Xiao et al. (2025) Flood Prediction with Sentinel-1 Synthetic Aperture Radar from Hurricane Helene

Identification

• Journal: Journal of Purdue Undergraduate Research
• Year: 2025
• Date: 2025-12-19
• Authors: Jingyu Xiao, Jessica Li
• DOI: 10.7771/2158-4052.1801

Research Groups

• Purdue University (Civil Engineering)

Short Summary

This study develops a flood prediction and mapping framework using Sentinel-1 Synthetic Aperture Radar (SAR) data and a Support Vector Machine (SVM) classifier. The approach successfully identifies flood extents from Hurricane Helene by integrating SAR backscatter with topographic and meteorological variables.

Objective

• To improve the spatial resolution and accuracy of flood prediction by relating Sentinel-1 SAR backscatter parameters and auxiliary environmental data to flood extents in regions with limited ground-based sensor networks.

Study Configuration

• Spatial Scale: Regional scale focusing on the southeastern United States, specifically parts of North Carolina, Tennessee, and Georgia, centered around 10 USGS streamgage locations.
• Temporal Scale: Short-term event analysis of Hurricane Helene (September 24–27, 2024), utilizing a data window from two weeks before to two weeks after the storm.

Methodology and Data

• Models used: Support Vector Machine (SVM) classifier.
• Data sources:
  • Sentinel-1 C-band Synthetic Aperture Radar (SAR) imagery (via Alaska Satellite Facility’s Vertex).
  • USGS National Water Information System (NWIS) stream stage data.
  • Precipitation rate data.
  • Land use/land cover (LULC) maps.
  • Digital elevation models (Elevation data).

Main Results

• The SVM classifier successfully related SAR backscatter parameters to flood occurrence.
• Model accuracy was significantly improved by incorporating multi-source inputs, including precipitation rates, land use, and elevation, rather than relying on SAR data alone.
• The 6-day revisit cycle of Sentinel-1 provided sufficient temporal resolution for monitoring the impacts of the hurricane and subsequent flash flooding.

Contributions

• Demonstrates an effective workflow for high-resolution flood mapping in all-weather conditions using SAR, which overcomes the limitations of optical sensors and sparse ground-based soil moisture networks.
• Provides a case study application of machine learning for rapid natural disaster response using the 2024 Hurricane Helene event.

Funding

• Not explicitly mentioned (Research conducted at Purdue University; published in the Journal of Purdue Undergraduate Research).

Citation

@article{Xiao2025Flood,
  author = {Xiao, Jingyu and Li, Jessica},
  title = {Flood Prediction with Sentinel-1 Synthetic Aperture Radar from Hurricane Helene},
  journal = {Journal of Purdue Undergraduate Research},
  year = {2025},
  doi = {10.7771/2158-4052.1801},
  url = {https://doi.org/10.7771/2158-4052.1801}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)