Development of new spectral indices for mapping of crop residue burning using satellite remote sensing

Crop residue burning (CRB) is a widespread agricultural practice particularly in the Indo-Gangetic Plain (IGP) of India contributing significantly to air pollution and smog formation. However, accurate mapping using satellite data remains challenging due to small fire scales and short-lived burn scars. Ten spectral indices (CSI, MCRC, NDSVI, NDI5, NBR, NBR8a, STI, NDVI and NDTI) were evaluated to map CRB, and the accuracy of burning was defined using mean spectral plots, Spectral Discrimination Index (SDI) and Jeffries-Matusita (JM) distance. These indices showed poor separability to spectral discrimination of burnt fields. A new spectral indices was developed, i.e. Agricultural Burnt Discrimination Indices (ABDI) to enhance CRB detection using Sentinel-2 imagery. Various performance evaluation techniques were used, such as Burnt Area Classification using SVM techniques. Results show existing indices, e.g. ABDI, derived from the Tassel Cap Brightness component (TC1) and NDSVI achieved better class separation than existing indices and classification accuracy of 92.26% using Support Vector Machine (SVM), outperforming NDTI (70.83%), NDVI (81.25%) and STI (77.68%). Compared to global products MCD64A1 and C3SBA10, ABDI detected 35% and 98% more burnt area and identifying small, scattered fires. ABDI provides accurate emission estimates, which help in fire management strategies in agricultural landscapes.