Scene-Level Meteorological Predictors Are Insufficient for AOD550 Estimation: A Baseline Study toward Hyperspectral Data Fusion

Abstract:

Atmospheric aerosols are a major component of the Earth system because they affect the radiative balance, cloud microphysics, atmospheric chemistry, visibility, air quality, and human health[1]. One of the most widely used integrated indicators of aerosol burden is aerosol optical depth (AOD), which quantifies the attenuation of solar radiation in the atmospheric column caused by scattering and absorption by aerosol particles[2]. As a column-integrated optical property, AOD has become a key variable in climate studies, air pollution assessment, and environmental monitoring[3]. Over the past two decades, satellite remote sensing has made large-scale aerosol monitoring operationally feasible[4]. Multispectral instruments and associated retrieval frameworks have enabled global or near-real-time observation of aerosol-related products and substantially advanced research on aerosol climatology, transport, and pollution events. At the same time, aerosol retrieval from space remains inherently difficult. The top-of-atmosphere signal depends not only on aerosol loading, but also on surface reflectance, cloud contamination, viewing geometry, aerosol type assumptions, and vertical aerosol distribution. As a consequence, retrieval uncertainty remains regionally variable, and the interpretation of AOD products requires caution, especially when they are used in downstream environmental analyses[5]. This broader methodological context is well established in the aerosol remote sensing literature and motivates continued efforts to improve the representation and estimation of aerosol optical properties. In parallel with the development of satellite retrievals, atmospheric reanalyses and composition services have become an important source of aerosol-related information. Products generated within frameworks such as CAMS and ERA5 provide globally consistent spatio-temporal