Ship tracking data and satellite images: what they reveal about commodities

In depth12 min read · 19 sources

Ship data from AIS reveals where tankers and cargo ships are sailing right now, at least where someone is receiving their radio signals. Researchers use it to build real-time measures of world trade, investors count tankers at chokepoints such as Hormuz, and satellite images even estimate oil inventories from the shadows on tank roofs. Here you will learn how this works, what freely available data can do, and why forecasts based on it are harder than they sound.

What AIS is and who has to transmit

The Automatic Identification System (AIS) is a radio system for safety at sea. Every equipped ship automatically broadcasts over VHF who it is, where it is and where it is going; other ships and shore stations receive this. According to IMO (2015), the SOLAS convention requires AIS on all ships of 300 GT (gross tonnage) and above on international voyages, on cargo ships of 500 GT and above on domestic voyages too, and on all passenger ships. Large tankers and freighters therefore almost always transmit, as long as nobody switches the device off.

The IMO guidelines distinguish three groups of information:

• Static: identifier (MMSI), name, IMO number, length, beam and ship type (entered once on installation).
• Dynamic: position, time, course and speed (automatic); only the navigational status, such as “at anchor”, is set manually by the watch.
• Voyage-related: draft, hazardous cargo, destination and estimated time of arrival (entered manually).

How often a ship reports depends on its speed: according to ITU (2026), a Class A ship under way transmits its position every 2 to 10 seconds, and every 3 minutes at anchor; according to IMO (2015), static and voyage data are sent every six minutes. Position and type are therefore fairly reliable, while draft and destination depend on the crew.

Terrestrial or via satellite

AIS is VHF radio and does not reach far: according to IMO (2015), a range of 20 to 30 nautical miles is typical at sea, depending on antenna height. Receivers on land therefore pick up coasts, straits and ports. On the open sea, only reception from space helps, with receivers on satellites. Cerdeiro et al. (2020) describe both routes and point out that satellites cover the deep oceans less well.

Can you track commodity ships worldwide in near real time? On coasts and at chokepoints, yes, even with freely accessible data; on the high seas, only with paid satellite data. Portfolio Lab therefore limits itself to chokepoints.

Measuring seaborne trade in real time

Official trade statistics appear with a delay of weeks or months, AIS data immediately. That is why statisticians try to use them for nowcasting, an estimate of the present. Using Malta as an example, Arslanalp et al. (2019) developed indicators from port calls and compared them with official statistics: with suitable filters, such data could complement the statistics and show economic turning points earlier.

Cerdeiro et al. (2020) took a global approach. They analyzed more than a billion messages from more than 50,000 ships from 2015 to April 2020, identified port boundaries with machine learning, assembled port-to-port voyages and estimated cargo from the draft. For many countries and for the world as a whole, the results matched official trade statistics well. Using ship tracking data, Verschuur et al. (2021) estimated that seaborne trade shrank by 7.0 to 9.6% in the first eight months of 2020, around 206 to 286 million tonnes.

For a single commodity, the picture is more mixed. Adland et al. (2017) compared crude oil exports estimated from AIS with customs data: in total they matched well, but by country and period there were marked deviations, because oil also flows through pipelines and is transshipped en route.

Chokepoints of seaborne trade

A great deal of oil passes through a few straits and canals, the chokepoints. According to EIA (2024), around 77.5 million barrels of oil a day were transported by sea in 2023, a good three quarters of the global supply of 101.9 million barrels. The table shows the volumes passing through the major chokepoints:

The shares are our own calculation (volume divided by 77.5) and cannot be added up, because the same oil often passes through several chokepoints, for example the Bab el-Mandeb and the Suez Canal.

Chokepoints matter to investors because disruptions show up there first. Hormuz is the most sensitive: according to Dunn and Barden (2025), an average of 20 million barrels a day flowed through it in 2024, about 20% of global oil consumption and more than a quarter of seaborne oil trade, plus about a fifth of global LNG trade. The Red Sea showed what happens when a chokepoint fails: according to Barden et al. (2024), after the attacks on merchant ships from November 2023, oil flows through the Bab el-Mandeb fell by more than half in the first eight months of 2024, while 9.2 million barrels a day now passed the Cape of Good Hope instead of 6.0, an increase of 53%.

Worked example: how many tankers are 20 million barrels?

According to EIA (2014), a VLCC (very large crude carrier) holds 1.9 to 2.2 million barrels of crude oil. This allows a rough estimate of what the volume through Hormuz means in ships:

Freight rates as a business cycle indicator

Not only the ships but also their price tells you something. Kilian (2009) measures global real economic activity with an index of single-voyage freight rates for dry bulk cargo (grain, oilseeds, coal, iron ore, fertilizer and scrap metal) in order to separate demand shocks from supply shocks in the oil market. The idea: the supply of shipping capacity is almost fixed in the short term, because new ships take years to build; if demand for commodities rises, rates rise. The rates are expressed in real terms and adjusted for a linear trend.

Using this measure, Kilian (2009) finds that the rise in oil prices in the 2000s was driven mainly by global demand. Bakshi et al. (2011) show that the growth of a well-known dry bulk freight rate index predicts stock and commodity returns as well as global economic activity, even out of sample.

Satellite images: oil tanks and parking lots

On land, crude oil is often stored in floating roof tanks: the roof rests on the oil and sinks as the fill level falls. Seen from above, the tank wall casts a shadow on the roof. The emptier the tank, the lower the roof and the wider the shadow. From the width of the shadow, the position of the sun and the tank dimensions, the fill level can be estimated.

Mukherjee et al. (2021) examined what such estimates mean for the US government's weekly inventory data. Clouds served as a natural experiment: when important tank farms are covered, the satellites see nothing. The result: some satellite estimates are now so good that the government report surprises the market less. The authors see this as a sign that governments are partly losing control over economic data.

It is similar with retailers: Katona et al. (2025) analyzed car counts in the parking lots of 44 US retailers from 2011 to 2017. Investors with access to this data were able to trade profitably, especially ahead of poor quarterly results. Short sales became better informed, purchases by retail investors less so, and liquidity around the reports fell. According to the authors, access cost hundreds of thousands of dollars a year. That was an advantage for a few.

Why this does not work with freely available images

Freely available images include those from the European Sentinel-2 program. According to ESA (2015), its sharpest bands have a resolution of 10 m, others 20 or 60 m, and two satellites pass over the same location every five days under cloud-free skies. That is not enough for the shadow in a tank.

Then there are clouds and the revisit rate: one image every five days under clear skies is tight for weekly inventory data. With freely available images, the shadow method is therefore not feasible.

The shadow fleet and switched-off transponders

AIS only works if ships transmit properly. According to IMO (2015), the master may switch the device off if operating it could endanger the safety of the ship. Others switch it off because they do not want to be seen. Tankers that circumvent sanctions are regarded as the shadow fleet: Cardoso et al. (2025) describe how, after the price cap on its oil, Russia built up a fleet that manages without Western services. Anyone sailing like this has a reason to switch off AIS or transmit false information.

Paolo et al. (2024) measured how big the gap is: they compared satellite images with public ship tracking in coastal waters from 2017 to 2021. Of the industrial fishing vessels, 72 to 76% could not be tracked publicly, and for transport and energy, between a fifth and almost a third of the activity was missing:

On top of this come errors in the information itself. Harati-Mokhtari et al. (2007) examined the reliability of AIS data and the role of human error on board, since the crew types in some of the information manually. The draft, from which the cargo load is inferred, may be out of date or wrong.

For advanced readers: from positions to tonnes

If you want to estimate volumes from AIS, you have to identify ports, build voyages, add the deadweight of each ship and estimate the cargo from the draft. Cerdeiro et al. (2020) take the last draft before arrival in port, because crews presumably report it more carefully at that point:

Further pitfalls are pipelines and transshipments (Adland et al. 2017) and ships that lie at anchor for weeks as floating storage. According to EIA (2020), the volume of crude oil stored on tankers rose in 2020 from 66 million barrels on March 1 to 222 million barrels on July 9, more than tripling; the count covered laden tankers that had been at anchor or drifting slowly for at least seven days. This pays off when storage on land is full and futures prices are well above the spot price (contango).

Limits for forecasts

• Gaps look like signals: if a receiver fails, the number of ships falls without there being less traffic.
• Ambiguous patterns: many ships waiting outside a canal can mean congestion, disruption or high demand. In 2024, fewer transits through the Red Sea mainly meant detours around the Cape of Good Hope.
• Already priced in: if satellite data anticipate the inventory report (Mukherjee et al. 2021), the information is reflected in the price early on. If you only read it from freely available data, you are late. This fits the efficient market hypothesis.
• Many attempts, few hits: if you test enough metrics from ship data, you will find one by chance that would have worked in hindsight (see backtest overfitting).

What this means for investors

Ship data show early what only appears in statistics weeks later. For private investors, they are above all background: why the oil price reacts to news from Hormuz, and how vulnerable supply chains are at a few straits. They do not promise a reliable trading advantage: the fastest and most complete data cost a lot of money, and those who have it trade earlier (Katona et al. 2025).

If you hold commodity risk in your portfolio through energy stocks or ETCs, you can use such data as a situation report, not as a signal to buy or sell. This article is not investment advice.

How Portfolio Lab calculates it

The Ships page (beta) receives AIS from a freely accessible network of terrestrial receivers and counts tankers and cargo ships in twelve areas, from Hormuz via Suez and Malacca to Panama, Dover and off Rotterdam and Houston. The class comes from the AIS ship type (70–79 cargo ship, 80–89 tanker), the size roughly from the length, and “probably laden” or “probably in ballast” from the draft relative to the largest draft seen and to a reference value derived from the length. A ship counts as under way from 2 knots. The page shows positions and the number of distinct ships over the last hour, per hour and per day, but no volumes and no forecasts. All the details are in the methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.