Credit: SZ
Our method is based on so-called SAR coherence: a measure of how stable the received phase remains between two SAR acquisitions of the same area at different points in time. Unlike classical backscatter, which primarily captures the amplitude or intensity of the reflected radar signal, coherence evaluates phase stability — that is, how similar the microstructure of a surface is between two acquisitions. Intact buildings and stable urban structures produce consistent phase patterns over time — coherence remains high. Even small changes in surface structure — caused by destruction, displacement, or debris accumulation — produce phase changes; the signals lose their coherence between the two points in time. We harness this effect through the LT-CCD approach (Long-Term Coherence Change Detection): instead of comparing individual image pairs, we first compute a stable reference state from multiple acquisitions taken before the event. Anything that subsequently deviates significantly and unusually, we flag as a potential damage indicator. For Tehran, we set the reference date roughly one year before the attacks began; from 25 coherent image pairs, we built a baseline that describes the typical dynamics of the urban area. Our destruction analysis then covered several time steps between February 28 and April 8 — making not only the extent of the damage but also its temporal sequence visible. On our maps, larger circles indicate particularly extensive destruction.
The types of targets Israel and the United States struck most often during these weeks can be read from the combined evaluation of the ISW data and our radar analysis: in Tehran, the attacks centered on military targets, security apparatus facilities, and civilian and political institutions. As already seen in the Twelve-Day War of June 2025, both armed forces also targeted Iran's nuclear program — and not only uranium enrichment plants, but also research centers and the researchers themselves. Both Sharif University and Shahid Beheshti University, which conduct research for the Islamic Republic's nuclear program, appear on the EU sanctions list.
Methodology in comparison: Scher, Bellingcat, and Vertical52
In parallel with our analysis, the researcher Corey Scher and the investigative platform Bellingcat published their own assessments of the damage in Tehran — likewise based on SAR data, but with different methodological emphases. What all three approaches share is their radar-based starting point: they operate independently of lighting conditions and cloud cover — where optical sensors go blind in clouds or darkness, SAR satellites keep delivering data. Methodologically, however, the approaches differ markedly. Scher's approach is very similar to ours: he too works with coherence analysis; the differences lie primarily in the interpretation of results, in the thresholds used, and in the selection of baseline and comparison images. Bellingcat, by contrast, relies more heavily on SAR backscatter — measuring changes in the amplitude of the reflected signal rather than in its phase stability.
From this, different strengths emerge: backscatter analyses respond sensitively to large, high-energy scattering changes — such as exposed debris fields or craters — and deliver particularly clear signals in those cases. Coherence analyses also capture more subtle changes in surface structure, but in return are more sensitive to vegetation, weather, and temporary effects. Regardless of the specific SAR method, the same caveat applies: we do not deliver images of destroyed buildings, but statistical anomalies in the radar signal — indicators, not proof. We see this difference not as a weakness, but as a methodological characteristic that enables a different kind of statement: not the individual documented impact image, but the spatial pattern across an entire urban area. Combined with high-resolution optical satellite data and OSINT verification, this produces a more robust overall picture than any single method could deliver on its own.