The data included in the Atlantic synthesis data set come from diverse sources, and the intercomparisons are complicated by the dynamic nature of the Atlantic Ocean; shifts in bottom-water characteristics have been documented. Therefore, we are being conservative and only recommend adjustments in the synthesis data set when there are clear biases of greater than 4 µmol/kg for DIC and 6 µmol/kg for TAlk. The recommendations are solely for the purpose of this analysis, to produce an internally consistent basin-wide data set of DIC and TAlk that can be used to quantify large-scale characteristics. Data sets that were deemed of insufficient quality for this analysis can have significant intrinsic merits for other applications.
Overall, the data appear to be of high quality, and only a few cruises showed consistent biases for each method. Variability within profiles and between stations often was of the same magnitude as the cruise-to-cruise differences (see Table 4). If adjustments had to be applied, the assessments were somewhat subjective because the statistical evidence of the different methods of comparison had to be weighed in an arbitrary manner. Our approach, as outlined in Sect. 4.1 and Sect. 4.2, consisted of first looking at the crossover results (see Table 4 and ) for offsets between cruises. From this comparison it was determined whether an offset appeared consistent for a particular cruise at all crossovers with other cruises. Table 5, which has the crossovers ordered per cruise, is particularly useful for this check. DIC anomalies of greater than 4 µmol/kg and TAlk anomalies greater than 6 µmol/kg prompted closer scrutiny. If cruises showed a consistent bias, it was determined whether a similar bias was apparent in the regional MLR, internal consistency, and overlapping cruise regressions. Cruise-by-cruise narratives for crossovers with offsets of DIC greater than 4 µmol/kg (Sect. 4.1) and TAlk greater than 6 µmol/kg (Sect. 4.2) provide a rationale for the proposed adjustments.
In cases of overlapping cruises when only one of the cruises is used, the repeat cruise is recommended. Thus A05R, A01WR, A12BR, and A16N in the equatorial overlap section with A16S are suggested. The repeat cruises are also often closer in time to the other cruises in the synthesis, thereby decreasing the possibility of a temporal bias. However, these overlapping cruises are invaluable for many applications, and all of the individual cruises are retained in the data holdings. Also, the cruise tracks of overlapping cruises sometimes diverge for part of the tracks, and these segments should be included in basinwide analyses.