Commodity flow intelligence from port vessel and yard activity
Fusing AIS vessel tracking with optical and SAR imagery of port berths, anchorages, and stockpile yards lets analysts estimate bulk commodity flows before official trade statistics are published. The method is probabilistic and the uncertainty is real, but so is the lead time.
Sensors
- Spire Global AIS (space-based): Spire's constellation of over 100 LEO satellites collects Automatic Identification System messages globally, with typical message latency under 30 minutes and near-global coverage that eliminates the coastal-only blind spots of terrestrial receivers. Vessel MMSI, position, speed, draught, and declared destination are the raw inputs for flow inference.
- exactEarth AIS (space-based): An alternative or complementary space-based AIS provider, useful for cross-validation and for filling gaps when a single constellation has a coverage hole. Combining two AIS feeds materially reduces the probability of missing a vessel departure or arrival event.
- Planet Dove (optical): Planet's Dove constellation provides daily 3-metre optical imagery of most port facilities. At that resolution, vessel presence and approximate size class are identifiable; stockpile footprint area is measurable. Cloud cover remains the primary operational constraint, particularly at tropical and high-latitude ports.
- Sentinel-1 SAR (C-band): The European Space Agency's Sentinel-1A and 1B satellites carry a C-band synthetic aperture radar operating at 5.405 GHz. SAR penetrates cloud and operates day and night, making it the primary sensor for vessel detection in persistently overcast regions. Ground range detected mode gives 10-metre resolution with a 250-kilometre swath; revisit at mid-latitudes is typically six days for a single satellite, three days when both are operational.
What the AIS record actually contains, and what it conceals
Every vessel over 300 gross tonnes on international voyages is required under SOLAS to broadcast AIS: position, speed, heading, MMSI, vessel name, draught, and a free-text destination field. Space-based AIS receivers from Spire and exactEarth collect these messages globally, giving analysts a near-complete record of vessel movements that updates continuously. The draught field is particularly useful. A Capesize bulk carrier loaded with iron ore sits around 18 metres deep; the same vessel in ballast sits around 9 metres. Declared draught, cross-checked against vessel class and deadweight tonnage from Lloyd's or similar registers, gives a first-order estimate of cargo quantity before the vessel arrives anywhere.
The limitations are equally important to state plainly. Destination fields are entered manually by crew and are frequently wrong, abbreviated, or left as the previous voyage's entry. AIS spoofing, where a vessel transmits a false position, is documented in several regions, particularly in sanctioned-trade corridors. Dark-vessel behaviour, where transponders are switched off deliberately, is not rare among tankers and bulk carriers moving cargo under sanctions scrutiny. A space-based AIS feed catches most commercial traffic but it does not catch everything, and the analyst who treats it as ground truth will be wrong in exactly the cases that matter most.
SAR and optical imagery close the gaps AIS leaves open
A vessel that has turned off its AIS transponder cannot turn off its radar cross-section. Sentinel-1 SAR detects vessels as bright point targets against the darker sea background, with a minimum detectable vessel length of roughly 20 to 30 metres in standard Interferometric Wide Swath mode. Comparing the SAR-detected vessel count in an anchorage against the AIS-reported count gives a direct measure of dark-vessel activity at that location on that date. Planet Dove optical imagery adds a visual confirmation layer on cloud-free days, and at 3-metre resolution the vessel class, approximate length, and loading state are distinguishable for vessels over about 100 metres.
Port yard imagery fills a different gap. AIS tells you a bulk carrier departed a Brazilian iron ore terminal loaded. Optical imagery of the Ponta da Madeira or Tubarão terminals can show whether the stockpile behind the berth shrank after that departure, which is corroborating evidence of actual loading rather than a repositioning voyage with ballast. Stockpile footprint change is measurable at Planet resolution; volumetric change requires either a digital elevation model or a shadow-geometry approach, which is covered separately in the coal and grain stockpile pages in this library.
Building a flow estimate: the inference chain and its error budget
A commodity flow estimate is a chain of probabilistic inferences, not a measurement. The chain runs roughly as follows. Vessel class and deadweight tonnage come from a vessel register. Draught at departure gives a cargo-weight estimate with an uncertainty of perhaps plus or minus 10 to 15 per cent depending on draught-reading accuracy and trim. Commodity type is inferred from vessel class, origin terminal type, and, where available, bill-of-lading data or port authority disclosures. Origin and destination are confirmed by AIS waypoints and, where the declared destination is implausible, by the actual port of arrival derived from subsequent AIS messages.
Each step introduces uncertainty that compounds. A Capesize arriving at a Chinese port could be carrying Australian coal or South African coal or, in some periods, Russian coal. Without additional signals, the origin is probabilistic. The honest approach is to publish a probability distribution over commodity origins rather than a single point estimate, and to update that distribution as subsequent imagery and AIS data arrive. Aggregated over dozens of voyages per month, the law of large numbers reduces the portfolio-level error considerably, even when individual voyage attribution remains ambiguous.
Official customs statistics, when they arrive, serve as a calibration benchmark. The published literature on satellite-derived trade flow estimation, including work using AIS and optical data for oil and dry-bulk markets, generally finds that satellite-derived monthly flow estimates precede official statistics by four to eight weeks and correlate with them at the portfolio level, though individual voyage misclassification rates vary significantly by commodity and region.
Where the method works well and where it does not
The method is most reliable for large, single-commodity terminals with limited vessel type diversity: iron ore at Ponta da Madeira, coal at Newcastle (Australia), crude oil at Ras Tanura. At these facilities, vessel class almost uniquely identifies commodity, terminal layout is stable, and high vessel traffic provides statistical power. It is least reliable at mixed-cargo ports, at facilities with persistent cloud cover that limits optical confirmation, and in regions with documented AIS manipulation.
Grain is harder than iron ore. Grain terminals handle multiple crop types on the same berth, and vessel class does not distinguish wheat from soya. Seasonal patterns and origin-port identity help, but the commodity-type uncertainty is genuinely higher. Crude oil tanker flows are well-studied and the AIS-plus-SAR method is mature for that commodity; the sibling page on LNG terminal throughput covers the gas-specific variant. For sanctioned-origin flows, where the incentive to manipulate AIS is highest, the dark-vessel detection from SAR becomes the primary signal, and the uncertainty bands widen accordingly.
From data to a decision-relevant product
Commodity traders, sovereign wealth funds, and government trade ministries use these estimates differently. A trader wants a weekly flow number for a specific commodity corridor with a confidence interval, delivered before the price-sensitive official release. A government ministry wants to know whether a trading partner's declared import volumes are consistent with what satellites observe at their ports, as a check on trade-balance reporting. The analytic product has to be designed around the decision it informs.
Satellize structures commodity flow analytics as a recurring feed rather than a one-off study, because the value is in the time series. A single month's estimate is interesting; twelve months of consistently produced estimates, calibrated against subsequent official releases, is a validated signal. The Satellize analytics team applies the same approach to agricultural commodity estimation that it uses in the Kingdom of Tonga crop programme: define the inference chain, quantify the uncertainty at each step, and report both the estimate and the error honestly. Subscribers to the Overhead column will recognise the format.
Typical figures
| Optical spatial resolution (Planet Dove) | 3 metres per pixel (SuperDove); vessel class identifiable above ~100 m length |
| SAR spatial resolution (Sentinel-1 IW mode) | 10 metres ground range; minimum detectable vessel ~20–30 m length |
| Optical revisit (Planet Dove) | Daily at most latitudes; cloud cover reduces effective revisit at tropical ports |
| SAR revisit (Sentinel-1) | 6 days single satellite; ~3 days with both 1A and 1B operational |
| AIS message latency (space-based) | Typically under 30 minutes from transmission to receipt (Spire, exactEarth) |
| AIS coverage | Near-global for vessels >300 GT on international voyages; dark vessels not detected by AIS |
| Cargo weight estimation uncertainty | ±10–15% per voyage from draught inference; reduces at portfolio level |
| Lead time over official customs statistics | 4–8 weeks typical, based on published AIS trade-flow literature |
| Archive depth | Sentinel-1 from 2014; Planet Dove from ~2017; AIS commercial archives vary by provider |
| Delivery formats | Structured JSON or CSV flow feed, GIS vessel-track layers, PDF monthly summary report |
Analytics Satellize can run
| Weekly commodity flow estimates by corridor | AIS draught-based cargo-weight inference combined with vessel-register classification and origin/destination confirmation from waypoint analysis | Recurring structured data feed (CSV/JSON) with point estimate and confidence interval per commodity corridor |
| Dark-vessel detection at target anchorages | Sentinel-1 SAR vessel detection via constant false alarm rate (CFAR) processing, cross-referenced against concurrent AIS vessel list | Alert layer (GIS polygon) flagging anchorage locations where SAR-detected vessel count exceeds AIS-reported count by a configurable threshold |
| Vessel dwell-time and queue analysis | AIS time-series segmentation to identify arrival, anchoring, berth occupation, and departure events; optical imagery cross-check on cloud-free days | Port congestion time series and berth utilisation report, updated weekly |
| Port yard stockpile footprint change | Multi-date Planet Dove optical imagery with supervised change detection on stockpile polygon footprints; area change correlated with vessel loading events | Monthly stockpile-area change layer (GeoTIFF) and tabular summary per terminal |
| Sanctioned-origin flow probability scoring | Bayesian origin attribution combining AIS last-known port, dark-vessel gap analysis, vessel-history flags, and terminal-type priors | Per-voyage probability score report with methodology note and uncertainty statement |
| Official-statistics calibration audit | Retrospective comparison of satellite-derived monthly flow estimates against subsequently published customs data to quantify systematic bias and adjust forward estimates | Quarterly calibration report with bias-correction factors by commodity and corridor |
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.