A Cloud Screening Preprocessing Framework for GOCI‐II Aerosol Optical Depth: Development and Impact on Retrieval Errors

Abstract Accurate retrieval of satellite‐derived aerosol optical depth (AOD) is critical for air quality forecasting, especially when AOD is assimilated into models. However, if retrieval errors in the satellite‐derived AOD are not corrected or characterized, they can lead to false analysis increments and ultimately degrade the quality of data assimilation (DA). The issue also arises in the Geostationary Ocean Color Imager (GOCI)‐II AOD product, which remains susceptible to cloud contamination. To address the issue, this study developed a cloud screening framework for GOCI‐II AOD that combines two processes: (a) cloud removal using cloud type information from Himawari‐8/9 Advanced Himawari Imager (AHI) and (b) a statistical cloud filtering based on AOD retrieval characteristics. Comparisons with Aerosol Robotic Network (AERONET) AOD demonstrated that the cloud screening framework effectively mitigated overestimation of GOCI‐II AODs compared to AERONET AODs. Consequently, the cloud‐screened GOCI‐II AOD showed a better agreement with AERONET AOD, improving Pearson's correlation coefficient from 0.77 to 0.82 and decreasing root‐mean‐square error from 0.175 to 0.110. Despite these improvements, the normalized biases of the cloud‐screened GOCI‐II AOD against AERONET AOD still exhibited diurnal variations, with GOCI‐II AODs being overestimated in coastal regions and underestimated in inland sites, under low‐AOD conditions. Nonetheless, the enhanced consistency of the cloud‐screened GOCI‐II AOD with AERONET AOD and its more Gaussian‐like error distribution supports the utility of the cloud screening framework in improving the quality of GOCI‐II AOD for DA systems.