Modelling of near-surface NO2 and O3 concentration over Germany using machine learning 

<p>Chemical transport models (CTMs) are commonly used to model air pollutant concentrations. CTMs, on the other hand, require a lot of computing power and sometimes yield biased findings that result from emission inventories and chemical mechanisms employed. Machine learning algorithms are used in a wide range of fields, including Earth system science. Its popularity stems from its ability to learn complex non-linear relationships. As a follow-up of our previous study [1], we attempted to deduce the capability of Machine Learning (ML) in modelling air pollutant concentrations.</p> <p>In this study, we employed the Gradient Boosted Tree (GBT) algorithm to model near-surface NO₂ and O₃ over Germany at 0.1 degree resolution and daily intervals. The GBT model is trained using TROPOMI satellite column NO₂, O₃, HCHO data, as well as meteorology and&#160;road density as an information for NO_X emission sources. Government air quality (NO₂ and O₃) observations from urban, suburban, and background stations are used as target variables; 321 stations are considered for NO₂ ML model training and 256 stations are considered for O₃ ML model training. The GBT model trained for near-surface NO₂ explains 68-88% of observed concentrations, whereas, for near-surface O₃, the GBT model explains 74-92% of observed concentrations.&#160;</p> <p>Road density and TROPOMI NO₂ data are the most important features in the fitted model for near-surface NO₂. This is due to the fact that road density (a proxy for traffic) is the main source of near-surface NO_X emission, and the TROPOMI tropospheric NO₂ column is a good representation of near-surface NO₂ concentration. The downward UV radiation (DUV) at the surface and temperature are the most important features in the fitted model for near-surface O₃. Since O₃ is formed from the photolysis of NO₂, DUV plays an important role in the fitted model for O₃. Temperature is the driver of biogenic Volatile Organic Compounds (VOCs), which are an important precursor to O₃.&#160;</p> <p>In all cases, the GBT model outperforms feed-forward neural networks. Furthermore, the developed GBT model for near-surface O₃ is reliably transferable to other locations and countries (R²=0.87-0.94), whereas the developed model for near-surface NO₂ is moderately transferable (R²=0.32-0.68). The reason could be that the road density is not the best representative of traffic NO_X emissions and can be improved in a future study. Overall, we developed a new machine learning model to cost-effectively model the near-surface NO₂ and O₃ concentrations, which could help us to better understand the air pollution distribution at a moderate resolution.</p> <p><strong>References:</strong></p> <p>Balamurugan, V., Balamurugan, V. and Chen, J., 2022. Importance of ozone precursors information in modelling urban surface ozone variability using machine learning algorithm.&#160;<em>Scientific reports</em>,&#160;<em>12</em>(1), pp.1-8.</p>