A novel algorithm for full-coverage daily aerosol optical depth retrievals using machine learning-based reconstruction technique

The ubiquitous missing values in the satellite-derived aerosol optical depth (AOD) products have always been a challenge for spatial and temporal analysis. To address this concern, we propose a novel data-driven model to attain the full-coverage daily AOD dataset with 0.01° spatial resolution in the Guangdong-Hong Kong-Macao Greater Bay Area (hereafter GBA) from 2010 to 2021. Firstly, the missing values of top-of-atmosphere (TOA) reflectance and surface reflectance of Moderate Resolution Imaging Spectroradiometer (MODIS) caused by cloud contamination, were reconstructed using the Data Interpolating Empirical Orthogonal Functions (DINEOF). Subsequently, a new model was developed for the estimation of AOD which integrates the geographical and temporal encodings into the Light Gradient Boosting Machine (LightGBM) with the inputs of reconstructed TOA/surface reflectance and other influencing variables like meteorological and geographical factors. Results showed that the derived gap-free AOD dataset outperforms the MODIS Multi-Angle Implementation of Atmospheric Correction (MAIAC) AOD product and agrees well with the ground-based observations, achieving an index of agreement (IOA) of 0.88, R of 0.84, root mean square error (RMSE) of 0.19 and mean absolute error (MAE) of 0.14. Moreover, the derived AOD dataset presents consistent temporal patterns with in-situ measurements, but with more spatial details than other gapless AOD datasets, i.e., Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). Overall, this study has developed a promising meteorological framework for the estimation of full-coverage AOD, which can also be applied over other regions. The derived long-term full-coverage daily AOD dataset can also be used for other applications related to climate change, air quality and ecosystem assessment.