Researchers from Technical University of Munich and LMU Munich have developed a new method for automatically detecting building destruction in conflict zones using publicly available satellite data. The project was supported by the Munich School for Data Science, and the results have been published in PNAS Nexus. The paper is available here.

The method builds on an interferometric technique using synthetic aperture radar (SAR) data from the European Sentinel-1 mission. By applying InSAR to repeated satellite observations, changes in signal coherence can be quantified, enabling the detection of structural damage to buildings shortly after it occurs. Since radar imagery is independent of cloud cover and daylight, the approach is particularly well suited for monitoring destruction in war and crisis regions.

To distinguish actual destruction events from normal fluctuations, the observed changes are assessed statistically at the pixel level and aggregated to individual buildings using footprint data from OpenStreetMap. This allows both the extent of damage and associated uncertainty to be estimated.

The method was validated using case studies including the Beirut port explosion in 2020, the destruction of Mariupol following the Russian invasion of Ukraine in 2022, and the ongoing war in Gaza. Overall, the approach provides a fast and cost-effective tool for humanitarian assessments, academic research, and post-conflict reconstruction planning.

This work was featured in an official press release by LMU.
Read the full article on the LMU website: German | English

Refernece: Racek, D., Zhang, Q., Thurner, P. W., Zhu, X. X., Kauermann, G. (2025). Unsupervised Detection of Building Destruction during War from Publicly Available Radar Satellite Imagery. PNAS Nexus, 4. Link.