- 1Peking University, College of Environmental Sciences and Engineering, China (xuanli@stu.pku.edu.cn)
- 2School of Environmental Science and Engineering, Tiangong University, Tianjin, China (canye@tiangong.edu.cn)
Gaseous nitrous acid (HONO), a critical precursor of hydroxyl radicals (OH), plays a key role in the atmosphere’s oxidizing capacity, driving the production of secondary pollutants. However, large uncertainties in its formation and removal mechanisms impede accurate simulation of HONO levels using chemical transport models (CTMs). In this study, a deep neural network (DNN) model was established based on routine air quality data (NO2, CO, O3, PM2.5) and meteorological parameters (temperature, relative humidity, solar zenith angle and season) collected from four typical megacity clusters in China. The DNN model exhibited robust performance on both train sets (slope = 1.0, r2 = 0.94, RMSE = 0.29 ppbv) and two independent test sets (slope = 1.0, r2 = 0.79, RMSE = 0.39 ppbv). It demonstrated excellent capability in reproducing the spatial temporal variations of HONO and outperformed an observation-constrained box model incorporated with newly proposed HONO formation mechanisms. Nitrogen dioxide (NO2) was identified as the most impactful features for HONO prediction using the SHapely Additive exPlanation (SHAP) approach, highlighting the contribution of NO2 conversion in HONO formation. The DNN model was applied to predict future change of HONO levels under different NOx mitigation scenarios, which is expected to decrease 27-44% under 30-50% NOx reduction, consistent with the box model outputs. These results suggest a dual effect brought by NOx abatement, leading to not only reduction of O3 and nitrate precursors but also decrease in HONO levels and hence primary radical production rates. The model was further employed to construct an hourly-resolved nationwide HONO dataset for China spanning 2015-2023, offering a valuable tool for constraining ozone production in the CTMs.
The construction and application of the DNN model has been published on Environ. Sci. & Tech. (Environ. Sci. Technol. 2024, 58, 29, 13035–13046).
How to cite: Li, X., Ye, C., Lu, K., Wang, C., and Zhang, Y.: Accurately Predicting Spatiotemporal Variations of Near-Surface Nitrous Acid (HONO) Based on a Deep Learning Approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3673, https://doi.org/10.5194/egusphere-egu25-3673, 2025.
