We compare the estimates of CO sources in Asia obtained using satellite observations versus aircraft observations. The figure at the left shows mean CO concentrations observed over the Western Pacific during the TRACE-P campaign in spring of 2001. Simultaneously, the MOPITT instrument aboard EOS-Terra was globally measuring CO. The figure on the right shows mean CO columns over Asian and the Western Pacific as measured by MOPITT during TRACE-P, and illustrates the main outflow pathway from Asia.

Bayesian synthesis inverse methods were applied to estimate Asian sources in 2001 based on these two sets of observations. The GEOS-CHEM CTM was used as the forward model and customized a priori emission estimates were obtained from a year 2000 anthropogenic emission inventory by Streets et al. [2003], and a year 2001 biomass burning inventory using fire count data by Heald et al. [2003a]. The central figure compares the a posteriori emission estimates from TRACE-P (red) and MOPITT (blue) to a priori (black). MOPITT observations provide greater information to regionally disaggregate regions (10 degrees of freedom), compared to the aircraft (4 degrees of freedom). This is due to the ability of the satellite to make observations over the source regions and over the Indian Ocean. The aircraft and satellite provide generally consistent estimates of Asian sources, where regions dominated by anthropogenic emissions, such as China are underestimated, and regions dominated by biomass burning emissions, such as SE Asia are overestimated. MOPITT observations support a more modest decrease in emissions in SE Asia than suggested by the aircraft observations, likely due to the satellite's ability to observe over the continental region and thus better characterize SE Asian outflow over the source region.

The inverse solution is sensitive to error specification and data selection. In particular, averaging the MOPITT data degrades the information which can be used to constrain regional sources. By conducting ensemble modeling with the MOPITT observations the uncertainty on retrieved sources is estimated to be 10-40%.

For a detailed account please see Heald et al. [2004].