A Practical Approach to Improve the MODIS MCD43A Products in Snow-Covered Areas

The MODerate Resolution Imaging Spectroradiometer (MODIS) MCD43A products have been extensively applied in the remote sensing field, but recent researchers have demonstrated that these products still had the potential to be further improved by using the latest development of the kernel-driven model [RossThick-LiSparseReciprocal-Snow (RTLSRS)] in snow-covered areas, since the MCD43A product algorithm [RossThick-LiSparseReciprocal (RTLSR)] needed to be improved for the accurate simulation of snow bidirectional reflectance distribution function (BRDF) signatures. In this paper, we proposed a practical approach to improve the MCD43A products, which used the Polarization and Directionality of the Earth’s Reflectance (POLDER) observations and random forest algorithm to establish the relationship between the BRDF parameters (MCD43A1) estimated by the RTLSR and RTLSRS models. We applied this relationship to correct the MCD43A1 product and retrieve the corresponding albedo (MCD43A3) and nadir reflectance (MCD43A4). The results obtained highlight several aspects: (a) The proposed approach can perform well in correcting BRDF parameters [root mean square error (RMSE) = ~0.04]. (b) The corrected BRDF parameters were then used to retrieve snow albedo, which matched up quite well with the results of the RTLSRS model. (c) Finally, the snow albedo retrieved by the proposed approach was assessed using ground-based albedo observations. Results indicated that the retrieved snow albedo showed a higher accuracy as compared to the station measurements (RMSE = 0.055, bias = 0.005), which was better than the results of the MODIS albedo product (RMSE = 0.064, bias = −0.018), especially at large angles.