Journal Article

Machine Learning‐Driven Identification of Factors Governing Secondary Organic Aerosol Formation During Autumn in Beijing

Jiang, Jingkun, Zhang, Peng, Chen, Tianzeng, Li, Hao, Mu, Yujing, Li, Shuying, Ma, Qingxin, He, Kebin, Liu, Yuan, Zhang, Chunyan, He, Hong, Huang, Wei, Worsnop, Douglas R, Liu, Quan, Liu, Jun, Chu, Biwu, Wang, Shuxiao, Wang, Yonghong

2025

Overview

Organic aerosol (OA) and its constituent particulate organic nitrate (pON) are critical factors affecting air quality and climate, yet their sources and transformation processes remain poorly understood. Machine learning (ML) excels at identifying nonlinear relationships among features, and in this study, interpretable ML is employed to identify the key factors governing OA and pON formation during an autumn field campaign in Beijing. Results demonstrate that both aerosol liquid water content (ALWC) and aerosol surface area are two primary factors governing the formation of OA and pON. Specifically, OA formation was predominantly driven by ALWC that is associated with aqueous‐phase processes or gas‐liquid partitioning, particularly during severe pollution episodes. pON formation was constrained by aerosol surface area, indicating the vital contribution of gas‐to‐particle partitioning from low volatility vapors or interface processes of precursors. Our results provide new insights into OA formation mechanisms.

• Publisher: John Wiley & Sons, Inc
• Subject: Aerosol content / Aerosol formation / Aerosols / Air quality / Algorithms / Autumn / Humidity / Learning algorithms / Liquid water content / Machine learning / Moisture content / Nitrates / Oxidation / Partitioning / Secondary aerosols / Surface area / Vapors / VOCs / Volatile organic compounds / Water / Water content