Comparison of Atmospheric xCO₂ Data

Measurements of the atmospheric xCO₂ were carried out by all underway fCO₂ systems except system "F" (see Table 2 in Sect. Participating Underway fCO₂ Systems). As will be shown, the atmospheric xCO₂ data - while not immediate focus of this exercise - may still provide additional information for identifying likely sources of error in the surface fCO₂ profiles. All xCO₂ data are given (in ppmv) for dry air and shown in Fig. 10.

Of the six data sets, four show good agreement to within ±1 ppmv throughout the exercise: Profiles "C" and "D" show virtually identical values (except for a few data points), whereas profile "E" tends to values that are lower by -0.5 to -1 ppmv. Profile "B" is characterized by a somewhat variable behavior: For most of the time, "B" is in very good agreement with "E." However, from June 10, 08:30 UTC, to June 11, 14:30 UTC, "B" shows a positive offset of about 1 ppmv from "E," hereby agreeing perfectly with "C" and "D." In contrast, "B" deviates by -0.5 to -1.0 ppmv from "E" during the period from June 14, 14:30 UTC, until June 13, 18:00 UTC, which is equivalent to an offset of -1.0 to -1.5 ppmv with respect to profiles "C" and "D." We have calculated a mean xCO₂ (air) of 366.21 ± 0.72 ppmv for the period of the exercise where data from all six systems are available from the means of profiles "B" through "E" which is shown in Fig. 10 (red line). Only results from this restricted period are used in the following comparison.

Two profiles ("A" and "G") are characterized by a much larger scatter which obscures the pattern of the atmospheric CO₂ contained in the other four profiles. This scatter is not a real property of the sampled air, as proved by profiles "B" through "E," and thus indicates an analytical problem associated with these systems. All air intakes were located on the same spot above the wheelhouse of R/V Meteor approximately 20 m above sea level, thus making differences in the properties of the sampled air very unlikely. The majority of measurements of "G" show a positive deviation of up to 8 ppmv which is consistent with the rather large positive offset of 3 to 6 ppmv determined during the checks of the CO₂ calibration performance. The mean of "G" (368.27 ppmv) is 2.06 ppmv higher than the combined mean of "B" through "E" (366.21 ppmv). With a mean value of 362.89 ppmv, the xCO₂ measurements of "A" are clearly marked by a negative offset of 3.32 ppmv with respect to the mean of "B" through "E."

Fig. 11 shows the individual mean and standard deviation of each data set as well as an overall mean calculated from the mean of profiles "B" through "E," all for the restricted period of time only. The individual standard deviations reflect the averaging interval in the case of laboratories "B" through "E," where the smaller scatter is associated with the longer averaging intervals of 4 to 5 min (laboratories "C" and "D") and the somewhat larger scatter reflects averaging intervals of 1 min (laboratories "B" and "E"). In the case of laboratories "A" and "G" the scatter is no obvious function of the averaging interval but an expression of an analytical problem.

In the comparison of surface fCO₂ data in next paragraph, these results, sometimes referred to as the general trends of agreement or disagreement between the xCO₂ (air) data sets, will largely be retained in the fCO₂ data. The combination of both results provides much of the argument for the discussion of the overall results. We will demonstrate that the three laboratories "C," "D," and "E" show the same high degree of agreement in surface fCO₂ data as they do in xCO₂, and a strong case will be made that these systems represent the "best" values of xCO₂ (air) and fCO₂.