Impact of differences in spatial coverage of multiple GOSAT-based CO2 datasets on regional flux estimates

摘要: The inference of regional CO2 fluxes with the top-down approach, as introduced in Chapter 1, relies solely upon atmospheric CO2 observations. As part of characterizing this inherent nature, several studies were conducted in the past to see the sensitivity of flux estimates to the choice of data-providing sites [eg Law et al., 2003; Yuen et al., 2005; Gurney et al., 2008] and to the expansion of surface monitoring networks over time [Bruhwiler et al., 2011]. These studies showed that changes in the geographical distribution of surface data have a large impact on regional-scale flux estimates. With the advent of GOSAT in early 2009, CO2 measurement by the surface monitoring networks is significantly augmented with the spaceborne XCO2 retrievals. As mentioned in Chapter 4, there exist five independent XCO2 retrieval datasets, and their precisions have been reported to be below 2 ppm level [Oshchepkov et al., 2013]. Where they coincide over land, the five XCO2 retrievals (bias corrected) were found to agree well within one standard deviation of about 1 ppm [Takagi et al., 2014]. Different from CO2 measurements at fixed surface monitoring stations, success in the retrieval of satellitebased XCO2 is highly affected by the existence of light-scattering clouds and aerosols in the local sky, and therefore the chance that the XCO2 retrievals can be obtained again at the same location over the surface in the satellite’s repeat cycle is not guaranteed. Also, in attempts to obtain better retrieval results, the five retrieval algorithms adopt different approaches in, eg, modeling the vertical distribution of clouds and aerosols and screening low-quality XCO2 retrievals …