Floating storage tanker inventory tracking for oil markets
Extended anchorage dwell times combined with laden freeboard signatures let satellite analysts estimate crude oil volumes held off-market in floating storage. Aggregated globally, the metric is an independent leading indicator for supply dynamics that no price feed or shipping report can replicate.
Sensors
- Planet PlanetScope: 3 m optical resolution with near-daily global revisit. Sufficient to measure hull freeboard visually and confirm vessel identity by deck geometry. Cloud cover is the primary limit; persistent overcast over key anchorages such as the Strait of Malacca or Gulf of Mexico can break the time series for days at a time.
- Maxar WorldView-3: 0.31 m panchromatic resolution, enabling precise freeboard measurement and hull-marking readout for identity confirmation. Tasked rather than systematic; best used for spot-verification of ambiguous cases flagged by coarser sensors. Latency from task to delivery is typically same-day to 48 hours depending on orbit geometry.
- Sentinel-1 SAR (C-band, ESA): 5 × 20 m resolution in Interferometric Wide Swath mode, 6-day repeat at the equator with both satellites active. Radar penetrates cloud and darkness, making it the backbone sensor for anchorage monitoring in monsoon-affected regions. Backscatter intensity and vessel length can be measured; freeboard cannot be read directly from SAR backscatter alone.
- Spire Global spaceborne AIS: Spire operates over 100 LEO satellites carrying AIS receivers, providing sub-hourly position updates for vessels broadcasting MMSI identifiers. AIS supplies vessel identity, speed-over-ground and draught as self-reported, linking optical and SAR detections to named ships. Self-reported draught is frequently inaccurate or stale, which is why freeboard measurement from imagery is the independent check.
Why a tanker's waterline is worth more than its manifest
A very large crude carrier (VLCC) in ballast rides roughly 10 metres above the waterline amidships. Fully laden, that freeboard drops to around 2 to 3 metres. The difference is visible in 3 m optical imagery and measurable with reasonable confidence from 0.31 m WorldView-3 imagery. Operators can and do falsify cargo declarations, delay port calls, and suppress AIS transmissions. They cannot alter the physics of displacement.
Floating storage is not a marginal phenomenon. During the 2020 contango event, independent analysts estimated that over 160 million barrels were held in floating storage at the peak, equivalent to roughly 1.6 days of global consumption. Tracking that inventory in near-real time requires watching hundreds of anchorages simultaneously. No human team can do that from ship-visit reports alone.
The two-signal method: dwell time plus freeboard
A vessel becomes a floating storage candidate when two conditions hold simultaneously over multiple satellite passes: it remains stationary at anchor beyond a threshold that typically ranges from 5 to 10 days depending on the analyst's model, and its freeboard signature indicates a laden state. Either signal alone is insufficient. A laden tanker waiting for a berth looks identical to one in deliberate storage from a single pass; the dwell duration resolves the ambiguity. A long-anchored tanker in ballast is probably waiting for a cargo assignment, not storing one.
The operational workflow chains three data types. Spire AIS establishes the dwell history and provides MMSI-linked identity. PlanetScope daily mosaics confirm position continuity and give a first-pass freeboard read. WorldView-3 is tasked selectively to produce a precise freeboard measurement for vessels where the PlanetScope read is ambiguous, typically ships anchored at angles that reduce the visible hull stripe. Sentinel-1 fills the cloud gaps, confirming the vessel is still present even when optical imagery is unusable, though it cannot independently confirm laden status.
Turning vessel counts into barrel estimates
Converting a freeboard observation into a volume estimate requires knowing the vessel's deadweight tonnage and hull form. Both are available from Lloyd's Register, Equasis, and similar public maritime databases, cross-referenced by MMSI or IMO number. A VLCC with a deadweight of 300,000 tonnes carrying crude at roughly 0.85 tonnes per cubic metre holds approximately 350,000 cubic metres, or about 2.2 million barrels, when fully laden. Partial loads introduce uncertainty; a vessel riding at half-freeboard might carry anywhere from 40 to 60 per cent of capacity depending on trim management.
Aggregate global estimates carry compounding uncertainty. Analysts typically report floating storage totals with an error band of plus or minus 10 to 20 per cent, which is honest given the freeboard measurement limits, partial-load ambiguity, and the possibility that some vessels are missed in cloudy regions. That said, even an imprecise global aggregate updated weekly is substantially more current than official inventory data, which in many jurisdictions lags by weeks or months.
Where the method breaks down
Persistent cloud is the most straightforward failure mode. The Strait of Malacca, the Gulf of Thailand, and West African anchorages all experience multi-week overcast periods during which optical freeboard measurement is impossible. Sentinel-1 maintains presence detection through cloud but does not substitute for the freeboard read. In practice, the last known laden status is carried forward with increasing uncertainty the longer the optical gap persists.
AIS spoofing and transponder suppression introduce identity risk. A vessel that turns off its AIS mid-anchorage can be tracked positionally by SAR and optical imagery but loses its MMSI link. Hull-geometry matching against a reference library of known vessels is the fallback, and it works reasonably well for distinct vessel classes such as ULCCs or FSOs, but is unreliable for mid-size tankers where hull profiles are similar across hundreds of ships. The method is also blind to vessels anchored in territorial waters where commercial satellite tasking is restricted or where the operator withholds imagery.
From raw detections to market intelligence
The analytical value is not in any single vessel observation but in the global aggregate time series. A rising floating storage count, particularly concentrated in regions distant from refining centres, signals that crude is being withheld from the market, either because producers are waiting for higher prices or because logistics bottlenecks have backed up supply. A rapid drawdown of floating storage, especially combined with port call activity, suggests a supply release that may not yet be visible in official inventory figures.
Satellize runs this type of anchorage-monitoring analysis on open constellations supplemented by commercial tasking, producing structured outputs that feed directly into commodity desk workflows. The Tonga crop-estimation programme established the organisation's approach to systematic multi-pass change detection, and the same temporal stacking logic applies here: what matters is not the snapshot but the trajectory. Clients receive a weekly global floating storage estimate broken down by region and vessel class, with confidence intervals and flagged anomalies where identity is uncertain.
The archive question: how far back can you look?
Sentinel-1 data is publicly archived from 2014. PlanetScope systematic archive coverage begins around 2016 for most regions. Spire AIS data is commercially available with a historical depth that varies by subscription. Together these archives allow retrospective analysis of floating storage behaviour across multiple market cycles, including the 2015 to 2016 oversupply period, the 2018 Iranian sanctions wave, and the 2020 demand collapse. Retrospective calibration of a floating storage model against known price events is the standard way to validate the method before deploying it as a live market signal.
Typical figures
| Optical spatial resolution | 0.31 m (WorldView-3 pan), 3 m (PlanetScope) |
| SAR spatial resolution | 5 × 20 m (Sentinel-1 IW mode) |
| Optical revisit | Near-daily (PlanetScope); tasked on demand (WorldView-3) |
| SAR revisit | 6-day repeat at equator with Sentinel-1A and 1B combined |
| AIS update frequency | Sub-hourly for most ocean regions (Spire Global LEO network) |
| Minimum detectable vessel length | Approximately 50 m in SAR IW mode; shorter vessels may be below noise floor |
| Freeboard measurement accuracy | ±0.5 to 1.5 m depending on vessel angle and imagery resolution; translates to ±10 to 25% volume uncertainty for partial loads |
| Dwell-time threshold for storage classification | Typically 5 to 10 days stationary at anchor, analyst-configurable |
| Archive depth | Sentinel-1 from 2014; PlanetScope from ~2016; AIS commercially variable |
| Delivery formats | GeoJSON vessel detections, CSV time-series aggregates, weekly PDF intelligence summary |
Analytics Satellize can run
| Global floating storage vessel count by region | Dwell-time filtering of AIS tracks combined with optical position confirmation across multi-pass PlanetScope mosaics | Weekly GeoJSON layer with vessel positions, MMSI, dwell duration, and laden/ballast flag |
| Aggregate crude volume estimate in floating storage | Freeboard-to-draught conversion using publicly registered deadweight tonnage from Lloyd's/Equasis, with confidence intervals reflecting partial-load ambiguity | Weekly CSV time series by region and vessel class, with upper and lower volume bounds |
| Cloud-gap continuity tracking via SAR | Sentinel-1 SAR vessel detection using CFAR (constant false alarm rate) algorithms to confirm anchorage presence when optical imagery is unavailable | SAR detection layer appended to main vessel feed, flagged as presence-only (no freeboard read) |
| Identity-uncertain vessel alert | Hull-geometry matching against reference vessel library when AIS is absent or spoofed; flagged with confidence score | Alert feed with candidate vessel identities, geometry match score, and recommended WorldView-3 tasking trigger |
| Floating storage trend signal for commodity desks | Rolling 4-week change in global laden-vessel dwell count, normalised by region and compared against historical baseline from Sentinel-1 and PlanetScope archive | Weekly one-page intelligence brief with directional signal (building / drawing / flat) and anomaly flags |
| Retrospective floating storage reconstruction | Batch reprocessing of Sentinel-1 archive (from 2014) and PlanetScope archive (from ~2016) using consistent dwell and freeboard criteria | Historical time-series dataset for model calibration, delivered as CSV with methodology note |
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.