Hotaki et al. (2026) Coherent and incoherent change detection for improved flood mapping: A Sentinel-1 SAR time-series approach

Identification

• Journal: Remote Sensing Applications Society and Environment
• Year: 2026
• Date: 2026-01-01
• Authors: Sulaiman Fayez Hotaki, Mahmud Haghshenas Haghighi, Mahdi Motagh
• DOI: 10.1016/j.rsase.2026.101961

Research Groups

• Institute of Photogrammetry and GeoInformation, Leibniz University Hannover, 30167 Hannover, Germany
• GFZ Helmholtz Centre for Geosciences, Department of Geodesy, Section of Remote Sensing and Geoinformatics, 14473 Potsdam, Germany

Short Summary

This study proposes a methodology for improved local-scale flood mapping and coherence-based damage proxy maps using Sentinel-1 SAR time-series data, combining incoherent and coherent change detection, and demonstrates its effectiveness in Afghanistan by identifying significantly more inundation areas compared to global flood products.

Objective

• To develop a methodology for improved local-scale flood mapping and coherence-based damage proxy maps using Sentinel-1 SAR time-series data.
• To assess the effectiveness of combining incoherent (SAR intensity) and coherent (interferometric coherence) change detection for flood detection in different land cover types (rural vs. urban).
• To validate the proposed method against high-resolution optical imagery and compare its performance with existing global flood products.

Study Configuration

• Spatial Scale: Local-scale floods in six study areas across five provinces in Afghanistan (Faryab, Panjsher, Sar-e-Pol, Parwan, and Baghlan). Sentinel-1 GRD data at 10 m x 10 m pixel spacing, Sentinel-1 SLC data at 5 m x 20 m resolution. PlanetScope imagery at 3.7 m spatial resolution for validation.
• Temporal Scale: Flood events between 2018 and 2024. Sentinel-1 data with a 12-day temporal baseline for coherence. Baseline periods for incoherent detection were typically one year.

Methodology and Data

• Models used:
  • Incoherent Change Detection (ICD) using Sentinel-1 GRD intensity data (VV polarization).
  • Coherent Change Detection (CCD) using interferometric coherence from Sentinel-1 SLC data.
  • Z-score anomaly detection (Z = (xi - mean(xbaseline)) / std(xbaseline)) with a threshold of Z < -3.
  • Radiometric Terrain Normalization (RTN) for GRD images.
  • Hybrid Pluggable Processing Pipeline (HyP3) for SLC data processing and coherence map generation.
  • Refining steps: JRC Global Surface Water dataset for permanent water body masking, SRTM DEM for slope masking (>10%), majority filter for noise reduction.
• Data sources:
  • Satellite: Sentinel-1 Ground Range Detected (GRD) (VV polarization), Sentinel-1 Single Look Complex (SLC).
  • Validation: PlanetScope imagery (high-resolution optical data, 3.7 m spatial resolution).
  • Auxiliary: JRC Global Surface Water dataset (Landsat-based, 1984-2021), SRTM Digital Elevation Model (DEM), ICIMOD land cover data (2018).
  • Platform: Google Earth Engine (GEE) for GRD and auxiliary data processing, Alaska Satellite Facility (ASF) for HyP3.

Main Results

• Incoherent change detection (ICD) effectively identified floods in irrigated agriculture and bare lands, achieving F1 scores ranging from 79% to 83%.
• Coherent change detection (CCD) revealed flood extent and coherence-based damage proxy maps in built-up areas, with F1 scores ranging from 69% to 73%.
• The proposed methodology identified approximately 5 times more inundation areas in some study areas compared to the Global Flood Monitoring (GFM) system, particularly for local-scale floods.
• The total flooded area across the six study sites amounted to 2220 hectares. Irrigated agriculture (36.3%) and bare land (28%) were the most affected land cover types.
• Coherence-based damage proxy maps in urban areas (e.g., Parwan) identified 115 hectares of flood-related surface changes, demonstrating its utility where intensity data fails.
• The VV polarization band was found to be more reliable for flood detection than the VH band due to its greater stability and less variability in backscatter.

Contributions

• Proposes a novel combined approach of incoherent and coherent change detection using Sentinel-1 time-series SAR data for improved local-scale flood mapping and damage proxy assessment.
• Demonstrates the complementary nature of incoherent (for rural/agricultural areas) and coherent (for urban areas) methods, addressing limitations of intensity-only approaches in complex environments.
• Provides a robust methodology for near-real-time local-scale flood mapping, particularly valuable in data-scarce regions like Afghanistan.
• Highlights the significant underestimation of local-scale flood extents by existing global flood products (e.g., GFM) compared to the proposed method.
• Introduces the concept of a coherence-based damage proxy map for flood-induced changes in urban areas, building on earthquake/typhoon damage assessment techniques.

Funding

• German Academic Exchange Service (DAAD)

Citation

@article{Hotaki2026Coherent,
  author = {Hotaki, Sulaiman Fayez and Haghighi, Mahmud Haghshenas and Motagh, Mahdi},
  title = {Coherent and incoherent change detection for improved flood mapping: A Sentinel-1 SAR time-series approach},
  journal = {Remote Sensing Applications Society and Environment},
  year = {2026},
  doi = {10.1016/j.rsase.2026.101961},
  url = {https://doi.org/10.1016/j.rsase.2026.101961}
}