Dataset: CO2 - Discrete pCO2
Deployment: RB-08-02

Readme: Data reduction discrete pCO2 for the Southern Ocean Gas Exchange study
Date of submission: April 26, 2009
Ship: Ronald H. Brown
Dates: March 7, 2008 to April 5, 2008
Study: A Lagrangian study focused on air-sea gas exchange
Study Area: 50 S, 38W about 200 miles North of South Georgia island.
Analyst: Robert Castle
Data Reduction: Rik Wanninkhof & Robert Castle

Analyses details:
Analyses were done by equilibrating water samples in 500-ml glass volumetric flaks at a fixed temperature of 20 °C. Headspace gas of known concentration was bubbled through the water and circulated through a LiCor 6252 infrared analyzer (IR). Details of the analysis and instrumental setup can be found in Wanninkhof and Thoning (1993). The analysis was done shipboard in the main laboratory that had poor temperature control which is believed to contribute to the variability in results.

A typical sampling sequence involves running six compressed gas standards, 8 samples, and ending with six more standards. The standards were run in following sequence:
Standard 1378.71 ppm
Standard 2792.51 ppm
Standard 31036.95 ppm
Standard 41533.7 ppm
Standard 5593.64 ppm
Standard 6205.7 ppm

Several times during the cruise the zero and span on the LiCor IR analyzer were set using nitrogen and the 1533.7-ppm CO2 standard, respectively, such that the instrument output was close to standard values. However, the sample concentrations were determined from the standard values run directly before and after the sample sequence.

Data reduction details:
A new data reduction program was created as an Excel macro to replace the GW-Basic program used from 1991- 2007. The program was set up to take the output from the IR (ppm CO2, dry and with band broadening correction applied). To calculate the fCO2 the following steps were applied by the macro:

- The XCO2 in the headspace was calculated from a linear interpolation with time of the two bracketing standards. This assumes that instrument drift is linear between standard runs.

- From the XCO2 in headspace after equilibration and the XCO2 in headspace before equilibration along with water and headspace volumes the small change in DIC of water is calculated

- The fCO2 is calculated in the headspace from the XCO2 accounting for pressure and non-ideality of CO2

- From the DIC and fCO2 after equilibration the TAlk of the water is calculated
- Since the TAlk does not change during equilibration, the TAlk and DIC before equilibration (obtained from the group analyzing DIC) are used to calculate fCO2 in water before equilibration.

- A small correction is applied to normalize data to 20°C if the bath temperature differed from 20 °C. The maximum deviation from 20 °C during the study was less then 0.02 °C which corresponds to a correction in fCO2 of less then 0.1 %.

The constants of Mehrbach as refit by Dickson and Millero (1987), along with other constants referenced therein are used in the calculations using the code based on Lewis and Wallace (1998) as ported to Excel by Pierrot. Silicate and phosphate values as provided by the nutrient group were used to determine the alkalinity. The final fCO2 values are not sensitive to the range of reported errors in DIC, SiO2 or PO4 that are used as input to calculate the fCO2 values.

Data reduction issues:
The IR setup was incorrect for this cruise in that the values reported by the IR and logged by the computer were not corrected for water vapor but they did have the pressure broadening correction applied. A second issue was that the standard appeared noisier then in the past. This is attributed, in part, to fluctuating laboratory temperatures. The standards were dry but after running samples, the first two standards had appreciable water vapor in them from flowing through tubing that is used for circulating the moist headspace. Therefore a linear regression was created for all standard sequences using standards 3-6 that were dry and from this linear relationship the corresponding (dry) standards were calculated. The linear regressions though the 4 standards always had a correlation coefficient (r2) of 0.99998 or better.

Since the water vapor channel was not spanned, we used the calculated water vapor value at 20°C (the temperature at which the samples were run) to correct the sample concentrations in the
following manner:

XCO2(dry) = XCO2 measured (1 + pH20) where

pH2O is the water vapor at 20°C of 0.0226 atm

Using the recalculated standards and the recalculated sample values the data reduction procedure listed above was executed.

Precision:
Precision of analyses was evaluated from the duplicates that were run every cast. Of the 46 duplicates the average precision was 0.3 % with a standard deviation of 0.2 % where the average precision is (ABS((sample1-sample2)*2/(sample1+sample2))*100). The duplicate values plotted against station are shown in Figure 1.
(BCO-DMO Note: Figure 1 Not Included)

Quality checks:
After data reduction fCO2 values at all stations were plotted versus depth and versus DIC to flag outliers that were assigned a quality control flag of 3 (questionable). The quality control flags of 3 (questionable) and 4 (bad) applied during analysis were retained. Values flagged as bad are not reported. In a few cases there were obvious input errors of sample number that were corrected. A multi-linear regression (MLR) was performed between fCO2(20) and DIC, T, and S. The resulting r2 was 0.98 and the RMS error in calculated fCO2(20) was 30. The large rms error precluded the use of this MLR for QC purposes.

Values reported:
A total of 629 values were reported from 51 stations. 554 were (single) good values; 47 were averages of duplicate (good) values; 11 were deemed questionable, and 17 values were bad and assigned a default value of -999

The data is reported with following columns:
JD= year day of analysis (UTC time)
Station=
Sample ID = Station # * 10000+ cast # *100 + sample #
Depth (m)
fCO2(20) = fugacity of CO2 at 20 °C (≈ analysis temperature)
fCO2 (situ) = fugacity of CO2 at in situ temperature. This is a calculated value and dependent on dissociation constants or empirical equation used
QC= quality control flag; 2= good; 3= questionable; 4= bad (and not reported); 6 = duplicate
TAlk(fCO2,DIC) = Calculated alkalinity based on fCO2 and DIC using the dissociation constants presented in Dickson and Millero (1987)

References:
Dickson, A. G., and F. J. Millero (1987), A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media, Deep-Sea Res., 34, 1733-1743.

Lewis, E., and D. W. R. Wallace (1998), Program developed for CO2 system calculations, Oak Ridge National Laboratory, Oak Ridge. http://cdiac.ornl.gov/oceans/co2rprt.html

Wanninkhof, R., and K. Thoning (1993), Measurement of fugacity of CO2 in surface water using continuous and discrete sampling methods, Mar. Chem., 44, 189-205.

BCO-DMO Processing Notes
- Generated from original file SOGasEx_fCO2disc_submitted.xls

BCO-DMO Edits
- event, station, date, time, lon, lat inserted from CTD headers file
- '-999' (No data flag in original) changed to 'nd'
- decimal places padded to 1 or 2 places as appropriate for consistency
- parameter names modified to conform to BCO-DMO convention