Feizbahr et al. (2025) Flood Susceptibility Mapping Using Machine Learning and Geospatial-Sentinel-1 SAR Integration for Enhanced Early Warning Systems

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-10-17
• Authors: Mahdi Feizbahr, Nicholas Brake, Homayoon Arbabkhah, Hossein Hariri Asli, Kolby Woods
• DOI: 10.3390/rs17203471

Research Groups

Not specified in the provided text.

Short Summary

This study develops a comprehensive framework for flood susceptibility mapping by integrating thirteen geospatial factors with statistical and machine learning models, finding that the XGBoost model achieves superior performance with an Area Under the Curve (AUC) of 0.92.

Objective

• To develop and evaluate a comprehensive framework for flood susceptibility mapping by integrating geospatial factors with statistical and machine learning models, and to identify the most influential factors for flood susceptibility.

Study Configuration

• Spatial Scale: Regional/Local (implied by flood susceptibility mapping and local task force validation, but no specific geographic area is given).
• Temporal Scale: 2018 to 2023 (5 years).

Methodology and Data

• Models used: Frequency-based statistical model, Random Forest (ML), XGBoost (ML), Convolutional Neural Network (CNN) (ML). SHapley Additive exPlanations (SHAP) for feature importance analysis in ML models.
• Data sources:
  • Features: Thirteen flood-related geospatial factors, including Digital Elevation Model (DEM), slope, Topographic Wetness Index (TWI), Normalized Difference Vegetation Index (NDVI).
  • Target Variable: Historical flood data derived from Sentinel-1 SAR imagery.
  • Validation Data: High-flood-risk locations monitored by flood sensors, BLE inundation models, and flood-prone areas suggested by a Local Community Task Force.

Main Results

• The XGBoost model demonstrated the highest performance for flood susceptibility mapping, achieving an AUC of 0.92.
• The frequency-based statistical model showed the weakest performance with an AUC of 0.65.
• SHAP value graphs identified elevation, slope, and TWI as the most influential features across all models.
• Susceptibility maps generated by the machine learning models showed strong agreement with BLE inundation maps and high-flood-risk areas identified by the local Community Task Force.

Contributions

• Presents a comprehensive framework for flood susceptibility mapping by integrating diverse geospatial factors with both statistical and advanced machine learning models (Random Forest, XGBoost, CNN).
• Demonstrates the superior performance of the XGBoost model (AUC 0.92) for flood susceptibility mapping compared to other tested models.
• Utilizes SHAP values to provide explainability for machine learning models, highlighting key influential factors such as elevation, slope, and TWI.
• Validates the generated susceptibility maps using multiple independent sources, including flood sensors, BLE inundation models, and input from a local Community Task Force.

Funding

Not specified in the provided text.

Citation

@article{Feizbahr2025Flood,
  author = {Feizbahr, Mahdi and Brake, Nicholas and Arbabkhah, Homayoon and Asli, Hossein Hariri and Woods, Kolby},
  title = {Flood Susceptibility Mapping Using Machine Learning and Geospatial-Sentinel-1 SAR Integration for Enhanced Early Warning Systems},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17203471},
  url = {https://doi.org/10.3390/rs17203471}
}