The spatial diversity of secondary organic carbon aerosols at the city level: insights from explainable machine learning

Abstract

Secondary organic carbon (SOC) is a ubiquitous component of atmospheric fine particles but its spatial diversity at the city level has rarely been characterized. This study is the first of its kind that examine the spatial diversity of SOC at the city level, utilizing hourly carbon analyzer data of 2022 from four representative observation sites in the megacity of Guangzhou, China (GYQ, DXC, NSPZ and CHLK) encompassing urban, suburban, and rural areas. The minimum R-squared (MRS) method was adopted for SOC estimation. This study pioneers the use of explainable machine learning for SOC reconstruction with the CatBoost algorithm and Shapley Additive Explanations (SHAP) to quantify the contribution of input factors. The 12 input factors were grouped into four categories: photochemical reactions, aqueous-phase reactions, primary emissions, and meteorological conditions. Photochemical reactions dominate variations in SOC concentrations at urban and suburban sites. The annual average contribution of photochemical reactions gradually decreases from urban (∼50%) to rural sites (∼30%). The aqueous phase and meteorological conditions both contribute ∼20% of SOC variations. Primary emissions contribution to SOC variations is less than 10% for suburban and urban sites but the contribution increased to 33% for the rural site. Further analysis suggests that the high contribution of primary emissions at the rural site reflects the influence of aged urban plumes. SOC at the suburban site is heavily influenced by shipping emissions due to its coastal location. SOC at the two urban sites exhibits different dependence on Non-methane hydrocarbon (NMHC) and O₃ levels, which is related to the uneven influence of transported SOC at the two urban sites. SOC from local, regional, and background contributions was quantified at the urban site DXC using multiple-site observations. It was found that approximately 50% of the SOC originated from outside Guangzhou city. This indicates that future efforts to control SOC may require a shift from relying solely on local mitigation measures to a more comprehensive approach that considers the spatial diversity of SOC for effective air quality management.