Geospatial AI in earth observation for environmental monitoring: Analysing crop spectral responses with air quality variations during stubble burning episodes
Year: 2026
Type: article
Abstract: A comprehensive spectral library was developed to characterize the reflectance behaviour of key agricultural land cover types, cropland, fallow land, and built-up areas using Sentinel-2 Level-2A imagery and MODIS time-series data. Critical spectral bands (B4, B8, B11, B12) and vegetation indices (NDVI, NBR, SAVI, EVI) were analysed to distinguish burnt and non-burnt agricultural fields during stubble burning episodes. These spectral signatures were used as inputs to supervise machine-learning classifiers, including Random Forest and Support Vector Machine, to generate accurate multi-year burn area maps for 2024. Spectral separability was rigorously validated using Spectral Angle Mapper and Mahalanobis distance, ensuring reliable discrimination among land cover classes. The derived burnt-area datasets were integrated with air-quality indicators (PM₂.₅, PM₁₀, NO₂, CO, SO₂) obtained from the Copernicus Atmosphere Monitoring Service and ground-based CPCB stations. Temporal analysis revealed strong concurrence between peak burning periods (October–November) and elevated Air Quality Index values, particularly across major urban centres such as Ludhiana, Amritsar, and Delhi. Regression analysis and Pearson correlation coefficients quantified the relationship between burn intensity and air-quality deterioration. Additionally, NDVI time-series analysis indicated delayed vegetation recovery in areas subjected to repeated or intense burning, highlighting persistent ecological impacts. This integrated Geospatial AI–driven framework demonstrates the effectiveness of combining spectral libraries with air-quality data for monitoring residue burning and its environmental consequences. The findings support the need for strengthened residue-management policies, targeted farmer awareness programs, and sustained satellite-based monitoring to enhance agricultural sustainability and regional air quality across the Indo-Gangetic Plain.
Institutions: CEPT University, Texas Tech University
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7160702086