LNG carrier laden versus ballast state discrimination from SAR freeboard geometry
High-resolution spotlight SAR can classify LNG carriers as laden or ballast by measuring freeboard geometry against hull markings, enabling cargo flow inference that bypasses AIS declarations entirely.
Sensors
- ICEYE X-band SAR (spotlight mode): Achieves approximately 0.25 m azimuth resolution in spotlight, sufficient to resolve hull markings and waterline position on vessels longer than 200 m. Revisit at any given ocean point varies by tasking priority but can reach sub-daily with the full constellation.
- Capella Space SAR-X (spotlight mode): Delivers 0.35 m resolution in its highest-resolution spotlight collect. X-band penetrates light rain but is sensitive to sea-surface roughness, which can obscure the waterline in sea states above Beaufort 4-5.
- COSMO-SkyMed Second Generation (spotlight mode): X-band system offering approximately 0.35 m resolution in Enhanced Spotlight. Operated by the Italian Space Agency and defence establishment; commercial tasking available. Useful for cross-checking ICEYE or Capella collects.
- Sentinel-1 (IW and EW modes): Free and open C-band SAR with 5 x 20 m resolution in Interferometric Wide Swath mode. Insufficient resolution to discriminate freeboard at the waterline, but useful for vessel detection, heading, and length estimation to cue higher-resolution tasking.
What the waterline actually reveals
A fully laden LNG carrier displaces substantially more water than the same vessel in ballast. For a typical Q-Flex carrier of around 210,000 cubic metres capacity, the difference in mean draught between laden and ballast conditions is roughly 6 to 9 metres. That translates directly into a visible freeboard change: the painted load-line markings and Plimsoll disc on the hull sit measurably higher or lower relative to the waterline.
In sub-metre spotlight SAR, the waterline appears as a sharp boundary between high-backscatter hull metal and the specular (low-backscatter) ocean surface. The position of that boundary relative to the vessel's deck edge and visible hull features is the primary signal. This is geometry, not radiometry, which makes it relatively insensitive to sensor calibration drift across different satellites.
Why sub-metre resolution is not optional
The freeboard difference between laden and ballast states is large enough to see, but the hull markings that anchor the measurement are painted stripes and characters roughly 0.5 to 1 metre tall. Resolving them demands pixel sizes well below one metre. Sentinel-1's 5 x 20 m pixels make this impossible: you can confirm a vessel is an LNG carrier by length and radar cross-section, but you cannot read its load state.
ICEYE and Capella spotlight modes sit comfortably inside the required resolution envelope. COSMO-SkyMed Second Generation does too. The practical floor is around 0.5 m pixel spacing, and even then the analyst is working with a small number of pixels across the freeboard zone. Automated classifiers trained on manually labelled imagery have been shown in published remote-sensing literature to achieve useful accuracy on tanker load-state discrimination at this resolution, though performance degrades on vessels shorter than roughly 150 m where the freeboard zone is narrower.
One honest caveat: the method produces a binary classification, laden or ballast, not a continuous cargo volume estimate. Partial loads, which occur during ship-to-ship transfers or when a carrier departs a terminal before full capacity, are difficult to distinguish from a laden state without additional contextual data.
Sea state is the principal enemy
Ocean swell and wind-driven chop scatter radar energy unpredictably near the hull. In sea states above roughly Beaufort 4 (significant wave height above 1.5 m), the waterline boundary becomes noisy and the freeboard measurement degrades. X-band SAR is more sensitive to this than C-band because shorter wavelengths interact more strongly with small-scale surface roughness.
Practical mitigation involves two things. First, collect scheduling should favour low sea-state windows where forecast data permits. Second, analysts should flag collects acquired in poor conditions rather than force a classification. A confident null result is more useful to a trading desk than a low-confidence binary call.
Connecting the classification to cargo flow
A single laden-or-ballast classification on a single vessel is a data point. The intelligence product is built by tracking the same vessel across multiple collects and correlating state transitions with port calls inferred from AIS, Spire Global's AIS feed, or Aireon's space-based AIS. A vessel that is classified as ballast leaving a known discharge terminal and laden arriving at a loading terminal has, with high probability, completed an unloaded repositioning voyage. The reverse transition confirms a cargo delivery.
This matters most when AIS declarations are absent, manipulated, or simply delayed. Vessels engaged in sanctions-sensitive trades have been documented turning off or spoofing AIS transponders. SAR freeboard classification provides an independent physical check. It does not identify the cargo origin or contract terms, but it does confirm whether a cargo moved.
Combining freeboard classification with terminal activity data from Sentinel-1 vessel detection and optical imagery of loading arms and jetty occupancy builds a more complete picture of LNG flows than any single data stream alone. Satellize integrates these layers for clients running commodity flow programmes, drawing on the same multi-source fusion methodology applied in its Tonga crop-estimation work.
Tasking geometry and archive limits
Commercial SAR constellations do not provide continuous coverage of open ocean. Tasking a specific vessel requires knowing its approximate position in advance, which is why AIS remains useful even when its declarations are suspect: it provides a search area. ICEYE's constellation of more than 30 satellites as of 2024 can revisit a defined ocean area within hours under tasking priority, but latency from collect to delivered image is typically 30 to 90 minutes after acquisition depending on ground-station geometry.
Archive depth for sub-metre spotlight imagery is limited. ICEYE and Capella began building commercial archives from around 2019 and 2020 respectively, so historical reconstruction of cargo flows beyond roughly five years is not currently possible from these sensors. Sentinel-1's archive extends to 2014 and covers most ocean areas systematically, but its resolution precludes freeboard classification. It can, however, provide historical vessel detection and routing context.
Honest limits and what to do about them
Three failure modes deserve explicit acknowledgement. First, very-large carriers (Q-Max class, around 266,000 cubic metres) have longer hulls and larger freeboard differences, making classification easier. Smaller LNG carriers, including some used in regional Asian trades, have freeboard differences that approach the noise floor at sub-optimal viewing angles. Second, simultaneous ballasting operations, where a vessel takes on seawater ballast while discharging cargo, can produce intermediate freeboard states that confuse binary classifiers. Third, ice-class vessels with reinforced hulls and different hull-marking conventions require separate classifier training.
None of these limits are fatal to the method. They argue for combining freeboard classification with vessel-type databases, AIS-derived port call histories, and, where available, optical imagery of cargo manifold connections at terminals. The freeboard signal is best treated as a strong prior, not a definitive verdict.
Typical figures
| Minimum SAR resolution for freeboard classification | 0.5 m pixel spacing or better (spotlight mode required) |
| Typical ICEYE spotlight resolution | ~0.25 m azimuth, ~0.5 m range |
| Typical Capella SAR-X spotlight resolution | ~0.35 m |
| Sentinel-1 IW resolution (vessel detection cueing only) | 5 x 20 m |
| Revisit (commercial tasking, ICEYE constellation) | Sub-daily at priority; typically 3-6 hours for defined ocean areas |
| Image-to-delivery latency (commercial SAR) | 30-90 minutes post-acquisition (ground-station dependent) |
| Radar frequency | X-band (9.6 GHz) for ICEYE, Capella, COSMO-SkyMed; C-band (5.4 GHz) for Sentinel-1 |
| Sea-state reliability limit | Beaufort 4 or below (significant wave height < ~1.5 m) for reliable waterline detection |
| Commercial SAR archive depth | ICEYE and Capella from ~2019-2020; Sentinel-1 from 2014 |
| Minimum vessel length for reliable classification | ~150 m (shorter vessels have insufficient freeboard zone pixel count) |
Analytics Satellize can run
| Laden/ballast state classification per vessel collect | Waterline boundary detection in SAR amplitude imagery via edge-detection algorithms; freeboard zone pixel analysis relative to hull-feature anchors | Per-vessel binary classification report with confidence score and sea-state flag, delivered as structured JSON or CSV within two hours of image acquisition |
| Voyage cargo-state timeline | Temporal fusion of repeated SAR classifications with AIS-derived port call events; state-transition logic to infer loading and discharge events | Vessel-level cargo event log covering defined monitoring period, delivered as a GIS-linked timeline or spreadsheet |
| LNG flow inference between terminal pairs | Aggregation of vessel-level cargo-state timelines across a defined fleet; origin-destination inference from laden departure and ballast arrival events at named terminals | Monthly or weekly flow matrix (volume in vessel-voyages, not tonnes) between monitored terminal pairs, delivered as a structured report |
| AIS dark-vessel cargo-state check | SAR vessel detection on Sentinel-1 to identify AIS-dark vessels; spotlight tasking on identified targets; freeboard classification on acquired imagery | Alert report on AIS-dark LNG carriers with laden/ballast classification and last-known position, delivered within six hours of SAR acquisition |
| Fleet-wide ballast ratio monitoring | Statistical aggregation of classified collects across a defined fleet or trade route; ballast ratio computed as fraction of observed vessel-days in ballast state | Weekly dashboard metric indicating fleet utilisation and implied cargo demand, delivered as a data feed or PDF summary |
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.