Crude tanker cargo volume estimation from waterline draft

High-resolution optical and SAR imagery can infer crude cargo load by measuring freeboard against published draft-displacement tables for known tanker classes, cross-referenced with AIS vessel identity. Accuracy depends on class identification and is honest about intermediate-load ambiguity.

Sensors

Pléiades Neo: 30 cm panchromatic resolution enables direct pixel-level measurement of the painted freeboard band on a tanker hull. Revisit at a given point is roughly 1 to 3 days with constellation tasking. Cloud cover remains the primary constraint.
WorldView-3: 31 cm panchromatic, 1.24 m multispectral. The SWIR bands (1195 to 2365 nm across eight channels) can help discriminate hull paint from waterline foam, reducing freeboard measurement error in moderate sea states.
Sentinel-1 SAR (C-band, 5.405 GHz): Interferometric Wide Swath mode delivers 10 m ground range resolution with 6-day repeat at mid-latitudes, all-weather. Backscatter intensity and incidence-angle geometry allow laden versus light-ballast discrimination at the binary level; sub-class draft gradations require optical confirmation.
AIS vessel registry data: Automatic Identification System transponder broadcasts give MMSI, vessel name, IMO number, reported destination and speed. Cross-referencing AIS identity with Lloyd's or Equasis class records provides the deadweight tonnage (DWT) and draft table needed to convert freeboard measurement to cargo volume estimate.

What the waterline actually encodes

Every commercial tanker is built to a known displacement curve. For a VLCC (Very Large Crude Carrier, typically 200,000 to 320,000 DWT), the difference between a light-ballast draft of roughly 9 to 10 metres and a fully laden draft of 20 to 22 metres is visible from space as a change in freeboard of 10 metres or more. That is not subtle. At 30 cm resolution, Pléiades Neo resolves the painted load-line marks on the hull directly, and the waterline itself sits within a few pixels of a measurable edge.

The physics is straightforward Archimedes: displacement equals the mass of water displaced, and for salt water at standard density (1,025 kg/m³) the relationship between draft and displacement is published for every class in Lloyd's Register and equivalent classification society tables. Cargo volume in barrels follows from displacement minus the vessel's lightship weight, divided by crude oil density, which varies by grade but sits between roughly 820 and 870 kg/m³ for most traded crudes. The uncertainty introduced by oil density alone is on the order of 5 to 7 percent of cargo volume, which is the honest floor for any remote-sensing estimate regardless of image quality.

Optical measurement: reading the freeboard band

The freeboard band, the painted section of hull between the waterline and the main deck, is the measurement target. In a nadir or near-nadir Pléiades Neo or WorldView-3 image, the analyst identifies the vessel outline, locates the waterline edge using a contrast gradient between hull paint and water, and measures the vertical freeboard in pixels scaled to ground sample distance. Off-nadir collection angles introduce a geometric foreshortening correction that depends on the vessel's heading relative to the sensor azimuth and the collection elevation angle. This correction is well-understood and documented in maritime photogrammetry literature, but it adds uncertainty of roughly 0.5 to 1.5 metres in draft estimate for collection angles beyond 25 degrees off nadir.

Sea state matters. Wave action at Beaufort 4 and above creates enough hull motion and spray to blur the waterline edge in a single-pass image. Pléiades Neo's 30 cm imagery is collected in under a millisecond of integration time, which largely freezes wave motion, but the waterline position itself oscillates. Analysts should treat freeboard measurements from single images in moderate sea states as carrying an additional uncertainty of plus or minus 0.5 metres of draft, equivalent to roughly 5,000 to 15,000 tonnes of cargo uncertainty depending on vessel class.

SAR backscatter: coarser but cloud-immune

Sentinel-1 C-band SAR does not resolve individual load-line marks. At 10 m pixel spacing, a VLCC is roughly 35 pixels long and 5 pixels wide in Interferometric Wide Swath mode. What SAR does offer is a reliable binary: a laden vessel sits low and presents a different radar cross-section geometry than the same vessel in light ballast, where the high freeboard creates a taller vertical face. Studies published in Remote Sensing (MDPI) have demonstrated that C-band backscatter intensity and the aspect ratio of the vessel's bright return are statistically separable into laden and ballast states for VLCCs and Suezmax vessels, though the discrimination degrades for Aframax and smaller classes where the freeboard difference is smaller relative to the resolution cell.

The practical workflow combines both sensors. Sentinel-1 provides all-weather, frequent screening across wide ocean areas to flag vessels of interest and classify them as laden or ballast. Optical tasking via Pléiades Neo or WorldView-3 is then triggered selectively for vessels where a quantitative cargo estimate is required, not just a binary state. This keeps commercial tasking costs proportionate to intelligence value.

AIS cross-referencing and the class-identification problem

The draft-to-cargo conversion only works if you know which vessel class you are measuring. A 30 cm image shows you freeboard; it does not directly tell you the vessel's lightship weight or its hull form coefficient. That information comes from AIS cross-referencing. The IMO number broadcast via AIS links to classification society records that specify DWT, summer load-line draft, and in many cases the full hydrostatic table.

Two complications arise. First, vessels engaged in sanctions evasion or cargo obfuscation sometimes transmit false identities or go dark on AIS entirely. A vessel with a spoofed MMSI can be misclassified by class, introducing a systematic error in the cargo estimate that can exceed 20 percent. Second, even with correct identity, intermediate load states, say 40 to 70 percent of capacity, are harder to bracket precisely because the freeboard change per unit of cargo is smaller in the mid-draft range where hull form is most curved. The honest position is that remote freeboard analysis distinguishes laden, light-ballast and near-full states reliably, but intermediate loads carry wider confidence intervals, typically plus or minus 15 to 25 percent of cargo volume.

What the method cannot do

It cannot see through cloud. Optical freeboard measurement is unavailable during persistent overcast, which is common in the North Sea, the Gulf of Guinea and parts of the South China Sea for weeks at a time. SAR fills the gap for binary state detection but not for quantitative volume estimates.

It cannot identify cargo grade. Freeboard tells you mass, not chemistry. Whether a laden VLCC carries Arab Light, Urals or West African crude requires corroboration from port call history, loading terminal identification, and cargo documentation, none of which are derivable from imagery alone. The method is also silent on partial transfers: ship-to-ship transfers at anchor can change cargo volume without a port call appearing in AIS history, and if the imagery cadence misses the transfer event, the estimate will be wrong. Satellize incorporates AIS dark-period detection and anchorage-pattern analysis, drawing on the same optical and SAR archive, to flag these events as a separate analytic layer rather than pretending they do not happen.

Putting the estimate to work

The practical output is a cargo-volume bracket, expressed in barrels with a stated confidence interval, attached to a specific vessel at a specific time and position. For commodity traders, the value lies in aggregation: tracking a fleet of VLCCs departing a loading region over a month gives a flow estimate that is independent of reported export statistics. For compliance teams, a single vessel showing a laden state at a position inconsistent with its declared route is a flag for further investigation.

Satellize runs this analytic as part of its broader vessel-activity intelligence suite, combining open Sentinel-1 screening with commercial optical tasking on client licence. The approach is the same one underlying our crop-estimation work in Tonga, where the principle is identical: measure a physical quantity that correlates with an economic one, state the uncertainty honestly, and let the client decide what the signal is worth. Analysts who want to explore a specific vessel list or trade route can request a scoped trial run covering a defined period and geography.

Typical figures

Optical spatial resolution	30 cm (Pléiades Neo panchromatic), 31 cm (WorldView-3 panchromatic)
SAR spatial resolution	10 m ground range (Sentinel-1 IW mode); 5 m in Stripmap mode (narrower swath)
Optical revisit (tasked)	1 to 3 days for Pléiades Neo constellation; 1 to 4.5 days for WorldView-3 depending on latitude and off-nadir tolerance
SAR revisit	6-day exact repeat for Sentinel-1 at mid-latitudes; combined ascending and descending passes reduce effective gap to 3 days over many ocean areas
Freeboard measurement uncertainty (optical, calm sea)	±0.3 to 0.8 m draft equivalent; improves with multiple passes
Cargo volume uncertainty (laden/ballast binary, SAR)	State classification only; no volume figure
Cargo volume uncertainty (quantitative, optical)	±5 to 25% depending on load state and vessel class confidence; worst at intermediate loads
Minimum detectable freeboard change	Approximately 1 to 2 m draft change reliably detectable at 30 cm optical resolution; equates to roughly 20,000 to 60,000 tonnes cargo change on a VLCC
Archive depth	Sentinel-1: from 2014; Pléiades (1A/1B): from 2012; WorldView-3: from 2014; AIS historical: varies by data provider
Delivery formats	GeoJSON vessel report, CSV cargo-estimate time series, georeferenced image chips, PDF intelligence brief

Analytics Satellize can run

Laden/ballast state classification	Sentinel-1 C-band backscatter intensity and vessel aspect-ratio analysis against class-specific thresholds derived from published maritime SAR studies	Daily GeoJSON feed with per-vessel state flag and confidence score
Quantitative cargo volume bracket	Optical freeboard measurement (pixel-edge detection on Pléiades Neo or WorldView-3) converted via published Lloyd's/classification society hydrostatic tables, corrected for off-nadir geometry	Per-vessel cargo estimate in barrels with stated confidence interval, delivered as structured report or API record
Fleet-level trade-flow aggregate	Aggregation of per-vessel cargo estimates across a defined loading region and time window, benchmarked against reported export statistics where available	Monthly or weekly CSV time series of estimated export volumes by region and vessel class
AIS dark-period and spoofing flag	Gap detection in AIS broadcast history cross-referenced with SAR vessel detection positions; identity inconsistency scoring against IMO registry	Alert report listing vessels with anomalous AIS behaviour during the observation window, with supporting image chips
Ship-to-ship transfer event detection	Anchorage-pattern analysis from AIS combined with optical or SAR imagery of two vessels in close proximity; freeboard change comparison before and after proximity event	Event log with timestamps, positions, estimated volume transferred and image evidence
Vessel class identification confidence score	Optical hull-dimension measurement (length, beam) compared against IMO-registered class specifications; AIS MMSI cross-check via Equasis or equivalent open registry	Per-vessel class confidence rating appended to all cargo estimate records

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.