Ta et al. (2026) SAR Flood Anomaly Mapping Through Statistical Time-Series Feature Classification and Pixel-Wise TCEV Modeling

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

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-26
• Authors: Liangyu Ta, Qi Liu, Chen Yu, Javier Valdés-Abellán
• DOI: 10.3390/rs18193323

Research Groups

Not specified in the provided text.

Short Summary

This study introduces a novel method combining statistical time-series feature classification and pixel-wise Two-Component Extreme Value (TCEV) modeling to map flood anomalies using long-term Sentinel-1 SAR data, demonstrating high accuracy in characterizing extreme flood events.

Objective

• To develop a statistical time-series feature classification and pixel-wise TCEV modeling method for SAR flood anomaly mapping by integrating extreme value theory, specifically for characterizing rare extreme flood events.

Study Configuration

• Spatial Scale: Pixel-wise analysis across regional areas, evaluated over four extreme flood events including New South Wales.
• Temporal Scale: Long-term Sentinel-1 SAR time series for constructing statistical features and calculating event-period anomaly probabilities.

Methodology and Data

• Models used: Tabular Prior-data Fitted Network (TabPFN) for flood classification, Two-Component Extreme Value (TCEV) model for anomaly probability calculation.
• Data sources: Long-term Sentinel-1 Synthetic Aperture Radar (SAR) time series observations.

Main Results

• The proposed method generated Temporal Flood Anomaly Maps (TFAMs) that effectively characterized flood-related temporal anomalies.
• Achieved highest ROC-AUC of 0.87 and average precision of 0.81 for New South Wales.
• The TFAM-derived binary flood extent map achieved an F1-score of 0.81, an Intersection over Union (IoU) of 0.69, and a Kappa coefficient of 0.72.

Contributions

• Proposes a novel statistical time-series feature classification and pixel-wise TCEV modeling method for SAR flood anomaly mapping.
• Integrates extreme value theory to better characterize rare extreme flood events, addressing limitations of existing approaches in temporal context and anomaly representation.
• Demonstrates high performance in mapping flood anomalies and deriving flood extents across diverse environmental conditions.

Funding

Not specified in the provided text.

Citation

@article{Ta2026SAR,
  author = {Ta, Liangyu and Liu, Qi and Yu, Chen and Valdés-Abellán, Javier},
  title = {SAR Flood Anomaly Mapping Through Statistical Time-Series Feature Classification and Pixel-Wise TCEV Modeling},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18193323},
  url = {https://doi.org/10.3390/rs18193323}
}