Effectiveness of data manipulation techniques in improving classification performance : simple layer stacking, principal component analysis, and three targeted spectral indices

The study investigated long-term land use/land cover (LULC) dynamics in Mbabane City and evaluated their implications for sustainable urban development. Medium-resolution satellite imagery was processed on the Google Earth Engine (GEE) platform to generate multi-temporal composite datasets by integrating optical bands and derived spectral indices to improve class separability. The principal component analysis (PCA) technique was applied prior to ensemble machine learning classification to reduce dimensionality and enhance model performance. This integrated framework enabled the production of high-accuracy LULC maps and robust detection of urban expansion patterns over time.The research further quantified Sustainable Development Goal (SDG) indicator 11.3.1 (Land Consumption Rate to Population Growth Rate), between 1997 and 2017, providing a significant assessment of urban land-use efficiency for Mbabane City. The findings offer critical decision-support information for urban planners and policymakers seeking to promote sustainable urban developmentOverall, the thesis presents a scalable and reproducible framework for urban monitoring in Eswatini and demonstrates the value of integrating multi-sensor satellite data and machine learning within cloud-based platforms to support evidence-based SDG reporting and urban policy, particularly in sub-Saharan Africa.