SAR backscatter oil spill and marine slick detection
Synthetic aperture radar detects surface oil by the dampening effect it has on centimetre-scale sea roughness, producing dark patches in C- and X-band imagery. The method works within a narrow wind-speed window and cannot, on its own, distinguish mineral oil from fish-school slicks or rain cells.
Sensors
- Sentinel-1 IW / EW modes (C-band, 5.405 GHz): Interferometric Wide swath gives 10 m resolution across a 250 km swath; Extra Wide gives 40 m across 400 km. Dual-pol (VV+VH) is standard. Revisit is 6 days at the equator for a single satellite, 3 days with two, though the constellation has operated with gaps. Free and open archive from 2014.
- RADARSAT-2 (C-band): Commercial Canadian satellite with selectable polarisation modes including full quad-pol, which is essential for advanced oil-type discrimination. Fine-beam modes reach 3 m resolution. Tasking is on-demand but incurs cost; revisit depends on orbital geometry and latitude.
- ICEYE constellation (X-band, ~9.65 GHz): Stripmap mode delivers 3 m resolution; Spot mode reaches sub-1 m. The constellation of more than 30 satellites (as of 2024) can achieve same-day revisit over a target area, sharply reducing the drift window between spill detection and response vessel arrival.
- Capella Space (X-band): Spotlight mode at approximately 0.5 m resolution. Useful for confirming slick boundaries and identifying source vessels once a spill has been flagged by wide-area surveillance. Tasking latency can be under an hour in favourable orbital windows.
Why radar sees what optical sensors miss
Optical sensors cannot see through cloud, and most significant spill events occur in weather that makes optical tasking unreliable. SAR illuminates the sea surface with microwave energy and measures what bounces back. On an uncontaminated sea surface in moderate wind, short gravity-capillary waves, the Bragg-resonant waves at centimetre scales, scatter C-band energy efficiently back to the sensor. Backscatter is high; the sea looks bright.
A surface oil film, even a very thin one, suppresses the surface tension and viscous damping that sustain those small waves. The film does not need to be thick crude: a monomolecular layer of biogenic surfactant produces the same dampening effect. The SAR image records a dark patch where the sea appears unusually smooth. That contrast against the surrounding roughened surface is the detection signal. It is a physical consequence of fluid dynamics, not a spectral signature, which is why it works day and night and through cloud.
The wind-speed window, and what falls outside it
Detection is only reliable within a wind-speed range of roughly 3 to 12 metres per second. Below about 3 m/s, the whole sea surface becomes smooth and dark; there is no contrast against which to see a slick. Above roughly 12 to 14 m/s, wave-breaking turbulence overwhelms any film damping and the slick disappears into a uniformly bright, rough sea. These thresholds are not sharp lines; they shift slightly with sea state, fetch and the viscosity of the oil itself. Any operational system must ingest co-located wind-field data, from scatterometers such as MetOp ASCAT or from SAR-derived wind retrievals, and flag imagery acquired outside the detection window as unreliable.
The fundamental ambiguity is harder to resolve than the wind constraint. Biogenic slicks from phytoplankton, fish schools and natural seeps produce backscatter signatures indistinguishable from mineral oil in single-polarisation intensity images. Rain cells create circular dark patches. Low-wind shadow zones behind islands or ships look identical in amplitude. No single SAR image can resolve this unambiguously. That is an honest limit, and anyone claiming otherwise is selling something the physics does not support.
What polarimetry adds, and what it cannot guarantee
Full or compact polarimetric SAR, available on RADARSAT-2 and some ICEYE modes, measures not just backscatter intensity but the polarisation state of the return signal. Parameters derived from the polarimetric covariance matrix, particularly the conformity coefficient and the degree of polarisation, have been shown in peer-reviewed literature to differ between mineral oil, plant-derived biogenic films and clean water. The separation is probabilistic, not categorical. Thick emulsified oil produces more distinctive polarimetric signatures than a thin sheen, which may be nearly indistinguishable from a biogenic film even with quad-pol data.
The practical implication: polarimetric SAR narrows the ambiguity and raises the confidence of a mineral-oil classification, but it does not eliminate false alarms. Operational systems combine polarimetric features with contextual layers, including vessel AIS history, known shipping lanes, pipeline routes and wind-field plausibility, to assign a confidence score rather than a binary classification. That score, not a definitive label, is what a responsible analytic product delivers.
Sentinel-1 for surveillance, commercial SAR for response
Sentinel-1's Extra Wide swath mode covers 400 km in a single pass at 40 m resolution. At that scale, a slick of a few hundred metres across is detectable, and the free, open archive allows retrospective analysis of spill histories going back to 2014. For routine maritime zone surveillance, particularly for exclusive economic zones running to 200 nautical miles, Sentinel-1 is the cost-effective backbone. Its 3-to-6-day revisit is adequate for monitoring chronic low-level discharge patterns over weeks.
The problem is response. A spill detected in a Sentinel-1 pass may be 12 to 72 hours old by the time the next overpass confirms its position. Oil drifts. A slick that appeared 40 km offshore may have reached a sensitive coastline before a response vessel arrives based on stale coordinates. Commercial constellations from ICEYE and Capella, with same-day or sub-day revisit and resolutions below 3 m, allow responders to track slick movement in near-real time and direct vessels to the actual current position rather than the historical detection point. The two tiers are complementary, not competing.
From dark patch to actionable intelligence
Raw detection of a dark patch is the beginning of the analytic chain, not the product. An operational pipeline adds wind-field screening to reject low-wind false alarms, morphological analysis to distinguish elongated slicks from circular rain cells, vessel proximity analysis from AIS data to identify probable sources, and drift modelling using ocean current and wind forecasts to project slick position forward in time. The output of that chain is a confidence-ranked alert with a probable source attribution and a projected drift envelope.
Satellize runs this pipeline on open Sentinel-1 acquisitions and can add commercial tasking from ICEYE or Capella under client licence for time-critical confirmation passes. The analytic approach follows published methods for SAR-based oil detection, including the dark-object segmentation and feature-classification frameworks documented extensively in the Remote Sensing journal literature. Delivery formats are GIS layers and structured alert feeds, not PDF reports that arrive after the slick has moved on.
Honest limits, stated plainly
Thin sheens below roughly 0.1 micrometres thickness may fall below the detectable contrast threshold, particularly in moderate-to-high wind. Slicks smaller than the sensor's spatial resolution, below 10 m for Sentinel-1 IW or below 3 m for ICEYE Stripmap, will not be resolved as discrete objects. Cloud has no effect on SAR, but heavy rain does introduce noise and can create false dark patches. Polarimetric discrimination of oil type remains a probabilistic tool; published false-alarm rates in operational studies range widely depending on sea state and slick type.
Archive depth for Sentinel-1 runs from 2014, giving a decade of baseline. RADARSAT-2 archive access is commercial and selective. ICEYE archive coverage depends on tasking history. No spaceborne SAR system currently provides the sub-hour revisit that would make real-time spill tracking possible without multiple satellites being tasked simultaneously, which is expensive. These constraints belong in any honest procurement conversation.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 10 m range × 10 m azimuth (multi-looked product) |
| Spatial resolution (ICEYE Stripmap / Spot) | 3 m / sub-1 m |
| Swath width (Sentinel-1 EW) | 400 km |
| Revisit (Sentinel-1, two-satellite constellation) | 3 days at mid-latitudes; 6 days single satellite |
| Revisit (ICEYE constellation) | Same-day tasking feasible over most maritime zones |
| Radar frequency | C-band: 5.405 GHz (Sentinel-1, RADARSAT-2); X-band: ~9.65 GHz (ICEYE, Capella) |
| Operational wind-speed window | Approximately 3 to 12 m/s; detection unreliable outside this range |
| Minimum detectable slick area (Sentinel-1 IW) | Approximately 0.1 km² under favourable wind conditions; smaller with commercial X-band |
| Sentinel-1 archive depth | 2014 to present (open access) |
| Typical detection-to-alert latency (Sentinel-1 NRT) | 1 to 3 hours after acquisition via Copernicus NRT processing |
Analytics Satellize can run
| Slick detection and extent mapping | Adaptive thresholding and dark-object segmentation on calibrated sigma-nought backscatter, with wind-field masking to reject low-wind false alarms | GIS polygon layer (GeoJSON / Shapefile) of detected slick boundaries with area estimate and acquisition timestamp |
| Confidence-ranked mineral-oil classification | Polarimetric feature extraction (conformity coefficient, degree of polarisation) combined with contextual scoring from AIS vessel proximity and pipeline proximity layers | Alert feed with per-slick confidence score (high / medium / low) and probable source attribution |
| Slick drift projection | Lagrangian particle drift modelling using CMEMS ocean current forecasts and ECMWF wind fields, initialised from SAR-detected slick centroid | Projected slick position envelope at 6, 12 and 24 hours, delivered as polygon layer for response vessel routing |
| Chronic discharge pattern analysis | Multi-temporal stacking of Sentinel-1 detections over 30 to 90 day windows to identify recurring slick origin points indicative of operational discharge | Heatmap layer of slick frequency by location, with candidate source vessel or infrastructure list |
| Commercial confirmation tasking | On-demand ICEYE or Capella Spotlight acquisition triggered by Sentinel-1 detection alert, with sub-3 m resolution follow-up image | High-resolution SAR image chip of slick with updated boundary polygon, typically within 4 to 12 hours of tasking request |
| EEZ surveillance report | Automated processing of all Sentinel-1 passes over a defined maritime zone, with monthly aggregation of detection events | Monthly PDF and GIS summary of all detected slick events, wind-validity flags, and confidence classifications within the client's jurisdiction |
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.