Open-loop scrubber washwater discharge plume detection in coastal waters

Open-loop scrubbers clean exhaust by flooding it with seawater, then discharge that acidic, PAH-laden washwater overboard. Sentinel-2 coastal and blue bands, fused with AIS tracks, can detect and attribute the resulting plumes in ports and restricted zones.

Sensors

Sentinel-2 MSI: 10 m resolution in the coastal blue (Band 1, 443 nm) and visible bands; 5-day revisit at the equator with both satellites. The coastal aerosol band is specifically sensitive to chromophoric dissolved organic matter (CDOM) and turbidity increases near the surface, making it the primary detection channel for scrubber plumes in clear coastal water.
Landsat 9 OLI: 30 m resolution across coastal/blue/green/NIR bands; 16-day single-satellite revisit, reducing to 8 days when combined with Landsat 8. Provides archive continuity back to 2013 (Landsat 8) for historical baseline construction. Coastal Band 1 (435–451 nm) responds to CDOM and suspended particulate matter.
Sentinel-3 OLCI: 300 m resolution with 21 spectral bands from 400 to 1020 nm and roughly daily revisit. Insufficient resolution to isolate individual ship plumes in busy ports, but useful for detecting persistent discharge accumulation in semi-enclosed bays and for flagging anomalous water-leaving radiance fields that trigger tasking of higher-resolution assets.
AIS (Automatic Identification System): Vessel position, heading, speed, MMSI and ship type broadcast every 2–10 seconds at close range. Terrestrial AIS covers ports reliably; satellite AIS (from operators such as Spire) extends coverage to open coastal waters. Track interpolation between fixes allows back-calculation of a ship's position at the time a plume pixel was imaged.

What the washwater actually contains, and why it shows up optically

An open-loop scrubber takes raw seawater, blasts it through the exhaust stream to absorb SO2 and particulate matter, then discharges the spent water overboard at rates that can exceed 45 cubic metres per megawatt-hour of engine output. The discharged water is acidic (pH as low as 2.5 before dilution), elevated in polycyclic aromatic hydrocarbons, heavy metals, and suspended particulate matter stripped from the exhaust. It is also warmer than ambient seawater.

The optical consequence is a two-part signal. First, suspended particulate matter raises water-leaving radiance in the blue and green bands, producing a turbidity signature detectable in Sentinel-2 Band 1 (443 nm) and Band 2 (490 nm). Second, PAHs and other dissolved organic compounds absorb strongly at ultraviolet and short-blue wavelengths, increasing CDOM absorption and altering the spectral shape of the water-leaving reflectance. In clear, low-turbidity coastal water, a freshly discharged plume can produce a reflectance anomaly of a few percent above background, which is within Sentinel-2's detection capability under calm, cloud-free conditions. The signal decays as the plume disperses and dilutes, typically over hours to a day depending on tidal mixing.

The honest limits of optical plume detection

Cloud cover is the dominant operational constraint. Sentinel-2 and Landsat 9 are passive optical sensors: a single layer of cloud over a port renders the acquisition useless. In the Baltic Sea, where scrubber discharge is prohibited under HELCOM regulations, cloud-free acquisition probability on any given day is well below 50% across much of the year. Detection programmes therefore depend on statistical accumulation across many passes rather than guaranteed daily surveillance.

Plume persistence is the second constraint. A ship moving at 12 knots discharges continuously but the plume disperses rapidly in tidal or wind-driven currents. By the time a satellite passes over, the plume may be undetectable or displaced kilometres from the ship's current position. Detection is most reliable when a vessel is slow-steaming, manoeuvring in port, or anchored. The 10 m resolution of Sentinel-2 helps resolve narrow near-wake plumes, but 30 m Landsat pixels can miss thin features entirely.

Natural turbidity variation, river outflow, algal blooms, and sediment resuspension from wind and waves all produce water-leaving radiance signals that overlap with the scrubber plume signature. Distinguishing anthropogenic discharge from background requires a well-characterised local baseline and, ideally, multi-date confirmation. A single anomalous pixel is not evidence of a violation; a coincident AIS track from a vessel type known to carry scrubbers is what converts an optical anomaly into an attribution candidate.

Fusing AIS tracks with plume geometry to attribute discharges

The published approach, applied in Baltic and North Sea monitoring studies, proceeds in three steps. First, candidate plume pixels are identified by comparing water-leaving reflectance in Band 1 and Band 2 against a multi-year seasonal baseline for the same location and sun angle. Pixels exceeding a threshold anomaly, typically expressed as a z-score against the baseline distribution, are flagged. Second, AIS records are queried for all vessels within a search radius around each flagged pixel at the satellite overpass time, accounting for plume drift using a simple surface current model or tidal atlas. Third, vessels are ranked by proximity, heading alignment with the plume axis, and ship type (vessels fitted with scrubbers are identifiable from IMO ship type codes and publicly maintained scrubber installation registries).

The result is a probabilistic attribution rather than a legal proof. A vessel whose track passes through the plume origin point within the plume's estimated age window, and whose ship type is consistent with scrubber installation, is a strong candidate. Multiple coincident detections of the same vessel across different passes substantially increase confidence. Port state control authorities in Finland, Sweden, and Belgium have used exactly this framework to initiate inspections, with physical water sampling then providing the evidentiary standard for enforcement.

Where detection works best, and where it does not

Semi-enclosed, low-turbidity waters with slow tidal exchange are the best environment. The Baltic Sea is the canonical case: low natural turbidity, a clear prohibition on open-loop discharge under HELCOM rules, and high vessel traffic density. Norwegian fjords and some Mediterranean ports share similar optical clarity. Detection in the southern North Sea, the English Channel, or estuarine ports such as Rotterdam is substantially harder because natural suspended sediment loads are high and the background signal is variable.

Ports are better than open sea for a different reason: vessels spend more time stationary or slow-moving, giving plumes time to accumulate and persist long enough for a satellite to catch them. Anchorage areas just outside port limits are a particular concern, since vessels sometimes discharge while waiting for a berth, beyond the jurisdiction of port reception facility requirements but within the discharge-prohibited zone of regional seas agreements.

Sentinel-3 OLCI's daily revisit at 300 m is genuinely useful for monitoring large anchorage areas and bay-scale accumulation, even if it cannot resolve individual ship wakes. Combining OLCI daily screening with Sentinel-2 targeted follow-up on flagged areas is a practical two-tier architecture.

From detection to enforcement: what the data product looks like

A working monitoring product delivers three things to a port state or maritime authority: a georeferenced plume polygon with spectral anomaly magnitude and estimated age; a ranked list of candidate vessels with MMSI, ship name, flag state, and track geometry; and a confidence score based on the number of independent lines of evidence (optical anomaly strength, AIS track alignment, ship type, historical violation record).

Satellize runs this fusion workflow on open Sentinel and Landsat imagery combined with satellite AIS feeds, producing attribution reports as GIS layers and structured alerts. The workflow is the same class of method applied in the Tonga crop-estimation programme, adapted from area-based anomaly detection to point-source plume attribution. Authorities using the output still need to commission physical water sampling for prosecution, but the satellite product identifies which vessels warrant inspection and reduces the cost of enforcement to a manageable level.

The archive depth of Sentinel-2 (from 2015) and Landsat 8/9 (from 2013) means historical compliance can be reconstructed for specific vessels or routes, which is useful in civil penalty proceedings where a pattern of behaviour is relevant to the penalty calculation.

Typical figures

Primary spatial resolution	10 m (Sentinel-2 Band 1/2); 30 m (Landsat 9 OLI Band 1/2); 300 m (Sentinel-3 OLCI)
Revisit at a given coastal point	5 days (Sentinel-2A+B combined); 8 days (Landsat 8+9 combined); ~1 day (Sentinel-3 OLCI)
Key spectral bands	Coastal aerosol (~443 nm) and blue (~490 nm) for CDOM/turbidity; green (~560 nm) for suspended particulate contrast; NIR for water masking
Minimum detectable plume width	Approximately 20–30 m in Sentinel-2 under low-turbidity background; broader threshold in turbid coastal water
Plume detection latency	6–24 hours from satellite overpass to processed anomaly layer, depending on cloud screening and processing pipeline
AIS positional accuracy	Typically <10 m for Class A transponders; satellite AIS message latency 30–90 minutes depending on provider and orbit
Cloud-free acquisition probability	Highly variable: ~30–50% per pass in Baltic winter; higher in Mediterranean summer. No optical detection through cloud.
Archive depth	Sentinel-2 from mid-2015; Landsat 8 from 2013; Landsat 9 from late 2021
Coverage	Global coastal zones; all regional seas with discharge prohibition regulations
Delivery formats	GeoTIFF anomaly rasters, GeoJSON plume polygons, CSV attribution candidate tables, PDF enforcement summary reports

Analytics Satellize can run

Plume anomaly map	Coastal/blue band water-leaving reflectance z-score against multi-year seasonal baseline; water masking via NIR threshold	GeoTIFF raster layer with per-pixel anomaly score, clipped to port or restricted zone boundary
Plume polygon and age estimate	Connected-component labelling of anomaly pixels; plume age estimated from vessel speed, heading, and surface drift model	GeoJSON polygon with metadata: area, peak anomaly magnitude, estimated age window, acquisition timestamp
Vessel attribution ranking	AIS track interpolation to overpass time; spatial intersection with plume origin zone; ship-type scrubber probability weighting	CSV table of candidate vessels ranked by attribution confidence, including MMSI, flag, ship type, and track geometry
Historical compliance profile	Retrospective plume detection across full Sentinel-2 and Landsat archive for a defined vessel or route corridor	Time-series report of detected anomalies with associated AIS candidates; suitable for use in penalty proceedings
Port-area discharge heatmap	Temporal accumulation of plume anomaly detections across all cloud-free passes over a defined period	GeoTIFF density raster showing spatial concentration of discharge events; supports zonal risk prioritisation
Sentinel-3 screening alert	Daily OLCI water-leaving radiance anomaly detection over anchorage and bay areas; threshold exceedance triggers Sentinel-2 tasking request	Automated alert with flagged coordinates and anomaly magnitude; feeds into next available Sentinel-2 acquisition queue

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.