Methodology
How this is built, what it counts, and what it does not claim.
What this is
An open-source record of damage to United States, Israeli and Gulf state military and economic infrastructure during the ongoing conflict. Every incident names the asset struck, carries a verification tier proportionate to its evidence, and links to the material it rests on.
Maintained independently and published pseudonymously. Not affiliated with any government, military or media organisation. All source material is public.
How an incident is built
- Public reporting is indexed in full, then filtered — so a change in what counts can be re-applied rather than losing what was already collected. Multi-image posts are expanded, because before-and-after pairs are often published as groups whose later frames carry no caption.
- Reposts are collapsed by text and perceptual image hashing, so one photograph circulated by eight accounts counts once. Amplification is the most common thing mistaken for confirmation.
- Every image is opened and examined, not judged from its caption. Text burned into the image — including Arabic and Farsi — is transcribed, because it often names the correct facility while the surrounding post does not. Dates burned into imagery outrank the date posted. Most reviewed imagery is rejected.
- Where redistributed imagery is unusable, before-and-after comparisons are built from Copernicus Sentinel-2 — composited from the raw 10 m red, green and blue bands rather than the pre-rendered true-colour product, which saturates over bright desert, with one contrast stretch across both dates so rendering differences cannot pass as damage.
- Coordinates are checked against the named installation rather than trusted from the report; those derived from a place name are marked approximate.
Verification tiers
Cost figures
Limits worth knowing
- Aircraft and drones shot down are excluded. This records damage to fixed assets on the ground. An aircraft lost in the air leaves nothing to assess, and mixing the two would make the totals uninterpretable.
- Imagery access is constrained. Commercial high-resolution coverage of the region is restricted, so overhead evidence leans on free Sentinel-2 at 10 m per pixel. Small or transient damage is not always resolvable.
- Reporting is thin in some theatres. Several countries have criminalised posting about strikes on their territory — often exactly where the most significant damage occurred.
- Absence is not evidence. A facility with no incidents may simply be one nobody photographed.
Credits and corrections
Imagery from Conflict Mapping / Soar Atlas is used under CC BY 4.0. Comparisons built here use modified Copernicus Sentinel-2 data (2026). Base map tiles by CARTO and Esri; map data © OpenStreetMap contributors.
Corrections are welcome and acted on — a misidentified facility, a bad coordinate, a duplicated strike, an image that does not show what it is captioned as. Write to damagetracker@proton.me.