CO2 Calibration Analytical Method

Updated February 2021

Historically, NOAA GML transferred the WMO CO2 mole fraction scale by non-dispersive infrared absorption (NDIR) spectroscopy. In 2016 we began using a new CO2 analysis system based on three laser spectroscopic techniques; cavity ring-down spectroscopy (CRDS), off-axis integrated cavity output spectroscopy (off-axis ICOS), and quantum cascade-tunable infrared laser differential absorption spectroscopy (QC-TILDAS). The new calibration system makes use of the ability of these analyzers to measure individual CO2 isotopologues to fully account for isotopic differences among members of the primary standard set and between the primary standards and subsequent levels of standards in the calibration hierarchy.

A full description of the analytical system and the isotope accounting method is presented in Tans et al. (2017) so only a brief description is given here. The overall calibration strategy is to decompose the total CO2 mole fraction assigned to the primary standards by manometric measurement (Hall et al., 2020; Zhao et al., 1997; and Zhao et al., 2006) into individual isotopologue mole fractions based on their assigned δ13C and δ18O values and use these to calibrate the analyzer responses for each individual isotopologue. Assigned δ13C and δ18O values are from analysis by the Stable Isotope Laboratory at the Institute of Arctic and Alpine Research, University of Colorado, Boulder (INSTAAR SIL)on the JRAS-06 realization of the VPDB-CO2 scales for δ13C and δ18O of CO2. The isotopologue mole fractions of the standards are used to simultaneously calibrate the CRDS instrument for 16O12C16O and either the off-axis ICOS or the QC-TILDAS for the 16O13C16O and 18O12C16O isotopologues. Secondary standards are measured relative to these isotopologue specific calibration curves to determine isotopologue mole fractions of the secondary standards. The three isotopologue mole fractions are then converted back into total CO2, δ13C and δ18O values accounting for the unmeasured rare isotopologues. A similar method is used when tertiary standards are calibrated against secondary standards. Several publications have described this technique (Wehr et al., 2013; Flores et al., 2017; Tans et al., 2017; and Griffith, 2018) All CO2 calibration data in the database and reported to the user are total CO2, not the isotopologue specific mole fractions.

The new calibration system offers a significant advantage over the NDIR because it is calibrated over the whole range of the scale and the response to the major isotopologue (16O12C16O) is linear within uncertainty. The NDIR system used groups of 3-4 bracketing primary standards to calibrate secondary standards and the response is not linear. This made the NDIR system susceptible to mole fraction dependent biases as discussed below.

The estimated reproducibility of total CO2 measurements on the new system is ± 0.01 ppm (1 sigma) based on approximately 5 years of target tank calibrations. The δ13C and δ18O measurements agree well with measurements by INSTAAR SIL of flasks filled from cylinders (average NOAA – SIL is 0.0 ± 0.2‰ for δ13C and 0.0 ± 0.2‰ for δ18O (each at 1 sigma)). The isotopic measurements are not to be used as a substitute to having cylinders directly measured by IRMS when isotopic standards are required. They are designed to be used only for making isotopic corrections to measurements of atmospheric CO2 on instruments that are sensitive to isotopic differences between standards and samples.

The new calibration system ran in parallel with the NDIR system from April 2016 to August 2017. This comparison shows good agreement between the two systems on the current WMO CO2 X2019 scale. On the now obsolete WMO CO2 X2007 scale there was a significant mole fraction dependent bias between the two systems (Figure 1, top panel). These offsets can be traced to the effects of calibrating the NDIR system with subsets of the primary standards coupled with mistakes in the implementation of the X2007 scale. Using subsets in this way makes the results from the NDIR system sensitive to incorrectly assigned values of individual primary standards. The mole fraction dependent bias between the laser spectroscopic and NDIR calibrations was resolved with the release of the WMO CO2 X2019 scale (Figure 1, bottom panel). The X2019 scale revision was retroactively propagated to all past calibrations of tertiary standards.

Figure 1
Comparison results (NDIR minus Laser Spectroscopy) from running the two calibration systems in parallel from April 2016 through Aug 2017 as a function of mole fraction. Top panel: results on the now obsolete WMO CO2 X2007 scale; bottom panel: results on the current WMO CO2 X2019 scale. Open symbols are cylinders with δ13C < -20‰. Dashed lines are the reproducibility of the NDIR system (±0.03 1 sigma). The differences on X2007 show a strong mole fraction dependency that can be traced to the NDIR sensitivity to mis-assigned primary standards. These biases are resolved with the X2019 scale revision. The isotopically depleted tanks show a slightly larger difference indicating the current NDIR analyzer is biased by the isotopic composition of the standards relative to the samples. This should be taken into account by the user when evaluating drift in isotopically depleted cylinders and comparing historical NDIR measurements with the new calibration system measurements.

The results from the new calibration system also show the NDIR system to be sensitive to differences in the isotopic composition of the tertiary cylinders calibrated. Figure 1 (open symbols on both scales) shows a larger offset between the two systems for iostopically depleted cylinders (δ13C < -20‰). Users are cautioned when comparing historical NDIR measurements of isotopically depleted tanks with measurements on the new calibration system. The new calibration system is much better at eliminating biases due to isotopic differences, but drift determinations may be compromised by the sensitivity of the NDIR to the isotopic differences.

References