The potentials of uncertainty analysis and Bayesian optimization in HONO source modeling diagnosis and improvement

Abstract

Nitrous acid (HONO) plays a critical role in atmospheric chemistry, significantly influencing hydroxyl radical (OH) production and the formation of secondary pollutants. However, current atmospheric chemical transport models (CTMs) still underestimate HONO formation, due to uncertainties in source parameterizations. This study proposed a new framework that combines uncertainty analysis with Bayesian optimization (RFM-BMC) to diagnose and reduce uncertainties in HONO source parameterizations, using the North China Plain (NCP) as a case study. The results show that uncertainties in source parameterizations cause HONO simulation concentrations varying by 8–20 times the baseline values. The primary contributors to uncertainties in HONO simulations include heterogeneous reactions on aerosol (33–59 %) and ground surfaces (18–30 %), vehicle emissions (12–33 %), and nitrate photolysis (26–30 %). By optimizing these parameters using observational data, the accuracy of HONO simulations significantly improves, reducing the normalized mean bias by 59 %. Additionally, this study identifies soil emissions, light-induced NO₂ heterogeneous reactions and underestimated nitrate as important underrepresented HONO sources in CTMs. These sources contribute to the systematic underestimation of HONO concentrations during midday (08:00–14:00). This work provides valuable insights for refining HONO source parameterizations and improving air quality simulations. Furthermore, the RFM-BMC framework can be applied to optimize parameterizations of other atmospheric chemical processes, such as sulfate and secondary organic aerosol formation.