A frequency-based approach to improve the geometric accuracy of FY4B/AGRI geostationary satellite observations

The Advanced Geostationary Radiation Imager (AGRI) onboard the FengYun-4B (FY4B) satellite—a new-generation geostationary (GEO) platform—offers spatial and radiometric resolutions comparable to those of polar-orbiting satellites such as EOS-MODIS, but with substantially higher temporal resolution. This enhanced temporal capability expands the potential of GEO observations beyond meteorology into terrestrial sciences. Precise geometric accuracy is essential for quantitative remote sensing, as the reliability of any downstream retrieval algorithm depends on accurate geolocation. Operational correction of geometric errors is challenging due to the scarcity of ground control points and large data volumes. Here, we evaluated the geolocation accuracy of FY4B/AGRI imagery using a full year of data and developed an integrated geometric correction workflow combining the Phase-Only Correlation method based on Fast Fourier Transform (FFT-POC) with a ray-tracing orthorectification process. In the original imagery, significant geometric instabilities were observed: east-west offsets (COFF) frequently fluctuated between ±5 and ± 10 pixels (reaching ±15 pixels) due to diurnal thermal deformation and operational maneuvers, whereas north-south offsets (LOFF) remained comparatively stable within ±5 pixels. These systematic errors were fully corrected by the FFT-POC step, while the subsequent orthorectification effectively eliminated terrain-induced parallax distortions exceeding 3 pixels in high-altitude regions. The corrected FY4B/AGRI data offers accurate geolocation to support operational hyper-temporal applications such as disaster monitoring and carbon cycle sciences.