A Machine Learning based Random Forest Model for Surface Ozone in India and its impact on Health and Economy

Surface ozone is a major photochemical oxidant that negatively impacts air quality, public health and ecosystems. This study investigates the spatiotemporal variability of surface ozone utilising ground-based data from the Central Pollution Control Board (CPCB) for the year 2022. Nevertheless, these measurements are confined to metropolitan and semi-urban regions in India, leading to conflicting data across numerous areas, and thus, limit precise evaluation. To overcome this data deficiency, we utilise available observations and implement a Random Forest (RF) machine learning technique to generate a spatially resolved dataset of surface ozone in India. Cross-validation results provide a robust predictive efficacy, with an average correlation coefficient of 0.95 and a root mean square error (RMSE) of 5.024 µg/m³ across 38,527 samples. Regional accuracy varies, where the Indo-Gangetic Plain (IGP) and Central India (CI) show a greater correlation and smaller RMSE, whilst North-East India (NEI) shows the lowest accuracy. Surface ozone concentrations reach 40–50 µg/m³ in the pre-monsoon season (March-April-May) and decline to 10–30 µg/m³ during monsoon, whereas the urban regions show elevated levels in winter. Our study indicates that the variability of surface ozone is predominantly influenced by meteorological factors and regional sources. The population risk index for surface ozone is notably higher in IGP, particularly in West Bengal. These findings emphasise the need for region-specific air quality strategies and stress the significance of continuous programs such as the National Clean Air Program (NCAP) to successfully address and reduce surface ozone pollution throughout India.