Flood Radar: Multi-Sensor SAR-Based Flood Mapping and Evacuation Modeling — A Case Study of the July 2025 Texas Flood

Antonika Shapovalova

Floods remain among the most destructive natural hazards worldwide, causing an average of USD 40 billion in annual damage and affecting more than 2.5 billion people between 1994 and 2014. The Central Texas flood of July 2025 was one of the most catastrophic in recent decades, triggered by the remnants of Tropical Storm Barry that delivered over 508 mm of rain within two days. This study presents Flood Radar, an integrated multi-sensor system designed for near-real-time flood mapping and evacuation planning, demonstrated through this extreme event. The system combines C-band Sentinel-1 synthetic aperture radar (SAR) data, L-band UAVSAR and ALOS-2/PALSAR-2 imagery, NASA GPM IMERG precipitation fields, and digital elevation models (SRTM and Copernicus DEM) with infrastructure layers from OpenStreetMap. Standardized preprocessing, including orbit correction, radiometric calibration, speckle filtering, and DEM-assisted geocoding, prepares inputs for a pretrained deep-learning segmentation model (U-Net/FCN) that classifies water and land surfaces at 10 m resolution. Change-detection and hydrodynamic modeling using HEC-RAS further estimate water depth, flow velocity, and potential road inundation.
The resulting flood-extent maps accurately delineated both open and sub-canopy inundation zones, revealing the rapid ≈ 9.8 m rise of the Guadalupe River and identifying ~740 acres of flooded cropland and pasture in Kerr County. Integration with OpenStreetMap enabled automatic evaluation of road passability and generation of optimal evacuation routes. The public web interface (https://evacuation-map-sar.vercel.app/) demonstrates the operational output of the system. The study highlights the advantages of multi-sensor fusion, SAR’s cloud-independent imaging, L-band’s vegetation penetration, and near-continuous IMERG rainfall monitoring, while noting limitations such as speckle noise, sparse revisit intervals, and misclassification in urban environments. The July 2025 case underscores the necessity of coupling advanced Earth-observation tools with effective early-warning and communication systems. Flood Radar exemplifies a scalable framework for rapid disaster intelligence that supports timely evacuation and post-event recovery planning in flood-prone regions.