Sentinel-1 Polarization Comparison for Flood Segmentation Using Deep Learning

Flood is one of the most damaging natural hazards, and timely detection of it is very important to save human lives and assess the level of damage. Floods usually occur in certain weather conditions such as excessive rainfall, which makes the presence of clouds in the sky of the region very likely. For this reason, radar-based sensors are the most suitable choice for real-time flood mapping. In the present study, the ETCI 2021 flood event detection competition dataset, organized by the NASA Advanced Concepts and Implementation Team in collaboration with the IEEE GRSS Geoscience Informatics Technical Committee, has been used. Moreover, we have utilized the U-Net architecture as a segmentation model to map flooded regions. This study aims to identify flooded areas from radar images of the study area in two different polarizations. By examining and comparing the obtained results, it was observed that the network designed to identify flooded areas in VV polarization has made better predictions and the Intersection Over Union (IOU) score has improved from 64.46 to 67.35 compared to VH polarization.