Sentinel-2-Based Land Use and Land Cover Classification Using Random Forest in Google Earth Engine

Land Use and Land Cover (LULC) classification is essential for environmental monitoring and land management. This study employs Sentinel-2 imagery and Random Forest (RF) classification in Google Earth Engine (GEE) to map LULC in Rangpur, Bangladesh. Sentinel-2 images from January to June 2024 were preprocessed using cloud filtering and median compositing. Spectral indices— NDVI, NDBI, BSI, and NDWI—were computed to improve class separability.A Random Forest classifier was trained using five LULC classes: water, cropland, vegetation, bareland, and built-up. The dataset was split into 80% training and 20% testing, and the model was optimized by testing different numbers of trees (20 to 600). The classification achieved an overall accuracy of 92.5% and a kappa coefficient of 0.89, confirming its reliability. Area estimation was performed to quantify the spatial distribution of each LULC class.The results demonstrate that integrating Sentinel-2 spectral indices with machine learning in GEE enables efficient and accurate LULC mapping. The use of GEE’s cloud-based processing eliminates hardware constraints, allowing scalable, real-time analysis. This study highlights the potential of remote sensing and machine learning for automated LULC classification with high accuracy.