Machine learning-optimized interpretability analysis for identifying key drivers of NO₃ lifetime variability

Nitrate radical (NO₃), a critical nocturnal oxidant, was measured continuously from January 2022 to December 2023 in Shanghai Dongtan Wetland Park, China. The nocturnal NO₃ levels ranged from approximately 4.7 to 300 pptv, with daytime measurements below the detection limit. A steady-state analysis, based on the observed NO₃, nitrogen dioxide (NO₂), and ozone (O₃), was employed to calculate NO₃ lifetime. >50 % of lifetime was shorter than 120 s, with only 6 % exceeding 10 min. Machine learning-assisted SHAP interpretability method was applied to elucidate the underlying drivers of NO₃ lifetime. NO₂, relative humidity (RH), and wind direction (WD) were identified as the predominant drivers. The relationship with WD suggested that NO₃ lifetime varied markedly across air masses, spanning from only seconds within polluted industrial ports air masses to approximately 13 min in clean ocean air masses, with inland air masses exhibiting median lifetime. Meanwhile, the loss mechanisms of NO₃ also differed significantly depending on air mass origins. The higher NO₃ lifetime during Clean air masses can be principally due to significantly reduced NO₂ levels (< 2 ppbv), a transition in the RH effects from negative to positive, and decreased PM_2.5. Our study uniquely explores nocturnal NO₃ lifetime within distinct air masses over coastal region, revealing how shifts in anthropogenic emissions and atmospheric transport processes drive marked variations in NO₃ loss mechanisms. We offer new insights into how environmental variables influence NO₃ dynamics, providing new perspectives for understanding the complex nitrogen chemistry at coastal sites and advancing future atmospheric studies.