Oil slick and illegal bilge-dump detection using SAR
Synthetic aperture radar detects oil on the sea surface by the way it kills capillary waves, creating a dark patch against surrounding rougher water. This page explains the physics, the false-alarm problem, and how Sentinel-1 makes coastal enforcement operationally viable.
Sensors
- Sentinel-1 (ESA): C-band SAR at 5.405 GHz. Interferometric Wide Swath mode covers 250 km at 5 x 20 m resolution; Extra Wide Swath covers 400 km at 20 x 40 m. Six-day repeat at the equator (three days with both satellites active) makes it the primary operational tool for coastal-state enforcement. Free and open archive from 2014.
- RADARSAT Constellation Mission (CSA): Three C-band satellites providing daily revisit over Canada's coastal zones and near-daily global coverage. Medium-resolution mode delivers 16 m resolution over a 500 km swath. Designed explicitly for maritime surveillance and oil-spill response.
- Capella Space (commercial): X-band SAR (9.65 GHz) with spotlight modes down to approximately 0.5 m resolution. Shorter wavelength increases sensitivity to fine surface texture differences, but also increases susceptibility to wind-induced false alarms. Useful for high-confidence discrimination of slick boundaries and for tasking within hours of an incident report.
- COSMO-SkyMed (ASI): X-band constellation of four satellites, with spotlight modes to approximately 1 m. Used operationally by the European Maritime Safety Agency (EMSA) CleanSeaNet service for oil-spill detection in European waters, providing a documented public benchmark for X-band performance.
What a thin film of oil actually does to radar
The sea surface returns a strong radar signal because wind generates centimetre-scale capillary waves. These waves scatter energy back toward the sensor through Bragg scattering: the radar wavelength resonates with the periodic surface roughness, and the backscatter is predictable from the wave spectrum. C-band Sentinel-1 is sensitive to waves with a Bragg wavelength of roughly 5.6 cm, which is squarely in the capillary-wave regime.
Oil, even in a film only micrometres thick, suppresses those capillary waves by increasing surface viscosity. The dampened surface scatters far less energy back to the sensor, appearing as a dark patch against the surrounding grey ocean. The contrast can be striking: slicks from illegal bilge dumps, which are typically thin and elongated, can appear as dark streaks tens of kilometres long. Larger spills from casualties or subsea seeps produce broader, more irregular shapes.
The look-alike problem is not a footnote, it is the central difficulty
Any process that smooths the sea surface produces a dark patch in SAR imagery. Natural biogenic films from phytoplankton blooms are chemically similar enough to oil to create near-identical signatures. Low-wind areas, where there are simply not enough capillary waves to generate strong backscatter, are indistinguishable from oil slicks at first glance. Rain cells dampen the surface and create circular dark patches. Ship wakes can occasionally mimic elongated slicks.
This is why wind speed is the first diagnostic. Below roughly 2 to 3 m/s, the entire scene goes dark and oil detection becomes unreliable. Above approximately 10 to 12 m/s, wave breaking re-roughens the surface and thin films are masked. The operational window is narrow. Analysts combine SAR backscatter intensity with contextual layers: wind fields from ECMWF or ASCAT scatterometry, vessel detections from the same SAR scene, AIS track history, and knowledge of known shipping lanes. A dark patch with a vessel at its head and no AIS transmission is a very different proposition from a dark patch in a known upwelling zone.
Polarimetric SAR adds discriminating power. Sentinel-1 acquires dual-polarisation data (VV and VH in IW mode). Oil slicks dampen both channels, but the ratio and the degree of polarisation differ from biogenic films in ways that published research has shown to be statistically separable, though not perfectly so. Honest assessment: even with polarimetric features and contextual data, false-alarm rates in operational systems remain non-trivial, and human analyst review of candidate detections is standard practice.
Sentinel-1's six-day repeat and what it means for enforcement
A coastal state trying to prosecute an illegal bilge dump faces a fundamental evidence problem. The vessel has moved on, the slick has dispersed, and the incident happened at night. SAR changes the geometry of that problem. Sentinel-1's six-day repeat (three days with two satellites) means that for a coastal zone covered by multiple ascending and descending passes, effective revisit can be shorter. EMSA's CleanSeaNet service, which is publicly documented, has processed Sentinel-1 imagery for oil-spill detection across European waters since the mission's launch, providing a real-world benchmark for what operational coverage looks like.
The key workflow is retrospective correlation. When a slick is detected, analysts reconstruct which vessels were in the vicinity using the AIS record and SAR-derived vessel detections from the same image. Drift modelling, using surface current and wind data, can back-propagate the slick to estimate its origin point and time. That narrows the candidate vessel list. It does not constitute proof on its own, but it is the kind of specific, documented chain of evidence that port-state control authorities and flag-state prosecutors can work with.
Bilge dumps versus accidental spills: the shape tells a story
Accidental spills from casualties tend to be point-source events: a roughly circular or lobate patch that grows and fragments over time. Illegal bilge dumps are different. A vessel pumping oily water while underway produces an elongated streak whose length encodes the duration of pumping and whose orientation matches the vessel's track. Widths of a few hundred metres and lengths of tens of kilometres are documented in the literature for single dump events.
This morphological signature is useful but not conclusive. Currents and wind shear distort the streak over hours. A slick detected six to twelve hours after the event may be fragmented, curved, or partially dispersed. Thickness estimates from SAR are not reliable: the radar contrast saturates quickly and cannot distinguish a very thin widespread film from a thicker concentrated one. For thickness and volume estimates, optical sensors or airborne fluorometry are needed, neither of which is available from space in operational conditions with the same coverage.
Building a monitoring programme that holds up in court
Detection is only the first step. Coastal enforcement agencies need a documented, repeatable process whose outputs can survive legal scrutiny. That means calibrated imagery with known acquisition parameters, a transparent detection algorithm (threshold-based or machine-learning classifiers trained on labelled data), documented wind conditions at the time of acquisition, and a clear chain of custody for the data products.
Satellize structures analytics pipelines to produce outputs with full provenance metadata, which matters when a detection report becomes evidence. The Sentinel-1 archive back to 2014 means that a newly identified vessel of interest can be checked against years of historical passes over its known operating areas. Repeat offenders leave patterns. For clients who need higher revisit or finer resolution on specific areas of interest, commercial SAR tasking through Capella or similar providers can be added on client licence.
One practical note on latency: Sentinel-1 Near Real-Time products are available within one to three hours of acquisition for designated areas, which is fast enough to vector airborne or vessel-based response to a fresh slick. Standard products arrive within twenty-four hours. For retrospective analysis, the full archive is accessible through the Copernicus Data Space Ecosystem.
Typical figures
| Primary sensor frequency | C-band, 5.405 GHz (Sentinel-1, RADARSAT); X-band, ~9.65 GHz (COSMO-SkyMed, Capella) |
| Sentinel-1 IW mode resolution | 5 m range x 20 m azimuth (ground range detected, multi-look) |
| Sentinel-1 EW mode resolution | 20 m range x 40 m azimuth; 400 km swath |
| Sentinel-1 repeat cycle | 6 days (single satellite); ~3 days effective with two satellites; shorter for high-latitude coastal zones |
| Operational wind window for detection | Approximately 3 to 10 m/s; detection unreliable below 2 m/s or above ~12 m/s |
| Minimum detectable slick area | Approximately 0.1 km² in favourable wind conditions (published EMSA CleanSeaNet operational experience); smaller features possible with X-band spotlight |
| Product latency (Sentinel-1 NRT) | 1 to 3 hours post-acquisition for designated NRT zones |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); continuous |
| Polarisation (Sentinel-1 IW maritime) | Dual-pol VV+VH standard over ocean |
| Delivery formats | GeoTIFF, NetCDF, shapefile/GeoJSON detection polygons, PDF enforcement report |
Analytics Satellize can run
| Slick detection and classification | Thresholded backscatter segmentation combined with polarimetric feature extraction (VV/VH ratio, degree of linear polarisation); look-alike filtering using coincident ECMWF wind fields | GeoJSON polygon layer of candidate slicks with confidence score and wind-condition flag, updated per Sentinel-1 pass |
| Vessel-to-slick attribution | Spatial and temporal correlation of SAR-derived vessel detections and AIS track history with slick head position; Leeway drift back-propagation using ERA5 wind and CMEMS surface current fields | Attribution report naming candidate vessels with positional evidence, timeline, and drift-model uncertainty radius |
| Coastal zone persistent monitoring | Automated processing of every Sentinel-1 pass over a defined exclusive economic zone or coastal corridor; alert triggered when candidate slick exceeds area and confidence thresholds | Email or API alert within 4 hours of acquisition, with image chip and metadata; monthly summary statistics for enforcement planning |
| Historical pattern analysis | Retrospective search of Sentinel-1 archive over defined area and vessel operating history; frequency, location and morphology of past detections aggregated by grid cell | Heat-map GIS layer of historical slick frequency; vessel-specific incident log where attribution is possible |
| Slick morphology classification | Shape-index and elongation metrics applied to detected polygons to distinguish point-source spills from elongated bilge-dump streaks; orientation compared to vessel track azimuths | Classification tag (probable dump / probable spill / ambiguous) appended to each detection polygon, with supporting morphometric data |
| Commercial SAR gap-fill tasking | When Sentinel-1 revisit is insufficient for an active incident or high-priority area, Capella or COSMO-SkyMed tasking requested on client licence; X-band imagery processed through the same detection pipeline | Merged multi-sensor detection layer with provenance metadata per image; tasking coordination 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.