Comparison of Surface fCO₂ Data

As described in Sect. Checks and Calculation Routines, the following main steps in the calculation of final fCO₂ values constitute the general procedure that was applied identically to all underway fCO₂ data sets:

• Calculation of xCO₂ in dry sample air (final data product received from every group)
• Synchronization of daily CTD and equilibrator profiles based on standard deviation of the temperature offset
• Calculation of fCO₂ in equilibrator (at T eq, 100% humidity)
• Correction of fCO₂ to in situ seawater temperature based on corrected temperature readings

In order to gain better interpretability of any differences in the final fCO₂ data sets, we tried to exclude as many controllable sources of error as possible. This was accomplished by carefully addressing the following points:

• Temperature readings can be a significant source of error as shown in Sect. Check of Equilibrator Temperature Sensors However, on the basis of the checks of the temperature probes against a reference probe we were able to remove this error and assure consistent temperature measurements.
• The choice of the parameterizations for calculating the saturation water vapor pressure and for the temperature correction of fCO₂ also introduces some kind of uncertainty, which, however, in our case seems to be rather small compared with the errors of the temperature measurements. Again, the common calculation procedure (Sect. Calculation of fCO₂ Results) excludes inconsistencies based on the use of different equations.
• Finally, the common infrastructure (i.e., the seawater and calibration gas supply) assured a physically identical background for all systems.

It should be emphasized that none of these consistent conditions are usually present in typical fCO₂ measurements in the field (i.e., temperature probes are sometimes used uncalibrated or at least not calibrated to the same standard; calculation procedures vary; calibration gases are of different origin and likely quality, too; the seawater sources may be quite different or even inadequate for gas measurements etc.).

In the interpretation of the results, any differences of >2 µatm (up to >5 µatm in the highly variable regime) in the final fCO₂ data can be attributed either to differences in the equilibration process itself and/or to differences in the subsequent measurement of CO₂. A tool to separate these two possible sources of error is measurements of the atmospheric xCO₂, which were discussed in some detail in Sect. Comparison of Atmospheric xCO₂ Data. Unlike the calibration gases, atmospheric air is comparable to the seawater equilibrated air in that it has a wet sample matrix. Thus atmospheric air undergoes the same procedure of (physical or arithmetical) drying. If, for example, differences between seawater fCO₂ data from two systems were also present in the atmospheric xCO₂ data, this is indicative of problems associated with the infrared CO₂ measurement and/or the drying procedure. If, in contrast, the atmospheric xCO₂ data turned out to be identical while seawater fCO₂ was different, the source of error must be attributed to the equilibration process and/or the way of handling of the seawater equilibrated air.

Whereas these reasons add to the interpretability of the results, it should be pointed out that any observed differences cannot per se be attributed to a particular data set or system. As a superior reference method was not available, the "true" fCO₂ values are simply not known. Given the still remaining uncertainty about the valid set of dissociation constants of carbonic acid in seawater, even consistency checks based on the other three parameters of the CO₂ system in seawater (i.e., CT, AT, pH)-although to be carried out later on-will not provide an unambiguous means of finding "true" fCO₂ values.

However, we found three data sets (systems "C," "D," and "E") to be very close in seawater fCO₂ and atmospheric xCO₂ values throughout the cruise, whereas the other data sets show variable offsets to these three profiles and some of them are also associated with significantly larger scatter. Because the general design of the three systems "C," "D," and "E" is significantly different (showerhead equilibrator-thin film equilibrator, small equilibrator volume-large equilibrator volume, small flow rates-large flow rates, equilibrator vented-equilibrator not vented, wet CO₂ measurement-dry CO₂ measurement, etc.), this agreement cannot simply be attributed to an essentially identical design. This is by no means a sufficient argument to regard the three consistent fCO₂ profiles as the "truth," although we feel that this marked agreement is at least a strong indication of this. However, with the lack of a superior method, this sort of discussion is to some extent futile and cannot be solved here.

When preparing the following figures, we wanted to discuss the fCO₂ results not only as absolute numbers but also as deviations from a reference. Because a superior method was not available and the choice of a single "true" fCO₂ profile was not feasible, we decided to calculate the deviation of every single fCO₂ data point from an 11-min running mean calculated from the three most consistent profiles (labs "C," "D," and "E"). In the light of the arguments given in the foregoing discussion, this choice remains arbitrary, but it nevertheless seems to be the most reasonable choice. However, it should be kept in mind that these deviations are, of course, dependent on the choice of the reference and are therefore not independent results. We are fully aware that this is a somewhat critical step in the interpretation, which seems only justified by the better visualization of the differences and the enhanced interpretability of the data set.

A further problem associated with calculating deviations from an 11-min running mean stems from the fact that this reference represents a strongly smoothed profile, whereas the original fCO₂ data represent significantly smaller averaging intervals (minimum 1 min). Thus, all temporal variability on the minute-scale as contained in the fCO₂ data with higher temporal resolution (e.g., profiles "B" and "E") translates into a larger scatter than that of the deviations from the smoothed reference profile. This artifact has to be kept in mind, because it is of a different magnitude for the various fCO₂ data sets. This effect is more strongly obvious in the strong gradient regime (e.g., June 10). The main message of these deviation figures therefore has to be the general offset rather than the scatter of a profile.

Table 3 provides an overview of the minima, maxima, and differences of measured in situ temperature, salinity, and fCO₂ (11-min running mean from profiles "C," "D," and "E") on a daily basis. As intended with the choice of the cruise track and as already documented in the daily profiles of temperature and salinity, the encountered conditions of the surface waters along the cruise track varied between a smooth regime with low variability during the second half of the cruise and a strong gradient regime with much higher variability in the area close to the northern turning point off Newfoundland (June 9-June 12) during the first half.

Table 3. Overview of minimum, maximum, and difference of measured values of temperature T (°C), salinity S, and the fugacity of CO₂ (fCO₂, 11-min running mean from profiles "C," "D," and "E"). The strong gradient regime is shaded.