Integration of Sentinel-1 SAR and Sentinel-2 Multi-Spectral Data for Cloud-Resilient Monitoring of Sediment Plume Persistence in Macajalar Bay

Monitoring sediment plumes is needed to better understand coastal degradation and transport of pollutants. This is a very effective task using optical satellite imagery but is most limited by the significant amount of cloud cover associated with storm events. This study combines Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical data to create a cloud-resilient monitoring system for Macajalar Bay, Philippines, following the devastating flooding of the Iponan River in October 2025. We explore supervised classification methods in traditional GIS, such as Maximum Likelihood and Minimum Distance, and compared them to Random Forest under a cloud-based environment, performed in Google Earth Engine (GEE). We found the "Accuracy Paradox" in traditional pixel-based radar classification, obtaining an Overall Accuracy of 90.53% and a Kappa coefficient of 0.0741 because of the great class imbalance in the evaluation of unmasked scenes. The multi-sensor Random Forest approach, on the other hand, was able to use the physical changes in the water's di-electric constant and surface roughness to achieve high classification accuracy and provide a continuous timeline of plume persistence and decay, even in the presence of clouds. This integrated approach demonstrates the potential of multi-sensor remote sensing for supporting near-real-time coastal water quality assessment and disaster response planning in tropical environments.