Shore-Based Replicate Measurements
During the expedition, 322 duplicate samples were collected and returned to SIO for shore-based measurements in the laboratory of C. D. Keeling. A total of 309 TCO2 and 314 TALK measurements were performed on these samples. The 13C/12C isotopic ratio of the carbon comprising the TCO2 was also measured (but not reported in this numeric data package). Comparisons between the shore-based measurements of TCO2 and TALK and those made at sea on water from the same Niskin bottles provide further quality control information on the carbon data set for WOCE Section P13.
Shore-based measurements of TCO2 were made by vacuum extraction/manometry using the procedures established for the DOE/WOCE ocean CO2 program (Guenther et al. 1994b). Results are tabulated in Appendix B. This table also lists the corresponding SOMMA TCO2 values and the differences between the shipboard and shore-based values. Shipboard data are identified as "SIO" and shore-based as "S.I.O." The repeatability of the shore-based results themselves can be estimated from the agreement of the duplicate samples measured (DOE 1994). The sample standard deviation, si, of an individual shore-based result represents the short-term imprecision of the laboratory analysis, together with imprecision introduced by sampling and storage. The si calculated for the set of 140 pairs of data was 0.95 µmol/kg. Twelve pairs were rejected from this calculation, as shown by the flags in Appendix B. This "replicate imprecision" is approximately average for DOE/WOCE program cruises.
Of the 140 ship - shore differences corresponding to the "good" pairs of shore-based data, two were rejected for being more than 3si from the average (-17.17 and 20.21 µmol/kg). The average difference for the remaining 138 comparisons was 1.37 µmol/kg, with the shore-based being higher, and the si of an individual difference was 3.11 µmol/kg. The average difference was typical for DOE/WOCE cruises during the 1991-1994 period, but the si is rather large. A reason for the increased scatter is the presence of a depth-dependent bias between the ship - shore differences. The usual sampling depths for shore-based replicate samples on DOE/WOCE cruises were surface and deep (nominally 3000 m). Differences for WOCE Section P13 are plotted in Figure 6 for this subset of comparisons. "Surface" samples are the shallowest on a station, ranging from 10 to 75 m in depth, and "deep" samples are the deepest, ranging from 1000 to 3200 m. The average surface - deep bias for the subset of surface and deep samples in Figure 6 (18 differences between "good" replicate pairs) is 3.5 µmol/kg (si = 2.5 µmol/kg). A surface - deep bias has been evident for only a few other cruises and usually is smaller. On this cruise, shore-based replicate samples were also collected in profile from 9 to 12 Niskin bottles from the surface to nominally 3000 m on 10 stations. Ship - shore differences for the top several depths of these stations change toward the more negative deep differences. From 400 m down, the differences are relatively constant.
The surface-deep bias results agree fairly well with measurements made at sea. Shipboard measurements for surface comparisons between shore-based and shipboard measurements were made early in the measurement runs, while those for deep comparisons were made late in the runs. Use of the lower calibration factors measured late in the runs resulted in a high bias for measurements made early in the runs. On average, CRM measurements made early in the runs were 2.6 µmol/kg higher than those made late in the runs. Also, deep samples measured early in the runs on Leg 1 on average were 2.3 µmol/kg higher than their duplicates measured late in the runs. However, this pattern was far less apparent for Leg 2.
Shore-based measurements of TALK were made by essentially the same potentiometric titration system as the measurements made at sea. The primary difference was that the aliquots for shore-based titrations more often were dispensed gravimetrically into the titration cell, instead of volumetrically. The aliquots were removed from the sample bottles after those for shore-based TCO2 had been removed. Results are tabulated in Appendix B. This table also lists the corresponding shipboard TALK values and the differences between shore-based and shipboard values. As described for the shore-based TCO2, the replicate imprecision of the shore-based TALK measurements is estimated from the agreement of the duplicate measurements. For samples with analyses from both gravimetric and volumetric systems, analyses separated by more than a week of elapsed time were rejected. For one set of titrations made within a few days on both systems, the gravimetric data were chosen over the volumetric. The si was 1.90 µmol/kg for 154 pairs of measurements, with four pairs rejected as shown by the flags in Appendix B. The apparent imprecisions of the shipboard TALK results and the shore-based results are similar, ~2 µmol/kg.
The average ship - shore difference for TALK is calculated from 147 of the total of 150 comparisons of "good" shore-based duplicates with corresponding shipboard values. Three comparisons with differences of 18.78, 15.63, and 23.01 µmol/kg (greater than 3si) were rejected. The average difference is 3.35 µmol/kg (shipboard higher). The si of an individual difference is 4.11 mol/kg. Both the average ship shore difference and its imprecision are likely to change after the anticipated adjustments to the TALK calibration scale are made, so further analysis and plotting of the data will not be presented at this time.