Sentinel-1 SAR and LiDAR to detect extent and depth flood using Random Forests machine learning

This research was carried out to identify the extent and depth of flooded areas using Sentinel-1 SAR, the Digital Elevation Model generated with LiDAR and Random Forest machine learning. Training and cross-validation was performed on a set of backscatter value samples obtained from Sentinel-1. The results indicate that out of five combinations, the Random Forest algorithm had the best performance when using the four combinations (RF + Polarization VH+VV + MDE) with F1m = 0.977, AUC = 0.998 and Kappa = 0.955. SAR images have potential advantages that allow rapid and efficient diagnosis of the extent of flooding caused by excess rainfall in many regions around world.