Exploring Aerosol Vertical Distributions and Their Influencing Factors: Insight from MAX-DOAS and Machine Learning

Understanding aerosol vertical distribution is crucial for aerosol pollution mitigation but is hindered by limited observational data. This study employed multiaxis differential optical absorption spectroscopy (MAX-DOAS) technology with a coupled radiative transfer model-machine learning (RTM-ML) framework to retrieve high-resolution aerosol optical properties in Shanghai. Retrievals indicated vertically decreasing aerosols, peaking in the upper atmosphere in the summer and in the lower atmosphere in the winter. Aerosol hygroscopicity followed similar seasonal patterns but increased with the altitude. Multifactor driving ML models and Shapley additive explanations (SHAP) were used to investigate the drivers to aerosol variation. Results indicated that emissions, east-west transport, and atmospheric oxidation were the main drivers of aerosols below 0.5 km. Above 0.5 km, humidity and atmospheric oxidation became dominant, suggesting that hygroscopic growth and secondary aerosol formation were more prominent. North-south transport also significantly influenced aerosol distribution within 0.5 to 1.6 km. Meteorological normalization emphasized that emission reduction can effectively lower aerosols in the lower atmosphere, while enhanced atmospheric oxidation promoted secondary aerosol formation, particularly in the upper atmosphere. These findings advance the understanding of multiple factors in shaping the vertical aerosol distributions and highlight that emission reduction strategies for addressing compound pollution should be conceived with a multidimensional and multifactorial understanding.