Development of a multi-module data-driven integrated framework for identifying drivers of atmospheric particulate nitrate and reduction emissions: An application in an industrial city, China

Atmospheric particulate nitrate (pNO₃^-), a crucial component of fine particulate matter, significantly contributes to haze pollution. The formation of pNO₃^- is driven by multiple factors including meteorology, emissions, and atmospheric chemistry. Understanding the key drivers of pNO₃^- formation and developing an accurate and physically meaningful method for the timely assessment of the direct causes of pNO₃^- pollution are essential. In this study, we propose a multi-module data-driven integrated framework that incorporates and improves four distinct machine learning modules. This framework enhances the physical interpretability of the statistical outcomes of the driving factors of pNO₃^-, quantifies the impacts of multiple factors on pNO₃^-, and reveals emission reduction trends. Our findings show that meteorology and emissions affect pNO₃^- by 35.3 % and 64.7 %, respectively, while atmospheric chemistry (48.0 %) and humidity (17.1 %) are the key drivers of its formation. Photochemistry promotes the formation of pNO₃^- in summer, whereas liquid-phase reactions dominate in winter at higher humidity levels (>60 %). The industry source (IS) (14.3 %), combustion source (CS) (12.8 %), and transportation source (TS) (11.8 %) are the main emission sources. The formation of pNO₃^- by the primary emissions and the transformation of NOx emitted from CS and TS is more sensitive to the changes of meteorological conditions, and controlling CS has the greater benefits to reduce pNO₃^-. The proposed framework could provide a reliable method for identifying drivers of pNO₃^- pollution at different haze events, supporting the formulation of control measures.