Integrated Wetland Mapping and Surface Water Quality Assessment Using Sentinel-1/2 and Machine Learning: A Case Study from Sidi Moussa–Oualidia, Morocco

Wetlands are vital ecosystems as they play a key role in agriculture, providing essential resources such as water for crops, livestock, and aquaculture, while also serving as habitats for a diverse range of species, particularly wild birds. Our study focuses on the Sidi Moussa–Oualidia wetland complex, integrating Sentinel-1 and Sentinel-2 imagery with spectral indices, radar backscatter (VV/VH ratio), and topographic features to classify land cover and monitor surface water quality. A Random Forest classifier optimized via Recursive Feature Elimination (RFE) achieved 91% accuracy and a macro F1-score of 0.90 across six classes, including permanent water, salt marshes, artificial marshes, hypersaline zones, oyster farming areas, and others. Surface water quality was also evaluated using the turbidity index as a proxy, revealing notable spatial degradation near aquaculture and agricultural activity hotspots. Our first-of-its-kind study in the Moroccan context proposes a scalable, reproducible methodology for simultaneous wetland land cover mapping and water quality monitoring, reinforcing the importance of remote sensing for integrated wetland-agriculture management in data-limited regions.