Answers are grounded in this paper and cite their sources.
Study estimates ozone and AQI in India, indicating the effectiveness of satellite data.
The study develops a CPCB-compliant, dual framework for modeling the Air Quality Index (AQI) under the National Air Quality Index (NAQI) system in India, with a clear distinction between the deterministic ozone sub-index and the multi-pollutant AQI forecast. Firstly, Arm-S constructs a lightweight surrogate model for the ozone sub-index using statistical descriptors derived from Sentinel-5P TROPOMI observations, specifically PM 2.5 and the interquartile range (IQR), along with temporal features. The Arm-S component is formulated as a surrogate mapping framework that evaluates whether satellite-derived statistical descriptors can approximate the regulatory ozone AQI sub-index transformation defined by CPCB guidelines, rather than serving as an independent predictive model. The ozone sub-index target is computed using the CPCB breakpoint transformation and validated against ground measurements obtained from the CPCB monitoring station at Jawaharlal Nehru Stadium, New Delhi. Both validation and cross-validation experiments demonstrate strong agreement, achieving an R² value greater than 0.98, indicating a near-linear scaling relationship between ozone concentration and its corresponding sub-index. Secondly, Arm-C generates a composite AQI forecast based on CPCB multi-pollutant ground data spanning 2018 to 2024. The modeling framework incorporates lag-based features, rolling statistical measures, and meteorological variables, evaluated using a strict forward-chaining validation protocol. Random Forest and XGBoost models achieve an R² score of approximately 0.86 and a band-level classification accuracy of 71.5% for the 2024 test year, outperforming the naive persistence benchmark. Correlation analysis between satellite-derived and surface-monitored data reveals moderate and statistically significant correlations for NO ₂ ( r ≈ 0.44 ) and CO ( r ≈ 0.43 ), whereas comparatively weaker correlations are observed for particulate matter and ozone.
No takes yet. Share an insight, caveat, or question.
Hasani et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: