AUTHOR=Gebrehiwot Asmamaw , Hashemi-Beni Leila 

TITLE=3D Inundation Mapping: A Comparison Between Deep Learning Image Classification and Geomorphic Flood Index Approaches

JOURNAL=Frontiers in Remote Sensing

VOLUME=Volume 3 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/remote-sensing/articles/10.3389/frsen.2022.868104

DOI=10.3389/frsen.2022.868104

ISSN=2673-6187

ABSTRACT=Remote sensing has been an effective tool for interpreting and analyzing water bodies and detecting floods over the past decades. Flood detection is one of the main remote sensing themes because of its nature of destructiveness and the need of timely and accurate geospatial data for hazard mitigation, emergency management. This research aims to investigate the performance of a deep learning-based flood depth mapping method that only employs flood extent and topographic data. The approach was applied to investigate the flooding due to Hurricane Florence in Princeville along the Tar River region in 2018. Our deep learning approach involves three steps: (1) flood extent mapping, (2) three-dimensional (3D) water surface generation, and (3) floodwater depth estimation. A flood extent map is generated using deep learning from optical UAV images. Then, the outputs are validated using the USGS gauge water level data and compared with the Geomorphic Flood Index (GFI) for floodwater depth map. The deep learning method demonstrated a better performance with an RMSE of 0.26 m for water depth. This approach is efficient on creating a 3D flood extent map at different scale to support emergency response and recovery activities during a flood event.