Comparative modelling assessment of land use and land cover data for the last five decades in Kumaun Lesser Himalaya

Abstract Accurately land use and land cover (LULC) mapping in mountainous regions is inherently challenging due to the complex topography, terrain-induced spectral distortions, and heterogeneous landscape characteristics. Overcoming these challenges requires the integration of advanced remote sensing techniques, GIS, terrain correction algorithms, and careful field-based validation. This study systematically evaluated and compared multiple pixel-based and object-based classification approaches using Sentinel-2 imagery (2020) to identify the most suitable method for accurate LULC mapping in the Kumaun Lesser Himalaya, India. Six land cover classes: water, barren land, settlement, agriculture, forest, and grassland were classified and assessed. Among the assessed approaches, the object-based Random Tree classifier produced the highest classification accuracy, demonstrating its effectiveness for mapping in discriminating heterogenous land cover types within complex mountainous landscapes. The optimal classifier model was subsequently employed to analyse long-term LULC dynamics from 1972 to 2020. The analysis showed a decline in forest (11.08%), agriculture land (17.04%), and water bodies (0.95%), while barren land, settlements, and grassland increased by 1.76%, 0.93%, and 26.40%, respectively. The marked expansion of grassland is likely associated with agricultural land abandonment, forest degradation, and rural-to-urban migration. These long-term landscape transformations indicate increasing anthropogenic pressures and environmental change in the fragile Himalayan ecosystem, with potential consequences for biodiversity conservation, ecosystem stability, and watershed functions. The findings provide a scientific basis for long-term environmental monitoring, biodiversity conservation, sustainable land-use planning, ecological restoration, and informed environmental management in the Kumaun Lesser Himalaya.