Improved 1-km-Resolution Hourly Estimates of Aerosol Optical Depth Using Conditional Generative Adversarial Networks

Aerosol Optical Depth (AOD) is a crucial parameter for various environmental and climate studies. Merging multi-sensor AOD products an effective way to produce with more spatiotemporal integrity accuracy. This study proposed conditional generative adversarial network architecture (AeroCGAN) improve the estimation of AOD. It first adopted MODIS Multiple Angle Implication Atmospheric Correction (MAIAC) data training initial model, then transferred trained model Himawari obtained 1-km-resolution, hourly products. Specifically, generator encoder–decoder preliminary resolution enhancement. In addition, three-dimensional convolutional neural (3D-CNN) was used environment features extraction connected residual improving Meanwhile, sampled were designed as conditions generator. The spatial distribution feature comparison quantitative evaluation over area North China Plain during year 2017 have shown that this approach can better accuracy help local patterns.