Land Cover Mapping of Boracay Island, Philippines Using Remote Sensing and Machine Learning

Urban expansion has been a threat to the global world with rapid urbanization threatening environmental resources, water supplies, waste management, and public health. Maps of Land Use/Land Cover (LULC) are used to manage this urbanization by visually mapping the anthropological use and land types in a region of interest. The use of supervised machine learning algorithms is one technique to generate LULC maps. This study shows three LULC maps generated for the island of Boracay in Aklan, Philippines using K-Nearest Neighbor (KNN), Random Forest (RF), and Gaussian Mixture Model (GMM). The data was acquired via the Copernicus Programme from Sentinel-2 satellites and were pre-processed, annotated, and trained using various machine learning models. Each model resulted in a high overall accuracy with the RF model achieving 98.95%, KNN model achieving 98.54%, and the Gaussian Mixture Model achieving 97.15%. In this application, the best performing model was the Random Forest model for this application. The built-up versus non-built-up maps also highlight the abundance of built-up areas in a globally renowned tourist location such as the Boracay Island.