Scalable Land Cover Classification with Linear Kernel SVM on Sentinel-2 Imagery via GEE: A Case Study

Performance evaluation of the linear kernel SVM for land cover classification using the GEE platform in Telangana, India (Longitude: 79.78E Latitude: 7.78N) is presented in this paper. High-resolution Sentinel-2 imagery of January 2020 to October 2024 was classified for four major land covers, namely Agriculture, Buildings, Bareland, and Water, by considering the training and testing based on the spectral bands B2, B3, B4, B8, B11, and B12. The overall accuracy with SVM is found to be 97.7%, and the Kappa coefficient is 0.968, which says that the model is reliable enough. The class-wise metrics— precision, recall, and F1 scores—are 0.95, 1.00, and 0.97 for the Agriculture class; 1.00, 1.00, and 1.00 for Buildings; 1.00, 1.00, and 1.00 for Bareland; and 0.88, 0.88, and 0.88 for the Water class. Overall, the Precision value is 0.96, Recall is 0.97, and the F1 score is 0.96. The present research has brought to light scalability and efficiency for geospatial analysis regarding GEE, opening up a whole new perspective on further applications in urban planning, environmental monitoring, and sustainable development. This study has also indicated that incorporating multi-temporal satellite data and nonlinear kernels increases classification accuracy significantly.