首页 > , Vol. , Issue () : -
The land surface reflection depends largely on the land surface albedo and interferes with retrieving the cloud geometrical thickness from the OCO-2 oxygen A-band observations due to its second-strongest reflection after the cloud. However, no product can provide the land surface albedo of the OCO-2 oxygen A-band required for the retrieval. Therefore, the accurate estimation of the land surface albedo is necessary and beneficial to the retrieval quality. In this study, we proposed an idea of estimating the land surface albedo in the oxygen A-band from the multi-channel black/white albedos from MODIS/MCD43C3 products. Although the estimation (MODIS → OCO-2) is land-cover-type-related, the comparison based on the Shannon entropy proved that the multi-channel albedo data contains the type information and sufficient to achieve the same accuracy as the land-cover-type-involved estimation. In addition, we implement the estimation model by BP neural network, and the accuracy is consistent with that of the analysis based on the Shannon entropy. We verified the multi-channel-based estimation model by the tests in different times and different spaces whose coefficients of determination were all over 0.9 and whose root-mean-squared errors were 0.026. In addition, the multi-channel-based model was always superior to the single-channel linear model on all land cover types, whether applied to the best performing type of the barren or sparsely vegetated, or the worse performing type of the snow and ice. The quality of the MODIS albedo data is the most important for the accuracy of estimation. The root-mean-squared error with the best inputs was slightly better than 0.02 and increased to more than 0.05 as the quality of the inputs decreased. The method of estimating the land surface albedo in the OCO-2 oxygen A-band from MODIS multi-channel black/white albedo data is feasible and can resist disturbance caused by unknown land cover type. The estimation accuracy depends mainly on the quality of the input MODIS albedo data.