Google Earth Engine Based Comparative Assessment of Supervised Classifiers for Land Cover Mapping

This study examined the effectiveness of supervised machine learning algorithms for land use/land cover (LULC) classification in the Ajmer region, using Landsat 8 imagery from December 1 to December 31, 2023. Six algorithms were tested: Classification and Regression Trees (CART), k-Nearest Neighbors (KNN), Random Forest (RF), Naive Bayes (NB), Smile Gradient Tree Boost (SGB), and Minimum Distance. Google Earth Engine (GEE) facilitated data processing, enhancing workflow efficiency and reproducibility. The study evaluated overall accuracy, with CART achieving the highest accuracy at 80.57%, followed closely by KNN, RF, NB, SGB, and Minimum Distance with accuracies of 78.68%, 79.89%, 76.79%, 77.14%, and 64.68% respectively. CART emerged as the standout performer for accurately classifying land cover types in Ajmer. This study emphasized the application of contemporary machine learning algorithms to develop precise land cover mapping for urban planning, environmental monitoring, and climate change studies, promoting informed decision-making for sustainable development in Ajmer other areas.