Comparison of halocarbon measurements in an atmospheric dry whole air sample.

The growing awareness of climate change/global warming, and continuing concerns regarding stratospheric ozone depletion, will require continued measurements and standards for many compounds, in particular halocarbons that are linked to these issues. In order to track atmospheric mole fractions and assess the impact of policy on emission rates, it is necessary to demonstrate measurement equivalence at the highest levels of accuracy for assigned values of standards. Precise measurements of these species aid in determining small changes in their atmospheric abundance. A common source of standards/scales and/or well-documented agreement of different scales used to calibrate the measurement instrumentation are key to understanding many sets of data reported by researchers. This report describes the results of a comparison study among National Metrology Institutes and atmospheric research laboratories for the chlorofluorocarbons (CFCs) dichlorodifluoromethane (CFC-12), trichlorofluoromethane (CFC-11), and 1,1,2-trichlorotrifluoroethane (CFC-113); the hydrochlorofluorocarbons (HCFCs) chlorodifluoromethane (HCFC-22) and 1-chloro-1,1-difluoroethane (HCFC-142b); and the hydrofluorocarbon (HFC) 1,1,1,2-tetrafluoroethane (HFC-134a), all in a dried whole air sample. The objective of this study is to compare calibration standards/scales and the measurement capabilities of the participants for these halocarbons at trace atmospheric levels. The results of this study show agreement among four independent calibration scales to better than 2.5% in almost all cases, with many of the reported agreements being better than 1.0%.


Introduction
Stratospheric ozone depletion has been linked to the presence of halogenated trace gases that include chloroluorocarbons (CFCs) and hydrochloroluorocarbons (HCFCs) (Montzka et al., 2011). CFCs and HCFCs together with HFCs are strong greenhouse gases (Forster et al., 2007;Prinn et al., 1998). Research groups around the globe continuously measure halocarbons and monitor their growth/decline in the atmosphere (Blake et al., 2003;Montzka et al., 1996Montzka et al., , 1999O'Doherty et al., 2004;Artuso et al., 2010;Prinn et al., 2000). Measurements from the National Oceanic and Atmospheric Administration (NOAA) can be found at www.esrl.noaa.gov/gmd; from the Advanced Global Atmospheric Gases Experiment (AGAGE) at http://agage.eas.gatech.edu/ and http://agage.mit.edu, and from the University of California Irvine (UCI) at http://cdiac.ornl.gov/trends/otheratg/blake/blake.html and http://ps.uci.edu/~rowlandblake/. hese measurements of the atmospheric abundance of halocarbons assist in eforts to determine their sources and sinks, their global distributions, and how these change over time in response to natural and anthropogenic processes.
Factors such as sampling or analytical methods/techniques may cause diferences in abundance reported by researchers. However, the most important and inluential factors usually are the calibration standards or scales used to determine those abundances. Gravimetric preparations of gas phase volatile organic compound (VOC) standards, including hydrocarbons and halocarbons, that are either in the gas or liquid phase at room temperature, have been reported (Montzka et al., 1993;Happell and Wallace, 1997;Rhoderick and Dorko, 2004;Rhoderick, 2006;Rhoderick et al., 2010). Additional literature cites the gravimetric preparation of standards/scales for methane and nitrous oxide which use the same or similar preparation techniques (Dlugokencky et al., 2005;Hall et al., 2007;Rhoderick et al., 2012;Kelley et al., 2014). In order to track and control the global and regional emissions of these species, and relate data from diferent laboratories, it is necessary to demonstrate measurement compatibility at the lowest levels of uncertainty. his requires reliable, accurate and precise measurements maintained over time. While there are agreed upon uncertainty requirements for precision measurements for some key atmospheric gases such as carbon dioxide, methane, carbon monoxide and nitrous oxide (World Meteorological Organization, 2013), there are no agreed-upon levels for halocarbons. However, those preparing their standards always strive to improve their accuracy and reduce the uncertainties, as well as reducing measurement uncertainties. By reducing uncertainties, trends assessment becomes easier and more reliable from the measurement data.
hese factors are of particular importance to the atmospheric monitoring and measurement communities such as the NOAA Global Monitoring Division (GMD), AGAGE and UCI. heir main objectives are to: 1) maintain adequate stability of their laboratories' internal calibration scales and thereby ensure that the atmospheric records they produce are internally consistent for determination of trends and gradients in atmospheric mole fractions over decades, and 2) maintain close links with other laboratories so that atmospheric data may be reliably merged across multiple laboratories and methods for interpretation, including atmospheric models.
Most nations have a National Metrology Institute (NMI) whose mission is to promote innovation and industrial competitiveness through the advancement of measurement science, standards and technology to improve the quality of life and economic security of that nation; https://net.nist.gov/fo/nistmission, http://www.bipm.org/en/worldwide-metrology/. NMIs have established a quality system through which each NMI establishes credentials related to their individual needs. hrough this process the NMIs demonstrate equivalence to each other's measurement processes and reference standards and therefore measurement accuracy is important. his established equivalence allows users around the world to obtain reference materials and calibrations from another country's NMI, if needed, thus supporting global commerce. hrough the Consultative Committee for Amount of Substance -Metrology in Chemistry (CCQM), the NMIs participate in key comparisons in order to demonstrate the equivalence of measurement capabilities. Degrees of equivalence are calculated from these key comparisons in order to support NMI Calibration and Measurement Capabilities (CMCs). For each key comparison a Key Comparison Reference Value (KCRV) is calculated and the degrees of equivalence calculated relative to the KCRV. Only CCQM member participants may participate in a key comparison. he NMIs have much experience with the preparation of gas standards, but very little, if any experience analyzing whole air samples.
here have been several reported comparisons between researchers in the halocarbon measurement community (Rasmussen, 1978;Jones et al., 2011;Hall et al., 2014). Additionally an important role of the quadrennially-published World Meteorological Organization (WMO) Ozone Assessment Reports allow for comparison results for ozone-depleting substances and their substitute gases within the atmospheric measurement community (Montzka et al., 2011). However, there have been very few opportunities between the NMI's and measurement/monitoring communities to compare standards and scales (Rhoderick and Dorko, 2004;Hall et al., 2014). herefore, this comparison was developed to include other laboratories and agencies regularly measuring these halocarbons in the atmosphere. his study evaluates and compares the reported results of participants, representing NMIs and the atmospheric community on a smaller scale than many reported studies, to a comparison reference value. his report describes the results of a comparison for several of the most abundant halocarbons at atmospheric mole fraction levels including the chloroluorocarbons (CFCs) dichlorodiluoromethane (CFC-12; ≈ 530 pmol mol -1 ), trichloroluoromethane (CFC-11; ≈ 240 pmol mol -1 ), and 1,1,2-trichlorotriluoroethane (CFC-113; ≈ 75 pmol mol -1 ); the hydrochloroluorocarbons (HCFCs) chlorodiluoromethane (HCFC-22; ≈ 225 pmol mol -1 ) and 1-chloro-1,1-diluoroethane (HCFC-142b; ≈ 22 pmol mol -1 ); and the hydroluorocarbon (HFC) 1,1,1,2-tetraluoroethane (HFC-134a; ≈ 65 pmol mol -1 ) in a real, dry, air sample. his study is viewed as an assessment of how each participant compares to each other.

Sample mixture for the comparison
he single gas mixture circulated as the comparison study sample was prepared by the Global Monitoring Division, Earth System Research Laboratory, NOAA, in Boulder, Colorado, US. Whole air was sampled, dried and pumped at Niwot Ridge, Colorado, into a new, nominal 30 L, aluminum cylinder that had been treated with the Aculife IV TM process (Air Liquide America Gases, Plumsteadville, Pennsylvania, US) to passivate the cylinder walls. Previous unpublished data has shown that many halocarbons are unstable in untreated aluminum cylinders. Some data demonstrating two years storage of several halogenated trace gases in nitrogen contained in Aculife IV TM treated aluminum cylinders have been reported (Miller and Rhoderick, 1995). he mole fraction of CFC-12, CFC-11 and CFC-113 in that stability study were at slightly higher pmol/mol (ppt) levels (9 to 35%) than the levels in this current sample. Although data have not been published, the National Institute of Standards and Technology (NIST) have observed stability for these three species, in a dry air matrix, for 15 years in similarly-treated cylinder types. (he samples must be dried as moisture will destroy the treatment). he cylinder was pressurized to nominal 12.5 Mpa. he Niwot Ridge facility is used by NOAA to prepare cylinder mixtures of dried whole air for various uses, including supplying mixtures for calibration of greenhouse and related trace gas measurements for laboratories which do not prepare their own primary standards.

Stability study of the comparison sample
NIST analyzed the comparison sample two times to determine its stability over the comparison time scale. Initially, NOAA analyzed 3 of the halocarbons followed by an analysis of all 6, and then a inal analysis after each participant had analyzed the sample. hese multiple analyses were used to demonstrate stability of the halocarbons in the mixture over the time period of this comparison. he mole fractions and expanded uncertainties (U) for those analyses are given in Table 1. Expanded uncertainties are calculated using the equation: (1) where u c is the uncertainty of known measurement results including those in the calibration standards and k is the coverage factor. he data in this study are reported using k=2 which implies a level of conidence of approximately 95% (http://physics.nist.gov/cuu/Uncertainty/coverage.html). he stability data are also displayed graphically in Figure S1. CFC-12, CFC-11, and HCFC-22 stability data statistically indicate that there has been no drift over time for each laboratory's individual set of data points, as all error bars overlap. While the NIST and NOAA data for CFC-113 indicate a small bias in reported values, the individual sets of data show stability, as all error bars overlap for each data set. NOAA reported values were determined using gas chromatography with either electron capture detection (GC-ECD) and or coupled to a mass spectrometer (GC-MS) showing a slight bias between the two instrumental methods. However, each data set for the two methods indicates stability. Of note is that the NOAA uncertainties shown in Table 1, k=2, are only based on their measurement precision and not measurement reproducibility, which would be a more appropriate uncertainty for detecting drift based on few samples. However, each limited data set by itself appears to indicate no drift of the halocarbons in the comparison sample. (hroughout the text k=2 represents the approximate 95% uncertainty conidence interval.) he NIST k=2 error bars for the two HFC-134a stability values overlap which indicates that the HFC-134a has remained stable. Even though the uncertainties overlap the diference between the two NIST data points is 2.3 pmol/mol. his suggests the HFC-134a has increased in the cylinder which is not a likely scenario based on past experience with similar mixtures at low nmol mol -1 levels. However, both NOAA data points and the second NIST value agree within the k=2 error bars, thus suggesting stability. An explanation for the larger diference between the 1 st and 2 nd NIST data points is most likely related to instrumental issues. he 1 st NIST data point was determined using GC-FID-preconcentration of the sample. he HCF-134a eluted on the tail of a rather large unknown peak in the chromatogram making the peak area determination diicult. he 2 nd NIST measurement was taken using a GC-MSD system which was not available for the 1 st analysis. While the standard uncertainty, 1.8 pmol/mol, is much larger for the measurement by GC-MSD, the column used for this measurement yielded better separation of HFC-134a from the other halocarbons, so the 2 nd NIST stability data point is most likely a more accurate representation of the mole fraction of the HFC-134a in the comparison sample.
he HCFC-142b stability data show results very similar to HFC-134a in that the 1 st NIST data point was determined using GC with a lame-ionization detector (GC-FID) and preconcentration, and the 2 nd data point using GC-MSD. In both cases the HCFC-142b peak was baseline-to-baseline separated with no interferences. Even though the 2 nd NIST data point is 0.7 pmol/mol (3.2% relative) higher than the 1 st , it is not a likely scenario that HCFC-142b is increasing. Peaks for the irst NIST data points for both the HFC-134a and HCFC-142b were very small, making peak area determinations relatively imprecise compared to other gases. he two NOAA data points are in agreement, suggesting stability.

Comparison of dry whole air sample
Participating laboratories in this comparison, listed in Table 2, analyzed the dry, whole air sample contained in aluminum cylinder AAL073358, in between stability measurements by NIST and NOAA. A time table for the comparison is shown in Table S1. Each laboratory was allowed appropriate time as needed for measurements and shipping of the cylinder to the next participant. Each laboratory was requested to provide their value determination and uncertainty for each halocarbon as a mole fraction from at least 3 individual determinations. An uncertainty budget, description of their analysis procedure, and their calibration methods were also requested. hese individual measurement reports can be found in the Supplemental Materials: Text S1, Text S2, Text S3, Text S4, and Text S5. We note here that the pressure in sample cylinder AAL073358 after all laboratories analysis and all stability measurements were completed was ≈ 4.8 Mpa (700 psi).
Instrumentation and methods development for the measurement of halocarbons has been previously documented (Montzka et al., 1993;Simmonds et al., 1995;Miller et al., 2008). Methods used for this comparison were solely at the laboratories discretion, and relect their normal measurement procedures as listed in Table 3.

Results
A small group of participants were selected for this initial "pilot" study coordinated by NIST in order to keep it manageable as only one sample was being circulated. Two NMIs, the Korea Research Institute of Standards and Science (KRISS) and NIST, and three atmospheric research laboratories, NOAA, Scripps Institution of Oceanography (SIO), and the Swiss Federal Laboratories for Materials Science and Technology (Empa), participated in this halocarbons comparison. Each participant develops and perpetuates their own calibration scales with the exception of Empa, as they use the SIO scales used by AGAGE and other projects for the species under discussion. Since the comparison was to be used to determine the agreement between diferent scales, the Empa data were not used in determining a comparison 'halocarbon (x) reference value (RV)' ( x RV ). However, their data are extremely useful since they are using the SIO scales and those results should be comparable within scale propagation uncertainties. Additionally SIO and Empa use the same type of instrument for analysis, the GC/MS Medusa system (Miller et al., 2008). herefore, any diferences in reported values between these two laboratories may be due to factors such as introduction of the sample into the instrument or other artifacts.

Data comparison among participants maintaining own standards/scale (KRISS, NIST, NOAA, SIO)
All measurement data were reported to NIST with the reported value (x i ) and expanded uncertainty (U xi ) (approximate 95% conidence interval) in pmol/mol given in Table 4. A standard uncertainty (u xi ) of a reported value was calculated by each laboratory for a reported value. he u xi includes the precision (measurement) uncertainties (p xi ) and accuracy (standards or calibration scale) uncertainties (a xi ) and is calculated using equation 2: he expanded uncertainty, U xi , is then calculated using equation 3:   where the coverage factor (k) equals 2 for an approximate 95% conidence interval. he reference values (x RV ) for the comparison were calculated from the values reported by the participants, excluding Empa, for each halocarbon. he x RV were calculated using the DerSimonian-Laird random efects model, a simple and wellestablished non-iterative weighted-mean estimator that includes a between-participant variance component as well as the participant-reported uncertainties in the weights (DerSimonian and Laird, 1986). Figures 1-6 display the results of the DerSimonian-Laird model for each halocarbon excluding the Empa data. (Empa data will be compared to the SIO results later). he open circles represent the reported x i for each participant and the vertical bars span the k =2 expanded uncertainty interval, x i ±U xi , reported by the participants. he horizontal black line represents the x RV with the red lines representing the k =2 expanded uncertainty interval of the x RV , x RV ±U(x RV ). Also given in Table 4 is the diference (d i ) between the RV and the participant's value, and % relative diference between the participants reported value and the x RV for each of the six halocarbons, except for Empa. he U(x RV ) are listed as both absolute pmol/mol and percent relative, 100*U(x RV )/x RV .
Uncertainties assigned to the halocarbon values are on the same order of magnitude for all participants with a few exceptions. hose reported by KRISS for HFC-134a, HCFC-22 and HCFC-142b are larger relative to NIST, NOAA and SIO. KRISS used a preconcentrator (Gerstel Co.) coupled to an Agilent Gas Chromatograph (7890A) equipped with an Agilent mass spectrometer (5975C) for their analyses. he large analytical uncertainties are a result mainly because of a non-uniform recovery rate of cryogenic adsorption and desorption at the preconcentrator. NIST assigned a lower uncertainty for CFC-11 and CFC-113 relative to NOAA, KRISS and SIO. he NIST method used (multi-step dilutions) resulted in a low uncertainty for the NIST gravimetric standards. Additionally, the analytical uncertainty was lower compared to the other halocarbons resulting in a lower expanded uncertainty (k = 2). he uncertainty for CFC-12 reported by SIO is noticeably larger relative to those reported by the other participants.
Dichlorodiluoromethane : he DerSimonian-Laird model statistics results for the CFC-12 are illustrated in Figure 1. All participants reported uncertainty bars intercept the x RV line (solid black) demonstrating good agreement with the x RV . he SIO value is slightly outside the upper uncertainty limit of the x RV (red line), however their uncertainty covers the upper and lower uncertainty bands. All participant d i are ≤ 0.6% relative to the x RV as shown in Table 4, accentuating the good agreement for CFC-12.
Trichloroluoromethane (CFC-11): Figure 2 shows the DerSimonian-Laird results for CFC-11. he reported values for SIO, NIST and NOAA all lie within the uncertainty bands of the x RV . he KRISS value
CFC-11 laboratory reported values in relationship to the DerSimonian-Laird reference value (RV) represented by the solid black line. he error bar represents the expanded uncertainty, k=2, reported by participants.
lies outside the lower x RV uncertainty, but their uncertainty bands intercept the x RV line indicating agreement. As with the CFC-12, all participant d i are ≤ 0.6% relative to the x RV as shown in Table 4, demonstrating the good agreement for 1,: he CFC-113 DerSimonian-Laird results are depicted in Figure 3 and appear to show the largest variations for any of the halocarbons in this comparison. KRISS, SIO, and NOAA data points all are within the x RV uncertainty bands. All three of these participants' uncertainty bands for these points intersect the x RV line indicating agreement. he NIST value, with very small uncertainty, is the only one in question as it is not within the x RV uncertainty estimates. Since the NIST CFC-113 value is signiicantly diferent it is probably biasing the RV. If we exclude the NIST value from the calculations, as shown in Figure S2 (NIST value in red not included in RV calculation), then the RV = (75.17 ± 0.65) pmol/mol (0.86% relative) compared to an RV of (75.9 ± 1.3) pmol/mol (1.7% relative) when including all 4 participants.
NIST used their CFC-113 2004 standards, prepared for the International Halocarbons in Air Comparison Experiment (IHALACE) (Hall et al., 2014), which will be discussed later, and newly prepared 2011 standards for this comparison. he 2004 and 2011 standards showed consistency predicting the CFC-113 in the comparison sample to within 0.1% using both suites. While the reported NIST uncertainty is small, all known sources of uncertainty were included in their calculations. Since NIST is predicting the CFC-113 high, it is possible that there is consistent loss in the aluminum cylinders containing the standards. However, this has not been observed by NIST in the past. It is more plausible that the analytical method is not optimized and there may be a small contribution from another component, CH 3 Cl, in the sample. All participant d i are < 2.3% relative to the x RV as shown in Table 4.
1,1,1,2-Tetraluoroethane (HFC-134a): Results for the HFC-134a data are shown in Figure 4. he NIST (low) and KRISS (high) values both lie outside the x RV uncertainty; however their uncertainties overlap the x RV uncertainty bands. he low NIST value is attributed to analytical issues, in particular the sensitivity and repeatability. SIO and NOAA data points intersect the x RV line with the d i < 1.0% relative to the x RV as shown in Table 4. he d i for NIST is 2.5% and 6.0% for KRISS.
Chlorodiluoromethane : he results for HCFC-22 are in agreement as shown in Figure 5. All participants' data points and uncertainties are on or within the x RV uncertainty bands. All participant d i are ≤ 0.9% relative to the x RV as shown in Table 4.
1-Chloro-1,1-diluoroethane (HCFC-142b): he results for HCFC-142b are very similar to the HCFC-22, but with larger diferences from the x RV . Agreement to the x RV is observed as shown in Figure 6. All participants' data points and uncertainties are on or within the x RV uncertainty bands. All participant d i are < 1.8% relative to the x RV as shown in Table 4.

Comparison of Empa data to SIO scale
In the case of the Empa data in Table 4, the diference is calculated versus the SIO value: x Empax SIO . Since Empa uses the SIO scale (used by AGAGE and other projects for the halocarbons discussed here) their uncertainties propagate from those of the SIO scale and their own measurement uncertainties. Empa is not  in bold indicate that the uncertainty bounds of the submitted values do not cover the % diference of the reported values and may be signiicant. Additionally, inter-calibration factors to the NOAA values for each of the comparisons are given in Table S2.
In general, NIST-NOAA diferences show consistent results for CFC-12 and CFC-11 compared between the agencies over a 15 yr. time period from the 2004 comparison to this work. (Measurements were actually taken during 1998-1999 followed by publication in 2004). he -1.5% diference for the CFC-12 in the IHALACE study is only slightly signiicant as the uncertainty bounds are minimally smaller and do not cover the diference. However, the uncertainties given for the IHALACE study are the standard deviation (sd) of the measurements. If we expanded those uncertainties as an approximate 95% conidence interval, as so done in Table 5, then they do cover the -1.5% diference. Considering both the NIST and NOAA uncertainty bounds for the CFC-11, they overlap, therefore the -0.9% diference is not signiicant. Considering all three studies, the average diferences for CFC-12 are -0.7% and 0.4% for CFC-11, while there is a larger diference for CFC-113 (average 2.5%) but consistent. he 3.8% diference for the CFC-113 in the IHALACE study is signiicant as the uncertainty bounds (sd) do not cover the diference. his present study was the irst between NIST-NOAA for the HCFCs and HFC-134a, resulting in diferences for HCFC-22 (-1.4%), HFC-134a (-2.9%), and HCFC-142b (-1.8%). An interesting observation is that NIST predicts lower values than NOAA for all but one of the halocarbons. Based on the associated % relative uncertainties of the reported values, there is no signiicant diference in the reported values.
his study represents the irst comparison that KRISS has participated in involving the measurement of halocarbons in a whole air sample. he diferences are within 1.5% of the NOAA values for all halocarbons except HFC-134a (5.5%) that is relatively high compared to the other participants. NIST and KRISS, the two NMI participants, were in agreement to ≤0.7% for CFC-12, CFC-11, HCFC-22 and HCFC-142b, but had larger disagreements for CFC-113 (3.3%) and HFC-134a (8.7%). he NIST-NOAA agreement (-2.9%) is closer with the other participant-NOAA diferences for HFC-134a, while the KRISS-NOAA diference is much larger at 5.5%. KRISS-NOAA agreement (-0.3%) is in line with SIO-NOAA (-0.3%) and Empa-NOAA (-1.5%) for CFC-113, while the NIST-NOAA agreement is much larger at 3.0%.

Discussion
his halocarbon comparison study has served to bring together two communities involved in supporting a broader group of scientist needing standards for measurements of atmospheric halocarbons. Each of the two communities has a diferent need and approach. he atmospheric community needs a stable scale based on precise measurements and continuity in order to establish trends of these gases in the atmosphere, and to relate many independent measurement sets. he NMIs need to establish degrees of equivalence between those NMIs in order to make CMC claims to support a global commerce where any NMIs standards may be used by others universally. As a requirement, accuracy is of the highest importance, as are the associated uncertainties in those standards. his study serves only as a measure of how well those two communities standards and scales for trace level halocarbons agree, and not as to what is necessarily the truth.
he atmospheric community has much more experience in actual measurements of whole air samples, and has reined their analytical instrumentation and methods. he NMIs have many years of experience developing and analyzing standards, in "clean synthetic air", of some of these halocarbons at trace levels, but not the level of experience measuring whole air samples. Whole air samples are much more complex than synthetic air in that they include a matrix of many compounds including permanent gases, hydrocarbons, halocarbons, volatile organic compounds and unknowns. his makes for a very complex chromatographic/mass spectral analysis where it can and is diicult to separate many components. Considering these factors, the results of this pilot study/comparison indicate rather good agreement between the participants. All reported participants values or associated uncertainties, k=2, lie within or at least intercept the DerSimonian-Laird x RV k=2 uncertainty limits for 97% of the data. All participant diferences from the x RV are within 2.5% except one and 66% are within 1.0%. his demonstrates that the diferent methods and techniques used to prepare standards/scales, and the measurement systems and techniques used to assign mole fractions to halocarbons in a dry whole air sample are consistent within the uncertainties reported. he only participant that does not make their own standards for the discussed species is Empa; it uses the SIO scales. heir values are in agreement within the uncertainties of the SIO data demonstrating that the analytical systems are in good control. he Empa uncertainties are large for all the halocarbons making it diicult to draw any conclusions on scale transfer.
While most of the diferences are not signiicant based on the k=2 uncertainties, changes are probably signiicant from study to study for some. For example, the NOAA CFC-12 scale is stable to about 0.3% over time, so even though the overall uncertainty is 0.7%, they expect to be able to reproduce results on a given scale to ∼1.8 ppt (0.3%), (95% conidence interval). On this basis, the NIST-NOAA diferences that move from -0.4% to -1.5% to -0.2% could be signiicant, depending on stability of the NIST scale. So, to compare results from one study to the next, one would have to assume that there have been no scale changes (or all are consistent) and also know the long-term reproducibilities. his is a topic which is better addressed in separate paper closely assessing these long-term reproducibilities.
NIST has prepared and analyzed standards for all of the halocarbons studied here for more than 30 years but at nmol/mol (ppb) to µmol/mol (ppm) levels. NIST has experience preparing and analyzing standards for CFC-12 and CFC-11 at atmospheric levels, but not for HFC-134a, HCFC-22 and HCFC-142b. Some of the diferences for NIST and KRISS are most probably due to lack of experience in analyzing halocarbons at the very low pmol/mol (ppt) levels. Continued eforts should assist in honing those capabilities resulting in improved agreement. Anticipated future comparisons coordinated by NIST will reach out to other willing participants.

Funding information
NOAA work was supported in part by the NOAA Climate Program Oice's AC4 program. Measurement at SIO was supported by grants from NASA's Upper Atmosphere Research Program as part of the Advanced Global Atmospheric Gases Experiment (AGAGE). Measurements at Empa were supported by the Swiss Federal Oice for the Environment (FOEN) within the project HALCLIM and European FP7 Infrastructure Project "Integrated Non-CO 2 Greenhouse gas Observing System" (InGOS).

Competing interests
he authors declare no competing interests.
Disclaimer: Certain commercial equipment, instruments and materials are identiied in order to specify experimental procedures as completely as possible. In no case does such identiication imply a recommendation or endorsement by the National Institute of Standards and Technology or other participants in this study nor does it imply that any of the materials, instruments or equipment identiied are necessarily the best available for the purpose.