Crude oil tanker cargo estimation from vessel draught and AIS
Fusing AIS position data with very-high-resolution optical and SAR imagery lets analysts estimate vessel draught and infer crude oil, LNG, or dry-bulk cargo loads independent of reported manifests. The physics is simple; the execution requires honest accounting of resolution limits, spoofing risk, and hull-class uncertainty.
Sensors
- Planet SkySat: Optical imagery at approximately 0.5 m native resolution (50 cm ground sample distance after processing), capable of resolving the painted load-line marks and waterline position on large tankers. Revisit is on-demand tasking; with the full constellation of around 21 satellites, same-day collection over a given port or anchorage is achievable on most days. Cloud cover is the principal operational constraint.
- Capella Space SAR: X-band synthetic aperture radar operating in spotlight mode at resolutions down to approximately 0.35 m. SAR is unaffected by cloud or darkness, making it the primary sensor for night departures and overcast anchorages. Freeboard measurement from SAR requires careful calibration: the bright double-bounce return from the hull-water interface provides a detectable edge, but sub-metre accuracy in draught estimation demands well-characterised incidence angles.
- Spire AIS: Spire operates a constellation of more than 100 small satellites receiving Automatic Identification System (AIS) transmissions globally, with typical message latency of a few minutes for open-ocean vessels. The feed provides MMSI, position, speed, heading, reported draught, and voyage data. Reported draught in AIS is self-declared by the master and is notoriously unreliable; it is used here as a cross-check, not a ground truth.
- Sentinel-1 SAR: ESA's Sentinel-1A and 1B (when operational) provide C-band SAR with a 10 m resolution Interferometric Wide Swath mode covering large ocean areas every 6 to 12 days at mid-latitudes. Insufficient resolution for individual vessel draught measurement, but valuable for fleet-wide vessel detection, anchorage monitoring, and confirming ship-to-ship transfer events that may precede AIS manipulation.
What a waterline actually encodes
Every commercial vessel is required under SOLAS and the International Load Line Convention to carry painted Plimsoll marks on its hull. These marks define the maximum legal draught for different sea conditions and seasons. The gap between the waterline and the deck edge, the freeboard, shrinks as cargo is loaded. For a VLCC (Very Large Crude Carrier) with a deadweight tonnage of around 300,000 DWT, the difference between a fully loaded draught of roughly 22 metres and a ballast draught of around 10 metres is visible from space as a colour band change on the hull, typically from red anti-fouling paint below the waterline to black or grey topsides above it.
This physical relationship is the foundation of the method. Cargo volume is not directly measured; it is inferred from draught via the vessel's hydrostatic tables, which are class-specific and publicly available for standard hull forms. A 1-metre error in draught translates to roughly 10,000 to 20,000 tonnes of cargo uncertainty on a VLCC, depending on hull form. That uncertainty range matters enormously for anyone using this data to estimate cargo value or verify a bill of lading.
Reading the waterline from orbit
In optical imagery at 0.5 m resolution, the waterline on a large tanker spans roughly 4 to 8 pixels in the across-track direction. A trained edge-detection algorithm, or a human analyst with experience, can place the waterline to within approximately one pixel, giving a freeboard estimate with an uncertainty of around 0.5 to 1 metre under favourable conditions. Glint, sea state, and vessel motion during the integration period all degrade accuracy. Vessels at anchor in sheltered water produce the cleanest measurements; vessels underway in moderate swell produce the worst.
SAR freeboard measurement works differently. The bright return from the hull-water interface creates a detectable edge in high-resolution spotlight imagery, and the geometry of the incidence angle allows a height estimate. Capella Space's 0.35 m spotlight mode is currently the most capable commercial option. Published research on SAR-based freeboard measurement for sea ice (where the technique is well-established) suggests centimetre-level precision is achievable in controlled conditions, but vessel hulls are more complex scatterers than ice sheets. Practical draught uncertainty in commercial SAR at today's resolutions is likely 0.5 to 2 metres, though peer-reviewed validation specific to tanker hulls remains sparse in the open literature.
AIS as context, not ground truth
AIS data from satellite receivers like Spire's constellation provides the positional and identity layer that makes vessel-specific draught measurement possible. Without AIS, a satellite image shows a ship; with AIS, it shows a named vessel with a known hull class, allowing the correct hydrostatic table to be applied.
The problem is that AIS is trivially manipulated. Vessels engaged in sanctions-evasion routinely transmit false positions, false destinations, or simply go dark. A vessel that has disabled its transponder will not appear in the AIS feed but may still appear in SAR imagery. The reverse is also possible: a vessel can transmit a valid AIS position while physically located elsewhere, a spoofing pattern detectable by comparing the AIS-reported position against the actual pixel location of the hull in a satellite image collected at the same time. This cross-validation is one of the more operationally valuable things that imagery-AIS fusion enables, though it requires near-coincident collection, which is a scheduling challenge given that most commercial tasking windows are planned hours in advance.
Revisit rates and the slow-vessel problem
A fully laden VLCC travels at roughly 13 to 15 knots. At anchor it is stationary. The practical challenge is collecting imagery at the right moment: after loading is complete but before departure, or after a ship-to-ship transfer in international waters. Anchorages are easier; a vessel may sit for days, giving multiple collection opportunities across optical and SAR sensors. Open-ocean transfers are harder. A transfer event can complete in 12 to 24 hours, and if cloud cover prevents optical collection during that window, the opportunity is lost until the next SAR pass.
Sentinel-1's 6 to 12 day revisit is too slow for tracking individual events but adequate for monitoring known anchorages and flagging unusual clustering of vessels. Commercial tasking of SkySat or Capella at 1 to 2 day revisit over a specific anchorage is the practical solution for time-sensitive monitoring. The cost of that tasking rises quickly when covering large ocean areas, which is why analytic prioritisation, using AIS anomaly detection to cue imagery collection, is essential rather than optional.
Uncertainty budgets and what they mean for users
Buyers of this analysis need an honest uncertainty budget before they act on it. The chain of inference runs: pixel-to-waterline edge (±0.5 to 1 m in good optical conditions), waterline to draught (adds tide and sea-state corrections, perhaps ±0.3 m), draught to displacement via hydrostatic tables (well-characterised for standard hull classes, less so for non-standard or modified vessels), displacement to cargo mass (requires subtracting lightship weight and bunker fuel, both of which carry their own uncertainty). Stacking these, a cargo mass estimate for a VLCC in good conditions might carry a total uncertainty of ±3 to 8 per cent of deadweight. For a 300,000 DWT vessel carrying crude at around $80 per barrel, that range spans tens of millions of dollars in cargo value.
LNG carriers are a harder case. Their cargo is cryogenic and stored in insulated tanks; the hull draught still reflects load, but the relationship between draught and cargo volume depends on boil-off rates and tank geometry that are not always publicly documented for a specific vessel. Dry-bulk carriers are somewhat easier because their hydrostatic tables are well-standardised and their hulls are less complex.
Satellize applies this draught-inference method as part of its broader commodity-flow analytics work, alongside the kind of physical-proxy approaches it uses in programmes such as the Kingdom of Tonga crop-estimation engagement. The method is the same in structure: replace a reported number with a physical measurement, then quantify honestly what the measurement cannot resolve.
Practical limits the analyst must state upfront
Cloud cover over tropical anchorages can persist for days, blocking optical collection entirely. Night-time departures are invisible to optical sensors; SAR is the only option, and commercial SAR tasking over open ocean is not always available on short notice. Hull fouling and repainting can alter the colour contrast at the waterline, reducing edge-detection confidence. Vessels in convoy or at crowded anchorages may be partially occluded in imagery, making waterline extraction unreliable for the obscured vessel.
None of these limits make the method useless. They make it a probabilistic intelligence product rather than a certified measurement. For commodity traders, hedge funds, and insurers using this data to inform positions or underwriting decisions, the right framing is: this is a physical cross-check on reported cargo, with a known uncertainty range, that is independent of the counterparty's incentive to misreport. That independence is the value, not the precision.
Typical figures
| Best optical resolution (SkySat) | ~0.5 m ground sample distance |
| Best SAR resolution (Capella spotlight) | ~0.35 m |
| Sentinel-1 SAR resolution (IW mode) | 10 m (vessel detection only, not draught measurement) |
| Revisit for targeted tasking (SkySat or Capella) | 1 to 2 days over a fixed anchorage; same-day possible but not guaranteed |
| AIS message latency (Spire satellite AIS) | Typically a few minutes for open-ocean vessels |
| Draught estimation uncertainty (optical, good conditions) | ±0.5 to 1 m |
| Cargo mass uncertainty (VLCC, stacked errors) | Approximately ±3 to 8% of deadweight |
| Minimum vessel size for draught measurement | Practical floor around Aframax class (~115,000 DWT); smaller hulls have insufficient freeboard contrast at 0.5 m resolution |
| SAR all-weather capability | Cloud and darkness do not affect SAR; sea state affects edge-detection quality |
| Archive depth (Sentinel-1) | From 2014 (Sentinel-1A launch); commercial SAR archives vary by operator |
Analytics Satellize can run
| Vessel draught estimate | Waterline edge detection in VHR optical imagery or SAR hull-water interface return; hull-class hydrostatic lookup | Per-vessel draught value with uncertainty bounds, delivered as a structured data record keyed to MMSI and timestamp |
| Cargo load estimate (tonnes) | Draught-to-displacement conversion via published hydrostatic tables; subtraction of lightship weight and estimated bunker fuel | Cargo mass estimate with stated uncertainty range, in a tabular report or API feed |
| AIS position verification | Cross-referencing AIS-reported position against actual pixel location of vessel hull in near-coincident satellite image | Spoofing flag (confirmed / suspected / unresolved) attached to each vessel observation record |
| Anchorage occupancy and vessel clustering alerts | Sentinel-1 IW SAR vessel detection over known anchorages; change detection against historical baseline | Automated alert when vessel count at a monitored anchorage exceeds a defined threshold, with image chip attached |
| Ship-to-ship transfer detection | AIS proximity analysis combined with SAR imagery confirming two or more large vessels stationary in open water | Event report with imagery evidence, vessel identities (where AIS is available), estimated timing, and cargo-state change before and after |
| Fleet-level cargo flow index | Aggregated draught observations across a monitored fleet or trade route, normalised by vessel class and deadweight | Weekly time-series GIS layer and summary statistics suitable for commodity-flow modelling |
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.