Floods are natural hazards with severe impacts, and their frequency and intensity are increasing due to climate change. Synthetic Aperture Radar (SAR) satellites are widely used for flood mapping, as they operate independently of weather and time of day. This thesis examines the potential of SAR-based flood monitoring through a case study of the July 2021 flood in Limburg, the Netherlands. Comparing Capella Space on-demand X-Band imagery (sub-meter resolution) with Sentinel-1 open-source C-Band imagery (5 m x 20 m resolution). The Water-Land Boundary was determined by estimating flood extent and water levels, with multiple methods evaluated for the distinct products from both data sources. Capella Space provided a single image acquisition at the flood’s peak, with a thresholding method used to classify flooded pixels. For Sentinel-1, an Amplitude Time Series Analysis (ATSA) was applied to data from 2017 to 2024 to identify flood-related outliers. The water levels are estimated from flood extent edges with the national LiDAR DEM (AHN4).
Evaluation of the modeled results using an error matrix at the acquisition time showed that 67% and 68% of the pixels were correctly classified from the flood extents derived from Capella Space and Sentinel-1, respectively. The maximum flood extent from Sentinel-1 data decreased to 45% correct classification when compared to the modeled results at the peak of the flood. This is consistent with the acquisition times, which were taken before and two days after the flood peak, missing the peak flood moment. SAR-based water levels showed an overall precision of 0.141 m for Capella Space and 0.156 m for Sentinel-1. Agreement with water level gauge measurements was better in flatter, less vegetated areas and lower in steep, vegetated areas. Achieving consistent 20 cm water level accuracy (as required by the Dutch Ministry of Infrastructure and Water Management), across the study area remains complex. Both methods are prone to false positives and negatives, especially in areas with steep slopes, narrow canals, high vegetation, or roads. False classifications result in inaccurate flood extents, thus decreasing water level accuracy. Higher resolution of Capella Space images provided better alignment with the maximum flood extent, while Sentinel-1 images have wider coverage but missed the timing of the flood peak. In the end, the choice of SAR-system depends on timing, surface characteristics, and mapping extent needs.