A prediction method of the global distribution map of CO2 column abundance retrieved from GOSAT observation derived from ordinary kriging

We propose a method to generate a global distribution map of carbon dioxide (CO₂) and methane (CH₄) column abundance retrieved from spectra on irregular observation points by GOSAT (Greenhouse gases Observing SATellite). Global distribution map is gridded by 1 degree for latitude and longitude. Kriging in spatial statistics is applied to the spatial data of CO₂ and CH₄ column abundance. We focus on CO₂ density in this study, the distance and difference of CO₂ column abundances between observation points for sample pairs at each observation points over ocean and over land on the earth´s surface are calculated. The relationship between the distance and the difference of CO₂ column abundances are represented by semi-variogram model. When semi-variogram is modeled, the difference of semi-variogram derived from the direction between observation points of sample pairs from North-pole direction is considered. And, we obtain the variogram model for each land cover. GOSAT was just launched, and CO₂ column abundance is not retrieved from spectra measured by GOSAT. Therefore, proposed method is applied to spatial data of XCO₂ instead of CO₂ column abundance. We set the observation points on the earth´s surface based on the GOSAT observation plan. Global distribution map of XCO₂ instead of CO₂ column abundance is used, XCO₂ value for each observation points are set. And we predict XCO₂ values on the grid in global distribution map from the set observation points and XCO₂. As a result, the standard deviation of prediction error (predicted value - actual value) is 0.324. This standard deviation, which is 0.1% of XCO₂ value, is enough small comparison with target accuracy (1%).