Predicting ozone formation in petrochemical industrialized Lanzhou city by interpretable ensemble machine learning

Revealing the Covariation of Atmospheric O2 and Pollutants in an Industrial Metropolis by Explainable Machine Learning

“…We trained the ML-based O 2 simulator using hourly air quality and meteorological parameters from January 2021 to December 2021. Hourly observations of air quality and meteorological parameters were obtained from the Lanzhou Atmospheric Components Monitoring Superstation (LACMS; 36.05°N, 103.87°E). ,, The air quality data include concentrations of particulate matter (TSP, PM 10 , PM 2.5 , and PM 1 , reported in micrograms per cubic meter), particulate carbon (TC = EC + OC, reported in micrograms per cubic meter), and trace gases (NO x , SO 2 , and CO, reported in micrograms per cubic meter). Instead of directly including TSP, PM 10 , and PM 2.5 , we introduced PM 1–2.5 , PM 2.5–10 , and PM >10 into our model.…”

“…O 2 is the only reactant in fossil fuel combustion processes [C x H y + ( x + y /4)­O 2 → x CO 2 + ( y /2)­H 2 O] that can be directly measured in the atmosphere. In addition, O 2 is essential for vital bodily functions such as nerve and brain function, muscle movement, digestion, and hormone synthesis. , Elucidating covariations of O 2 and pollutants reveals drivers of pollution episodes and informs mitigation strategies for air pollution impacts. − …”

“…Machine learning (ML) models have been demonstrated to be a powerful tool for reconstructing, simulating, and predicting atmospheric pollution, including PM 2.5 , − O 3 , , NO x , , etc., outperforming finely designed chemical transport models . The use of ML models provides greater flexibility and efficiency when utilizing real-world data and is especially adept at revealing complex and hidden nonlinear correlations , that might not be easily identified using traditional physical models, providing new insights into the underlying mechanisms of the studied phenomena .…”

“…The SHapley Additive exPlanations (SHAP) framework, for instance, offers insights into the impact of a feature on model outcomes . ML models combined with explainable tools have been extensively used in various aspects of air pollution modeling, ,,, yet their application in reproducing variations in O 2 concentrations, particularly in urban settings, has been rather scant.…”

“…Hourly observations of air quality and meteorological parameters were obtained from the Lanzhou Atmospheric Components Monitoring Superstation (LACMS; 36.05°N, 103.87°E). 9,26,30 The air quality data include concentrations of particulate matter (TSP, PM 10 , PM 2.5 , and PM 1 , reported in micrograms per cubic meter), particulate carbon (TC = EC + OC, reported in micrograms per cubic meter), and trace gases (NO x , SO 2 , and CO, reported in micrograms per cubic meter). Instead of directly Atmospheric O 2 concentrations were measured on the rooftop of a building on the campus of Lanzhou University (∼300 m from LACMS), using a gas chromograph (7890B, The orange shade indicates the potential dust source area (barren or sparsely vegetated), derived from the MODIS land cover data (MCD12C1 version 6.1).…”