The vertical distribution of atmospheric aerosols exhibits highly complex characteristics and spatiotemporal variability， making it a key factor influencing the improvement of aerosol retrieval accuracy in satellite remote sensing. This study systematically investigates the vertical distribution characteristics of aerosols using unsupervised clustering methods based on CALIOP （The Cloud-Aerosol Lidar with Orthogonal Polarization） Level 3 aerosol profile data from 2010 to 2020. Three clustering algorithms—Gaussian Mixture Model （GMM）， K-means， and spectral clustering—were compared using multiple evaluation metrics to assess their clustering performance. Based on the vertical distribution characteristics of the extinction coefficient， the aerosol profiles were classified into five representative types using the GMM clustering method： low-pollution composite type， high-pollution composite type， exponential decay type， low-pollution uniform type， and high-pollution oscillatory type. Furthermore， the spatiotemporal distribution characteristics of these profiles were analyzed across different seasons and in three typical regions： the Tibetan Plateau， the Beijing-Tianjin-Hebei region， and the Yangtze River Delta. The results indicate that the aerosol profiles obtained through GMM clustering exhibit significant seasonal and regional variations.