Machine Learning-Based Land Cover Classification and Environmental Change Detection for Dam Monitoring at Suleja Dam Using Remote Sensing Data

Abstract

Effective dam monitoring is essential for sustainable water resource management and environmental protection. This study integrates machine learning, remote sensing, and Geographic Information Systems (GIS) to assess environmental changes within the Suleja Dam catchment, Nigeria. Multi-temporal Landsat 8 and Sentinel-2 imagery acquired between May and October of 2024 and 2025, together with climatic variables including precipitation, surface temperature, soil moisture, evaporation, and terrain data, were analyzed to evaluate land-cover dynamics. The Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) were used to monitor vegetation and surface water changes, while Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) classifiers were employed for land-cover classification. Classification performance was evaluated using confusion matrices and standard accuracy metrics. RF achieved the highest overall accuracy (92.83%; Kappa = 0.904), outperforming SVM (88.67%; Kappa = 0.849) and ANN (85.00%; Kappa = 0.800). Environmental change detection revealed a 6.5% decline in vegetation cover and increases in built-up areas (5.8%), water extent (3.2%), and bare surfaces (2.1%). The results demonstrate that integrating machine learning with remote sensing provides a robust and cost-effective framework for continuous dam monitoring, environmental change detection, and sustainable watershed management.