Prediction of deforestation risk in north-east India: evaluating forest canopy density dynamics and spatial drivers through machine learning models

Rapid population growth in North-East India has led to urban sprawl, agricultural expansion, and industrialization, resulting in significant deforestation. Numerous studies have utilized Forest Canopy Density (FCD) coupled with remote sensing and GIS to assess forest cover changes. This study evaluates how FCD changed from 2000 to 2020, highlighting deforestation patterns in North-East India. Machine learning models, including Binary Logistic Regression (BLR), Random Forest (RF), REP-Tree, and XGBoost Regression (XGBR), were used to identify areas at risk of deforestation. Influencing factors considered were forest density, barren land, agricultural land, urban areas, distance to roads, and topographical characteristics. Results indicated that proximity to roads notably increases deforestation risk. Higher densities of agricultural and urban land also contribute to greater deforestation rates, whereas increased forest and barren land densities reduce this risk. Among the tested models, Random Forest showed superior performance with a high True Positive Rate (TPR) and low False Positive Rate (FPR), effectively identifying high-risk deforestation zones. Analysis also revealed a concerning shift from high-density to low-density forests, signifying substantial forest cover loss and potential threats to biodiversity and ecosystem services. The findings emphasize the need for integrated land-use planning and targeted conservation, specifically addressing road proximity, to effectively combat deforestation in North-East India and similar regions.