ANNOUNCEMENTS
This study focuses on assessing wheat stubble burning in Vidisha district, Madhya Pradesh, by integrating multi-temporal satellite data from Sentinel-2 and MODIS with machine learning techniques. Stubble burning is a common post-harvest practice that contributes significantly to air pollution, soil degradation, and climate change. The research aims to accurately map and quantify burned areas to provide valuable insights for environmental management and policy.
The methodology involved preprocessing Sentinel-2 Level-1C images to Level-2A for atmospheric correction and utilizing the MODIS MYD14A1 active fire product to detect thermal anomalies indicating fires. Normalized Difference Vegetation Index (NDVI) time series were generated to monitor wheat crop growth stages and improve the precision of wheat field delineation. Random Forest classifier is used for classifying the stubble burning area. To train the model a combination of points derived from visual interpretation of False Colour Composite (FCC) images from sentinel-2 and MODIS fire points, validation was done using ground truth points.
To enhance classification accuracy, non-agricultural areas were masked using auxiliary land cover data.
Results revealed significant stubble burning during the peak months of April and May, with up to 91% of harvested wheat fields affected in some years. Spatial analysis showed varying degrees of burning across different sub-districts, with Basoda and Gyaraspur exhibiting higher incidences. Accuracy assessments using confusion matrices and classification metrics confirmed the robustness of the approach, demonstrating high overall accuracy and reliability in burned area detection.
The integration of optical and thermal satellite data with machine learning provides a scalable, cost-effective solution for continuous monitoring of agricultural fires. This study underscores the importance of satellite-based approaches in supporting sustainable agricultural practices and informing policy to reduce environmental damage caused by stubble burning.
Keywords: Wheat stubble burning, Sentinel-2, MODIS MYD14A1, Random Forest Classification, NDVI.
