Authors: Dan Tian, Hongxing Liu, Sagy Cohen, Tantu Mandal – The University of Alabama; Lei Wang – Louisiana State University
Title: Rapid Automated Flood Extent and Depth Mapping with SAR and Optical Remote Sensing Imagery: Toolsets and Flood Map Products
Presentation Type: Poster Presentation
Abstract: Accurate, rapid, and standardized flood inundation mapping is critical for hazard assessment, flood forecasting, emergency response, and post-flood damage analysis. This project develops an end-to-end remote sensing framework for generating operationally ready flood extent and water depth products from multi-sensor satellite observations. The framework integrates satellite synthetic aperture radar (SAR), such as Sentinel-1, high-resolution optical imagery, such as PlanetScope, and terrain-derived information from high-resolution and national-scale DEMs, such as LiDAR-derived DEMs and USGS 3DEP products, to improve flood detection accuracy and produce hydrologically consistent inundation maps. It is organized around three complementary technical components: (1) seeded locally adaptive thresholding for automated flood pixel detection from SAR and optical imagery; (2) object-oriented commission-error filtering to remove false-positive water detections associated with terrain shadows, smooth surfaces, vegetation effects, permanent water bodies, and other non-flood features; and (3) hydrologically guided region growing to reduce omission errors while simultaneously estimating floodwater depth. Together, this three-step workflow mitigates both overprediction and underprediction in satellite-derived flood maps and enables the reproducible joint retrieval of flood extent and depth.
The framework has been implemented as a Python software package, a modular QGIS toolbox, and a cloud-based Google Earth Engine application, providing flexible pathways for research applications, regional-scale flood analysis, and near-real-time operational deployment. Using these toolsets, the project has produced a historical flood map repository that includes 476 Sentinel-1-derived flood extent and depth maps at 10 m spatial resolution and 24 PlanetScope-derived high-resolution flood extent and depth maps at 3 m spatial resolution. By combining multi-sensor remote sensing, terrain analysis, and hydrologically guided post-processing, this project delivers scalable tools, standardized flood map products, and benchmark datasets that strengthen NOAA’s capacity for data-informed flood analysis. The resulting products support hydraulic model evaluation, flood inundation forecasting, event-scale impact assessment, and broader operational flood mapping applications.