Carbon Measurements

The samples for TCO₂ were taken in 500-mL borosilicate glass bottles in accordance with the procedure specified in Handbook of Methods for the Analysis of the Various Parameters of the Carbon Dioxide System in Sea Water (DOE 1994), an earlier version of which was available at the time in manuscript version to the DOE Science Team. The samples were poisoned with mercuric chloride to minimize biological activity prior to analysis.

Two duplicate samples were taken and analyzed for each profile: one in surface water (near the top of the cast) and one in deep water (near the bottom of the cast). These are used to assist in the assessment of the measurement quality.

Analysis Technique

The samples were analyzed using a Single Operator Multiparameter Metabolic Analyzer (SOMMA) developed by K. Johnson (Johnson et al. 1985; 1987). The procedure using this specific instrument is described in detail in the SOMMA operating manual (Johnson 1991 - unpublished manuscript), and a description of the procedure is available in the DOE handbook (DOE 1994).

The principle behind this analysis is as follows: A known amount of seawater is dispensed into a stripping chamber where it is acidified and purged with an inert gas. The presence of solid carbonates, such as CaCO3, thus constitutes an interference in the method. The amount of CO₂ in the resulting gas stream is determined by absorbing the CO₂ in an absorbent containing ethanolamine and titrating coulometrically the hydroxyethylcarbamic acid that is formed. The pH of the solution is monitored by measuring the transmittance of a thymolphthalein indicator at approximately 610 nm. Hydroxide ions are generated by the coulometer circuitry so as to maintain the transmittance of the solution at a constant value. The relevant chemical reactions occurring in the solution are:

CO₂ + HO(CH₂)₂NH₂ --> HO(CH₂)₂NHCOO^- + H⁺

and

H⁺ + OH^- --> H₂O.

The hydroxide ions used are generated at the cathode by electrolyzing water:

H₂O + e^- --> ˝H₂(g) + OH^- ,

while silver is dissolved at the anode:

Ag(s) --> Ag⁺ + e^- .

The overall efficiency of the coulometric procedure is calibrated using known amounts of CO₂ gas, either from gas loops or from seawater-based reference materials.

Order of Analyses

The samples were analyzed in the order surface-to-deep. This order allowed the cooler deep samples to come to room temperature before they were analyzed. However, this means that it is not possible to ascertain from the analytical measurements alone if there is a systematic variation in the calibration with the life of the coulometric cell (see below).

Calibration of the Analyses

The calibration of the analyses reported here was problematic. The original plan was to use gas loops to calibrate the coulometer system and to check the performance of the analyses using certified reference materials (CRM Batch 13, certified TCO₂ value 2015.13 ľmol/kg). Unfortunately, a post-cruise examination of the results showed that the calibration factor calculated for gas loops was unexpectedly variable; an examination of the calibration factor that would have been calculated from the analyses of the CRMs also showed similar variability (equivalent to a standard deviation of measurement of 2.4 ľmol/kg).

A more detailed examination showed that the variability was restricted to those measurements that had been made in the early stages of a cell's lifetime; measurements on gas loops (Fig. 3) or on CRMs (Fig. 4) made later in the cell's lifetime were much more stable as well as being lower (counts/ľmol) than the initial measurements.

The reason for this variability appears to be that the cell was not adequately conditioned prior to being calibrated and used (Ken Johnson, BNL, personal communication). Consequently, measurements made early in the cell lifetime are suspect. These include all of the initial gas loop calibrations as well as the initial measurement of the reference material. The early measurements that were made on water from the upper ocean may also be somewhat degraded (see below).

The calibration approach used to calculate the results presented here was as follows:

The calibration of an individual coulometer was assumed to remain stable from day to day throughout its period of use. This assumption reflects the experience of most investigators (Dickson 1992) and is also borne out by the measurements from this cruise made later in the cell life (see Fig. 3 and Fig. 4). Note that a single coulometer unit was used throughout Leg 1 and for part of Leg 2; it was exchanged during Leg 2 on October 7, 1992, prior to measurement of samples from station 65.
Thus the measurements on reference materials were divided into two groups: one prior to station 65, the other from station 65 to the end of the cruise, and a mean calibration factor was calculated separately for each group of analyses (based on the measurements made on reference materials later in the cell lifetime).
This universal (coulometer dependent) calibration factor (i.e., based on the CRMs) was used to calibrate the measurements made on individual sea water samples.

Measurement Data Quality

Because of the difficulty in assigning a meaningful calibration to the analyses of total dissolved inorganic carbon made on this cruise, it is difficult to assess the data quality of the measurements presented here. Although it is apparent that analyses made later in the coulometric cell's lifetime are less variable, it is less clear when the measuring system settles down. Thus the measurements that are made early in the cell lifetime are also necessarily suspect (this is discussed in more detail below).

One indication of the potential accuracy of the measurement system is the degree of agreement between the calibration factors based on gas loops and those based on CRMs. The average difference is of the order of 0.1% (Leg 1: 0.14%, Leg 2: 0.06%), thus indicating that the gas loops had the potential of providing an accurate calibration if the cell had been adequately conditioned.

The precision of measurement is harder to assess. Duplicate samples were taken at each full station. These were typically a surface sample (in the top 10 m) and a deep sample (usually from one of the two deepest Niskin bottles). The duplicates were analyzed with the surface pair being analyzed at the beginning of a run and the deep pair being split between the beginning and end of a run.

The standard deviation of a single measurement calculated from these duplicates was 1.3 ľmol/kg for the surface samples (analyzed together); and 2.0 ľmol/kg for the deep samples (analyzed at the start and end of a run).

However, the standard deviation figures are somewhat misleading. The mean difference for the surface samples (first and second) is 0.4 ľmol/kg; that for the deep samples is 1.2 ľmol/kg. This suggests that even during the measurement of these duplicates the calibration of the cell is changing in the direction shown in Fig. 3 and Fig. 4. Hence, the measurements on the samples done in the first part of a run, those from the upper ocean, may, on occasion, be biased high by the use of a calibration factor more appropriate to the later measurements. An examination of the data on duplicates indicates that the extent of this bias is unlikely to exceed 4 ľmol/kg and may on many occasions be less than that (see Shore-Based Replicate Measurements Section for further discussion of bias). The measurements on the later (deep) samples would be expected to have a precision similar to that found for the later CRMs: a standard deviation of 1.1 ľmol/kg (i.e., a similar magnitude to that found for those duplicate measurements that were run side-by-side at the beginning of the run).

akozyr 11/2000