Leveraging Sentinel-2 Multispectral Data and Machine Learning Algorithms for Land Use Land Cover Mapping in Semi-Arid Regions

Land use and land cover (LULC) classification is a fundamental approach for landscape pattern mapping and monitoring of natural ecosystems and anthropogenic activities and is, therefore, an essential base for resource management to support sustainability. The land cover classes were classified in this study using multispectral data. The study utilizes Sentinel-2 data captured in February 2024, and a median composite image was generated using the Google Earth Engine (GEE) platform. Three different classification algorithms, random forest (RF), classification and regression tree (CART), and naive bayes (NB), were used to produce LULC maps. The algorithms are efficient in capturing different land cover patterns with spectral diversity. The performance with essential measures such as overall accuracy (OA) and Kappa coefficient (KC) has been evaluated to assess classification outcomes. Among these classifiers, the RF classifier with the maximum OA (96.91%) and KC (96.10%) exhibited optimal results and robustness for LULC classification. CART followed this, with OA of 95.88% and KC of 94.81%. NB achieved an OA of 85.57% and KC of 81.93%. It confirms the validity of the suggested method for obtaining LULC maps through Sentinel-2 data and ML algorithms, achieving maximum accuracy in semi-arid regions.