Automated Coherent Change Detection from Sentinel-1 SAR for Rapid Damage Assessment

Synthetic Aperture Radar (SAR) change detection is a key tool for rapid post-event damage assessment because it provides day-and-night, all-weather observations that optical sensors cannot guarantee. Coherent Change Detection (CCD) based on multi-temporal interferometric coherence is highly sensitive to structural disturbances caused by building collapse, military strikes, and wildfires. However, practical CCD workflows typically require complex manual processing across multiple desktop tools, resulting in substantial per-site processing times that limit operational deployment during time-critical emergencies.

This paper presents a fully automated Python-based pipeline for end-to-end CCD damage mapping from Sentinel-1 Single Look Complex (SLC) imagery. The pipeline automates data discovery and download from the Copernicus Data Space, interferometric preprocessing, generation of 90 multi-window coherence-change variables in dual polarization, signal-to-noise ratio screening, principal component analysis (PCA)-based fusion, and damage delineation through automated thresholding, with optional interactive refinement. The unsupervised, label-free workflow requires no labeled damage data for training and reduces per-site processing effort from approximately two days of manual operation to less than 40 minutes of unattended computation.

The pipeline was applied to sites affected by the 2026 Iran–United States and June 2025 Iran–Israel conflicts and to the July 2025 Latakia wildfire in Syria. Conflict-site results were validated against damage inventories manually digitized from PlanetScope imagery, while wildfire results were compared with an independent UNITAR rapid-mapping product. The best-performing conflict site achieved pixel-level F1 and IoU values of 0.73 and 0.57, respectively, with overall accuracy exceeding 0.98. Object-level F1 scores ranged from 0.75 to 0.92, detecting the large majority of damaged structures. For the Latakia wildfire, the mapped burned-area extent showed close agreement with the UNITAR product.

These results demonstrate that automated multi-scale coherence analysis combined with PCA-based fusion provides a robust, label-free damage indicator with potential to transfer across conflict and wildfire events. By substantially reducing manual processing without requiring event-specific labeled data, the proposed framework enables rapid, scalable, and operationally viable SAR-based damage assessment for disaster response, humanitarian operations, and post-event situational awareness.