Global climatology of aerosol layers from ICESat-2 lidar observations

The Advanced Topographic Laser Altimeter System (ATLAS) onboard the Ice Cloud and land Elevation Satellite-2 (ICESat-2) has been providing Earth observation data since October 2018, primarily for monitoring ice sheets, sea ice, and vegetation. Although ATLAS is optimized for altimetry, it also includes an atmospheric channel that measures atmospheric backscatter from 14 km down to the surface using the photoncounting technique and three 532-nm laser beams operating at 10 kHz with pulse energy of 500 µJ. ATLAS provides valuable data with high spatial (280 m) and vertical resolution (30 m) for aerosol and cloud measurements that can constrain various models. In this work, we present a global analysis of spatial and vertical distributions of aerosol layers over land and sea using multiple years of level-3 ICESat-2 ATL09 data. We employ a machine learning algorithm, a U-Net convolutional neural network (CNN) model, trained with traditional datasets, for cloud-aerosol detection and discrimination. The CNN model identifies aerosol and cloud with higher accuracy and finer spatiotemporal resolution compared to the traditional products. This approach allows us to investigate vertical aerosol structure at a small spatial scale and utilize lidar data under both clear-sky and cloudy conditions. A global depiction of vertical aerosol layer distribution with high spatial resolution is crucial not only for studies on air quality, cloud formation, and climate, but also for correcting the impact of aerosol-induced scattered radiation on passive satellite sensors.