Shifting cultivation induced burn area dynamics using ensemble approach in Northeast India

•Multiple spectral indices allow better identification of burned area than single proxy.•Ensemble modeling approach indicated profound accuracy in burned area mapping.•Indices on vegetation, burned area and leaf water indicated superior performance. Identifying shifting cultivation areas and assessing their spatio-temporal dynamics are essential in framing climate-adaptive policies for efficient forest management and agriculture practices for the benefit of people. The current study attempts to develop an alternative approach to classify the shifting cultivation areas using an ensemble technique, integrating multiple spectral indices in three states of northeast India (NEI), such as Assam, Manipur, and Meghalaya. The adopted approach integrates green cover and leaf water content changes during shifting cultivation land preparation in Landsat imagery. The deforested burned area patches were identified based on threshold values using Landsat data-derived indices on vegetation, burned area and leaf water, and digital elevation model (DEM). The ensemble approach provided shifting cultivation maps with good overall accuracy (> 83%). The maximum shifting cultivation area was observed in Assam (126.87 km2), followed by Meghalaya (51.53 km2) and Manipur (46.04 km2) in 2016. The normalized difference vegetation index (NDVI) and NDVI difference performed better than other vegetation indices. The ensemble approach can be applied in other regions with minor modifications in threshold values, thus having the potential for accounting to shifting cultivation dynamics on an operational basis. Future research may include blending local traditional knowledge and modern scientific solutions for improved forest and land resources planning for the benefit of inhabitants and the mountain environment under the climate change scenarios.