CSIRO ATMOSPHERIC RESEARCH ASPENDALE, VICTORIA, AUSTRALIA GASLAB CALIBRATION INFORMATION (June-2001) This document contains information on CSIRO's calibration strategies to minimize data uncertainties. In addition to the trace gases and d13C presently included in CDIAC DB1021, information on N2O and d18O is also given; data pertaining to those variables are expected to be added to CDIAC DB1021 at some point in the future. The CSIRO Atmospheric Research sample collection strategies have been described elsewhere [Francey et al., 1996]. GC CALIBRATION (R.L. Langenfelds, L.P. Steele, and R.J. Francey) There are numerous sources of experimental uncertainty in atmospheric trace gas measurements and various ways in which these uncertainties impinge on interpretation of data. A comprehensive calibration strategy consists of various parts that help to minimise and define errors/uncertainties and distinguish relative contributions from different sources. Here we provide a brief summary of CSIRO's calibration strategies, and assign realistic uncertainties to key experimental parameters (Table 1). These parameters are defined in a way to be directly applicable to field data. Uncertainty estimates are derived from various sources of information such as long-term statistics of instrument performance, intercomparisons with other laboratories (eg. of CO2 with NOAA/CMDL who are also the current WMO Central CO2 Calibration Laboratory, with the responsibility of maintaining and propagating the WMO Mole Fraction Scale [Zhao et al., 1997]) and results from laboratory and field tests. Key laboratory tests include evaluation of non-linearity in instrument response, regulator performance, preparation and analysis of test flasks (multiple flasks, of the same type used for field sampling programs, simultaneously filled from a high pressure cylinder) for characterisation of trace gas modification due to 1) analytical procedures tailored for limited sample volumes and 2) storage of air in these flasks [Cooper et al., 1999]. Data in Table 1 represent experimental uncertainty only and make no allowance for atmospheric variability. CO2 CO2 is analysed by gas chromatography (GC), involving conversion of the separated CO2 to CH4 on a heated nickel catalyst (400 degrees C) followed by flame ionisation detection (FID). Data are reported in the WMO CO2 Mole Fraction Scale. The link to this scale was established with 9 primary standards (of a suite of 10 synthetic mixtures of CO2, CH4 and CO in zero air) in high-pressure cylinders that were calibrated by NOAA/CMDL in 1992. They span a CO2 range of 291-377 ppm. NOAA recalibrated a subset of 4 of these in 1994 (mean differences -0.01, -0.04, 0.00, +0.05 ppm at 362, 349, 339, 326 ppm respectively). The primary suite retains 75% of original pressure (2000 psig) after 9 years. The scale is monitored using two broad approaches. Relative stability is monitored using ~15 assorted secondary standards with lifetimes of 4-10+ years. Stability against absolute scales is monitored (but the scale is never adjusted) by independent comparisons, including 6 Nippon Sanso standards (volumetrically prepared, calibrated against a gravimetric scale at Tohoku University), WMO Round-Robin (3 cylinders in 1994-1997 and again in 1998-1999), IAEA CLASSIC (5 cylinders: 1996-1998, 1999-2000) and other high pressure cylinder comparisons, and flask air sharing comparisons with several laboratories (including ~6 per month with NOAA/CMDL since 1992; Masarie et al., 2001). The best measure of how closely CSIRO has remained aligned to the WMO scale is obtained from results of 8 (most reliable) cylinder comparisons, but excluding the 9 primary standards. They imply a mean scale factor of 0.99986 +/- 0.00012 (CSIRO/NOAA), equivalent to a difference of -0.05 +/- 0.04 ppm (CSIRO-NOAA) at current ambient atmospheric CO2 levels. This difference reflects collective uncertainty from both laboratories. There is no significant drift in the difference, but there are insufficient statistics to resolve a constant offset from any drift. Uncertainties due to unaccounted variations of non-linearity in our instrument response are within 0.2% of mole fraction difference (eg. within +/- 0.04 ppm for a CO2 difference of 20 ppm) and apply to any measurement of concentration differences, for example seasonal cycle amplitudes. Additional uncertainties apply to measurements from flask samples due to experimental factors specific to flasks. This is exemplified by the Cape Grim flask intercomparison program involving CSIRO and NOAA/CMDL [Masarie et al., JGR, in press] where systematic differences of between 0.0 and 0.2 ppm have been observed, despite a much closer level of agreement demonstrated for high pressure cylinders. Resolution of such discrepancies must remain a high priority if CO2 records from flask sampling networks of different laboratories are to be merged into a self-consistent global database. Further details of analytical techniques and calibration for CO2 and other species measured by GC will be provided elsewhere [Langenfelds et al., in preparation]. CH4 CH4 is analysed by GC (FID). Data are reported in the CSIRO94 CH4 scale [Steele et al., 1996], which is derived from the CH4 scale maintained at NOAA/CMDL. The link to this scale was established with 2 high pressure cylinders containing dry, natural air that were calibrated by NOAA/CMDL between 1987 and 1990. Results from later exchange (1992-1997) of 12 high- pressure cylinders indicate a small, but measurable difference. The ratio of CSIRO/NOAA values for these samples is 1.00021 +/- 0.00010, implying a difference of +0.36 +/- 0.17 ppb (CSIRO-NOAA) at 1700 ppb. Stability of the CSIRO scale is monitored with ~25 assorted standards with lifetimes of 4-10+ years. Instrument response has been evaluated with a suite of 6 Nippon Sanso CH4-in-air standards (volumetrically prepared, calibrated against a gravimetric scale at Tohoku University) spanning the range 310-1845 ppb. The results show the response to be linear within confidence limits. A deviation from linearity of 0.2% of mole fraction difference was measured but is of similar magnitude to uncertainty of the gravimetric preparation technique. We thus treat the response as linear and allow for uncertainty of 0.2% of mole fraction difference, equivalent to only 2 ppb in a difference of 1300 ppb. CO CO is analysed by GC with "reduction gas detection" where the separated CO reduces heated (275 degrees C) mercuric oxide to mercury vapour that is subsequently detected by UV absorption. Data are linked to the gravimetrically-derived scale of NOAA/CMDL [Novelli et al., 1991] using a single high-pressure cylinder standard with CO mole fraction of 196 ppb. This standard is one of five synthetic mixtures of CO2, CH4 and CO in zero air, in the range 30-196 ppb, that were calibrated at NOAA/CMDL between 1992 and 1994. The reason that only the highest concentration standard is used to link the CSIRO and NOAA scales is that large discrepancies exist in the respective laboratories' determination of relative mole fraction among these standards and other high pressure cylinders exchanged since. The highest concentration is likely to give the smallest proportional error in linking the scales. We have established the instrument response characteristics of the CSIRO instrument by diluting air containing above- atmospheric CO and CH4 mole fraction with varying amounts of zero air. Precise dilution ratios are determined by analysis of CH4 for which non-linearity in instrument response is negligible by comparison to CO. The results of five such experiments between 1993 and 1999 show the relative CO mole fraction among these standards to consistently differ from that indicated by the NOAA assignments. Using 196 ppb as the fixed reference point for linking to the NOAA scale, differences of up to 4 ppb are found at lower concentrations. These discrepancies are most likely due to different treatment of instrument non-linearity, despite the fact that both laboratories employ similar analytical techniques. NOAA has used either cubic or quadratic functions to describe their instrument response. At CSIRO, we have found it necessary to use a different function (y = ax(sup 2) + bx + cx(sup d), where y is CO mole fraction, x is peak height counts and a, b, c, and d, are fitted parameters) that better captures sharper non-linearity in instrument response at low concentrations (especially below 100 ppb). Stability of the CSIRO scale and variations in instrument response are monitored with ~20 high-pressure cylinder standards, with lifetimes of 4-10+ years, spanning a CO range of 20-400 ppb. From our experience of maintaining a CO measurement program and extensive laboratory tests, we have identified several specific effects that can adversely affect data quality. For example, instrument "blanks" (i.e. appearance of CO peaks in chromatograms from air containing no CO) affect the shape of the instrument response function, especially at low CO concentrations, and are monitored by regular analysis of zero air (where necessary scrubbed of residual CO). It is common for CO to be produced by internal surfaces of regulators and/or cylinder valves of high-pressure cylinders. Avoiding significant errors requires careful selection of standards (especially working standards), and adherence to suitable gas handling procedures that limit the residence time of air inside the valves/regulators. We have also observed an instrumental memory effect that causes H2 and CO measurements to be affected by the CO mole fraction of preceding chromatograms. Knowledge of the existence of these effects allows us to implement analytical and/or data processing procedures that limits resulting uncertainty to acceptable levels. Further details will be provided elsewhere [Langenfelds et al., in preparation]. H2 H2 is analysed by GC on the same instrument and with identical techniques to that described above for CO. Data are reported in the CSIRO94 H2 scale which was defined by dilution of high purity H2 and CH4 (in a 1:3 ratio) with zero air to produce a mixture with H2 mole fraction close to atmospheric levels, and "bootstrapping" to a gravimetrically-derived, absolute CH4 scale. The relationship of the CSIRO94 H2 scale with the gravimetrically-derived H2 scale of NOAA/CMDL [Novelli et al., 1999] is not well-defined due to both time and concentration- dependent variations in the difference. Results from the Cape Grim flask air-s haring intercomparison show systematic, time-dependent differences of between 0 and 20 ppb between 1992 and 1998 [Masarie et al., JGR, in press]. CSIRO instrument response was determined by the same "dilution experiments" described above for CO and was found to be significantly non-linear and of similar shape to that of CO. Unlike CO, however, a constant response function is assumed because 1) we have too few well-behaved standards with mole fraction significantly different from ambient atmospheric levels to adequately describe variation in the response function (y = ax(sup 2) + bx + cx(sup d), although these standards do constrain the magnitude of any variations and 2) the proportional range of variation in the background atmosphere is much smaller for H2 than for CO so that minor changes in instrument response are less critical for atmospheric studies. Stability of the scale is monitored with ~20 high-pressure cylinder standards, with lifetimes of 4-10+ years, spanning a H2 range of 430-1000 ppb. Most of these cylinders are of electropolished, stainless steel construction. In our experience, the aluminium cylinders successfully used for long-term calibration of other GASLAB species are generally not reliable for H2. Apart from a single batch of 5 of these cylinders, all have been found to grow H2 at varying rates of up to hundreds of ppb per year. The suite of ~15 stainless steel and 5 aluminium cylinders are all stable against each other to better than +/-1 ppb yr-1. N2O N2O is analysed by GC with electron capture detection (ECD). The scale maintained at CSIRO was established using 6 high-pressure cylinder standards (of a high purity N2O/CO2 mixture diluted with varying amounts of zero air) that were gravimetrically prepared by NOAA/CMDL in 1993. Their N2O mole fraction values were assigned on the basis of the gravimetric preparation with a nominal uncertainty of +/-1 ppb, and span a range of 264-344 ppb. In 1995, two high-pressure cylinders were exchanged with NOAA/CMDL's Halocarbons and other Atmospheric Trace Species (HATS) group. Comparison of GC measurements, which are of higher precision than the gravimetric assignments, indicated a scale factor of 1.0014 (CSIRO/NOAA) at this time, equivalent to a difference of 0.44 ppb (CSIRO-NOAA) at 315 ppb. However, further cylinder intercomparisons will be necessary to reliably quantify the scale difference and identify any time variation. A difference of similar magnitude is observed in early results from the Cape Grim flask intercomparison conducted with NOAA/CMDL's Carbon Cycle Group (CCG), but the precise relationship between HATS and CCG scales is not known at this stage. Stability of the CSIRO scale is monitored with ~30 high-pressure cylinder standards, with lifetimes of 4-9+ years. Gas Chromatography References Cooper, L.N., L.P. Steele, R. L, Langenfelds, D.A. Spencer and M. P. Lucarelli, Atmospheric methane, carbon dioxide, hydrogen, carbon monoxide and nitrous oxide from Cape Grim flask air samples analysed by gas chromatography, in Baseline Atmospheric Program (Australia), 1996, edited by J. L. Gras, N. Derek, N. W. Tindale and A. L. Dick, pp. 98-102, Bureau of Meteorology and CSIRO Division of Atmospheric Research, Melbourne, Australia, 1999. Francey, R.J., L. P. Steele, R. L. Langenfelds, M. P. Lucarelli, C. E. Allison, D. J. Beardsmore, S. A. Coram, N. Derek, F. R. de Silva, D. M. Etheridge, P. J. Fraser, R. J. Henry, B. Turner, E. D. Welch, D. A. Spencer and L. N. Cooper, Global Atmospheric Sampling Laboratory (GASLAB): supporting and extending the Cape Grim trace gas programs, Baseline Atmospheric Program (Australia) 1993, edited by R. J. Francey, A. L. Dick and N. Derek, pp. 8-29, Bureau of Meteorology and CSIRO Division of Atmospheric Research, Melbourne, Australia, 1996. Langenfelds, R. L., L. P. Steele, L. N. Cooper, D. A. Spencer, D. M. Etheridge and M. P. Lucarelli, CSIRO GASLAB measurement of CO2, CH4, CO, H2 and N2O by gas chromatography, 1991-2001, CSIRO Atmospheric Research Technical Report, in preparation. Masarie, K.A., R.L. Langenfelds, C.E. Allison, T.J. Conway, E.J. Dlugokencky, R.J. Francey, P.C. Novelli, L.P. Steele, P.P. Tans, B. Vaughn and J.W.C. White, NOAA/CSIRO Flask-Air Intercomparison Experiment: A strategy for directly assessing consistency among atmospheric measurements made by independent laboratories, J. Geophys. Res., 106, 20445-20464, 2001. Novelli, P.C., J.W.Elkins and L.P. Steele, The development and evaluation of a gravimetric reference scale for measurements of atmospheric carbon monoxide, J. Geophys. Res., 96, 13109-13121, 1991. Novelli, P.C., P.M. Lang, K.A. Masarie, D.F. Hurst, R. Myers and J.W. Elkins, Molecular hydrogen in the troposphere: Global distribution and budget, J. Geophys. Res., 104, 3 0427-30444, 1999. Steele, L. P., R. L, Langenfelds, M. P. Lucarelli, P. J. Fraser, L. N. Cooper, D.A. Spencer, S. Chea and K. Broadhurst, Atmospheric methane, carbon dioxide, carbon monoxide, hydrogen and nitrous oxide from Cape Grim flask samples analysed by gas chromatography, in Baseline Atmospheric Program (Australia), 1994-95, edited by R. J. Francey, A. L. Dick and N. Derek, pp. 107-110, Bureau of Meteorology and CSIRO Division of Atmospheric Research, Melbourne, Australia, 1996. Zhao, C., P. Tans and K. Thoning, A high precision manometric system for absolute calibrations of CO2 in air, J. Geophys. Res. 102, 5885-5894, 1997. d13C and d18O ISOTOPE CALIBRATION (C.E. Allison and .R J. Francey) The CO2 d13C and d18O data are reported on the international VPDB-CO2 scale. Samples are measured using a dual-inlet Finnigan MAT252 mass spectrometer with an MT Box-C cryogenic pre-treatment attachment for the extraction of CO2 from air samples (normally dried at collection). First, d45 and d46 values of sample CO2 are obtained with respect to a pure reference CO2. The reference CO2 is one of 6 sub samples of an ultra-high purity high-pressure cylinder of CO2 (HC453) maintained in large-volume glass containers. HC453 has been the sole source of reference CO2 for use at CSIRO since 1977 and HC453 sub samples were measured against NBS-19 in the 1980's resulting in an assignment of VPDB-CO2 values of d13C = -6.396%, and d18O = -13.176% [Francey and Goodman, 1987]. The link to VPDB-CO2 has been maintained by comparisons between these sub-samples. Corrections to convert d45 and d46 values to preliminary d13C and d18O are applied using methods described by Allison et al. [1995]. This includes correction for the presence of nitrous oxide, co-trapped with the CO2, using measured concentrations of N2O and CO2 in each sample [Francey et al., 1996]. Final d13C and d18O values on the VPDB-CO2 scale are obtained after a correction based on comparison of measured and assigned values in air standards (high-pressure cylinders of air) that are processed every 4 samples. The initial assignment of VPDB-CO2 isotopic values to air standards was referred to as CG92 [Allison and Francey, 1999]. A revised assignment, CG99, was developed during 1999-2000, which takes into account recently identified and independently quantified systematic biases [e.g. Francey and Allison, 1994; Meijer et al., 2000]. The CG99 assignment is used here. CSIRO expressions of the VPDB-CO2 scale are monitored using a number of high-purity CO2 standards (GS-19, GS-20, OZTECH-3, OZTECH-30, OZTECH-40) and a number of surveillance standards (high-pressure cylinders of air). The very small sample requirements mean that all high-pressure cylinder air standards used since 1991 remain in the surveillance suite. (Note: Measurements made on all surveillance gases are used solely for diagnostic, not adjustment, purposes). CSIRO monitors its expression of the VPDB-CO2 scale relative to those of other laboratories, using a range of samples that includes the NIST high-purity CO2 SRMs [Verkouteren, 1999], the IAEA CLASSIC cylinders [2 four-laboratory circulations of 5 cylinders, Allison et al., in press], flask air sharing comparisons with several laboratories (for example, ~six flasks of air per month with NOAA/CMDL since 1992) and through participation in other comparison exercises. Isotope References Allison, C.E. and Francey, R.J. (1999). d13C of atmospheric CO2 at Cape Grim: The in situ record, the flask record, air standards and the CG92 calibration scale. Baseline A tmospheric Program (Australia) 1996, edited by J.L. Gras, N. Derek, N.W. Tindale and A.L. Dick, Bureau of Meteorology and CSIRO Atmospheric Research, Melbourne, Australia, pp. 45-56. Allison, C.E. and Francey, R.J. (in preparation), Calibration of stable isotope measurements of atmospheric CO2 at CSIRO Atmospheric Research. Allison, C.E., Francey, R.J. and Meijer, H.A.J. (1995). Recommendations for the reporting of stable isotope measurements of carbon and oxygen in CO2 gas. Reference and Intercomparison Materials for Stable Isotopes of Light Elements, IAEA-TEDOC-825, edited by K. Rozanski, Vienna, pp. 155-162. Allison, C. E., Francey, R. J., and Steele, L. P. (in press). The International Atomic Energy Agency Circulation of Laboratory Air Standards for Stable Isotope Comparisons: Aims, preparation and preliminary results. IAEA-TECDOC-xxx, Edited by M. Groening and H.A.J. Meijer, Vienna. Francey, R.J. and H.S. Goodman (1988). The DAR stable isotope reference scale for CO2. Baseline Atmospheric Program (Australia) 1986. (Eds. B.W. Forgan and P.J. Fraser). Department of Science/Bureau of Meteorology with CSIRO/Division of Atmospheric Research, Australia. p 40-46. Francey, R. J., and Allison, C. E. (1994). The trend in atmospheric d13CO2 over the last decade. In: Isotope variations of carbon dioxide and other trace gases in the atmosphere: final research coordination meeting, coordinated research programme: final report, Vienna, Austria, K. Rozanski (editor). [Vienna]: International Atomic Energy Agency. 7 p. Francey, Roger J., Paul (L.P.) Steele, Ray L. Langenfelds, Marco Lucarelli, Colin E. Allison, David J. Beardsmore, Scott A. Coram, Nada Derek, Fred de Silva, David M. Etheridge, Paul J. Fraser, Reg J. Henry, Brian Turner and Emily D. Welch (1996). Global Atmospheric Sampling Laboratory (GASLAB): supporting and extending the Cape Grim trace gas programs, Baseline Atmospheric Program (Australia) 1993. (Eds. R.J. Francey, A.L. Dick and N. Derek) Bureau of Meteorology and CSIRO Division of Atmospheric Research, Melbourne, pp 8-29. Meijer, H.A.J., Neubert, R.E.M., and Visser, G.H. (2000), Cross contamination in dual inlet isotope ratio mass spectrometers. International Journal Of Mass Spectrometry 198(1-2), 45-61. Verkouteren, R.M. (1999). Preparation, Characterization, and Value Assignment of Carbon Dioxide Isotopic Reference Materials: RMs 8562, 8563, and 8564. Anal. Chem. 71, 4740-4746. Table 1. Measurement uncertainties and related information applicable to CSIRO flask sampling data. CO2 d13C d18O CH4 CO H2 N2O Raw measurement +/-0.09 +/-0.01% +/-0.02% +/-2.3 +/-0.6 +/-1.5 +/-0.3 precision (1sigma)(a ppm ppb ppb ppb ppb Long-term scale +/-0.007 +/-0.03 +/-0.2 +/-0.3m +/-0.02 stability (b ppm yr-1 ppb yr-1 ppb yr-1 ppb yr-1 ppb yr-1 Alignment of 0.99986 1.00021 ? ? ? CSIRO's internal +/-0.00012 +/-0.00010 scale with established, independent scales (c Random uncertainty +/-0.13 +/-0.02% +/-0.1% +/-1.7 +/-0.7 +/-1.5 +/-0.3 on individual flask ppm ppb ppb ppb ppb samples (1sigma)(d Uncertainty relative +/-0.1 +/-0.4 +/-0.4 +/-1 +/-0.1 to other CSIRO ppm ppb ppb ppb ppb network sites (e Uncertainty due to +/-0.2% +/-0.2% +/-1% +/-2% +/-1% non-linearity of instrument response, expressed as a percentage of mole fraction (or delta) difference (f High-precision 290-380 d45: d46: 300-1850+ 20-400 430-1000 260-340 calibration ppm +/-2% +/-2% ppb ppb ppb ppb range (g a) Based on the long-term, mean standard deviation of repeat aliquots from high-pressure cylinder working standards. Of the listed species, CO shows greatest variation of precision over the range of mole fraction measured in the background atmosphere. The value shown here relates to a CO mole fraction of 100 ppb. b) Based on drift rates implied by long-term standards in high-pressure cylinders, the degree of relative stability among many such standards, and for CO2, also from results of intercomparisons with NOAA/CMDL. c) The CO2 scale factor is with respect to the manometrically-defined WMO CO2 mole fraction scale. The CH4 scale factor is with respect to the CH4 scale maintained by NOAA/CMDL. Scale factors represent CSIRO/NOAA values of exchanged cylinder air. Alignment of CO and H2 scales cannot be easily quantified due to time and concentration-dependent differences between CSIRO and NOAA/CMDL. Preliminary estimates of the N2O scale factor are given in the text but are not yet reliably quantified and may not be consistent for NOAA's HATS and CCG groups. d) Based on raw instrumental precision, results from test flasks showing noise associated with flask sample analysis and dependence on storage time, and flask pair differences from Cape Grim which also include a contribution from noise associated with flask sampling procedures. The values shown here relate to use of CSIRO's glass, 0.5 litre flasks with dual PFA O-ring valves and are representative of a typical network site, although values for specific sites differ according to mean storage times. Similar values apply to other flask types except where storage-related drifts are significantly higher (eg. CO in glass flasks with Viton O-rings and 1.6 litre, stainless steel "Sirocans"). e) Based on uncertainty due to possible systematic error in allowance for storage drift, gauged from laboratory tests and from overlapping 0.5 and 5.0 litre glass flask data at other CSIRO sampling sites (South Pole and Macquarie Island). The values shown here relate to use of CSIRO's glass, 0.5 litre flasks with dual PFA O-ring valves and are representative of a typical network site, although values for specific sites differ according to mean storage times. Similar values apply to other flask types except where storage-related drifts are significantly higher (eg. CO in glass flasks with Viton O-rings and 1.6 litre, stainless steel "Sirocans"). f) Based on recognised uncertainty in assignments to primary standards (specifically in the accuracy of relative mole fraction or delta), variability of measurements from these standards (after correction for non-linearity, where applicable) as a function of mole fraction (or delta) difference from the working standard, and for CH4 also from analysis of 6 standards with gravimetrically-derived assignments provided by Tohoku University. g) The range for which instrument response is well constrained and routinely monitored with calibration standards. Higher uncertainties apply to values outside of this range.