Exploring Sentinel-1 and Sentinel-2 diversity for flood inundation mapping using deep learning

Identification of flood water extent from satellite images has historically relied on either synthetic aperture radar (SAR) or multi-spectral (MS) imagery. MS sensors are limited to cloud free conditions, whereas SAR imagery is plagued by noise-like speckle. Prior studies that use combinations and data overcome individual limitations these have not fully examined sensitivity mapping performance different derived spectral indices band transformations in color space. This study explores the diverse bands Sentinel 2 (S2) through well-established 1 (S1) along with their assess capability for generating accurate inundation maps. The robustness S-1 S-2 was evaluated using 446 hand labeled spanning across 11 events Sen1Floods11 dataset which highly terms land cover as well location. A modified K-fold cross validation approach used evaluate 32 S1 S2 a connected deep convolutional neural network known U-Net. Our results indicated usage elevation information improved produce more Compared median F1 score 0.62 when only bands, combined led an 0.73. Water extraction based statistically significant superior comparison S1. Among all combinations, HSV (Hue, Saturation, Value) transformation provides 0.9, outperforming commonly owing HSV’s transformation’s contrast distinguishing abilities. Additionally, U-Net algorithm able learn relationship between raw corresponding but relatively complex computation involved latter. Results paper establishes important benchmarks extension data-based efforts over large spatial extents.