LMU and TUM Researchers Develop Automatic Building Destruction Detection

Researchers from Ludwig-Maximilians-Universität München (LMU) and the Technical University of Munich have developed a method to automatically detect building destruction in conflict zones. This innovative approach utilizes freely accessible satellite data, offering a rapid and cost-effective tool for various applications.

Leveraging Radar Data for Detection

The method employs synthetic aperture radar (SAR) data from the Sentinel-1 mission, which is available at 12-day intervals. Notably, it operates without reliance on commercial satellite imagery or external training data, making it widely accessible.

The team applies the InSAR interferometric technique. This technique involves comparing repeated images of the same geographical region to compute a coherence measure, which reflects the similarity of backscattered radar signals. A sudden drop in this coherence often indicates structural changes, such as damage or destruction, within buildings.

Statistical Precision and Data Integration

To effectively distinguish these crucial signals from random fluctuations, the collected data undergoes rigorous statistical assessment. A 'normal' pattern of variation over time is estimated for each individual pixel. Deviations from this pattern are then quantified using p-value probabilities.

This statistical information is subsequently combined with existing building footprints from OpenStreetMap. This integration allows the results to be aggregated at the individual building level, providing not only detection but also an associated measure of uncertainty.

Dr. Daniel Racek, the study's first author and a former doctoral researcher at LMU’s Institute of Statistics, highlighted the method's significant advantage:

"Using freely accessible data enables tracking of destruction across space and time almost in real time."

Proven Accuracy in Conflict Zones

The method underwent successful testing through several case studies. These included the Beirut port explosion (2020), the widespread destruction in Mariupol following the 2022 Russian invasion, and the ongoing conflict in Gaza (from 2023). In each instance, the approach accurately reconstructed both the spatial patterns and the precise timing of building destruction.

A Rapid Tool for Humanitarian Aid and Planning

The researchers propose that this new methodology can serve as a rapid and cost-effective tool with broad utility. Its applications include humanitarian situation assessments, academic research into conflict impacts, and crucial planning efforts for post-conflict reconstruction.

The research was generously funded by the Munich School for Data Science and its findings were published in PNAS Nexus.