Abstract
True-color images synthesized from red, green, and blue (RGB) bands are essential for monitoring the Earth’s surface and atmosphere. However, the Advanced Very High-Resolution Radiometer (AVHRR), despite four decades of observations, lacks green and blue bands, limiting natural-color representation and feature discrimination. Cross-sensor spectral inconsistencies between AVHRR and the Moderate Resolution Imaging Spectroradiometer (MODIS), together with radiometric variations caused by atmospheric scattering, aerosol effects, and calibration drift, complicate true-color reconstruction. To address this limitation, we propose XBAER-RGB, a reconstruction framework within the eXtensiBle Atmospheric and surfacE Retrieval (XBAER) algorithm family. The framework removes Rayleigh scattering and ozone absorption while preserving aerosol-related signals, and reconstructs missing bands using an eXtreme Gradient Boosting (XGBoost) model trained on corrected MODIS corrected reflectance. XBAER-RGB was validated on MODIS by comparing reconstructed and observed green/blue bands, and then applied to spectrally adjusted AVHRR data. The reconstructed MODIS green/blue bands achieved peak signal-to-noise ratio/structural similarity index (PSNR/SSIM) values of 37.49 dB/0.9869 and 30.08 dB/0.9544, while AVHRR achieved 33.66 dB/0.8480 and 32.51 dB/0.8440. A global AVHRR true color record since 1981 was generated, capturing clouds, wildfires, dust storms, and land-ocean features, while supporting cloud detection, environmental monitoring, and long-term climate analysis.