Authors: Dinuke Munasinghe, Anupal Baruah, Dipsikha Devi, Yixian Chen, Sagy Cohen – University of Alabama
Presentation Type: Poster Presentation
Title: A Scalable Framework for Generating Flood Inundation Maps from High-Water Marks
Presentation Type: Poster Presentation
Abstract: Accurate observation-based flood inundation maps are essential for benchmarking hydrodynamic models, improving remote-sensing algorithms, and supporting hazard assessment. We present HighWaterFIM, a repository of ground-truth flood inundation maps for the United States spanning 2010-2025, derived from field-surveyed high-water marks (HWMs). HWMs provide event-specific measurements of maximum flood elevation and serve as authoritative indicators of flood presence. The dataset is generated using a reproducible geospatial workflow that integrates robust outlier filtering, semivariogram-informed neighborhood selection, and density-based clustering to minimize interpolation uncertainty. Water surface elevation (WSE) rasters are converted to binary flood extent maps using high-resolution Digital Elevation Models, with spatial masking applied to limit interpolation beyond data-supported regions. The resulting products represent conservative estimates of maximum flood extent and are accompanied by metadata describing processing parameters and event characteristics. HighWaterFIM provides a standardized, quality-controlled benchmark dataset for flood model validation, algorithm development, and long-term flood hazard research.