Error matrix construction method for atmospheric carbon dioxide Bayesian retrieval

Based on the retrieval of atmospheric carbon dioxide using infrared spectral data detected by GOSAT, a method of building error matrixes of Bayesian theory was proposed and validated. Firstly, the effect on retrieval results by different initial guesses, different building results of priori information error matrix Sa and measurement error matrix S was simulated, and then two combinations of Sa and S building results were validated in the retrieval using part of GOSAT measurements of Taklimakan desert during 2009. The result shows that the retrieved results are more concentrate in the case of bigger priori information variance or smaller measurement error, and the retrieved results are dispersed under the converse circumstance. It is difficult to get real error matrixes in atmospheric remote sensing, therefore, this study will be significant for getting more accurate error matrixes and improving retrieval precision.