Probabilistic Multi-Temporal Land Use and Land Cover Change Analysis Using Random Forest and Sentinel-2 Data in Haridwar, India

Multitemporal satellite images are useful for analyzing Land Use Land cover (LULC). It captures temporal images which are used to analyze the gradual transition which are difficult to detect with single date observations. This study focused on detecting the changes in LULC using multi-temporal approach for sentinel-2 images from 2017 to 2025.The approach obtains statistical results and different maps like the hard LULC maps, class wise probability maps, persistence, class to class change detection maps. Random Forest (RF) classifier has been used for classifying the stack of multispectral bands and different indices (Normalized Difference Water Index (NDWI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), Modified Normalized Difference Water Index (MNDWI), and Land Surface Water Index (LSWI), Normalized Difference Red Edge Index (NDRE)). Seasonal composites (January-April) were created to assess phenological regularity. Annual classification accuracy was high, ranging from 93.18% to 96.25% with Kappa coefficients of 0.92 to 0.96. Temporal Persistence analysis using a threshold greater than equal to six years to identify the stable classes. Statistics output indicated class specific temporal changes like vegetation phenology, hydrological dynamics, and land management techniques. The results show effectiveness of probabilistic multi-temporal analysis in capturing the gradual LULC changes.