MarineTraffic / Global Fishing Watch: Intelligence Source Guide

Two very different ways to see a vessel: MarineTraffic, a commercial AIS aggregator selling position and voyage data, and Global Fishing Watch, a non-profit that turns AIS into modelled fishing effort, encounters, loitering and port visits. Only one of them is free, and the catalogue does not say which.

At a glance

Source	MarineTraffic / Global Fishing Watch
Category	Maritime, Shipping & Fisheries › AIS & Vessel Tracking
Homepage	https://globalfishingwatch.org/
Machine interface	`https://gateway.api.globalfishingwatch.org/`
Format	`JSON`
Access	Free registration — API key at no cost
Disciplines	Maritime Intelligence
Mission domains	Maritime Security, Maritime Piracy, Environmental Crime

AIS vessel tracking & maritime intel. — as catalogued in the platform’s own source registry.

This entry bundles a commercial data vendor and a research non-profit under one code because both answer questions about where vessels are. They are not substitutes. MarineTraffic operates one of the largest terrestrial AIS receiver networks, supplemented by satellite AIS, and sells positions, voyage history, port calls, vessel particulars and photographs through a web platform and a credit-metered API. Global Fishing Watch is an independent non-profit, founded through a collaboration between Oceana, SkyTruth and Google, that ingests AIS and, in some products, vessel monitoring system data shared by partner governments, and applies machine-learning classifiers to the resulting tracks. Its outputs are not positions but inferences: apparent fishing effort, encounters between vessels at sea, loitering, port visits, and gaps in AIS transmission. The distinction is fundamental. MarineTraffic tells you a vessel reported being at a place at a time. Global Fishing Watch tells you a model believes a vessel was fishing, meeting another vessel, or had stopped transmitting. Its API is a documented REST interface at a gateway host, versioned, with endpoints for vessel search and identity, event types, aggregated effort reports and insights, and it requires a bearer token. The catalogue records that gateway as the machine interface, which is correct for the Global Fishing Watch half and has nothing to do with MarineTraffic.

The analytical job this pairing does is to convert self-reported navigational broadcasts into behavioural claims that can be argued about. AIS was designed for collision avoidance, not for surveillance, and everything in it about identity is typed in by a crew. What makes it intelligence rather than noise is the fact that behaviour is much harder to fake than identity: a vessel can broadcast any name it likes, but the shape of its track over three days — speed changes, turning patterns, time spent in a place — is a physical signature. Global Fishing Watch's contribution is to make that signature computable and public, so that a claim like 'this vessel fished inside a national exclusive economic zone' can be examined by anyone rather than asserted by a vendor. Its second contribution is the negative space: the encounter and gap products treat transhipment and AIS disabling as first-class observable events, which is what turns dark-fleet analysis from anecdote into a dataset. MarineTraffic's contribution is different and complementary — depth of historical position data, port call records and vessel particulars that are commercially curated. For MARINT work on illegal fishing, sanctions evasion, transhipment, forced labour at sea and piracy, the practical pattern is that you use the free modelled products to find the behaviour and the commercial data to establish the detail, and you never present either as proof of what happened on the water without a corroborating source.

Who publishes it, and why that matters

The two operators have opposite incentives and you should read their outputs accordingly. Global Fishing Watch is a non-profit funded by philanthropy and by partnerships with governments and research institutions, and its stated mission is transparency in ocean activity. That produces open publication, peer-reviewed methods and unusually candid documentation of limitations — the organisation publishes what its models cannot see, which almost no commercial vendor does. It also produces the risks that attach to mission-driven organisations: a dependence on funding cycles, a research agenda that shapes which products get built, and government partnerships that determine which countries' vessel monitoring data is available and which is not. MarineTraffic is a commercial business, originating in an academic project and now part of the Kpler group, selling to shipping, commodities and finance. Its incentive is coverage and reliability for paying customers, which is a good incentive for data quality and a poor one for transparency about method. Neither is going away. What changes is terms: Global Fishing Watch's licensing and API conditions have evolved as the organisation has grown, and MarineTraffic's product packaging has changed with ownership. Confirm current terms for either before building a dependency, and never assume that a dataset which was free for research last year is free for commercial use this year.

Provenance is the first question to ask of any dataset and the one most often skipped. Who collects it, what their incentive is, whether they publish a methodology, and whether they correct the record when they get something wrong all bear directly on how much weight a finding drawn from it can carry.

What a record actually contains

The fields you will be working with, what each one means, and whether it is something you can pivot on. Read the meanings carefully — more analysis is wrecked by misreading a field than by failing to find one, and a field that looks like an observation is often an inference.

Field	Type	What it means	Pivot value
`mmsi`	string	The nine-digit Maritime Mobile Service Identity broadcast by the transponder. The first three digits are the maritime identification digits indicating the flag administration that issued it. It is not a permanent vessel identifier: it changes with reflagging, it is reused, and duplicate or invalid MMSIs are common in some fleets.	Flag state via the identification digits, vessel registries, RFMO authorisation lists, IUU vessel lists
`imo`	string	The seven-digit IMO ship identification number, permanent for the life of the hull and unchanged by reflagging or renaming. Most fishing vessels do not have one, because eligibility was only extended to fishing vessels above a size threshold and uptake is partial.	FAO Global Record, classification societies, ownership records, sanctions lists
`shipname / callsign`	string	Self-declared identity fields typed into the transponder by the crew. They are the least reliable data in AIS and the most cited. Name changes are frequent, legitimate and also a concealment technique; a name that changes at the same time as a flag change is a signal, not a coincidence.	Historical vessel name records, registries, port state records, media reporting
`flag`	string	The registering state, derived from the identification digits or from a registry record. Flags of convenience concentrate in a small number of registries, and a flag change is one of the more reliable indicators of an attempt to change a vessel's regulatory exposure.	Country dashboards, RFMO membership, port state measures obligations, sanctions jurisdiction
`position (lat/lon) and timestamp`	timestamp	Where the transponder reported being and when. The timestamp is the transmission time, which is not the reception time and not the ingest time; conflating them is the most common temporal error in AIS analysis and produces tracks that are subtly wrong.	Exclusive economic zone and marine protected area boundaries, port and anchorage polygons, other vessels' tracks
`speed and course`	string	Speed over ground and course over ground from the transponder. These are the inputs the behavioural classifiers actually use — fishing, loitering and transhipment all have characteristic speed and turning profiles — and they are more informative than position alone.	Behavioural classification, gear-type inference, encounter detection
`apparent fishing effort`	string	A model output, not an observation: hours in which a classifier assessed a vessel's track as consistent with fishing, aggregated to a grid and a time period. The word apparent is load-bearing and should survive into every report that uses the field.	Effort maps, exclusive economic zone compliance analysis, fleet comparison over time
`encounter event`	object	Two vessels detected within a short distance of each other, at low speed, for a sustained period. It is the observable proxy for transhipment, and it is a proxy: an encounter is consistent with transferring catch, crew, fuel or nothing at all.	Reefer and supply vessel identity, port visit chains, RFMO transhipment authorisation records
`loitering event`	object	A single vessel moving slowly or stationary in open water for a sustained period. Loitering by a refrigerated cargo vessel in a fishing ground is a meeting waiting to happen with a partner that may not be transmitting, which is exactly why the event type exists.	Gap events in the same area and window, satellite imagery tasking, patrol prioritisation
`gap / AIS disabling event`	object	An interval in which a vessel that had been transmitting stopped, and later resumed. It may be deliberate switching-off, equipment failure, or simply an absence of receiver coverage. The event is real; the intent is an inference and frequently an unsupported one.	Satellite radar detections, reception-coverage rasters, prior behaviour of the same vessel
`port visit event`	object	A vessel entering, staying in and leaving a defined port or anchorage area, derived from an anchorage dataset built from vessel behaviour rather than from an official port register. Anchorage boundaries are modelled, so a visit near a boundary is a weaker claim than one at its centre.	UN/LOCODE port codes, port state measures inspections, port call records from commercial sources
`vessel type / gear`	enum	Classification of the vessel and, for fishing vessels, an inferred gear type such as trawler, longliner, purse seiner or squid jigger. Gear inference is a model output based on movement patterns and registry data, and it is more reliable for some gear types than others.	Fleet composition analysis, catch documentation schemes, RFMO gear-specific rules
`registry and authorisation references`	array	Links from a vessel to registry entries, RFMO authorised vessel lists and IUU listings where the source has matched them. Matching between AIS identity and registry identity is itself an inference and is the point where most vessel-identity errors are introduced.	FAO Global Record, RFMO vessel databases, combined IUU vessel lists, sanctions designations

Coverage — and what is not in it

Coverage is global in ambition and highly structured in practice. AIS reception is dense near coasts with terrestrial receiver networks and in areas well served by satellite constellations, and sparse in the high latitudes, in parts of the Southern Ocean and wherever satellite revisit is thin. It is worst, counter-intuitively, in the busiest places: in congested waters such as major straits and heavily fished regional seas, message collision means satellites receive a small fraction of transmissions, so the areas with the most vessels can have the least reliable coverage per vessel. Class A transponders, carried by larger commercial vessels under carriage requirements, transmit more powerfully and more often than the Class B units common on smaller vessels, so a fleet's visibility correlates with its size. Vast numbers of small-scale and artisanal vessels carry no transponder at all and are simply absent from every product built on AIS; in some regions that is most of the fishing fleet. Temporally, the AIS record is deep — Global Fishing Watch publishes effort data going back over a decade — but the early years are thinner and less comparable, so trend analysis across the full window must control for changing receiver coverage rather than reporting increased detections as increased activity. Vessel monitoring system data shared by partner governments extends coverage into fleets that AIS misses, but only for the countries that have chosen to share it.

Access, licensing and what you may do with it

Access model: Free registration — an account or API key, at no cost

The catalogue's free-with-a-key description fits one half of this entry and not the other, so read it as applying to Global Fishing Watch alone. Global Fishing Watch issues API tokens at no charge through a request form on its site; you state who you are and what you intend to do, and tokens are issued for non-commercial research, journalism, enforcement and analysis. Much of the underlying data is also published for download alongside the code that produced it, which is why the organisation's datasets appear in peer-reviewed work with reproducible methods. MarineTraffic is a commercial product. Some position information is visible on the public website with delays and limits, but historical tracks, port call records, detailed particulars and any API access are paid, credit-metered and licensed, and the terms restrict redistribution. There is no free MarineTraffic API tier that an analyst can build on. The practical consequence for anyone planning work from this catalogue entry is that the free, reproducible, citable half is Global Fishing Watch and everything you design should assume that; MarineTraffic is a procurement decision to be made when the free half has taken you as far as it can, typically when you need commercially curated vessel particulars or dense historical positions in a specific area.

Licence

The two halves diverge here as sharply as anywhere. Global Fishing Watch publishes much of its data under an open Creative Commons attribution-and-share-alike arrangement, with API use governed by separate terms of use that distinguish non-commercial from commercial application; the attribution requirement is not decorative, and derived works inherit obligations. That combination is what makes the data usable in academic publication, in NGO reporting and in journalism without a negotiation. MarineTraffic data is licensed commercially with explicit restrictions on redistribution, on public display, and on incorporation into derived products, and screenshots of its interface in a published report are a licensing question rather than a fair-use assumption. Because both organisations have revised terms as they have grown and changed ownership, the only safe practice is to read the current licence page before publication rather than relying on what a colleague did two years ago. Where a report will be read by a regulator, a court or a defendant's counsel, record which licence version you relied on and when — a licensing challenge to a published vessel-tracking finding is a real event, not a hypothetical one.

Rate limits and fair use

For the Global Fishing Watch API, the throughput discipline that matters is about query shape rather than request count. Aggregated effort and event queries over large areas and long time windows are expensive to compute and slow to return, and the way to work is to narrow the spatial and temporal bounds and iterate, caching each result. Requests for gridded effort at fine resolution across an ocean basin will time out or return volumes you cannot process interactively; request the coarsest grid that answers the question and refine only where something is interesting. Event queries should be filtered by vessel, by region and by date window rather than pulled wholesale. Respect the obvious etiquette of a non-profit service: this is philanthropically funded infrastructure, not a commodity API with a support contract, so back off on errors, do not parallelise aggressively, and where bulk data is published for download, download it rather than reconstructing it through thousands of API calls. For MarineTraffic the constraint is a purchased credit balance, which makes every call a spending decision and makes caching a budget control rather than a courtesy.

Licensing changes, and it changes without warning. A dataset that was free for research this year may not be free for commercial or evidential use next year. Confirm the current terms before you build a dependency on it, and record the terms you relied on alongside the data — the licence in force at the time of collection is part of the provenance.

Collecting it

How MarineTraffic / Global Fishing Watch is actually pulled, in the order you would set it up. Prefer the bulk or export interface over per-item lookups wherever one exists: it is kinder to the publisher, faster for you, and gives a reproducible snapshot rather than a series of point-in-time answers you cannot reconstruct later.

Method	Format	Cadence	Notes
Vessel identity search	JSON	On demand	Resolve a name, MMSI, IMO or call sign to the source's vessel records, including identity history. This is the first step in any vessel workflow and the step where identity errors are introduced, so record what you searched and what matched.
Event retrieval by vessel and window	JSON	On demand, refreshed as a case develops	Pull encounters, loitering, port visits, fishing and gap events for a vessel over a date range. This is the core investigative call and it produces a behavioural chronology rather than a track.
Gridded effort reports	JSON	Periodic, matched to the analysis window	Aggregated apparent fishing effort over a spatial region and time period, used for fleet-level and area-level questions rather than for individual vessels. Choose the coarsest grid that answers the question.
Bulk dataset download	bulk	On release	Global Fishing Watch publishes datasets for download alongside method documentation. For reproducible research this is preferable to the API: you get a versioned artefact you can cite and re-run against, rather than a query result that cannot be reproduced after the underlying data updates.
Commercial position and port call retrieval	JSON	Per purchased credit	MarineTraffic and comparable vendors for dense historical positions, port calls and curated particulars where the free half is insufficient. Budget-metered, licensed, and not redistributable.
Independent corroboration by imagery	HTML	Case-driven	Satellite radar and optical imagery detect vessels that are not transmitting at all, which is the only way to test a dark-vessel hypothesis rather than assert it. Treat this as part of the collection plan from the start rather than as an afterthought.

Ingesting it into the platform

Every step below is idempotent and cursor-based: interrupt one and it resumes from where it stopped rather than duplicating rows or losing progress. Collection is recorded per source, so a feed that quietly stops publishing shows up as a stale timestamp instead of silently thinning your coverage.

Register the source and hold the token centrally — Record the API token in sources.php so the credential and the last-collected state live in one place. Tokens are issued against a stated purpose; keeping them registered keeps that purpose visible to whoever inherits the pipeline.
Resolve vessel identity first and record the match evidence — resolve-everything.php binds MMSI, IMO, name and call sign into a vessel entity. Store what matched what, because identity resolution across AIS and registry data is an inference and it is where most downstream errors originate.
Collect events rather than raw positions — collect.php pulls encounters, loitering, gaps and port visits into the platform. Events are the analytic unit; raw position streams are large, expensive and rarely what an analyst needs once behaviour has been classified.
Preserve the model provenance — ingest.php stores the event with the dataset version and the query parameters that produced it. Model outputs change when models are retrained, and a stored event without its dataset version cannot be reproduced or defended.
Geofence against the boundaries that matter — enrich.php attaches exclusive economic zone, marine protected area and RFMO convention area context to each event. A fishing event is analytically inert until you know whose waters it happened in, and boundary datasets themselves carry uncertainty that should be recorded.
Build the vessel timeline — timeline.php sequences events, flag changes, name changes, port visits and gaps into a single chronology. Reading that chronology, rather than any individual event, is what produces defensible findings.
Correlate across vessels and infrastructure — correlate.php and link-analysis.php connect vessels through shared encounters, shared ports, shared operators and shared ownership, which is how a single suspicious vessel becomes a network. ais.php is the mission view for this work.
Export with the model caveat attached — export.php and reports.php carry the apparent-fishing and inferred-event language through to the output. No relationship or attribution in the platform is generated by a language model; Summarise (Copilot) writes prose about stored events, and the events themselves come from published classifiers with citable methods.

Registered sources and their last-collected state are listed in sources.php, and the scheduled chain that keeps them current is in automation.php.

How it is wrong, and how to tell

Every dataset is wrong in characteristic ways. Knowing which ways is the difference between using a source and being used by one, and it is the part of source evaluation most often skipped because it is the part that takes work.

Judge this source in two layers, because the reliability of the observation and the reliability of the inference are different questions. The observation layer — that a transponder broadcast a position at a time — is highly reliable when a message is received, and the failure modes are absence rather than error: messages lost to collision, to coverage gaps, or to a transponder that is off. The identity layer is weak, because name, call sign and even MMSI are typed in and are wrong often enough that identity resolution is a discipline rather than a lookup. The inference layer is the interesting one, and Global Fishing Watch is unusually honest about it: the classifiers are published, peer-reviewed and evaluated against labelled data, and the organisation states its error characteristics rather than hiding them. That makes the inferences defensible in a way that vendor black-box analytics are not, and it also means you can read the published evaluation and know roughly how often the model is wrong for the gear type and region you care about. What you cannot do is treat a model output as an observation. An apparent fishing event is a classifier's assessment of a track; an encounter is a geometric coincidence consistent with transhipment; a gap is an absence with several possible causes. Each of those is a good basis for investigation and a poor basis for an accusation, and the difference between a competent maritime analyst and a fast one is entirely located here.

Characteristic false positives

AIS gaps read as deliberate disabling. A vessel stops appearing for reasons including equipment failure, power loss, receiver coverage, satellite revisit timing and message collision in busy waters. Deliberate switching-off is one cause among several and is over-attributed, particularly in regions where coverage is objectively poor.
Encounters read as transhipment. The event detects two vessels close together and slow for a sustained period. That is consistent with transferring catch, and equally consistent with transferring fuel, crew or stores, with a mechanical assistance, or with two vessels working the same ground. Calling every encounter a transhipment inflates the phenomenon and discredits the finding when tested.
Identity spoofing and MMSI reuse. Transponder identity fields are typed in and are sometimes wrong by accident and sometimes wrong on purpose. Duplicate MMSIs, invalid MMSIs and identities borrowed from other vessels all occur, and a track that teleports across an ocean is usually two vessels sharing an identifier rather than one vessel doing something remarkable.
Apparent fishing where no fishing occurred. Classifiers key on movement patterns, and other activities produce similar patterns — surveying, cable work, search operations, and simply manoeuvring. The word apparent exists because the model's authors knew this, and it is routinely dropped by the time the finding reaches a headline.
Coverage mistaken for activity. More detections in a region can mean more receivers, better satellite revisit or a new data-sharing agreement rather than more vessels. Any time series over the full data window that does not control for changing coverage will show growth that is partly an artefact of observation.
Small-vessel invisibility read as absence. Fleets that carry no transponder do not appear at all. In regions where most fishing is small-scale, an effort map is a map of the industrial fleet only, and presenting it as a map of fishing misrepresents the fishery.
Boundary errors at exclusive economic zone edges. Both the vessel position and the maritime boundary carry uncertainty, and disputed or unsettled boundaries carry a great deal more. A fishing event a short distance inside a claimed zone is a much weaker claim than its map pin suggests, and in disputed waters the boundary itself is a political assertion.
Registry matching errors. Linking an AIS identity to a registry entry, an RFMO authorisation or an IUU listing is a fuzzy match on names and numbers that change. A wrong match attaches another vessel's history to your subject, and it is the single most consequential error available in this domain because it can defame a lawful operator.

None of these make the source unusable. They make it a source that requires corroboration before an assertion built on it goes into a product, which is true of every source and admitted by few.

Ageing

Position and event data age immediately in the operational sense and not at all in the archival sense. A vessel's current position is stale within the hour, which is why real-time tracking is a commercial product and why free and research feeds carry delays. A historical event does not decay: an encounter that occurred three years ago remains a fact about that date. What does age is everything about identity and status. Names change, flags change, ownership changes, transponders are replaced and MMSIs are reissued, so a vessel record assembled last year may describe a vessel that no longer presents itself the same way. Authorisation and IUU listing status change on RFMO and national timetables and are the fields most likely to be out of date in any cached copy. Model outputs age in a subtler way: when a classifier is retrained or a dataset version is superseded, previously published events can change, so a stored event without its dataset version is not reproducible. A stale record here looks entirely normal, which is why the discipline is to store the retrieval date and the dataset version with every event and to re-resolve vessel identity before any finding is published or acted on.

What this source feeds

A source is only worth what it lets you conclude. These are the disciplines that collect through it, the mission domains it serves and the data points it yields — every one is a tag, so you can follow any thread from here into the rest of the library.

Collected by these intelligence disciplines

Maritime Intelligence

Serves these mission domains

Yields these data points

How each sector uses MarineTraffic / Global Fishing Watch

The same dataset is worked very differently depending on who you are, what authority you hold, and what you are ultimately producing. A military analyst is supporting a commander’s decision; a journalist is meeting a publication standard; an NGO caseworker is protecting a person. The records are shared — the constraints, thresholds and outputs are not.

🎖 Military and defence

Maritime domain awareness work uses this pairing for pattern-of-life on vessels of interest, for identifying the behavioural signatures of transhipment and rendezvous, and for understanding where the AIS picture is blind — which is operationally as important as where it is not. The gap and loitering products are the ones that matter most, because they describe the boundary between the cooperative picture and the uncooperative one, and because they are the natural cue for tasking sensors that do not depend on a vessel's cooperation. Two cautions. AIS is a self-reported system and an adversary who understands that can shape what it shows, so treat a clean AIS picture as a hypothesis rather than as coverage. And the modelled products are research outputs with published error rates, appropriate for cueing and for context, not for any decision that requires certainty about what a vessel was doing.

🕵 National intelligence

For national intelligence work the value is in fusion and in the negative space. Sanctions-evasion analysis lives on exactly the behaviours these products model: encounters at sea, loitering in transfer areas, and transmission gaps around a voyage that should be continuous. Combining an encounter with a gap by the counterparty in the same window is a much stronger indicator than either alone, and both are free and citable. The structural limitation is that vessel monitoring system data reaches Global Fishing Watch only where a government has chosen to share it, so the fleets you most want to see are frequently the ones the sharing agreements do not cover. Treat the MARINT picture here as the open floor of the discipline, valuable precisely because it is public and can be shown to partners without disclosing sources.

👮 Law enforcement

Fisheries enforcement, coastguard and maritime police work is the use case the free half was built for, and it is used operationally by enforcement agencies. The practical products are: prioritising inspections under port state measures by looking at what a vessel did before it arrived, evidencing incursions into national waters, and identifying vessels whose behaviour warrants boarding. The evidential position needs care. A modelled event is investigative intelligence that justifies an inspection; the evidence is what the inspection finds, what the logbook and catch documentation say, and what the vessel monitoring system record held by the flag state shows. Build the case on those and use the modelled events to explain why you looked, recording the dataset version so the analysis can be reproduced at trial.

🔍 Private investigation and corporate security

Corporate investigators and maritime risk teams use this for counterparty due diligence: whether a chartered vessel has a history of encounters with sanctioned tonnage, whether a supplier's fleet operates where it claims, whether a vessel's transmission record has the discontinuities that precede a compliance problem. The free half supports all of that. The two constraints are licensing — do not redistribute commercial vendor data into a client deliverable without checking the terms — and the standard of proof a client will act on. A findings memo that says a vessel had encounters consistent with transhipment, with dates and a stated method, is professional. One that says the vessel was transhipping is a claim you cannot support and that a counterparty's lawyers will attack.

📰 Journalism and OSINT media

This is one of the most productive open datasets in investigative journalism, and it has supported major reporting on illegal fishing, forced labour at sea and sanctions evasion. It works for journalism because the method is published, the data is licensed for reuse with attribution, and a reader can in principle check the work. The discipline that separates the strong stories from the retracted ones is verbal: apparent fishing effort is not fishing, an encounter is not a transhipment, and a gap is not an act of concealment. Every strong maritime investigation in this space pairs the modelled events with something independent — satellite imagery, port records, crew testimony, corporate filings — and says plainly which part came from which. Name a vessel only when the corroboration supports it, because a wrongly named vessel is a livelihood.

🌍 NGO, humanitarian and human rights

Environmental, labour-rights and anti-trafficking organisations use this data for advocacy, for supply-chain scrutiny and for casework support. In fisheries, transhipment at sea is the mechanism that separates catch from its origin, and it is the same mechanism that keeps crews at sea for years without touching a port, which is why forced labour and illegal fishing appear in the same investigations. Modelled encounter and port-visit data lets an organisation ask which vessels are structurally capable of that and where the flag and port states with responsibility sit. The victim-centred constraint governs everything: where crew welfare is implicated, the pathway is the relevant labour inspectorate, the port state authority, and established seafarer welfare organisations, and analysis must never delay a referral or expose an individual crew member. Publish about vessels, companies and states; do not publish about identifiable crew.

🎓 University and research

Global Fishing Watch is a rare thing in this field: a dataset published with its methods, its code and peer-reviewed evaluation, which makes it usable in research that has to survive review. The standard obligations apply and are more than formalities here. Cite the dataset version, because model retraining changes historical outputs. Control for changing receiver coverage in any temporal analysis, because detection growth is confounded with activity growth over the full window. State explicitly that the fleet under study is the transponder-carrying fleet, which in many regions is a minority of vessels. And attribute as the share-alike licence requires, including in derived datasets. Researchers using MarineTraffic data alongside it must handle the redistribution restriction, which usually means publishing analysis and code rather than the underlying positions.

Playbook: working MarineTraffic / Global Fishing Watch end to end

A repeatable sequence from first pull to finished product. Each phase states what you are trying to establish, not merely what to click — the objective is a defensible chain of reasoning, not a completed checklist.

Phase 1 — Establish the vessel's identity before anything else

Resolve MMSI, IMO, name, call sign and flag together, and treat any disagreement between them as the first finding rather than as a data-quality nuisance. Record identity history: a vessel that changed name and flag in the same month has told you something, and a vessel whose MMSI appears on two hulls will otherwise corrupt every subsequent step.

Phase 2 — Map what you can and cannot see in the area of interest

Before drawing any conclusion from absence, characterise receiver and satellite coverage for the region and period. In congested waters and high latitudes, absence of transmission is the expected condition rather than a signal. Write this assessment down; it is the control that makes every later gap claim defensible.

Phase 3 — Build the behavioural chronology, not the track

Pull encounters, loitering, port visits, fishing events and gaps for the vessel across the window and sequence them. The chronology is the analytic object. A dense position track looks impressive and rarely changes a conclusion that the event sequence has not already suggested.

Phase 4 — Geofence every event against the boundaries that create obligations

Attach exclusive economic zones, marine protected areas, RFMO convention areas and national fisheries closures to each event, and carry the boundary dataset's own uncertainty forward. An event near a boundary is a weaker claim than an event well inside one, and in disputed waters the boundary is a contested assertion that your analysis should not silently adopt.

Phase 5 — Interrogate the gaps properly

For each transmission gap, establish duration, location, coverage quality in that place and period, the vessel's prior transmission behaviour, and what happened immediately before and after. A gap that begins as a vessel approaches a known transfer area and ends after it leaves, in a region with good coverage and by a vessel that otherwise transmits reliably, is a strong signal. A gap in the Southern Ocean is weather and geometry.

Phase 6 — Work the encounter from both sides

An encounter has two vessels and the counterparty is often the more informative one. Refrigerated cargo vessels and bunkering vessels have their own histories, ownership, authorisations and port patterns, and a carrier that meets many vessels is a hub worth mapping. Check whether either party held a transhipment authorisation for that area and period; the absence of one is a concrete regulatory fact rather than an inference.

Phase 7 — Test identity against the registries

Check the vessel against the FAO Global Record, the relevant RFMO authorised vessel lists and the combined IUU vessel listings. Presence, absence and mismatch are all informative: a vessel fishing in a convention area without appearing on the authorised list is a specific, checkable compliance question, and absence from a voluntary registry is not evidence of anything by itself.

Phase 8 — Bring in a source that does not depend on the vessel's cooperation

Satellite radar detects hulls whether or not they transmit, and optical imagery can confirm a rendezvous. Where the case turns on what happened during a gap, this is the only way to move from hypothesis to observation, and it should be planned into the collection rather than added when a reviewer asks.

Phase 9 — Follow the vessel to the shore side

Port visits connect a vessel to a jurisdiction, an inspection regime, an agent and a company. Port state measures create obligations that produce records, and the corporate layer behind a vessel — registered owner, beneficial owner, operator, manager, insurer — is usually where the accountable party actually sits. A vessel is a piece of equipment; the investigation is about who directs it.

Phase 10 — Separate observation, inference and allegation in the write-up

Structure findings in three explicit registers: what was transmitted, what a published model inferred from it, and what you assess. Name the model and dataset version for the middle register. This structure is what survives challenge from a well-resourced counterparty, and its absence is what causes maritime findings to be withdrawn.

Phase 11 — Route human-welfare indicators to the welfare pathway immediately

Long periods at sea without port calls, sustained transhipment dependence and crew transfers at sea are recognised indicators associated with forced labour. When they appear, the response is a referral to the appropriate labour inspectorate, port state authority or seafarer welfare organisation, in parallel with and never after the analysis. Publish about vessels and companies; never publish material identifying individual crew members.

The platform ships this as a step-checked workflow in playbooks.php, so progress is recorded against a case rather than held in someone’s head.

What to pair it with

No single source carries a finding. These are the datasets that corroborate, extend or contradict this one — and a source that contradicts is worth more than one that agrees, because it is the only thing that will tell you when you are wrong.

Source	Relationship	What it adds
FAO Global Record of Fishing Vessels	corroborates	The intergovernmental vessel registry keyed on unique vessel identifiers. The right place to test whether an AIS identity corresponds to a registered, authorised vessel, and to see flag and name history.
Combined IUU vessel list	extends	Consolidated listing of vessels designated by regional fisheries management organisations as engaged in illegal, unreported and unregulated fishing. Turns a behavioural suspicion into a regulatory status that already exists.
FAO Port State Measures Agreement	prerequisite	The legal framework that gives port states the power to inspect and refuse entry, and therefore the framework that determines what an enforcement finding can actually achieve.
UN/LOCODE	prerequisite	Standard codes for ports and logistics locations, which is how port visit events join to trade, customs and logistics data rather than sitting in a maritime silo.
ITU maritime identification digits	prerequisite	The registry of MMSI country prefixes, and the correct authority for deriving a flag from a transponder identity rather than guessing from a name.
IMO ship identification number scheme	prerequisite	Defines the permanent hull identifier and its extension to fishing vessels above a size threshold, which is why so many fishing vessels have no IMO number and why that absence is not suspicious in itself.
RFMO authorised vessel records	corroborates	Regional fisheries bodies publish lists of vessels authorised to fish in their convention areas. Fishing activity by a vessel absent from the relevant list is a concrete compliance question rather than an inference.
MarineTraffic	extends	Commercial position, voyage and port call data with deeper coverage in some areas and curated vessel particulars. Paid and licence-restricted; the right purchase when the free half has reached its limit.
Satellite radar and optical imagery	corroborates	Detects vessels regardless of whether they transmit, and is the only way to test a dark-vessel hypothesis rather than assert one. Plan it into the collection from the start.

Legal, ethical and operational constraints

Vessel tracking sits at the intersection of maritime law, data licensing and, increasingly, human rights obligations. AIS is broadcast openly and its reception is lawful; that does not make every use of aggregated tracking data lawful or wise. Licence terms bind: attribution and share-alike conditions on the open half are enforceable obligations, and redistribution restrictions on the commercial half apply to reports and to screenshots. Flag state jurisdiction is the governing principle at sea, and an assertion that a vessel violated a rule is an assertion about a specific legal regime — a national exclusive economic zone, an RFMO convention area measure, a port state requirement — that must be identified precisely rather than gestured at, and disputed maritime boundaries mean a single map pin can embed a contested territorial claim. Where analysis touches crew, the framing changes: seafarers are people, frequently in vulnerable positions, and indicators of forced labour trigger referral obligations under labour and anti-trafficking frameworks in most jurisdictions. Publish about vessels, operators, owners and states; do not publish material identifying individual crew, and route welfare concerns to the appropriate authority rather than into a report. Confirm current licence terms for both halves before any publication.

Operational security

The free half is comparatively benign: retrieving published research data over an authenticated API discloses your interest to a non-profit that holds a token issued against a stated purpose. That is still disclosure, and a token application naming your organisation and purpose is a record that exists. The commercial half discloses more, because credit-metered queries against specific vessels are exactly the pattern a vendor's account management can see, and vendors in this sector sell to shipping, commodities and finance — sectors with an interest in knowing who is looking at what. Concretely: a burst of queries against a single vessel, its owner's fleet and a counterparty is a legible statement about an investigation. Mitigate by querying broader sets than you need where budget allows, by pacing, by using organisational rather than case-specific accounts, and by downloading published bulk datasets and analysing them locally instead of interrogating an API vessel by vessel — which is both better opsec and better science, because a local dataset is reproducible. Note the asymmetry that matters most: none of this notifies the vessel. Tasking imagery, contacting a port agent or approaching an owner does, and those steps belong at a deliberate point in the plan, not at the start.

Two rules that hold regardless of jurisdiction. Collection that is lawful is not automatically proportionate, and a dataset assembled for one purpose does not carry consent for another. Where the records concern identifiable people, the question is not only whether you may hold the data but whether holding it serves the purpose you are accountable for.

Is it earning its place?

Sources accumulate. Feeds get added during an incident and are never reviewed again, and a decade later the pipeline is carrying dead weight that nobody dares remove. These are the measures that show whether MarineTraffic / Global Fishing Watch is contributing anything, and they are worth baselining now so the answer is available later.

Share of vessels of interest for which identity resolves consistently across MMSI, IMO, name and flag — a low share means your identity discipline, not the data, is the bottleneck.
Proportion of gap events for which you recorded a coverage assessment before interpreting them, which should approach one hundred per cent and rarely does.
Number of findings that were corroborated by a source independent of AIS, against the total number of findings published.
Rate at which registry matches are later found to be wrong, tracked deliberately, because a wrong match is the most damaging error available in this domain.
Time from event detection to enforcement or advocacy action, which is the only measure of whether behavioural intelligence is reaching anyone who can act.
Coverage of the fleet under study by transponder type and vessel size, stated in every product, so that readers know which fleet the analysis describes.
Fraction of published findings that record the dataset version and retrieval date, which determines whether your work can be reproduced or defended at all.

Beware of volume. Indicator counts rise easily and say almost nothing. Unique contribution — findings this source produced that no other source in your stack would have — is the measure that matters, and it is usually far lower than anyone expects.

Tradecraft notes

The distinctions that separate a competent analyst from a fast one:

Apparent fishing is a model output and the word apparent must survive into the final sentence of the report. Every serious dispute about maritime tracking analysis has turned on someone dropping it.
Behaviour is harder to fake than identity. Track shape, speed profile and time on station are physical; names and call signs are typed in. Weight your inference accordingly.
A gap is an absence of data, not a presence of intent. Establish coverage quality, the vessel's normal transmission behaviour and the surrounding context before the word disabling appears anywhere in your notes.
Encounters have two parties and the counterparty is usually more informative. Carriers and bunkering vessels are network hubs, and mapping the hub reveals the fleet in a way that following one fishing vessel never will.
The absence of an IMO number on a fishing vessel is normal, not suspicious. Eligibility was extended to fishing vessels only above a size threshold and uptake is partial; treating its absence as an indicator manufactures suspects out of the small-boat fleet.
Congested waters have worse per-vessel coverage than empty ones, because message collision destroys satellite reception where vessels are densest. This inverts the intuition most analysts start with.
Store the dataset version with every event. Models get retrained and historical outputs change, so an event retrieved last year may not reproduce today, and a finding that cannot be reproduced cannot be defended.
Follow the vessel ashore. Ownership, management, insurance and port agency are where accountability lives; the hull is equipment. An investigation that ends at the vessel has stopped one layer short of the answer.
Treat crew welfare indicators as a referral trigger rather than as an analytic finding. Long voyages, transhipment dependence and at-sea crew transfers are recognised forced-labour indicators, and the correct first response is a referral, not a paragraph.

Questions analysts actually ask

Is Global Fishing Watch data free?

The API tokens are issued at no charge on request, stating who you are and what you intend to do, and much of the underlying data is published for download with the code that produced it. Licensing distinguishes non-commercial from commercial use, so confirm the current terms if your output is commercial. MarineTraffic, bundled into the same catalogue entry, is a paid commercial service with no equivalent free API.

Does a gap in AIS mean the vessel switched off its transponder?

Sometimes, and you cannot tell from the gap alone. Equipment failure, power loss, satellite revisit timing, message collision in congested waters and simple absence of receivers all produce identical gaps. Establish coverage quality for that place and period and the vessel's normal transmission behaviour before you interpret it, and remember that gaps in poorly covered seas are the expected condition.

Can I say a vessel was fishing illegally based on this data?

You can say a published classifier assessed its track as consistent with fishing at a position and time, and that the position falls within a zone where the vessel does not appear to hold authorisation. Those are two checkable statements. The conclusion that it fished illegally requires the regulatory status, the boundary certainty and ideally an inspection or corroborating source, and asserting it without them is how findings get withdrawn.

Why do so many fishing vessels have no IMO number?

Because the IMO ship identification scheme was extended to fishing vessels only above a size threshold and on a voluntary basis, so smaller vessels are outside it and adoption among eligible vessels is incomplete. The practical consequence is that MMSI — which changes with reflagging and is sometimes reused — is often the only identifier available, which is precisely why identity resolution is hard in this domain.

What does an encounter event actually prove?

That two transmitting vessels were close together, at low speed, for a sustained period. That is consistent with transhipment of catch and equally consistent with transferring fuel, stores or crew, with assistance, or with working the same ground. Check whether either vessel held a transhipment authorisation for that area and period; that is a concrete regulatory fact rather than an inference.

Does this data show all fishing?

No, and the gap is enormous. It shows vessels carrying and operating AIS transponders, which skews heavily to larger industrial vessels. Vast numbers of small-scale and artisanal vessels carry nothing, and in some regions those are most of the fleet. An effort map is a map of the transponder-carrying fleet and must be described that way.

Can I use this in court or in a regulatory submission?

It is used to support enforcement, but the analysis must be reproducible: record the dataset version, the query parameters, the retrieval date and the model documentation, and expect the method to be examined. Treat modelled events as intelligence that justifies an inspection and build the evidential case on the inspection, the catch documentation and the flag state's vessel monitoring record.

How should I handle indicators of forced labour on a vessel?

As a referral, immediately and in parallel with any analysis. Long periods at sea without port calls, dependence on at-sea transhipment and crew transfers at sea are recognised indicators. Route to the relevant labour inspectorate, port state authority or seafarer welfare organisation. Publish about vessels, owners and states, never material identifying individual crew, and never let the analysis delay the referral.

Do I need MarineTraffic if I have Global Fishing Watch?

Usually not at the start. The free half answers behavioural and effort questions and is citable and reproducible. Buy commercial data when you need dense historical positions in a specific area, curated vessel particulars, or port call records that the free products do not provide — and check the redistribution terms before any of it reaches a published deliverable.

Standards, formats and interoperability

What this source speaks natively, and what it has to be translated into before a partner can consume it. Work that arrives in a recognised format is easier to defend, easier to hand over and easier to automate against:

ITU-R Recommendation M.1371 — the technical standard defining AIS messages, which determines what fields exist, how often they are transmitted and why Class A and Class B vessels have different visibility.
ITU maritime identification digits — the country prefixes embedded in MMSI numbers, and the authoritative route from a transponder identity to a flag administration.
IMO ship identification number scheme — the permanent hull identifier and its partial extension to fishing vessels, which explains most identity resolution difficulty in fisheries work.
FAO Port State Measures Agreement — the international instrument that gives port states inspection and entry-refusal powers, and the framework in which enforcement findings have effect.
UN/LOCODE — the standard port and location codes that let port visit events join to customs, trade and logistics datasets.
Regional fisheries management organisation authorised vessel and IUU listing schemes — the regulatory status layer against which behavioural findings are tested.
Creative Commons attribution and share-alike licensing — the terms under which much of the open half is published, carrying obligations that propagate into derived datasets.
STIX 2.1 and MISP — the platform's export formats, in which vessels appear as observables and events as timestamped relationships with their model provenance attached.

References

Primary documentation and authoritative references for this source. Publishers revise and retire material, so treat the retrieval date as part of the citation and re-check before relying on any of it in a formal product.

Global Fishing Watch — Global Fishing Watch. The organisation and its public map. The starting point for understanding what is modelled, what is observed, and how candidly the difference is documented.
Our APIs — Global Fishing Watch. The API programme overview, including which endpoints exist and what they are intended for. Read before designing any pipeline against the gateway host.
API documentation — Global Fishing Watch. The current endpoint reference, parameter semantics and dataset versions. Treat this as authoritative over any secondary description of the schema, including this guide.
API tokens — Global Fishing Watch. The request route for a free token, including the statement of purpose. This is the practical answer to how access actually works.
Data and downloads — Global Fishing Watch. Published datasets for download, which are preferable to the API for reproducible research because you get a versioned, citable artefact.
Fishing effort dataset and code — Global Fishing Watch. The apparent fishing effort data with its accompanying code — the concrete reason this source can be used in work that has to survive peer review.
Research — Global Fishing Watch. Peer-reviewed publications describing the classifiers and their evaluation, which is where you find the error characteristics you should be quoting alongside any model output.
Frequently asked questions — Global Fishing Watch. Unusually direct answers about what the data cannot show, including coverage limits and the meaning of apparent fishing. Worth reading before writing anything for publication.
MarineTraffic — MarineTraffic (Kpler). The commercial half of this catalogue entry. Useful for understanding what paid AIS aggregation offers and where its licensing restrictions bite.
Recommendation ITU-R M.1371 — International Telecommunication Union. The AIS technical standard. The definitive answer to what a message contains and why transponder class determines a vessel's visibility.
Maritime identification digits — ITU-R. The authoritative MMSI country prefix list, and the correct way to derive a flag administration from a transponder identity.
Port State Measures Agreement — Food and Agriculture Organization of the United Nations. The treaty that turns a behavioural finding into an enforceable inspection at a port, and therefore the framework any enforcement-directed analysis should be written against.

Link integrity: every reference above was verified with a live request when this page was generated. Where a publisher had moved or withdrawn a document, the link was repointed at a preserved copy in the Internet Archive and marked as archived. Anything with no reachable copy anywhere had its link removed rather than left to rot — the source is still credited, it simply cannot be linked.

Put it into practice

The Quantus Intel threat intelligence platform operationalises this source: ingesting encounter, loitering, gap and port-visit events with their dataset versions intact, geofencing them against exclusive economic zones and convention areas in enrich.php, and building vessel chronologies on ais.php that keep observation, model inference and analytic judgement in separate registers. Browse the full source catalogue, or follow any tag above into the rest of the library.