Parallel global atmospheric correction for FY3/MERSI data over land on multi-core and many-core architectures

For the accurate derivation of biophysical parameters based on surface reflectance, atmospheric correction is a necessary step to remove scattering and absorption effects by multiple atmospheric components. However, the huge amount of data and complex algorithms pose great computing challenges for massive operational tasks. Towards the global atmospheric correction for Medium Resolution Spectral Imager (MERSI) data onboard FY-3A and FY-3B, this paper describes an atmospheric correction algorithm considering the directional properties of the observed surface, and exploits its parallel implementations on multi-core and many-core architectures. The algorithm was developed with Open Multiprocessing (OpenMP) for multi-core processors and Compute Unified Device Architecture (CUDA) for Graphics Processing Units (GPU). Experimental results show the runtime was reduced from 187.19s to 42.63s and 10.11s when implemented on a multi-core processor and NVIDIA Tesla K80 respectively.