Most of what we know about the global carbon budget has been derived from the now 40-year-long record of atmospheric CO2 sampled in clean “background” locations. These data are interpreted with a high degree of confidence at the global scale, leading to a detailed record of the time series of the rate of increase. The rate of combustion of fossil fuels is known from econometric tabulations (Marland, 1989; Andres et al, 1996), so the history of atmospheric CO2 can be used to infer the integral of all other sources and sinks in the Earth system by subtracting this anthropogenic emission rate (Keeling et al, 1989a, 1995; Conway et al, 1994; Francey et al, 1995). Such analyses have led to the conclusion that about half of the CO2 from the emission of fossil fuels is removed from the atmosphere by sinks that vary in strength by about a factor of two from year to year. The cause of these interannual fluctuations is difficult to determine, as are the underlying sink mechanisms.
The nature and variability of the sinks of anthropogenic carbon have also been investigated using the spatial distribution of atmospheric CO2 as measured by flask samples collected in the remote marine boundary layer. Until recently, the distribution of these flask sampling sites was so sparse that the data were only sufficient to characterize the north-south gradient. The addition of new stations and recent efforts to combine sampling networks through intercalibration (Masarie and Tans, 1996, see Fig 1) has made it possible to analyze the longitudinal variations as well.
Quantitative interpretation of the spatial structure of atmospheric CO2 in terms of sources and sinks at the surface requires accounting for atmospheric transport upstream of the observing stations. This is done using numerical models in which CO2 is transported by winds derived either from analyzed observations or from general circulation models (GCMs). The process of inferring surface sources and sinks from observed concentration patterns using a transport model is referred to as “inversion” of the data (Enting and Mansbridge, 1989, 1991; Hartley and Prinn, 1993; Enting et al, 1995). Inversion of the interhemispheric gradient of CO2 concentration can be accomplished using a two-box model of atmospheric mixing; the more dense network of observing stations can only be interpreted in terms of continental fluxes with a full three-dimensional chemical tracer model (CTM).
Several groups are now using three-dimensional CTMs to perform time-dependent inverse calculations of the atmospheric carbon budget with continental or regional resolution (Rayner et al, 1998; Fan et al , 1998; Gloor et al, 1998; Peylin et al, 1998?). These inversions, made possible by the expansion of the global flask network, innovative mathematical techniques, and the use of multiple geochemical tracers, may yield valuable insights into the processes responsible for removing CO2 from the atmosphere. In addition, there is the possibility that such calculations may eventually prove useful for monitoring the rate of anthropogenic CO2 emissions from continents or even countries.
The current suite of carbon budget inversion studies produce results which are difficult to reconcile with one another. A recent calculation by Song-Miao Fan and colleagues at Princeton University (Fan et al, 1997) found that the carbon sink in the Northern Hemisphere between 1988 and 1992 was dominated by terrestrial uptake in North America. Their results, if correct, imply that the terrestrial sink approximately compensates for the anthropogenic emissions in this region. Peter Rayner and colleagues at Monash University in Australia perfoApril 28, 2006 3:42 PMa different mathematical method. They found that the northern terrestrial sink was distributed almost evenly across the northern continents, with North America acting as only a weak sink (Rayner et al, 1997).
Interpreting Satellite CO2 Data
Variations in atmospheric CO2 concentration contain information about sources and sinks at the Earth’s surface. Quantitative interpretation of this information by inverse modeling requires a knowledge of the atmospheric transport that links surface exchanges to concentrations in space and time. Such calculations have greatly added to our understanding of the global carbon cycle, yet have been limited by uncertainties in transport and the sparseness of concentration measurements (Gurney et al, 2001). The latter problem may be greatly alleviated over the next decade by the advent of a series of satellite missions to measure atmospheric CO2 globally at high spatial and temporal resolution.
Previous studies of the impact of satellite CO2 observations on the inverse problem have assumed perfect knowledge of atmospheric transport by using the same numerical model for the creation of the satellite “pseudodata” and the Jacobian used in the inversion. In this paper, we investigate the effects of vertical weighting of atmospheric observations and of transport error on flux inversions from hypothetical satellite retrievals of CO2.