Total Alkalinity Measurements
Total alkalinity was measured on 18,928 samples using two closed-cell automated potentiometric titration systems (hereafter designated as MATS) developed at the University of Miami. The MATS are described by Millero et al. (1993) and by Millero et al. (1998). The latter reprinted in Appendix C of this document, completely describes the Indian Ocean Survey TALK measurements and results. Briefly, the MATS consisted of three parts: a water-jacketed, fixed- volume (about 200 mL determined to ± 0.05 mL) closed Plexiglass sample cell, a Metrohm model 665 Dosimat titrator, and a pH meter (Orion, Model 720A), the last two controlled by a PC. The titration cell was similar to those used by Bradshaw and Brewer (1988), but had a greater volume to improve the precision of the measurements. The cell was equipped with flush-mounted fill and drain valves to increase the reproducibility of the cell volume. The cell, titrant burette, and sample container were held at a temperature of 25 ± 0.01°C using a constant temperature bath (e.g., Neslab, Model RTE 221).
A Lab Windows C program was used to run the titrators, record the volume of titrant added, and record the measured electromagnetic fields (emf) of the electrodes through RS232 serial interfaces. Two electrodes were used in each cell: a ROSS glass pH electrode (Orion, Model 810100) and a double-junction Ag/AgCl reference electrode (Orion, Model 900200). The specific electrodes used during the Indian Ocean survey were selected after careful screening for non- Nernstian behavior. Only those electrodes which gave TCO2 results in good agreement with TCO2, as determined coulometrically, were used.
Seawater samples were titrated by adding increments of HCl until the carbonic acid endpoint of the titration was exceeded. During a titration, the emf readings were monitored until they were stable (± 0.09 mV). Sufficient volume of acid was added to increase the emf by preassigned increment (~13 mV) in order to give an even distribution of data points over the course of a full titration, which consists of 25 data points. A single titration takes about 20 min. A FORTRAN computer program based on those developed by Dickson (1981) and by Johansson and Wedborg (1982) was used to calculate the carbonate parameters. The pH and pK of the acids used in the program are on the seawater scale, and the dissociation constants for carbonic acid were taken from Dickson and Millero (1987). For further details see Appendix C and DOE (1994).
The titrant (acid) used throughout the cruises was prepared prior to the cruise, standardized, and stored in 500-mL borosilicate glass bottles for use in the field. The 0.25-M HCl acid solution was prepared by dilution of 1-M HCl in 0.45-M NaCl to yield a solution with total ionic strength similar to that of seawater of salinity 35.0 (I = 0.7 M). The acid was standardized by coulometry (Taylor and Smith 1959; Marinenko and Taylor 1968), and was also checked by independent titration in A. Dickson's laboratory at SIO. The independent determinations agreed to ± 0.0001 M, which corresponds to an uncertainty in TALK of ~ 1 µmol/kg. The Dosimat titrator burettes were calibrated with Milli-Q water at 25°C to ± 0.0005 mL.
While CRM samples were available to the TCO2 analysts from the beginning of the measurement program in 1990, the Indian Ocean cruises were the first to have a certified alkalinity standard as well. Hence, the accuracy of the method was checked in the laboratory by analyzing CRM samples from batches 23, 24, 26, 27, 29, and 30 and comparing the analyzed values with the certified TALK determined by A. Dickson at SIO (in the same manner as for TCO2). These results are summarized in Table 11 (see also Appendix C). The mean difference between the MATS measurements in the laboratory and the certified TALK values was -0.8 µmol/kg for CRM samples with a concentration range approximately one-half as large as the range of a typical seawater profile. The excellent agreement indicated that the CRM concept for alkalinity was valid and that the methodology for TALK was ready for the Indian Ocean survey. The results for the at-sea measurements of the CRM samples have been extracted from Table 2 of Appendix C, summarized, and are given in Table 12.
The analytical differences are for the most part within the precision of the measurements (~ 2-5 µmol/kg) except for the I7N Section. The larger at-sea differences were attributed to operator error or procedures and to uncertainties in the volume of cells, especially after repairs due to leakage, breakage, or repositioning the electrodes after changing the inner filling solutions. Variations between different MATS systems used on a single cruise were corrected using the adjustments required to reproduce the values assigned for the CRM (see Table 11). The at-sea sample titrations were corrected using the results of the at-sea CRM analyses. For TALK, the calibration factor (CF) used to correct the at sea measurements was
CF = TALK (meas., CRM) - CRM (certified value),
and the corrected TALK (TALKc) was
(TALKc) = TALK (meas., Spl) x [ CRM / (CRM + CF)],
where CRM was the certified TALK and Spl was the measured sample TALK.
The overall precision of TALK determinations during the Indian Ocean survey was ± 4.2 µmol/kg. The precision of the potentiometric pH and TCO2 measurements are given in Table 3 of Appendix C.
TALK was also checked at the crossover locations of two cruises in the same way as TCO2. The agreement between the corrected TALK measurements made at the crossover locations (± 100 km) on different cruises was examined by assuming that the temporal and spatial variations of the deep-ocean TALK were small relative to measurement accuracy and precision. Hence, deep ocean waters should have the same TALK at different times in the absence of internal vertical motion, and because deep ocean motion probably occurs along constant-density surfaces (isopycnals), the comparisons of TALK measurements were made with reference to density and not depth.
AppendixesC and D give a description of the statistical procedures used to make the crossover comparisons. For water samples collected below 2500 m, a smooth curve was fit through the TALK data as a function of the density anomaly referenced to 3000 dbar (sigma3) using Cleveland's LOESS smoother (Cleveland and Devlin 1988). A separate fit was performed on the data collected at each of the two intersecting crossover points, with the same tension parameter being used for all of the crossovers so that the smoothing function was consistently applied. The difference between the two smoothed curves was evaluated at 50 evenly-spaced points covering the density range where the two data sets overlapped. Mean and standard deviations for the differences at the 50 points were calculated for each crossover point. For TALK, differences never exceeded 6 µmol/kg, and the overall mean and standard deviation of the differences was 2.1 ± 2.1 µmol/kg. The latter were consistent with the overall precision of the CRM analyses (± 4 µmol/kg).
Table 13 is a final summation of the inorganic carbon analytical work completed during the Indian Ocean CO2 survey from 1994 to 1996.
