Oil spill and illicit discharge detection at sea and near terminals

Surface oil films dampen short ocean waves, creating dark signatures in SAR imagery that persist through cloud and darkness. Sentinel-1, COSMO-SkyMed and ICEYE make systematic and rapid-response detection operationally viable.

Sensors

Sentinel-1 C-band SAR (ESA): 10 m ground range resolution in IW mode, 250 km swath, 6-day repeat at the equator and 2-3 days at mid-latitudes with both satellites. Systematic global acquisition makes it the baseline archive for spill history and routine patrol. C-band (5.405 GHz) is sensitive to the Bragg-scale capillary waves that oil suppresses, but wind speed must sit between roughly 2 and 14 m/s for reliable contrast; calm seas and very rough seas both produce ambiguities.
COSMO-SkyMed X-band SAR (ASI): Spotlight mode delivers 1 m resolution; Stripmap delivers 3-15 m over a 40 km swath. Constellation of four satellites supports tasking within hours for incident response. X-band is slightly more sensitive to thin films than C-band, though the wind-speed ambiguity window is similar.
ICEYE X-band SAR (commercial): Stripmap mode at approximately 3 m resolution, Spot mode at sub-1 m. Revisit to a specific point can be arranged within hours on a tasking basis. Useful for confirming slick boundaries and tracking drift when a spill is already located.
Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands, 20 m red-edge and SWIR bands, 5-day revisit with both satellites. On cloud-free days, oil on calm water produces characteristic colour differences in the visible bands and suppressed SWIR reflectance. Useful for extent mapping and distinguishing weathered oil from fresh, but entirely blind through cloud or at night.

What a slick actually looks like to a radar

Ocean radar backscatter is dominated by centimetre-scale capillary waves riding on longer gravity waves. Surface-active hydrocarbons, including crude oil, refined products and biogenic films, reduce surface tension and damp those short waves. The result is a patch of sea that returns less energy to the satellite antenna than the surrounding water, appearing as a dark region against a brighter background in SAR imagery.

The physics sets hard limits. Wind speed below about 2 m/s produces a naturally calm sea where everything looks dark; wind above roughly 14 m/s breaks up thin films and mixes them into the water column, erasing the signature. Between those bounds, the method works reliably. Slick thickness also matters: films thinner than a few microns may not produce enough damping to clear the noise floor, though even monomolecular layers of some compounds are detectable under good conditions.

The same dark-patch signature is produced by natural look-alikes: biogenic surface films from algae, rain cells, current boundaries, and low-wind shadow zones behind islands or vessels. Distinguishing anthropogenic oil from these requires shape analysis, contextual knowledge of vessel traffic and terminal locations, and ideally a second acquisition or optical confirmation.

Routine patrol versus incident response: two different acquisition strategies

Sentinel-1's systematic global acquisitions, archived continuously since 2014 for Sentinel-1A and 2016 for Sentinel-1B (before Sentinel-1B's end of operations in 2021), make it the natural baseline for patrol programmes. An operator monitoring a shipping lane or a terminal approach can run automated dark-object detection on every overpass, flag candidates, and queue them for analyst review. At mid-latitudes the revisit is short enough to track a spill's drift over successive passes and estimate its source timing.

Incident response is a different problem. Once a spill is reported, the priority is the highest-resolution, fastest-available image of the current extent. COSMO-SkyMed and ICEYE can be tasked to deliver sub-3 m imagery within a few hours of a request, depending on orbital geometry. That resolution resolves slick boundaries sharply enough to support containment-boom deployment decisions and legal documentation. The trade-off is cost per acquisition and the need for a pre-arranged tasking contract.

Optical imagery: useful when the sky cooperates

On a cloud-free day, Sentinel-2 and very-high-resolution commercial optical sensors add information that SAR cannot. Fresh crude oil appears dark brown to black in true colour; emulsified oil takes on an orange or reddish-brown hue. Weathered sheens produce iridescent colour in the visible bands. The SWIR bands (1610 nm and 2190 nm on Sentinel-2) show suppressed reflectance over oil-covered water, which helps separate thin sheens from sediment plumes that can look similar in visible light.

The honest caveat is that the North Sea, Gulf of Mexico, and most operationally significant spill zones are frequently overcast. SAR is not a backup to optical; for maritime spill detection it is the primary sensor. Optical confirmation, when available, is valuable for thickness estimation and for producing evidence-quality imagery that is easier for non-specialists to interpret in legal or regulatory proceedings.

Illicit discharge: the detection challenge is not physics, it is attribution

Operational discharges from tank washing, bilge pumping and cargo residue flushing are illegal under MARPOL Annex I but widespread. They are typically performed at night, in poor weather, or in areas with low patrol density. SAR detects the slick regardless of time of day or cloud cover. The attribution problem is harder.

Automatic Identification System (AIS) data, cross-referenced with the SAR acquisition time and the slick's estimated drift trajectory, can place a vessel at the origin point. The drift calculation requires a surface current model and a wind field estimate for the hours between discharge and detection, introducing uncertainty that grows with time elapsed. Slicks older than roughly 12-24 hours become difficult to attribute confidently because they have dispersed, fragmented, and drifted beyond the range that a simple back-trajectory can pin to a single source. This is the core evidentiary limitation that any serious enforcement programme must acknowledge.

Near-terminal monitoring: a different geometry

Pipeline ruptures and terminal spills produce slicks that originate at a fixed, known point rather than a moving vessel. That simplifies attribution but complicates detection: shallow water near jetties and breakwaters produces strong SAR returns from infrastructure and vessel wakes that can mask a thin film. Very-high-resolution X-band imagery from COSMO-SkyMed or ICEYE, with its finer spatial discrimination, performs better in congested nearshore scenes than Sentinel-1's 10 m IW mode.

Repeated baseline acquisitions from Sentinel-1 over a terminal area establish what the scene normally looks like, making anomalous dark regions easier to flag. This change-detection approach is more reliable than single-image thresholding in complex coastal geometry. Satellize applies this kind of baseline-and-anomaly framework in its analytics work, and the same logic that underpins its Tonga crop-estimation programme, comparing current state against a calibrated seasonal baseline, transfers directly to terminal-area spill monitoring.

One further limit: SAR cannot measure oil volume or thickness with useful accuracy from backscatter alone. Slick area can be mapped; the mass of oil discharged cannot be read directly from the image. Thickness estimation requires multi-polarisation analysis or optical spectral data, and even then the figures carry wide uncertainty bands.

What the archive is good for beyond live incidents

The Sentinel-1 archive back to 2014 is a historical record of surface anomalies over any ocean area the programme has systematically acquired. For a government establishing a new maritime enforcement regime, or an insurer assessing chronic discharge risk along a shipping lane, that archive is a starting point for baseline frequency analysis. How often do dark slicks appear near a given anchorage? Do they correlate with vessel traffic patterns from AIS? Are there seasonal patterns consistent with weather windows that make detection harder?

These are answerable questions from open data. The answers do not constitute legal proof of individual violations, but they do provide the statistical foundation for prioritising patrol resources and for demonstrating to regulators or courts that a systematic monitoring programme is in place.

Typical figures

Sentinel-1 SAR spatial resolution (IW mode)	10 m range × 22 m azimuth (ground range detected)
COSMO-SkyMed SAR spatial resolution	1 m (Spotlight) to 15 m (Stripmap)
ICEYE SAR spatial resolution	~3 m (Stripmap); <1 m (Spot mode)
Sentinel-1 revisit (mid-latitudes, single satellite)	6 days; 2-3 days with constellation coverage
Sentinel-2 optical revisit	5 days at equator with both satellites
Radar frequency	C-band 5.405 GHz (Sentinel-1); X-band ~9.6 GHz (COSMO-SkyMed, ICEYE)
Effective wind-speed window for slick detection	Approximately 2-14 m/s; detection unreliable outside this range
Minimum detectable slick area (SAR, IW mode)	Typically ~0.1 km² under good wind conditions; thin sheens may require larger extent
Archive depth (Sentinel-1)	Sentinel-1A from April 2014; continuous systematic acquisition
Delivery formats	GeoTIFF slick-extent polygons, GeoJSON alert feeds, PDF incident reports with drift trajectory overlay

Analytics Satellize can run

Routine slick-detection alerts	Automated dark-object segmentation on SAR backscatter (constant false alarm rate thresholding or neural-network classifier trained on labelled Sentinel-1 scenes), filtered by wind-speed mask from ERA5 or ECMWF operational fields	Near-real-time alert feed (GeoJSON) with candidate slick polygons, confidence score, and acquisition timestamp, delivered within 2-4 hours of satellite overpass
Look-alike discrimination report	Multi-feature analysis combining slick shape metrics (elongation, fractal dimension), proximity to vessel AIS tracks, sea-state context and, where available, cross-polarisation ratio (VV/VH) to separate biogenic films from petroleum	Analyst-reviewed PDF report classifying each candidate as probable petroleum, probable biogenic, or ambiguous, with supporting evidence
Drift back-trajectory and source attribution	Lagrangian particle back-tracking using slick centroid and CMEMS ocean surface current analysis combined with ERA5 wind drift correction; uncertainty envelope computed over elapsed time since estimated discharge	Map layer showing probable discharge origin zone with time-stamped uncertainty polygon; cross-referenced AIS vessel list within the zone at discharge time
Slick extent and area quantification	Polygon delineation from SAR segmentation; area computed in equal-area projection; optical Sentinel-2 SWIR composite added on cloud-free acquisitions for independent extent cross-check	GIS layer (Shapefile or GeoPackage) with area in km², perimeter, and centroid coordinates; versioned across successive acquisitions to show drift and dissipation
Historical discharge frequency baseline	Time-series analysis of Sentinel-1 archive over defined area of interest; slick occurrence frequency mapped per grid cell; correlation with AIS vessel density and seasonal wind climatology	Annual or quarterly frequency heatmap (GeoTIFF) and summary report suitable for regulatory submission or insurance underwriting
Terminal-area anomaly monitoring	Change detection against multi-temporal SAR baseline for fixed terminal geometry; anomalous low-backscatter regions flagged against expected scene statistics	Weekly monitoring report with flagged anomaly images and comparison against baseline; escalation alert on confirmed deviation

Sources

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.