During the R/V Thomas G. Thompson expedition along WOCE Section P10, 1072 TCO2 samples were collected according to methods outlined in the DOE Handbook (DOE 1994) and stored in 300-mL borosilicate glass bottles in a 20ºC water bath until the samples could be processed (maximum of 12 h). Samples were poisoned immediately after collection with 200 µL of saturated HgCl2 solution to minimize biological activity prior to analysis. Samples were analyzed using a computer controlled single-operator multiparameter metabolic analyzer (SOMMA) system and UIC model 5011 coulometer following standard methods (Johnson et al. 1985, 1987; DOE 1994).
The SOMMA temperature sensors (National Semiconductor, Santa Clara, Calif., model LM34CH) were calibrated against a certified mercury thermometer and thermistors certified to 0.01ºC (Thermometrics, Edison, N.J., part number CSP60BT103M). These sensors monitored the pipette, the gas sample loops, and the coulometer cell temperatures. The Digiquartz Transducer barometer (Paroscientific, Redmond, Wash., model 216B-101) was factory- calibrated prior to the cruise. The SOMMA sample delivery pipette volume was gravimetrically determined before the cruise to be 21.7758 ± 0.0016 mL using deionized water at a temperature of 20.32ºC. Post-cruise calibration confirmed that the pipette volume remained constant throughout the cruise. Sample weight was calculated from the pipette volume, the measured sample temperature, and the bottle salinity value measured by the WHOI CTD group on all but three samples. The bottle salinity values from station 18, Niskins 25, 26, and 27, had values significantly different (>0.1) from the CTD salinity values and from other bottle salinity values from equivalent depths of the surrounding casts. This difference was large enough to result in an error in the calculated density that was greater than the sample precision; therefore, the sample density was recalculated for those three samples using the CTD salinity. Cylinders of compressed ultrahigh-purity nitrogen and 350-ppm CO2 in air were used as the system's carrier gas and headspace gas, respectively.
Titration cells were prepared with fresh cathode and anode solutions at the beginning of each cast. A system blank was determined for each cell by keeping track of the total number of counts accumulated by the coulometer's voltage-to-frequency converter over a 10-min period. The counts used for determining the TCO2 of a sample were then determined by subtracting the blank counts (average blank value in counts per minute times the length of the titration in minutes) from the total counts registered for that titration.
Two methods were used to evaluate the calibration of the TCO2 system. The first method titrated a known volume of CO2 gas to determine a system efficiency. This method involved filling one of two different-sized gas loops with primary standard-grade CO2 gas (Scott Grade 5 CO2, 99.999% pure). Based on the loop volume, pressure, and temperature, a known amount of CO2 was introduced into the coulometer and titrated. Gas calibrations using both the large (1.5224 mL) and small (1.0586 mL) loops were run at the beginning and end of each titration cell. The second calibration method involved the titration of certified reference materials (CRMs) provided by Andrew Dickson of SIO at the beginning and end of every titration cell. The CRMs were analyzed in the same manner as a sample and the results compared to the certified value determined by vacuum extraction and manometry.
Although the gas loop and CRM calibration methods are very different (pure CO2 gas vs seawater), the results can be directly compared by examining the titration efficiency (TE; coulometer counts per mole of carbon titrated) determined for each sample. The TE for the CRM samples (TEcrm) was determined from the blank corrected coulometer counts, sample volume (vol), sample density (p) (Millero and Poisson 1981), and certified CRM value (Tcert; 2031.65 µmol/kg):
TEcrm = counts / (Tcert × vol × p) .
The TE for the gas loop calibrations was determined by dividing the blank corrected coulometer counts by the amount of CO2 (moles) introduced to the coulometer. The amount of CO2 was determined by dividing the loop volume by the molar volume of CO2 (VCO2) at the measured loop temperature (T) and pressure (P) using an iterative approach:
VCO2 = RT / × P [1 + B(T) / VCO2] ,
where B(T) is the first viral coefficient for pure CO2, and R is the gas constant.
Generally, the gas loop calibration is very reliable and accurate for SOMMA system calibrations (Johnson et al. 1987); however, a plot of the CRM and gas loop TE values shows that during the first half of the cruise, the gas loop efficiency was lower than the TEcrm values (Fig. 3). Because the loop calibration system was new and untested on this system, and there was no reason to think that the CRM values would not be stable over the length of the cruise, the TEcrm values were deemed to be more representative of the system efficiency. The gas loop TE values determined near the end of the cruise were more consistent with the TEcrm values. Despite post-cruise recalibration of the SOMMA at Brookhaven National Laboratory (BNL) and extensive conversations with Ken Johnson of BNL, the exact cause of the gas loop problem has not been determined.
A small increase in efficiency was observed in the TEcrm values during the cruise. The TE values used to calibrate the sample TCO2 values, therefore, were determined by fitting the TEcrm values with a linear regression as a function of time (Fig. 3). The TCO2 of samples in µmol/kg was determined using the following equation:
TCO2 = counts / (0.1763148 × day + 4746.161) × [1000 / (vol × p )] × 1.00067 ,
where counts are the blank corrected coulometer counts, day includes the fractional day determined from the titration time, vol is the "to deliver" volume of the pipette corrected for the thermal expansion of glass, p is the density of seawater, and 1.00067 corrects for the dilution of the sample by the addition of 200 µL of HgCl2 to the 300-mL sample bottle.
The analytical precision of the TCO2 analyses can be estimated from the standard deviation (SD) of the 83 CRMs analyzed throughout the cruise. The SD of the calibrated batch 15 CRM values was ±1.91 µmol/kg. The sample precision can be evaluated from the 70 sets of duplicate samples collected in shallow, mid-depth, and deep waters at every station. The average difference between duplicates was 0.16 ± 1.71 µmol/kg, suggesting that sample precision was not significantly different from the analytical precision of the CRMs. As a further check on the accuracy of the TCO2 analyses, duplicate samples from the surface and the 3000-m Niskins were collected from 10 stations along the cruise track and returned to SIO for analysis by vacuum extraction and manometry in C. D. Keeling's laboratory. Ten samples have been analyzed to date, giving a mean difference (shore - sea) of 0.64 ± 1.79 µmol/kg, which is not statistically different from zero (Table 1). Note that one sample was considered bad and was excluded from the calculation because it was more than three standard deviations from the mean. These replicates also further corroborate the use of the CRM calibration since the mean difference for the gas loop calibration values would have been 2.97 ± 2.49 µmol/kg.
Table 1. Comparison of shipboard TCO2 analyses to shore-based TCO2 analyses made by C. D. Keeling at SIO
|Station no.||Niskin no||TCO2 (PU) (µmol/kg)||TCO2 (SIO) (µmol/kg)||ΔCO2 (SIO-PU) (µmol/kg)|
*does not include sample from station no. 65, Niskin no. 12.