DB-SEN1FloodNet: A Deep Learning Based Approach for Flood Inundated Regions Detection using Sentinel-1 Synthetic Aperture Radar Satellite Datasets

Timely identification of floods is crucial for preserving lives and assessing damage. To ensure effective flood protection, it is imperative to have consistent real-time mapping of flooded regions. Floods often occur in specific weather conditions, accompanied by cloud cover due to excessive precipitation. Remote sensing, primarily using multi-spectral imagery from optical sensors and backscatter data from synthetic aperture radar (SAR), plays a vital role in delineating flood extents. SAR remote sensing, known for its all-weather capabilities, is widely adopted for flood mapping, providing continuous day and night coverage. This study utilized Sentinel-1 satellite datasets to map floods across 11 events, employing the DB-SEN1FloodNet model, achieving outstanding performance metrics.