Surface ozone (O$$_3$$ 3 ) is primarily formed through complex photo-chemical reactions in the atmosphere, which are non-linearly dependent on precursors. Even though, there have been many recent studies exploring the potential of machine learning (ML) in modeling surface ozone, the inclusion of limited available ozone precursors information has received little attention. The ML algorithm with in-situ NO information and meteorology explains 87% (R$$^{2}$$ 2 = 0.87) of the ozone variability over Munich, a German metropolitan area, which is 15% higher than a ML algorithm that considers only meteorology. The ML algorithm trained for the urban measurement station in Munich can also explain the ozone variability of the other three stations in the same city, with R$$^{2}$$ 2 = 0.88, 0.91, 0.63. While the same model robustly explains the ozone variability of two other German cities’ (Berlin and Hamburg) measurement stations, with R$$^{2}$$ 2 ranges from 0.72 to 0.84, giving confidence to use the ML algorithm trained for one location to other locations with sparse ozone measurements. The inclusion of satellite O$$_3$$ 3 precursors information has little effect on the ML model’s performance.

Surface ozone (O3) is primarily formed through complex photo-chemical reactions in the atmosphere, which are non-linearly dependent on precursors (NOX and VOC). Exploring the potential of machine learning (ML) in modeling surface ozone has received little attention, particularly when it comes to the inclusion of limited available ozone precursors information in the ML model. The ML model with past O3 , meteorology (relative humidity, temperature, boundary layer height, wind direction), season type and in-situ NO information explains 87 % (R2 = 0.87) of the ozone variability over Munich, a German metropolitan area. The ML model trained for the urban measurement station in Munich can also explain the ozone variability of the other three stations in the same city, with R2 = 0.88, 0.91, 0.63. While the same model robustly explains the ozone variability of two other German cities’ (Berlin and Hamburg) measurement stations, with R2 ranges from 0.72 to 0.84, giving confidence to use the ML model trained for one location to other locations with sparse ozone measurements. In all cases, including coarse CAMS model O3 simulations in the ML model slightly improves the ML model’s performance in predicting surface ozone.