An efficiency analysis for data synthesis of sunand geo-synchronous satellites in kernel-driven BRDF model

The land surface reflectance related to solar energy system parameters is important for interpreting the solar-earth energy balance system and climate change. Previous studies using time series of satellite-sensed images have reported that surface anisotropy effects can cause serious error in estimations of solar radiation parameters such as reflectance, albedo, vegetation indices, and insolation. To correct such BRD (Bidirectional Reflectance Distribution) effects, we use a BRDF (Bi-directional Reflectance Distribution Function) model based on a semi-empirical model. Recently, synthesizing data from polar-orbit and geostationary satellite images has been considered an alternative solution to the angle sampling difficulties caused by these different observation systems. The angular sampling distribution is examined using polar plots to identify differences between geostationary and polar-orbit satellites. We conduct comparative analyses of various BRDF kernels to determine the optimal BRDF geometric kernel (from the Li-Sparse model). To characterize prominent bidirectional effects, such as strong backward scattering, scatter plots in the principal plane are created for various solar zenith angle ranges. Then, the determined Li-Sparse geometry with Roujean volume scattering kernel is applied to estimate reflectance in a standard geometry over the study area. For all of the retrieved surface nadir reflectances, RMSE (Root Mean Square Error) variations are less than 0.3 during the study period, even in summer when severe weather conditions occur. Winter has the highest accuracy for the satellite data set. Combining MTSAT-1R (Multi-functional Transport Satellite-1R) and SPOT/VGT (Satellite Pour l'Observation de la Terre VEGETATION) data improve estimated normalized reflectance, compare with that for MTSAT-1R values alone over the study area.