Decoupling geometric and spectral discrepancies between GEO and LEO vegetation indices: A physical cross-calibration framework for FY-4B/AGRI and MODIS

New-generation geostationary Earth orbit (GEO) satellites, such as the Fengyun-4B Advanced Geostationary Radiation Imager (FY-4B/AGRI), acquire observations every 15 minutes that complement polarorbiting lowEarth orbit (LEO) sensors like theModerate Resolution Imaging Spectroradiometer (MODIS). However, directly combining FY-4B/AGRI and MODIS vegetation indices fails because of two basic physical differences. First, spectrally, FY-4B/AGRI uses a much broader red band (550–750 nm) than MODIS (620–670 nm) that extends into the bright vegetation red edge, systematically elevating red reflectance and depressing FY-4B vegetation indices across the growing season. Second, geometrically, FY-4B observes each ground location from a fixed slant view; around local solar noon, the sun aligns behind the sensor (minimal phase angle), eliminating canopy shadows and triggering a strong backscattering hotspot surge. To separate and correct these two effects, we develop a sequential two-stage physical crosscalibration framework that combines kernel-driven satellite view-angle mapping (SVA) with a land-coverspecific spectral band adjustment factor (SBAF), coupled with a dual-index quality-controlled midday compositing scheme. Evaluated across eight Eddy Covariance flux tower footprints and full-disk 0.05◦ grids (N = 689, 195 vegetated pixels) across Asia and Oceania for the full year 2023, our results show that the midday hotspot inflates red reflectance disproportionately relative to near-infrared reflectance. Consequently, ratio-based NDVI suffers a sharp midday drop (“double penalty” and “negative hotspot” depression), whereas EVI and EVI2 show a midday bulge because their additive canopy background constant (L = 1) stabilizes the denominator. At the seasonal scale, conventional nadir BRDF normalization (SNA) fails to resolve cross-orbit discrepancies. Forward-projecting MODIS BRDF parameters into the exact FY-4B viewing geometry via SVA synchronizes phenological waveforms and strips away angular scatter (raising full-disk R2 from 0.699 to 0.795), which unmasks the pure hardware spectral offset. Applying SBAF on top of SVA subsequently closes the spectral baseline gap, cutting pooled cross-site RMSE by more than half to 0.042–0.045 (MAD = 0.031–0.034) and achieving the lowest full-disk unbiased RMSE (ubRMSE = 0.059). Across all three indices, two-band EVI2 proves to be the optimal choice for cross-platform monitoring by avoiding both NDVI’s hotspot sensitivity and EVI’s blue-band atmospheric instability, providing a transferable physical foundation for seamless GEO–LEO virtual constellations across next-generation missions such as FY-4B/AGRI, Himawari-8/9, and GOES-R paired with MODIS and VIIRS.