Port infrastructure damage assessment after extreme weather or conflict
SAR coherence and very-high-resolution optical imagery can detect structural damage to quay walls, cranes, warehouses and aprons within hours of an event, even under cloud cover. Honest limits apply: pre-event archive gaps and debris ambiguity constrain confidence.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10-metre ground range resolution in IW mode, 6-day revisit at mid-latitudes (12-day per satellite). Coherence change detection between pre- and post-event passes identifies surface disruption regardless of cloud or darkness. Free and open archive from 2014.
- ICEYE SAR (X-band): Spotlight mode delivers approximately 0.5-metre resolution; tasking latency can be under 24 hours. X-band penetrates light rain and provides sharper structural detail than C-band, making it well suited to resolving individual crane booms or warehouse roof sections.
- Maxar WorldView-3 (optical/SWIR): 0.31-metre panchromatic resolution once skies clear. Eight-band multispectral and SWIR bands help distinguish wet debris, bare concrete and standing water. Revisit roughly 1-4.5 days depending on latitude and tasking priority.
- Airbus Pléiades Neo (optical): 0.3-metre native resolution, four-satellite constellation offering sub-daily revisit at many latitudes. Stereo and tri-stereo tasking allows surface-model differencing to quantify rubble volumes or collapsed roof heights.
What coherence loss actually tells you
SAR coherence measures how similar the radar backscatter phase is between two acquisitions of the same scene. When a quay wall collapses, a warehouse roof fails or storm surge rearranges surface material, the scattering geometry changes and coherence drops sharply. A coherence value near 1 means the surface is unchanged; values below roughly 0.3 in a port context are a reliable signal that something has moved or been destroyed. This threshold is not universal: loose cargo, tidal variation and vessel movements all reduce coherence without indicating structural damage, so interpretation requires masking out known-dynamic zones such as water surfaces and active berths.
Sentinel-1's C-band is the workhorse for rapid post-event passes because the data are free and the archive stretches back to 2014, giving analysts a stack of pre-event coherence baselines to compare against. ICEYE's X-band adds sharper spatial discrimination once tasked: at sub-metre resolution, a coherence anomaly can be localised to a specific crane foundation rather than a 10-metre pixel that might span both the crane and the adjacent apron.
Once the cloud lifts: optical damage classification
Coherence change detection tells you where something happened. Optical imagery tells you what. The standard workflow pairs a pre-event image with a post-event acquisition at the same or better resolution, then applies change detection across spectral bands. Collapsed roofing exposes insulation or steel substructure with a distinct spectral signature. Standing water on aprons that were previously dry shows up clearly in near-infrared. Scorched or fire-damaged surfaces have characteristic SWIR reflectance that differs from storm debris.
Damage classification typically follows a four-class scheme adapted from the Copernicus Emergency Management Service methodology: no damage, moderate damage (partial structural failure), severe damage (major structural loss), and destroyed. At 0.3-metre resolution, individual shipping containers, crane booms and warehouse bays are resolvable, which allows assessors to assign damage grades at the asset level rather than the facility level. That granularity matters when a port authority needs to know which berths are operationally viable within days of an event.
The honest limit here is time. WorldView-3 and Pléiades Neo can be tasked quickly, but cloud cover after a tropical cyclone or monsoon event can persist for several days. In that window, SAR is the only source of information, and SAR cannot reliably distinguish a collapsed roof from a roof covered in debris from a neighbouring warehouse.
The pre-event archive problem
Every change-detection method depends on a pre-event baseline of comparable quality. For major commercial ports in Europe, East Asia and North America, WorldView and Pléiades archives typically hold multiple cloud-free images from the preceding months. For smaller ports in the Pacific, Indian Ocean or West Africa, the commercial archive may be sparse or absent at the required resolution.
Sentinel-1's open archive partially compensates: C-band SAR coverage is global and systematic, so a coherence baseline can usually be constructed from the six or twelve months before an event. But if the pre-event SAR stack was acquired during a period of active construction or heavy cargo movement at the port, the baseline coherence will already be low, and post-event loss becomes harder to interpret. Analysts should flag this explicitly in any damage report rather than presenting confident damage polygons over areas with ambiguous baselines.
Debris versus cargo: the ambiguity that matters most
Port facilities are unusual damage-assessment targets because they are supposed to contain large quantities of heterogeneous material: containers, bulk commodities, vehicles, equipment. Storm debris and normal cargo can be spectrally and structurally similar at satellite resolution.
Several approaches reduce this ambiguity. Comparing against a pre-event optical image that shows the same area in an organised state helps: a container stack that was orderly before the storm and is now scattered at odd angles is a legitimate damage signal. Pléiades Neo stereo pairs can generate a surface model differential that distinguishes a pile of debris (irregular height variation) from a standing container stack (regular prismatic shapes). Still, in heavily loaded yards, residual ambiguity is real and should be communicated as a confidence interval rather than a binary damage classification.
Conflict damage adds another layer of difficulty. Secondary fires, unexploded ordnance and active hostilities can prevent ground-truth verification for weeks or months. In those cases, satellite assessment may be the only available evidence, which raises the stakes for being explicit about what the imagery can and cannot confirm.
Turning an assessment into an operational decision
A damage map is only useful if it connects to a decision. Port authorities and insurers typically need answers to a small set of questions: which berths can accept vessels immediately, which require inspection before use, and which are structurally compromised. Analysts can map those categories directly onto the damage classification output, overlaying berth geometry, crane coverage arcs and warehouse footprints from pre-event GIS data.
For humanitarian response, the relevant question is often whether a port can receive relief cargo at all. A single functional deep-water berth with crane access may be sufficient. SAR coherence can identify that berth within 24 hours of a Sentinel-1 overpass, even before optical confirmation is possible. Satellize structures its port damage assessments around these operational questions rather than producing generic damage density maps.
The Copernicus Emergency Management Service publishes activation reports for major disasters that include port infrastructure, and those reports are a useful benchmark for what satellite-derived damage grading looks like in practice. They also illustrate the latency achievable under real operational conditions: first products typically appear 12 to 48 hours after activation, depending on data availability.
Limits worth stating plainly
Spatial resolution determines the smallest detectable damage unit. At Sentinel-1's 10-metre IW resolution, a crack in a quay wall is invisible; only wholesale surface disruption registers. ICEYE spotlight at 0.5 metres resolves individual structural members but costs more per acquisition and requires tasking in advance of the event if pre-event imagery is needed at that resolution.
Revisit is a genuine constraint for optical sensors. A single satellite cannot guarantee a cloud-free image over a specific location within 24 hours. Constellation operators such as Maxar and Airbus can increase tasking priority after a declared emergency, but cloud persistence after tropical cyclones routinely delays first optical acquisition by two to five days.
Finally, satellite assessment is not a substitute for engineering inspection. It identifies where to look and prioritises ground teams. A berth that shows no coherence change and no optical damage signal may still have submerged quay-wall damage that no current satellite sensor can detect. That limit should appear in every damage report delivered to a port authority or insurer.
Typical figures
| SAR spatial resolution (Sentinel-1 IW) | 10 m range × 10 m azimuth (after multi-looking) |
| SAR spatial resolution (ICEYE Spotlight) | ~0.5 m |
| Optical spatial resolution (WorldView-3 / Pléiades Neo) | 0.31–0.3 m panchromatic |
| Sentinel-1 revisit (mid-latitudes, both satellites) | 6 days |
| Commercial optical revisit | Sub-daily possible with Pléiades Neo 4-satellite constellation; 1–4.5 days for WorldView-3 depending on latitude |
| First SAR product latency after event | 12–48 hours (subject to next overpass timing) |
| SAR frequency | C-band 5.405 GHz (Sentinel-1); X-band ~9.6 GHz (ICEYE) |
| Minimum detectable damage unit (optical) | Individual warehouse bay or crane boom at 0.3 m resolution; facility-level only at 10 m |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch) |
| Delivery formats | GeoTIFF damage classification raster, GeoJSON damage polygons by asset class, PDF assessment report with confidence grades |
Analytics Satellize can run
| SAR coherence change map | Interferometric coherence differencing between pre- and post-event Sentinel-1 or ICEYE acquisitions; coherence computed over 5×5 or 9×9 pixel windows | GeoTIFF raster with coherence loss magnitude; alert within 24 hours of first post-event pass |
| Optical damage classification layer | Object-based image analysis comparing pre- and post-event multispectral imagery; damage grades following four-class Copernicus EMS-style schema | GeoJSON polygon layer with damage grade per asset (berth, crane, warehouse, apron); delivered once cloud-free imagery is available |
| Berth operability assessment | Overlay of damage classification output with pre-event berth geometry and crane coverage GIS; binary operability flag per berth | Tabular report and GIS layer showing immediately usable, inspection-required and compromised berths |
| Debris versus cargo disambiguation | Stereo surface-model differencing (Pléiades Neo tri-stereo) to separate irregular debris piles from regular container stacks by height-profile shape | Confidence-graded damage polygons with explicit ambiguity flags where pre-event cargo density was high |
| Pre-event baseline coherence quality report | Statistical analysis of Sentinel-1 archive coherence over the facility to identify periods of low baseline coherence that would reduce post-event interpretability | One-page baseline quality memo flagging zones of inherent ambiguity before assessment is delivered |
| Change time series for recovery monitoring | Repeat coherence and optical classification at each new acquisition epoch to track debris clearance, crane reinstatement and structural repair progress | Weekly GIS layer updates and summary chart of recovery progress by asset class |
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.