Logo Lanfrica

Sezibra/conflict-satellite-damage

Domaine:

peace and securitygeospatial
Créateur:
Sez
Hôte:
Cross-context transfer of SAR damage detection from Mariupol to Tigray, with multi-source triangulation against UCDP events and Sentinel-2 optical change. # Satellite Imagery for Conflict Damage Assessment ## Research Question How reliably do satellite-based damage detection methods transfer from well-documented conflict zones to data-scarce environments, and what do multiple independent data sources reveal about the spatial pattern of destruction in the Tigray conflict? ## Navigation | Section | Description | |---------|-------------| | Motivation | Why satellite-based damage detection matters for conflict research | | Key Findings | Main results across all methods and study areas | | Data | Six data sources covering three cities across two conflicts | | Methods | Pixel-Wise T-Test, cross-context transfer, multi-source triangulation | | Results in Detail | Figures and interpretation for each analytical step | | Limitations | Honest assessment of constraints | | Notebooks | Five-notebook analytical progression | | How to Reproduce | Setup and replication instructions | | References | Academic sources | ## Motivation Most SAR-based damage detection methods are developed and validated on well-documented conflict zones like Ukraine, where UNOSAT provides verified building-level damage labels. Few researchers have tested whether these methods work when transferred to conflict zones without such labels. Yet data-scarce environments are precisely where satellite-based damage detection is most needed: places with communications blackouts, restricted humanitarian access, and incomplete event databases. This project tests that transfer. It calibrates a Pixel-Wise T-Test (PWTT) damage detection pipeline on Mariupol, Ukraine (where UNOSAT labels exist), then applies the same pipeline to Mekelle and Axum in Tigray, Ethiopia (where no labels exist). It measures the performance gap, analyzes what drives it, and triangulates the SAR results against UCDP conflict events and Sentinel-2 optical change to identify high-confidence damage zones. ## Key Findings **The Mariupol threshold does not transfer to Tigray.** Direct applic …