Assessing the impact of urban expansion, deforestation and long-term climate variations in the Himalayan foothills: A machine learning approach

This study assesses decadal climate-land cover interactions in Dehradun (2014–2025) using Landsat and CHIRPS/ERA5 data. Utilizing multi-index feature engineering (NDVI, SAVI, NDBI, MNDWI) to enhance class separability by 37%, the research compares three machine learning models: CART, Random Forest (RF), and K-Means. CART achieved the highest overall accuracy (99.88%), significantly outperforming K-Means (89.2%). Findings reveal urban expansion of 33.0% (150.96 to 200.73 km²), a 6.0% decline in water bodies, and a 1.6% loss in dense forest, with 420 hectares lost specifically in the Lacchiwala Range. Climate analysis shows that the temperature will rise by 1.2°C, and that warming will speed up to 0.21°C/year after 2020, which is twice the global average. These shifts have created a 2.3°C urban-forest thermal gradient and reduced evaporative cooling by 18%. The study advocates for adaptive governance and forest corridor restoration to mitigate hydrological vulnerability in the Himalayan foothills.