Rapid Coastal Flood Mapping with SAR data Using Random Forest Technique
Abstract
A rapid and accurate tool to delineate coastal flood extent is vital for decision makers to provide effective and protective measures which help save human lives and assets. Recent hazardous floods all around the world repeatedly highlight the pressing need for a state-of-the-art procedure for updating flood maps in the affected areas. Flood extent extraction from satellites imageries can dramatically expedite this process; however, images acquired by optical sensors are highly influenced by meteorological and diurnal cycles. This study aims at improving the flood extent extraction procedure by utilizing high-resolution Synthetic-aperture radar (SAR) data, regardless of weather condition and daytime. We use SAR imageries of before and during floods for generating a multi-temporal raster. This raster is segmented into polygons (objects) based on similar back-scattering properties and indices such as the Normalized Difference Flood Index (NDFI), and the Normalized Difference Flood in short Vegetation Index (NDFVI). Then, we develop a Random Forest (RF) model to classify flooding regions. This process is automated and tested for the 2020 floods in Mobile Bay area, Alabama during Hurricane Sally. We compared the performance of our proposed SAR-based flood mapping approach with alternative methods and found the current approach more effective. The predictive accuracy as well as computational efficiency of the proposed approach supports its applicability for reliable flood hazard risk mapping.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H35I1138H