Appendix: Critical Evaluation of the Repeat Station Data
When the ship's tracks of the present WOCE expeditions intersected previous WOCE stations, measurements were repeated at the crossover locations. The time intervals between pairs of repeat stations varied from about two months to about a year. This allowed us to evaluate reproducibility of measurements made for deep waters below about 2000 meters, where variabilities are expected to be small or negligible. In this section, the following nine properties measured at five pairs of repeat hydrographic stations are compared in deep waters below about 2000 meters or sigma-4 density exceeding 45.5: potential temperature, salinity, pCO2, alkalinity and the concentrations of oxygen, total CO2, silicate, nitrate and phosphate. The pCO2 values were measured at a constant temperature of either 4°C or 20°C, and hence only pairs of repeat station data for pCO2 obtained at the same temperature are compared. The alkalinity values were computed using the pCO2 and total CO2 concentration values according to the scheme described in Section 3d. The locations of these five pairs of stations are indicated in Fig. 1 with large filled circles and the data are presented in Figs. 6, 7, 8, 9, and 10. In each figure, the potential temperature () vs salinity relationship is shown in the top left panel, and other seven properties are plotted against the sigma-4 density (potential density computed at 4000 db). The position of the repeat stations and dates of observations are listed in Table 3.
Table 3. The positions and dates of repeat stations.
In Table 4, listed are the precision of measurements (expressed in terms of one standard deviation) estimated for a single station at sigma-4 densities greater than about 45.6. The precision for salinity measurements have been estimated assuming that temperatures are known, and the precision for all other properties have been estimated assuming that the sigma-4 density values are know. If the data for each pair of stations are indistinguishable, the standard deviation for combined stations are given. If systematic differences are observed between a pair of station data sets, the mean difference (the top station in the table minus the bottom station) is shown in the parentheses. The ± values listed in the column "MEAN" indicate the mean of precisions observed at five pairs of stations and represent the mean precision of measurements made at a single station. The mean precision of single-station measurements for the eight properties listed are considered to be at the state-of-the-art level. On the other hand, nagging systematic differences between pairs of measurements are observed more often in nitrate (5 out of 5 crossovers) and phosphate (4 out of 4 crossovers) than the others. These station-to-station differences often exceed single-station precisions by several folds. Since these systematic differences do not always correspond to those observed for the concentrations of oxygen and CO2 and salinity, and since the deviations observed for nitrate are not always to the same sign as those observed for phosphate, the observed systematic differences are most likely to be due to expedition-to-expedition calibration problems.
Several features shown in Figs. 6, 7, 8, 9, and 10 deserve additional comments. First, the nine quantities measured at each pair of stations a month to a year apart at a given location do not always agree each other. When temperature and/or salinity values, the two quantities measured most reproducibly and precisely, differ beyond the estimated errors for each station data, we are inclined to conclude that the characteristics of deep waters have changed with respect to a fixed geographical position on the earth. On the basis of the estimated errors of ±0.001°C for temperature and ±0.001 for salinity (these correspond to the size of data points shown in Figs. 6-10), it appears that three of the five pairs of repeat stations remained unchanged (Crossovers 1, 3, and 4), one changed partially (Crossover 3) and one changed (Crossover 5) clearly. The oxygen data appear to be most consistent with the temperature-salinity data. At Crossover 3 (Fig. 8), the salinity changed by about 0.005 in a temperatures range between 1.3 and 1.8°C or in a sigma-4 range between 45.8 and 45.9. The oxygen data as well as other chemical data appear to reflect this change. However, it is not clear whether the observed changes represent a real change or malfunction of the Rosette CTD samplers.
Secondly, while no systematic differences between 4 out of 5 pairs of stations have been observed, they are observed only at Crossover 2 (Fig. 7). However, it should be noted that the TCO2 values observed at the both stations at densities greater than about 45.85 are indistinguishable, whereas those obtained at TUNES 179 are considerably greater than those at JUNO 119 in waters with densities between 45.72 and 45.85. The salinity, oxygen, silicate and phosphate data appear to suggest that these waters had changed, whereas the pCO2 and phosphate data show no changes. Hence, no clear conclusion can be made. In Table 4, a mean difference in TCO2 of 4 µmol/kg is listed for a density range from 45.70 to 45.91.
Thirdly, at Crossover 5 (Fig. 10), the salinity values observed at S-4P Station 703 (Feb. 28, 1992) and JUNO Station 229 (Jan. 16, 1993) about a year later differ systematically by about 0.003, about 3 times the precision obtained at each station. The oxygen, pCO2, nitrate and phosphate data also change consistently with each other. On the other hand, TCO2 and silicate data show no change. Again, it is not clear whether the properties of deep ocean waters changed or the calibrations of instruments slipped.