Data Checks and Processing Performed by CDIAC

An important part of the NDP process at the Carbon Dioxide Information Analysis Center (CDIAC) involves the quality assurance (QA) of data before distribution. Data received at CDIAC are rarely in a condition that would permit immediate distribution, regardless of the source. To guarantee data of the highest possible quality, CDIAC conducts extensive QA reviews. Reviews involve examining the data for completeness, reasonableness, and accuracy. Although they have common objectives, these reviews are tailored to each data set, often requiring extensive programming efforts. In short, the QA process is a critical component in the value-added concept of supplying accurate, usable data for researchers.

The following summarizes the data QA checks and processing performed by CDIAC on the data obtained during the NOAA/PMEL R/V Malcolm Baldrige CGC-90 Cruise in the Southwest Pacific.

1. These data were provided to CDIAC as two ASCII-formatted files and accompanying printed documentation (NOAA Data Report ERL PMEL-42) (Lamb et al. 1993). A FORTRAN 77 retrieval code was written and used to reformat the original files.
2. To check for obvious outliers all data were plotted by use of a PLOTNEST.C program written by Stewart C. Sutherland, of the Lamont-Doherty Earth Observatory. The program plots a series of nested profiles, using the station number as an offset; the first station is defined at the beginning, and subsequent stations are offset by a fixed interval (Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, and Fig. 8)¹. Several outliers were identified and flagged after consultation with the principal investigators.
3. To generate a section profile plot of TCO₂ concentrations along the 170^o W, the SURFER program developed by Golden Software, Inc. for Windows version 5.0 was used (Fig. 9).
4. To identify "noisy" data and possible systematic methodological errors, property-property plots for all parameters were generated (Fig. 10), carefully examined, and compared with plots from previous expeditions in the Southwest Pacific.
5. To identify possible instrumentation drifts and methodological errors, the data intercomparison for reoccupied stations was provided(Fig. 11).
6. All variables were checked for values exceeding physical limits, such as sampling depth values that are greater than the given bottom depths.
7. Station locations (latitudes and longitudes) and sampling dates were examined for consistency with maps and with cruise information supplied by Lamb et al. (1993).
8. The designation for missing values, given as "-99.00" in the original files, was changed to "-999.90".

¹ All data were plotted with excluded questionable measurements flagged by quality flag 3.