Tropospheric Ozone Estimation Using Machine Learning Of Remote Sensing Data: a Case Study of Delhi NCR

Tropospheric ozone, an adverse secondary pollutant due to the photochemical reactions between nitrogen oxides and VOCs, is an arising health emergency. In 2021, ozone pollution caused more than 489,000 casualties globally, with the probability of increasing dynamically by 2050. In India, the death toll is 24.1 per 100,000 people, with Delhi NCR exceeding permissible limits multiple times- Gurugram, for instance, surpassed the 100 ppb limit in May 2024. CPCB provides valuable localized data on air pollutants. This study harnessed remote sensing to inscribe the need for continuous, wide-ranging monitoring. The study utilizes the Random Forest model, incorporating the TROPOMI-derived ozone, tropospheric NO2, HCHO, and MODIS LST to estimate the ozone. The model globally estimates tropospheric ozone accurately for different monitoring stations and best for Okhla Phase 2 station ($R^{2}=87 \%$) in Delhi, highlighting its ability to capture complex and non-linear relationships.