Assessing Flood Mapping with Pre- and Post-Flood SAR Imagery: Advances, Challenges, and Performance Evaluation

Synthetic Aperture Radar (SAR) imagery has become a crucial asset in real-time disaster response, providing uniform and reliable data acquisition regardless of weather or lighting conditions. This feature makes SAR especially useful for flood mapping even under dense cloud cover. This paper provides a detailed review of recent developments in SARbased flood mapping methods, comparing conventional thresholdbased approaches, machine learning methods, and deep learning architectures. By utilizing pre- and post-flood SAR images, we compare all methodologies in relation to accuracy, applicability to various flood situations and limitations. Close attention is paid to the influence of deep models like UNet and transformer models, along with hybrid techniques incorporating SAR and optical data. The comparative analysis above elucidates the advantages and disadvantages of each method, providing information on what is trending and where to proceed in future SAR-based flood mapping.