Data fusion of CO₂ retrieved from GOSAT and AIRS using regression analysis and fixed rank kriging

This paper proposes an improved statistical method for fusing carbon dioxide (CO₂) data retrieved from two major instruments, the Greenhouse gases Observing SATellite (GOSAT) and the Atmospheric Infrared Sounder (AIRS). These two datasets were fused to obtain CO₂ concentrations near the surface, which is a region that is especially important for studies on carbon sources and sinks. Overall, the CO₂ monthly average values from GOSAT are all lower than those from AIRS from 2010 to 2012. The datasets show the similar seasonal cycles of carbon dioxide and show an increasing trend with a determination coefficient of 0.45. A strong correlation was determined by adding the climatic factors as independent variables for regression analysis. The correlation coefficients between the CO₂ values from AIRS and GOSAT significantly increased in response. The true CO₂ data processes were then predicted using the fixed rank kriging method. This showed that the data-fusion CO₂ product provides more reasonable information and that the corresponding mean squared prediction errors are smaller than those from the single GOSAT CO₂ dataset.