached new highs in 2013. These maxima include (396.0 ± 0.1) μmol mol−1 for CO2, (1824 ± 2) nmol mol−1 for CH4, and (325.9 ± 0.1) nmol mol−1 for N2O [1, 2]. Compared to pre-industrial levels (prior to 1750), these are relative changes in concentration of 140, 250, and 120 %, respectively. A recent update by the National Oceanic and Atmospheric Administration (NOAA) reports the preliminary global CO2 level at 400.83 μmol mol−1 as of March 2015 [3]. All three species are long-lived greenhouse gases (LLGHGs) which absorb electromagnetic radiation and affect the budget of tropospheric and stratospheric ozone (O3). Thus, these LLGHGs contribute significantly to atmospheric radiative forcing and have an important role in atmospheric chemistry. As the most important anthropogenic GHG in the atmosphere, CO2 has contributed 65 % to radiative forcing by LLGHGs. Methane is the second most important contributor to radiative forcing by LLGHGs, with a contribution of ≈17 % followed by N2O at ≈6 % [1, 2].

In order to accurately establish trends in gas concentration, assess the role of LLHGLs in atmospheric chemistry, and relate measurement records from many laboratories and researchers, it is essential to have a stable, accurate, and internationally recognized source of calibration standards or scales. The longest available records of atmospheric CO2 measurements date back to those started in 1957 at the Scripps Institution of Oceanography (SIO), La Jolla, CA. These measurements were based on manometry [4]. NOAA began their CO2 measurement program in 1968 and eventually developed a manometric system for absolute calibrations of CO2 in dry-air mixture gas [5]. NOAA is the WMO Central Calibration Laboratory (CCL) for CO2, CH4, N2O, and some other key atmospheric species [6, 7]. Thus, NOAA maintains calibration standards and scales and disseminates that scale to WMO participants via tertiary calibration mixtures. They maintain their own measurement records for these LLGHGs which can be found at www.esrl.noaa.gov/gmd. Another organization that has a long history of measurements, beginning in 1978, is the Advanced Global Atmospheric Gases Experiment (AGAGE) at http://agage.eas.gatech.edu/. AGAGE does not focus on CO2 but does measure and track CH4, N2O, and halocarbons.

The ever-increasing number of measurements outside of the current atmospheric measurement networks had increased demand for standards. In 2009, the United States Congress funded the National Institute of Standards and Technology (NIST) to support climate change research. As part of this effort, NIST’s Gas Sensing Metrology Group (GSMG) began developing gas standards required for accurate measurements of GHGs for ground-based and ambient atmosphere applications. Initial efforts included the development of gas-cylinder-contained gravimetric primary standard mixture (PSM) suites of ambient atmospheric-level CH4 and N2O in a dry-air balance as previously described [8, 9]. A PSM suite for ambient-level CO2 was also produced but has not yet been described in the literature, although key issues that had to be addressed in the preparation techniques have been documented [10].

A key component of the NIST primary standards program involves maintaining strong international links with other National Metrology Institutes (NMIs) through key comparisons proposed by the Gas Analysis Working Group (GAWG) of the Consultative Committee for Amount-of-Substance: Metrology in Chemistry and Biology (CCQM). The CCQM’s mission is to establish global comparability of measurement results through promoting traceability to the SI [11]. CCQM key comparisons of ambient CO2, N2O, and CH4 have been conducted in 2006, 2008, and 2013, respectively [12–14]. The results of these comparisons show agreement within the reported measurement uncertainties between NIST, many other NMIs, and NOAA. We note that as a signatory of the Mutual Recognition Arrangement (MRA) of the Consultative Committee of Weights and Measures (CIPM), the WMO recommends the CCL for the species being compared as an official participant in a CCQM Key Comparison [6]. Thus, NOAA, as the WMO-CCL for CO2, CH4, and N2O, can participate in key comparisons for these three analytes.

The PSM suites developed at NIST are kept in-house and are used to certify other calibration mixtures such as reference materials (RMs) and certified reference materials (CRMs). CRMs certified by NIST are trademarked as Standard Reference Materials® (SRMs). After NIST developed and verified the PSM suites for CO2, CH4, and N2O through in-house processes, and bilateral and/or international key comparisons, these mixtures were subsequently used to certify SRM 1721 Southern Oceanic Air.

SRM 1721 is one of two new, first ever, natural air SRMs. It was developed to support those needing calibration standards: (a) state and local government agencies performing ground-level urban and remote measurements, (b) the automotive industry which measures N2O fugitive emissions from autos needing a baseline calibration point, (c) mega-cities projects needing a low-end calibration point, (d) academia performing research and atmospheric measurements, and (e) scientists in general measuring these GHGs in many atmospheric environments. In the remainder of this article, we describe the development and eventual certification of this new SRM.

Experimental procedure

Preparation of mixtures

The SRM candidate mixtures were prepared from southern oceanic air by the National Institute of Water and Atmospheric Research (NIWA) at their research station at Baring Head, New Zealand, as shown in Fig. 1. The facility, contained in three small buildings, is located at the coast on a cliff 79 m above the ocean at latitude of 41° 25′ S and longitude of 174° 52′ E (the windiest area in New Zealand). The prime conditions for pristine oceanic air occur during the months of March and April (white arrow in Fig. 1a) when the prevailing winds originate from Antarctica and subsequently traverse the southern Pacific Ocean to New Zealand. The sampling tube/line is mounted on a tower outside the building at a height of about 4.6 m as indicated by the white arrow in Fig. 1b.

Fig. 1
figure 1

a optimal wind direction from Antartica shown with white arrow and b Position of sample line of tower of NIWA Barring Head station buildings

Full size image

Thirty new aluminum compressed gas cylinders of 29.5 L internal volume, conditioned for ambient CO2 gas mixture use by Scott-Marrin, Riverside, CA, were pumped with air to a pressure of approximately 14.5 MPa. The Rix SA6E compressor used to pressurize the cylinders was equipped with a 0.01-μm pore-size filter (SMC, SFD200-N02) in order to remove salt and other particulates during the sampling process. Additionally, the sampled air was dried in three stages. The first two stages involved the use of two drying towers on the pump. In this process, air was spun inside each unit, with water being condensed and drained from below while the “dryer” air exited the top. These two stages were followed by a final high-pressure step which used magnesium perchlorate as a chemical desiccant. The high-pressure samples are collected with a factory-standard ring material, which imparts no significant change to the air composition.

NIWA used a Siemens Model 3A non-dispersive infrared (NDIR) spectrometer to measure CO2 concentration at the start of the sampling process for each cylinder. The fill date, start time (NZST), wind direction, wind speed, and the CO2 mole fraction at the start of sampling are given in Table 1. After pressurizing 30 cylinders, these samples were shipped to NIST in Gaithersburg, MD, USA, for commencement of the analysis and certification process.

Table 1 SRM Southern Oceanic Air sample data
Full size table

Analysis of SRM mixtures

One of the 30 cylinder samples, sample number 1721-AL-01 (cylinder # CC2598), was designated as the lot standard (LS). The LS was analyzed repeatedly and served as an analytical control to compensate for any instrument drift during a measurement sequence. One to three of the candidate SRM sample cylinders were analyzed between analyses of the LS. All analyses were accomplished using a computer-operated gas analysis system (COGAS) developed at NIST. COGAS consists of 12 ports + 1 common stainless steel gas sampling stream selection valve where one cylinder sample is connected to a single port. While one port is being sampled, the other 11 are dead ended with the sample stream flow through the common port and controlled by a needle valve. A low dead-volume two-stage regulator (Model Y12-C144D, Airgas Specialty Gases, Riverton, NJ) is attached to each cylinder with a section of 0.16-mm stainless steel tubing from the regulator to the port in the gas sampling valve. Up to 11 samples can be connected to COGAS, one of which is designated as a control (the LS in this instance). The 12th port is reserved as a “rest port” which is connected to some source of air, either a cylinder mixture or “house” air. After completion of all cycles, the COGAS stream selection valve is advanced to the rest port. The computer program allows the user to select the number of samples to run between two control samplings. A complete sequence represents each of the samples being analyzed, and up to ten sequences can be made in continuous operation

The COGAS was used to randomly sample from six to ten samples in a full cycle. Ten instrument response measurements were taken as each sample was analyzed, and an average was calculated. All measurement sets were equally separated in time. A ratio (r i ) of the ith candidate sample (cs i ) to that of the LS was determined by dividing the corresponding average instrument response \( {\overline{S}}_i \) by the interpolated value of the LS responses according to

$$ {r}_i=\frac{{\overline{S}}_i}{{\overline{S}}_{\mathrm{LS},1}+\frac{i\left({\overline{S}}_{\mathrm{LS},2}-{\overline{S}}_{\mathrm{LS},1}\right)}{i_{\mathrm{cs}}+1}} $$
(1)

where \( {\overline{S}}_{\mathrm{LS},1} \) and \( {\overline{S}}_{\mathrm{LS},2} \) are the averaged LS responses that bracket the candidate sample(s) and i cs is the number of candidate samples within a complete sequence. A typical analysis sequence would be LS, cs1, cs2, and LS, for which i cs = 2, representing an “LS set.”Footnote 1 (In some cases, i cs = 1.) Analyses were repeated for enough cycles in order to obtain a minimum of ten ratios of each sample to the LS. The Picarro Models G1301-c and G2101-I cavity ring-down spectroscopy (CRDS) analyzers were used to measure the CH4 and CO2 concentrations, respectively, and a Los Gatos Research Model 23d off-axis Cavity-Enhanced Absorption Spectroscopy (CEAS) system was used for measurement of the N2O. All three instruments measured the same sample stream by sampling the main sample line via internal instrument pumps as illustrated in Fig. 2.

Fig. 2
figure 2

Diagram of instruments sampling of gas stream

Full size image

All ratios determined for CO2, CH4, and N2O for each SRM candidate sample are listed in the Electronic Supplementary Material (ESM) Table S1. Each SRM candidate sample, which is assigned a “Sample Number,”Footnote 2 as listed in ESM Table S1, was analyzed and bracketed by analyses of the LS. Several SRM samples are analyzed between the LS analyses representing one “LSSet”1. Continuous analyses of the SRM samples, randomly chosen, representing several LSSet1 without interruption represent a “Break Set.”Footnote 3 There can be several Break Sets3 during a full “Day”Footnote 4 of analyses. The “Port Number” is the port on COGAS to which the sample gas is connected. This configuration may or may not have been the same for all ratio determinations.

Analysis of SRM LS to primary standards

The same instrumentation was used as above to analyze the LS against the suites of gravimetric primary standard mixtures (PSMs). Development of the CH4 in dry, whole-air (natural) suite (NIST_CH4_2011) and N2O in synthetic air (O2/N2) suite (NIST_N2O_2012) has previously been discussed [8, 9]. Preliminary analysis of the N2O using the CEAS system indicated that argon needed to be present when measuring whole-air samples, a known phenomenon due to the impact of the matrix composition on the pressure broadening coefficient of the absorption line used by the analyzer to deduce the N2O mole fraction [15–18]. Therefore, a new suite of N2O PSMs was developed in 2014 (NIST_N2O_2014) in a balance of synthetic air which included argon. A bias between assigned values to an air sample using the argon-free versus argon-included N2O PSM suites was confirmed and found to be ~.8 nmol mol−1 when using CEAS. The development of these new N2O PSMs has not yet been published. NIST is currently comparing the NIST_N2O_2012 (matrix of ~20.9 % O2/79.1 % N2) and NIST_N2O_2014 (matrix of ~0.93 % Ar/20.9 % O2/78.2 % N2) suites with a gas chromatograph/electron capture detector (GC/ECD) to make sure the suites are in agreement. (Theoretically, there should be no matrix affect due to argon when using GC/ECD.)

When using these spectroscopic instruments for analyses of natural air samples, the isotopic composition may also come into play. In the case of CH4, measured with the Picarro Model G1301-c instrument, the PSMs were prepared from pure methane (99.9993 % ± 0.06 %), with an isotopic composition of −43 ‰ 13δC (VPDB), in methane-free dry whole air. This then would mimic the whole-air SRMs, thus reducing both pressure broadening and isotopic composition concerns. However, a recent study suggests that this is not a significant issue for CH4 [19]. In the case of CO2, the authors realize the spectroscopic issues related to the isotopic compositions and different isotopologues that are involved when measuring ambient samples. However, while the authors were able to measure the 13C-CO2 (described later) to the total CO2, it is only considered a rough determination with higher uncertainties than required. Therefore, we cannot properly address these issues at this time. Methods development for accurate 13C and 12C measurements as well as the development of isotopic standards is in its beginning phases at NIST and will be part of future work in GHG standards development.

The NIST CO2 primary gravimetric standards were prepared from pure CO2 obtained from Airgas Specialty Gases (Riverton, NJ) and dry CO2 scrubbed whole air from Scott-Marrin (Riverside, CA). The same general procedures described for the preparation of CH4 and N2O [8, 9] were used to prepare the CO2 suite. The pure CO2 was recovered from the flue at the AES Warrior Run power plant fired from bituminous coal mined in Maryland. NIST obtained a large sample of this CO2 from Airgas who sourced it from the Linde Carbonic facility (Cumberland, MD). The GSMG at NIST analyzed the pure CO2 for other components of air and determined it to be (99.9994 % ± 0.0002) % mol mol−1 CO2. Further analyses to determine the relative amounts of 12C and 13C were made by Fourier transform infrared spectroscopy (FTIR) (Nicolet Nexus 670, Nicolet Instrument Corp., Madison, WI). A nominal 500 μmol mol−1 PSM, prepared from this pure CO2 and dry CO2 scrubbed dilution air, was analyzed. Using the same approach as is common in the atmospheric community [20–22], the relative 13C content was determined by ratio of the responses to each other; C13/12C . The resulting ratio was determined to be 0.01092 ± 0.00022 (1.092 ± 0.022) %, resulting in a δ13C-CO2 isotopic delta value of about −28 ‰ (VPDB scale) in the PSMs compared to ≈−8 ‰ (VPDB) determined in two of the SRM candidate mixtures. (Only two of the 30 samples were measured for the δ13C-CO2 as at this time NIST does not have PSMs for 13C-CO2 or the instrumental equipment necessary to make accurate measurements and suitable uncertainties).

The dilution air was analyzed for major components including O2, Ar, CH4, N2O, and CO, as well as trace CO2 (typical levels 0.2 μmol mol−1). Nitrogen was determined by subtraction. Atomic masses taken from IUPAC data [23, 24] were used to calculate the molecular mass and the uncertainty of the dry whole-air diluent gas for each cylinder of natural air from the sum of the analyzed component fractions. The isotopic content of the 12C and 13C in the pure CO2 was also taken into account in the calculation of the gravimetric CO2 mole fraction. The development of the NIST CO2 PSMs (NIST_CO2_2012) has not yet been published, although previous work describes issues that needed to be addressed such as absorption/desorption of CO2 within the cylinders used to prepare the PSMs [10]. The cylinders used to prepare each of the CO2 PSMs were tested for initial absorption, and the gravimetric mole fractions were adjusted accordingly.

The measurements of the LS to the CH4, CO2, and N2O PSM suites were carried out individually. COGAS was used to randomly sample all PSMs against the LS in a full cycle. The average of ten instrument response measurements, given by \( {\overline{S}}_{\mathrm{PSM}} \), was obtained for the analysis of each PSM. A ratio (r PSM) of each PSM to the LS was determined by dividing \( {\overline{S}}_{\mathrm{PSM}} \) by the average instrument response to the LS given by \( {\overline{S}}_{\mathrm{LS}} \). This definition gives

$$ {r}_{\mathrm{PSM}}=\frac{{\overline{S}}_{\mathrm{PSM}}}{\frac{1}{2}\left({\overline{S}}_{\mathrm{LS},1}+{\overline{S}}_{\mathrm{LS},2}\right)} $$
(2)

Only one PSM was randomly analyzed between LS analyses, within one Break Set for a full cycle and several cycles during a full “Day” of analyses. Analyses were repeated for enough cycles over several days in order to obtain a minimum of ten PSM-to-LS ratios for each PSM. The PSM cylinder numbers used in the calibrations for CO2, CH4, and N2O are listed in Table 2 along with the gravimetric mole fraction, its associated combined standard uncertainty (u) and expanded uncertainties (U) representing an approximate 95 % confidence interval. The u values can be considered as “one standard deviation” estimates that include all known components of measurement uncertainty.

Table 2 PSMs used to value assign mole fractions to SRM 1721 samples
Full size table

Other analytes

All the cylinders were analyzed for moisture content using a Meeco Aquavolt (P2O5) moisture analyzer and carbon monoxide (CO) by Cavity-Enhanced Absorption Spectroscopy (CEAS), Los Gatos Research. Molecular oxygen (O2) and argon (Ar) were determined in only two cylinders in order to conserve sample pressure in the SRM samples. The O2 analysis was accomplished using a Siemens Oxymat 6 paramagnetic analyzer calibrated with NIST SRM 2659a LS (71-DL-01) at 20.7236 % ± 0.0005 % (0.002 % relative). Argon was analyzed using an Agilent 6890 gas chromatograph equipped with a thermal conductivity detector (GC/TCD) calibrated with NIST SRM 1676 LS (19-BL-01) at 0.980 % ± 0.001 %.

Results and discussion

The analysis functions used to assign values to the SRM and the evaluations of uncertainty for the assigned values were computed as described. These functions have been built using methods that extend the methods described in ISO 6143 [25] to take into account the case when statistical evaluations of uncertainty are based on small numbers of observations and when Monte Carlo methods are used to evaluate the uncertainty associated with the certified values [26, 27]. These procedures are routinely used by the NIST Gas Sensing Metrology Group for value assignment and uncertainty evaluation of reference materials.

Final expanded uncertainties (U) are expressed and calculated from

$$ {U}_{95\%}=ku $$
(3)

with coverage factor k determined separately for each GHG species and described later. The true value for each analyte mole fraction, x analyte, is asserted to lie in the interval [x analyte ± U 95 %(x analyte)] with a level of confidence of approximately 95 %. Data treatment discussions follow for each of the greenhouse gas species being certified in this SRM 1721.

Function for carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O)

Four CO2, five CH4, and six N2O PSMs listed in Table 2 were analyzed against the SRM LS control, and instrument response ratios were calculated. The ratios, mole fractions, x analyte, and associated uncertainties u(x analyte) for each PSM are listed in ESM Tables S2, S3, and S4, respectively. Figure 3a, b depicts the analysis function (first-order polynomial) selected from among the first-, second-, and third-order polynomials for CO2 and CH4. (Second- and third-order polynomials were tested to assure that the instruments were responding in a linear manner as would be expected.) This expression gives the mole fractions, c j (ordinates) of the jth PSM (denoted by PSM j ), as a function of the measured response ratios (abscissae). The fitted first-order polynomial analysis function gives the unknown SRM CO2, (\( {c}_{{\mathrm{CO}}_2,j} \)), or CH4, (\( {c}_{{\mathrm{CH}}_4,j} \)), mole fraction in terms of the response ratio as

Fig. 3
figure 3

a, b Plots of first-order polynomial for CO2 and CH4 PSMs from ratiometric and gravimetric mole fraction data, and residuals between gravimetric and corresponding predicted values of the PSMs by the analysis function. c represents the mole fraction (concentration) in micromoles per mole for CO2 and nanomoles per mole for CH4. ĉ represents the predicted mole fraction (concentration) in micromoles per mole for CO2 and nanomoles per mole for CH4. r is the ratio of the peak areas of the PSM to the LS. The gray regions represent the 95 % coverage of the uncertainty in the values

Full size image
$$ {c}_{{\mathrm{CO}}_2,j}={\overline{r}}_{\mathrm{SRM},j}{b}_1+{b}_0 $$
(4)

where \( {\overline{r}}_{\mathrm{SRM},j} \) is the average SRM sample response ratio for CO2, or CH4, corresponding to the jth SRM. A residual plot, the difference between the gravimetric mole fraction, c, and the predicted, ĉ, for each polynomial, is also shown below the analysis function plot including a 95 % coverage region for such differences.

Figure 4 depicts the N2O analysis function (second-order polynomial) selected from among first- through fifth-order polynomials. This expression gives the mole fractions, c j (ordinates) of the jth PSM (denoted by PSM j ), as a function of the measured response ratios (abscissae). It should be noted that while the fifth-order polynomial, depicted in ESM Fig. S1, better captures a slight curvature at the low end, it has larger associated uncertainties (see gray uncertainty bounds). The second-order polynomial has more marked lack of fit but smaller associated uncertainties. Because the differences in the SRM sample \( {x}_{{\mathrm{N}}_2\mathrm{O}} \) are not significantly different (<0.03 nmol mol−1 in most cases), as shown in ESM Table S5, we chose to use the second-order polynomial fit. The fitted analysis function gives the unknown SRM N2O mole fraction in terms of the response ratio as

Fig. 4
figure 4

Plot of second-order polynomial for N2O PSMs from ratiometric and gravimetric mole fraction data, and residuals between gravimetric and corresponding predicted values of the PSMs by the analysis function. c represents the gravimetric mole fraction (concentration) in nanomoles per mole. ĉ represents the predicted mole fraction (concentration) in nanomoles per mole. r is the ratio of the peak areas of the PSM to the LS. The gray regions represent the 95 % coverage of the uncertainty in the values

Full size image
$$ {c}_{{\mathrm{N}}_2\mathrm{O}}={{\overline{r}}^2}_{\mathrm{SRM},j}{b}_2+{\overline{r}}_{\mathrm{SRM},j}{b}_1+{b}_0 $$
(5)

where \( {\overline{r}}_{\mathrm{SRM},j} \) is the average SRM sample response ratio corresponding to the jth SRM. A residual plot, showing the difference between the gravimetric mole fraction, c, and the predicted, ĉ, for the N2O polynomial with the 95 % coverage region, is also shown beside the analysis function plot. That all but one residual are positive suggests systematic lack of fit for the second-order polynomial, possibly owing to a non-linear response of the instrument used for measurement. Since a third-order polynomial would improve the residual pattern without reducing the magnitude of the residuals overall, we adopted the simpler model.

Table 3 lists the coefficients b 0 and b 1 for the CO2 and CH4 functions. This linear response equation was used to assign CO2 and CH4 mole fraction values and their combined standard uncertainties to the LS, 1721-AL-01, (391.338 ± 0.034) μmol mol−1 and (1771.06 ± 0.34) nmol mol−1, respectively. The x analyte and u(x analyte) were also calculated for each SRM sample using the respective ratio data and Eq. 4. The post hoc value for the 95 % expansion factor of Eq. 3 was computed using Monte Carlo method results as k = U 95 %/u: k(CO2) = 1.78 and k(CH4) = 1.97. The U 95 %(\( {x}_{{\mathrm{CO}}_2} \)) and U 95 %(\( {x}_{{\mathrm{CH}}_4} \)) for each SRM sample were calculated using Eq. 3. Table 3 also lists the coefficients b 0, b 1, and b 2 for the N2O second-order function. This quadratic response equation was used to assign an N2O value to the LS of 324.101 nmol mol−1 with a combined standard uncertainty (k = 1) of ±0.042 nmol mol−1. A value and combined standard uncertainty were also calculated for each SRM sample using their respective ratio data and Eq. 5. The post hoc value of the coverage factor was computed as the ratio between the 95 % expanded uncertainty and the combined standard uncertainty, both evaluated using the Monte Carlo method, k = 1.72. The expanded uncertainty (U 95 %) of \( {x}_{{\mathrm{N}}_2\mathrm{O}} \) for each SRM sample was calculated using Eq. 3. The individual x analyte, u(x analyte), and U 95 %(x analyte) for all three analytes in each SRM sample are given in Table 4.

Table 3 Function coefficients for CO2, CH4, and N2O
Full size table
Table 4 NIST mole fraction values and uncertainties for SRM 1721 Southern Oceanic Air samples
Full size table

Components of air

Water content ranged from (0.5 to 20.4) μmol mol−1 while CO values were (36 to 56) nmol mol−1. Due to instability in the NIST PSMs, in the form of growth of CO observed over time, and possibly the SRM samples themselves, it is not possible to certify the CO at this time. Further research is needed to assess the CO instability problem and determine if a suitable cylinder/treatment process can be developed to solve this problem. The H2O and CO (at the time of its determination) are given in ESM Table S6 but are provided only as information of potential interest. Cylinders 1721-AL-01 and 1721-A-14 were analyzed for molecular oxygen (O2) and argon. The mole fraction values were the same for both samples at 20.93 and 0.93 % for oxygen and argon, respectively.

Final certified values

The final certified mole fraction x analyte and U 95 %(x analyte) for CO2, CH4, and N2O corresponding to each of the 27 SRM samples are given in Table 5. The \( {U}_{95\%\left({\mathrm{CO}}_2\right)} \) range from (±0.07 to ±0.14) μmol mol−1 or (0.018 to 0.035) % relative to the certified value. The U 95 %(analyte) for the CH4 and N2O were rounded up to ±1.0 nmol mol−1 (0.06 % relative) and ±0.20 nmol mol−1 (0.06 % relative), respectively. The uncertainties associated with the mole fractions of the analytes in the SRM samples are often smaller than those associated with the PSMs. This is a consequence of the averaging effect of the regression function used for value assignment, and of the assumption that none of the uncertainty components expressed in the uncertainty evaluations for the PSMs reflect persistent errors (biases). In future versions of the procedure currently in use at NIST, we will make provisions to accommodate the presence of persistent errors that will not average out as more and more sample values are combined into derivative quantities. The Monte Carlo method described in [26] is sufficiently flexible to allow such accommodation.

Table 5 NIST-certified mole fraction values for SRM 1721 Southern Oceanic Air samples
Full size table

One of the goals in this SRM development was to provide standards with uncertainties that were equivalent or exceeded the WMO Data Quality Objectives (DQOs) [28]. The combined standard uncertainties, u, of the certified values as given in Table 2 are well within the WMO DQOs (which are one standard deviation specifications) for CH4 and N2O, the DQO for CH4 = 2 nmol mol−1 and N2O = 0.1 nmol mol−1. The standard deviation of CH4 values of standards from their reference values was 1.7 nmol mol−1 in the most recent international comparison [14] compared to the reported relative combined uncertainty, u (k = 1), of <0.5 nmol mol−1 for this work. This smaller u is owed to the use of the NIST statistical program assessing the variability of the ratios as previously mentioned in the previous section.

Final certified values

For 25 of the 27 SRM cylinders, the u(\( {x}_{{\mathrm{CO}}_2} \)) meets the southern hemisphere DQO of ±0.05 μmol mol−1. The NIST u includes both the uncertainty in the PSMs used for calibration and the measurement of the individual SRM sample.

Conclusions

This SRM will assist those scientists, researchers, and state and government agencies that need or require certified, traceable GHG concentration standards. It will allow the calibration of instrumentation used to provide GHG concentrations and to a wide variety of groups and monitoring communities. Additionally, this SRM provides the metrological basis required for researchers to accurately track the long-term growth rates of CO2, CH4, and N2O in urban, rural, and atmospheric environments. For those laboratories that do not need to report to the WMO/GAW (which requires traceability to the CCL (NOAA)), this SRM will be of particular interest. For instance, NIST has received verbal inquiries from the Automobile Industry/Government Emissions (AIGER) group about obtaining SRM 1721 specifically for the measurement of N2O. The certified N2O concentration in SRM 1721 will complement the recently developed SRM 1718 which contains N2O, in a synthetic air blend of Ar/O2/N2 mimicking whole-air mole fractions, at nominal mole fraction value of 1000 nmol mol−1, and a mixture specifically developed for AIGER members at ≈400 nmol mol−1 in O2/N2 synthetic air.

Future development of GHG primary standards at NIST will focus on well-characterized isotopic gas standards beginning with the carbon isotopes of CO2. Pure 12C-CO2 and 13C-CO2 starting materials will be blended that will have final mole fractions with different isotopic delta values. Those standards can then be used to calibrate the optical and mass spectroscopic instruments used to measure samples or atmospheres for CO2. Eventually, the isotopes of O2 will also be well characterized within those primary standards.