Land Use/Land Cover Classification of Rewa using Supervised Machine Learning ( 2025 )

This study presents a supervised machine learning–based Land Use/Land Cover (LULC) classification of Rewa district, Madhya Pradesh, for the year 2025, utilizing Sentinel-2 MSI Level-2A imagery within the Google Earth Engine (GEE) cloud computing environment. A Random Forest classifier was trained using labeled samples representing five land cover classes—Agriculture, Barren, Forest, Urban, and Water—derived from multispectral bands (B2, B3, B4, and B8) at 10 m spatial resolution. Preprocessing steps included cloud masking and median compositing to ensure data quality. The classified output was exported and further processed in QGIS for cartographic representation and spatial analysis.
Accuracy assessment yielded an Overall Accuracy of 94.49% and a Kappa coefficient of 0.927, indicating strong classification reliability. Results reveal that agriculture dominates the district, covering approximately 65% of the total area, followed by barren land (≈16%) and forest cover (≈15%), while urban and water bodies constitute minor proportions. Spatial patterns correspond closely with the district’s physiography, agro-climatic conditions, and settlement structure. The study demonstrates the effectiveness of integrating cloud-based geospatial processing with supervised classification techniques for generating high-accuracy thematic maps to support regional planning, resource management, and environmental monitoring.