Satellite-AOD and Machine Learning for Urban Air Quality and HealthRisk Assessment in the Indo-Gangetic Plain

Satellite based remote sensing techniques have proved effective in estimating critical pollutants concentration especially in regions which lack spatial coverage due to limited ground-based monitoring stations. This study presents a comprehensive, data-driven framework for evaluating urban air quality and associated health risks across the Indo-Gangetic Plain (IGP), a critically polluted region characterised by limited ground-based monitoring coverage. High-resolution satellite observations namely Aerosol Optical Depth (AOD) from MODIS-MAIAC and tropospheric ozone (O₃) from Sentinel-5P TROPOMI were integrated with meteorological parameters to estimate surface-level concentrations of PM_2.5, PM₁₀, and O₃ for 16 major cities across the IGP during the period 2020–2024. A Random Forest (RF) modelling approach demonstrated strong predictive performance (R² = 0.94 for PM_2.5/PM₁₀ and 0.84 for O₃; RMSE = 8.03–14.02 µg m^-3; Index of Agreement > 0.96). Estimates of the relative risk (RR) of mortality attributable to long-term PM_2.5 exposure indicated a substantial health burden in cities such as Delhi, Patna, and Kanpur, highlighting the pressing need for targeted mitigation and intervention strategies. The integrated satellite-machine learning framework effectively identifies pollution hotspots, enables robust exposure assessment, and addresses critical data gaps, thereby strengthening the scientific basis for informed decision-making. The findings provide actionable insights for the development of evidence-based, region-specific clean air action plans, contributing to enhanced urban liveability, improved environmental governance, and greater social equity. The novelty of this work lies in the combined use of satellite-derived AOD, TROPOMI-based O₃ observations, meteorological variables, and machine learning techniques to simultaneously predict PM_2.5, PM₁₀, and O₃ concentrations and assess associated health risks across the IGP. By advancing progress towards Sustainable Development Goals 3.9, 11.6, and 13, this research supports the transition towards healthier, more resilient, and sustainable urban environments in South Asia.

Keywords: Aerosols, Satellite Observations, Predictive Modelling Framework, Relative Risk