2.9. DATA INTEGRATION
Measurement records from the cooperative air sampling network are frequently used to constrain two- and three-dimensional transport models that derive plausible source/sink scenarios of CO [Tans et al., 1989; Ciais et al., 1995a,b] and CH [Fung et al., 1991]. Interpretation of these scenarios is limited to large spatial scales and relatively short time scales due to the sparsity of sampling sites and the lack of temporal continuity among observations from different locations. In an attempt to reduce uncertainty in modelderived carbon budget scenarios due to the limitations of the observations, a procedure was developed that addresses these issues. Data extension [Masarie and Tans, 1995] attempts to extend the knowledge gained during a limited period of measurements beyond the period itself producing a record with no discontinuities. Extrapolated and interpolated values are derived using information about (1) the behavior of the record itself that can be described by its average seasonal cycle and longterm trend, and (2) the average behavior of the record relative to other CMDL measurement records that are nearby in latitude. The result is a set of records containing measured data and extrapolated and interpolated values. To improve the spatial resolution of the observations, high precision CO measurement records from many laboratories were extended and integrated with considerable effort to ensure compatibility with respect to methodology and calibration.
During the past year, the Cooperative Atmospheric Data Integration Project for Carbon Dioxide was created with the ongoing aim of producing a globallyconsistent CO database with unprecedented spatial resolution and temporal continuity. Flask and continuous data from the CMDL programs comprise the bulk of the database, but measurement records contributed by other laboratories in Australia, Canada, China, France, Germany, Hungary, Italy, Japan, New Zealand, and the United States have enhanced the spatial and temporal coverage and have provided unique opportunities to compare overlapping independent records. This effort has resulted in the first release of GLOBALVIEWCO2, the most complete atmospheric CO database yet available (Internet access by way of anonymous FTP to ftp.cmdl.noaa.gov, Path: ccg/co2/GLOBALVIEW).
With the framework of a global CO
database in place, ways to enhance the methods used to extend and integrate
records from different laboratories are being explored. Data extension techniques
will become more robust with the addition of updated and new measurements records.
Data integration techniques will be enhanced by using information resulting
from the comparisons of overlapping measurement records. Furthermore, frequent
comparisons will improve the quality of the original data. Beginning in 1991,
for example, a subset of CMDL air samples collected at Cape Grim, Tasmania,
was routed through the CSIRO/DAR Global Atmospheric Sampling Laboratory in Aspendale
where they were analyzed for CO
and a host of other trace gas species [Peterson and Rosson, 1994]. This
subset was then returned to Boulder where the same air was analyzed for a suite
of trace gas species including CO.
Results from this ongoing intercomparison will complement calibration and methodology
considerations and further validate the integration of measurements from the
two laboratories. Similar intercomparisons with other laboratories may be a
necessary step to fully justify the merging of independent measurement programs.
Data extension and integration techniques are now being applied to other trace
gas species such as CH,
CO, and the stable isotopes of CO.
It is anticipated that GLOBAL-VIEW13CO2 and GLOBALVIEWCH4 will be
available in the near future.