Clouds and aerosols play a crucial role in the Earth’s atmospheric system, significantly impacting the Earth’s radiation balance, water cycle, and air quality. Space-borne lidar serves as a unique tool for the vertical simultaneous detection of aerosols and clouds, providing the advantage of all-weather operation. The cloud-aerosol lidar and infrared pathfinder satellite observations (CALIPSO) satellite represents the most notable example of this technology. However, due to its low signal-to-noise ratio, traditional lidar layer detection algorithms based on slope and threshold often miss optically thin layers of clouds and aerosols. Therefore, we propose a U-Net neural network classification model based on a two-dimensional hypothesis testing layer detection algorithm (2DMHT-UNet) to achieve high-precision detection and classification of these missed layers.

We initially employ a two-dimensional hypothesis testing (2D-MHT) algorithm for high-precision layer detection of CALIPSO observations. Subsequently, we construct a cloud and aerosol classification model based on the U-Net neural network, using RGB inputs of optical signals such as depolarization ratio, color ratio, and backscatter coefficient. This model aims to categorize atmospheric layers detected by the 2D-MHT but missed by official CALIPSO products. To ensure spatial consistency with CALIPSO products, we use long-term CALIPSO official classification products (VFM) as the training set, validating model performance with independent samples. Furthermore, we compare the combined classification results of 2DMHT-UNet (including both successfully detected and missed layers by CALIPSO) with Radar-Lidar joint observation products for validation.

The model, trained using CALIPSO VFM official products as ground truth and validated for accuracy based on independent samples from one month, indicates a classification consistency of 89.4% (land) and 90.2% (sea), with accuracy above 88% for both day and night (Fig. 2, Fig. 3 and Table 2). Comparative results based on Radar-Lidar joint observations demonstrate that the model effectively identifies cloud information missed by CALIPSO VFM official products due to low signal-to-noise ratio, reducing the relative error in cloud base detection by 21% (land) and 25% (sea) (Fig. 6).

The results demonstrate the excellent performance of 2DMHT-UNet in classifying atmospheric layers undetected by the CALIPSO official product. The 2DMHT-UNet algorithm significantly improves CALIPSO’s ability to detect boundary layer clouds, especially over land. However, due to the similarity in properties between marine aerosols and thin water clouds, accurately distinguishing between them remains challenging and may lead to misclassifications. Future efforts involve further optimizing the model to enhance classification accuracy and adding more validation experiments for aerosols based on airborne observations.