Generating Hourly Fine Seamless Aerosol Optical Depth Products by Fusing Multiple Satellite and Numerical Model Data

Due to cloud/snow contamination and retrieval method limitation at night, satellite aerosol optical depth (AOD) products often have many missing gaps not only at night but also during the day. In contrast, hourly seamless AOD data with coarse resolution from numerical models can usually be used to fill in the missing gaps in satellite products. However, current studies on seamless optimization of satellite AOD data only focus on fusion results during satellite overpass time and ignore the spatiotemporal complementarity of multiple satellite products. In this study, we propose a model for fusing multiple AOD datasets, which for the first time combines three satellite AOD products and two aerosol numerical model products to produce hourly seamless 1-km AOD products throughout day and night in the Beijing-Tianjin-Hebei urban agglomeration region. Compared with ground AERONET AOD data, the validated results not only achieve a promising accuracy, e.g., R-2= 0.91 (root-mean-square error (RMSE) = 0.09), in the region containing three satellite AOD retrievals, but also obtain a reasonable result, e.g., R-2= 0.83$ ( RMSE = 0.21 ), in the region without satellite AOD retrievals. The spatial information entropy evaluation results also indicate that the generated AOD data can capture more spatial details than the numerical model data. Our results demonstrate that the proposed model can generate reliable hourly seamless fine AOD data, having significant meanings in aerosol-related fields.