9:00 AM - 10:30 AM
[ACG36-P03] Estimation of Surface Reflectance of Fengyun-4A/AGRI for Terrestrial Monitoring
Keywords:Fengyun-4A/AGRI, Geostationary satellite, Surface reflectance, Atmospheric correction, BRDF correction, Terrestrial monitoring
The surface reflectance observed by remote sensors is an essential parameter for terrestrial monitoring in a quantitative way. Fengyun-4A (FY-4A), a new generation geostationary satellite, and its onboarding sensor, Advanced Geostationary Radiation Imager (AGRI) can monitor terrestrial vegetation with high temporal frequency across Asia and Oceania. In particular, the combination of FY-4A/AGRI and Himawari-8/AHI can cover most of Asia and Oceania regions. However, significant efforts including atmospheric correction, bidirectional reflectance distribution function (BRDF) correction, and their evaluation are required prior to use for land surface monitoring.
Purpose of this study is to estimate surface reflectance including BRDF correction using FY-4A/AGRI data. In this study, 71 images including full disk and the China regional image were used over 8 hours during daytime, where there was no cloud over the study area. The study area is located in Henan and Anhui provinces of China and the land cover is mainly forest and cropland. The Second Simulation of a Satellite Signal in the Solar Spectrum (6S) radiative transfer model (RTM) were used for atmospheric correction. A kernel-driven BRDF model was applied to the visible band, near-infrared band and short-wave Infrared band over 8 hours during daytime. Two types of volume kernel (Ross-Thick and Ross-Thin) and three types of geometric kernel (Li-Sparse-Reciprocal, Li-Dense and Roujean) were used for BRDF correction. The BRDF model parameters fiso, fvol, fgeo were adjusted by the number of samples observed during a certain period using multiple regression.
The goodness of fit for six combinations of model kernel of every pixel were compared. It was concluded that the average adjusted R2 of all bands' best kernel combinations are greater than 0.95. Kernel combinations and kernel parameters depended on wavelength, terrain, land cover type et al.. The best kernel combinations in the study area were the Ross-Thin/Li-Dense kernel for band1 and band2 and Ross-Thick/Li-Dense kernel for other bands. Inter-comparison of the result with the MODIS Surface Reflectance (MOD09GA) and Vegetation Index (VI) (MOD13A2) product has a higher correlation after atmospheric correction and BRDF correction. This results shows the BRDF model is the effective method for correction of view and illumination angle effects. Atmospheric correction and BRDF correction can improve data accuracy significantly in application of terrestrial monitoring using geostationary remote sensing images.
Purpose of this study is to estimate surface reflectance including BRDF correction using FY-4A/AGRI data. In this study, 71 images including full disk and the China regional image were used over 8 hours during daytime, where there was no cloud over the study area. The study area is located in Henan and Anhui provinces of China and the land cover is mainly forest and cropland. The Second Simulation of a Satellite Signal in the Solar Spectrum (6S) radiative transfer model (RTM) were used for atmospheric correction. A kernel-driven BRDF model was applied to the visible band, near-infrared band and short-wave Infrared band over 8 hours during daytime. Two types of volume kernel (Ross-Thick and Ross-Thin) and three types of geometric kernel (Li-Sparse-Reciprocal, Li-Dense and Roujean) were used for BRDF correction. The BRDF model parameters fiso, fvol, fgeo were adjusted by the number of samples observed during a certain period using multiple regression.
The goodness of fit for six combinations of model kernel of every pixel were compared. It was concluded that the average adjusted R2 of all bands' best kernel combinations are greater than 0.95. Kernel combinations and kernel parameters depended on wavelength, terrain, land cover type et al.. The best kernel combinations in the study area were the Ross-Thin/Li-Dense kernel for band1 and band2 and Ross-Thick/Li-Dense kernel for other bands. Inter-comparison of the result with the MODIS Surface Reflectance (MOD09GA) and Vegetation Index (VI) (MOD13A2) product has a higher correlation after atmospheric correction and BRDF correction. This results shows the BRDF model is the effective method for correction of view and illumination angle effects. Atmospheric correction and BRDF correction can improve data accuracy significantly in application of terrestrial monitoring using geostationary remote sensing images.