Conflict damage assessment at UNESCO and ICOMOS-listed heritage sites
Bi-temporal SAR coherence loss and very-high-resolution optical change detection can document structural collapse at listed heritage sites within days of an event, supporting legal evidence chains and post-conflict recovery planning.
Sensors
- Sentinel-1 C-band SAR (ESA): 20 m ground range resolution in Interferometric Wide Swath mode; 6-day exact repeat at mid-latitudes (12-day for a single satellite). Coherence between pre- and post-event interferometric pairs drops sharply where building fabric has been displaced or demolished, providing a cloud-independent collapse signal independent of optical visibility.
- WorldView-2 and WorldView-3 (Maxar): WorldView-3 delivers 31 cm panchromatic resolution and 1.24 m multispectral at nadir. WorldView-2 is nominally 46 cm pan. Both support sub-metre damage mapping of individual structures, rubble extent, and burn-scar boundaries. Revisit at a given point is typically 1 to 4.5 days depending on latitude and tasking priority.
- Pléiades-1A and 1B (Airbus): 50 cm panchromatic, 2 m multispectral, with a combined daily revisit capacity from the two-satellite constellation. Stereo tasking in a single pass produces DSMs accurate to roughly 1 m vertically, useful for quantifying rubble pile volume and detecting storey-level collapse.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands with a 5-day revisit. Too coarse for individual structure assessment but useful for mapping burn-scar extent across entire heritage buffer zones and for flagging areas that warrant higher-resolution tasking.
What the radar sees that the eye cannot
SAR coherence is the statistical correlation between the phase of two radar acquisitions over the same ground. Intact masonry, stone columns, and mudbrick walls return a stable phase signal across repeat passes. When a building collapses, the scattering geometry changes abruptly. The coherence value, which runs from 0 to 1, typically falls below 0.3 in collapsed zones, against background coherence of 0.6 to 0.8 for undisturbed urban fabric in C-band Sentinel-1 imagery. That drop is detectable regardless of cloud cover, smoke, or time of day.
The critical methodological point is that coherence loss is not unique to conflict damage. Vegetation growth, flooding, and even heavy rain can decorrelate a SAR pair. Distinguishing building collapse from a seasonal vegetation flush requires either a time series of coherence values (to establish the site's normal decorrelation pattern) or corroboration from optical imagery. UNOSAT's published methodology for Aleppo and Raqqa used exactly this combination: Sentinel-1 coherence to flag change, then WorldView imagery to classify the change type.
Bi-temporal optical analysis: rubble, burn scars, and what they do not tell you
Very-high-resolution optical change detection compares a pre-event image with a post-event image at the pixel level. At 30 to 50 cm resolution, the analyst can distinguish a collapsed roof (irregular pale rubble, loss of straight shadow lines) from a standing but fire-damaged structure (darkened walls, intact outline) from deliberate demolition (clean rubble pile, sometimes with machinery tracks). The AAAS Geospatial Technologies Project applied this approach to Palmyra after 2015, identifying the destruction of the Temple of Bel and the Temple of Baalshamin with sufficient spatial precision to document the footprint of individual structures.
Burn scars are spectrally distinct: charred organic material absorbs strongly in the near-infrared while reflecting modestly in the shortwave infrared, producing a characteristic spectral signature. WorldView-3's eight-band multispectral suite, covering 400 to 1040 nm, resolves this from surrounding dust and pale limestone rubble. Sentinel-2 can map the broader burn extent at 10 m. Neither sensor can determine whether a fire was caused by an airstrike, artillery, or deliberate arson. Attribution of cause is an intelligence question, not a remote-sensing one. Honest analysis stops at damage type and location.
The Palmyra, Mosul, and Aleppo record
These three sites are the most extensively documented cases in the published literature. At Palmyra, UNOSAT's 2015 to 2016 assessments used WorldView imagery to map damage to the Temple of Bel, the Arch of Triumph, and the tetrapylon. The coherence-loss approach was less central here because Palmyra's archaeological zone is largely open ground; optical change detection was the primary tool. At Mosul, the deliberate demolition of the Assyrian reliefs at the Mosul Museum and the Winged Bull statues at Nineveh's Nergal Gate was documented through before-and-after WorldView imagery, though the interior destruction of the museum required ground verification.
Aleppo's Old City, a UNESCO World Heritage Site, presented the hardest problem: dense urban fabric where conflict damage and pre-existing urban decay are difficult to separate. UNOSAT's 2013 to 2017 assessments classified damage into five severity levels, from no visible damage to destroyed, using a combination of Pléiades and WorldView imagery. The coherence approach was applied retrospectively by several academic groups using Sentinel-1 archives, which extend back to late 2014. The archive depth matters enormously: without a credible pre-conflict baseline, the damage assessment has no reference state.
The attribution limit and the evidence chain
Satellite imagery establishes that a structure was intact before a date and damaged after it. It does not establish who caused the damage, by what means, or whether the site had military use at the time. This is not a technical failure; it is the correct scope of the method. Legal instruments such as the 1954 Hague Convention for the Protection of Cultural Property in the Event of Armed Conflict require evidence of intentionality and military necessity, neither of which is readable from a coherence interferogram.
What imagery can support is the documentation of a pattern. A single collapse event might be ambiguous. A sequence of dated events showing systematic destruction of a specific site category, correlated with known military activity in the area, contributes to a body of evidence that legal and investigative bodies can use. The International Criminal Court's 2016 conviction of Ahmad al-Faqi al-Mahdi for the destruction of religious and historic buildings in Timbuktu drew on satellite imagery as part of its evidence base. The imagery was not the verdict; it was a precisely dated, spatially explicit record that other evidence could be measured against.
Practical limits analysts should state plainly
Cloud cover is the most obvious constraint in optical analysis. Syria and Iraq have relatively dry summers, but winter cloud can delay post-event optical acquisition by days or weeks. SAR fills that gap but cannot match optical resolution for structure-level classification. A Sentinel-1 pixel at 20 m covers a courtyard; a WorldView-3 pixel at 31 cm can resolve a single column drum.
Revisit latency matters for legal documentation. If a site is re-used or cleared between the damage event and the first satellite pass, the record is incomplete. Commercial tasking of WorldView or Pléiades can, in principle, achieve same-day or next-day collection over a prioritised target, but cloud and satellite geometry are not negotiable. Archive gaps before 2014 (the Sentinel-1 launch) mean that SAR-based coherence baselines for pre-2014 damage rely on other SAR missions with less consistent coverage. Satellize structures commercial tasking workflows around these constraints, adding Pléiades or WorldView collection to Sentinel-1 monitoring where the legal or institutional client requires sub-metre optical corroboration.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 20 m range × 22 m azimuth |
| Optical resolution (WorldView-3 panchromatic) | 31 cm at nadir |
| Optical resolution (Pléiades-1) | 50 cm panchromatic, 2 m multispectral |
| Sentinel-1 repeat pass interval | 6 days (two-satellite constellation); 12 days single satellite |
| WorldView-3 revisit | Typically 1 to 4.5 days depending on latitude and off-nadir angle |
| Coherence change detection sensitivity | Coherence drop below ~0.3 (from baseline ~0.6–0.8) flags probable structural displacement |
| Sentinel-1 archive depth | Late 2014 to present (Sentinel-1A launch October 2014) |
| Minimum detectable collapse footprint (optical) | Approximately 1–2 m² at WorldView-3 resolution; single-structure level |
| Typical post-event optical delivery latency | 24–72 hours for tasked commercial collection; archive imagery available immediately |
| Deliverable formats | GeoTIFF change maps, GeoJSON damage polygons, PDF site assessment reports, GIS-ready shapefiles |
Analytics Satellize can run
| Coherence-loss change map | Sentinel-1 interferometric coherence differencing between pre- and post-event image pairs; temporal baseline matched to site's decorrelation profile | GeoTIFF and GeoJSON layer showing coherence-loss magnitude per pixel, colour-classified by severity threshold |
| Structure-level damage classification | Bi-temporal very-high-resolution optical change detection (WorldView or Pléiades) with five-class UNOSAT-compatible severity schema: no damage, possible damage, moderate, severe, destroyed | GIS polygon layer with per-structure damage class and date of assessment; exportable to QGIS, ArcGIS, or web map |
| Burn-scar extent mapping | Normalised Burn Ratio (NBR) differencing on Sentinel-2 or WorldView-3 SWIR bands; pre- and post-event image pair | Classified raster and polygon boundary of burn extent within heritage buffer zone, with area statistics in hectares |
| Rubble volume estimate | Pléiades stereo DSM differencing against pre-event DSM or photogrammetric baseline; height-loss converted to volume using standard rubble density assumptions | Tabular report of estimated rubble volume per structure or zone, with uncertainty bounds stated |
| Damage timeline reconstruction | Multitemporal coherence and optical archive analysis to assign damage events to specific date windows; correlated against reported incident dates where available | Chronological event log with satellite acquisition dates, damage state at each epoch, and confidence rating per event |
| Site monitoring alert service | Automated Sentinel-1 coherence monitoring on 6-day repeat; optical tasking triggered on coherence anomaly exceeding defined threshold | Email or API alert with preliminary coherence map within 48 hours of anomaly detection; followed by optical assessment report |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.