Overview

The overall picture of agreement is characterized by a very good agreement of profiles "C," "D," and "E" essentially throughout the cruise. While also in good agreement for most of the time, profile "B" shows a 2-day period with a marked positive offset. Two profiles show a more or less constant sign of deviation, which is positive in the case of "F" and negative in the case of "A." The reason for this could not be identified easily. However, for system "A" we know of an instance of severe damage in the NDIR instrument toward the end of the exercise, which may well have started biasing the measurements in an early stage of the exercise. With respect to system "F," which is of a principally different design (see Participating Underway fCO₂ Systems, Table 2), the question of whether the different principle of measurement could be the reason for the rather large observed offset should be addressed carefully. Finally, system "G" shows anything from large negative offsets over periods of good agreement to rather strong positive offsets. These problems were also apparent in atmospheric xCO₂ measurements and checks of the CO₂ calibration performance probably because of an improper calibration technique. The calibration of the system appears to lack-at least during this exercise-the necessary reproducibility (i.e., it may be good in one case and bad in another one). This obvious problem of system "G" also needs careful checks.

In addition to the daily figures (Figs. 12-20) representing the full data set, we present three figures (Figs. 21, 22, 23) with enlarged views of shorter periods. These were chosen because they reveal more detail than is available in the daily figures. Furthermore they also cover the whole range of situations, from smooth to highly variable.

Fig. 21 shows a 3-hour period of measurements on June 9 that was characterized by very low variability in the surface seawater fCO₂ ( Fig. 13) as well as temperature and salinity ( Fig. 7). The total change in fCO₂ values during this period of time is about 6 µatm. This is uniformly seen in all profiles, which are almost perfectly parallel. Profiles "B," "D," "E," and "F" agree to within 1 µatm, while profiles "A" and "G" are characterized by a negative offset of about 8 µatm and 5 µatm, respectively. The scatter is smallest in profile "D" (averaging interval 5 min; large time constant, as shown) and highest in profile "A" (averaging interval 3 min; short time constant). The comparatively small scatter in profile "B" with 1-min averaging intervals shows that much of the scatter in profile "A" (also seen in the atmospheric xCO₂ data of "A") is not real and may thus indicate again the existence of a technical problem.

In contrast, Fig. 22 shows the much more variable situation of a 3.5-hour period of measurements during June 12. The total range of fCO₂ values covered during this period is about 35 µatm with gradients of up to 3 µatm/min^-1. Again, the agreement is good in profiles "B," "D," and "E." Profiles "F" and "G" show positive offsets, while profile "A" has a negative offset of a few µatm.

Individual time constants involved in the equilibration process going on in every system can be estimated rather precisely with step experiments carried out under well-defined conditions in a shore-based laboratory (Copin-Montegut 1988; Koertzinger et al. 1996b). This is definitely not the case in the present intercomparison exercise. We therefore do not try make any estimates of individual time constants. Nevertheless, in addition to the examination of offsets we do try to gain insight into the apparent time constants (i.e., we want to see whether there is any indication of differences in kinetic aspects of the equilibration processes). Because most time constants are on the order of a few minutes, this analysis is only feasible where fCO₂ was measured at rather short intervals of < or = 5 min (only profiles "A," "B," "C," and "D"), but even in these cases this is not a sound approach.

We have marked approximate relative minima and maxima observed in the enlarged periods shown in Fig. 22 and Fig. 23. The pattern of vertical lines observed in these groups is highly consistent: Extrema always occur first and simultaneously in profiles "A" and "B," while profiles "C" and "D" lag behind by 5 to 8 min and 2 to 5 min, respectively. The range is mainly a consequence of the different averaging intervals. These time lags cannot be attributed to a temporal mismatch of the profiles. They are, however, clearly related to differences in the general design of these systems. Systems "A" and "B" are similar with respect to volumes and flow rates of water and air. For example, the total air volume of the equilibrator is exchanged every 2.5 min and 0.5 min, respectively, hence the similar equilibration times. In system "C" the large volume of air in the equilibrator is only exchanged every 20 min, which explains the more sluggish response seen in Fig. 23. System "D" is of the thin film type (i.e., unlike in the other system no turbulent mixing occurs in the equilibrator). It is known that this equilibration concept is characterized by somewhat larger time constants.

We would like to point out that different time constants are no quality criterion per se but rather must be seen in the context of the application. A detailed process study would certainly require high spatial and temporal resolution and hence an fCO₂ system with rather small time constants to resolve small-scale features. This is not equally the case in a basin-wide assessment of the fCO₂ in surface seawater, where the large-scale averaging would eliminate the effect of different time constants. The main point here is simply to show that these different characteristics are clearly reflected in the fCO₂ data set.