Methodology

How the register is assembled, which vessels get a page, and where the data is known to be weak.

Every emissions figure on this site is reproduced from the published EU MRV returns without modification. Nothing is modelled, estimated or interpolated. Where a ship reported nothing for a year, the row is empty. An empty cell means the source is silent, never that the value is zero.

Which vessels get a page

Not every ship in the source data has enough behind it to justify a page. A page with a name and a single number is thin, and publishing tens of thousands of them would make the whole register less trustworthy, not more complete.

A vessel is published when it has verified MRV emissions, a ship type, a name, and either several reporting years or a filled-out set of particulars: tonnage, build year, flag, builder, photograph. That yields 22,271 pages out of a larger pool. Vessels with one reporting year and almost no particulars are excluded from indexing.

The rule lives in one place in the codebase and the site reads it. Pages do not decide for themselves whether they are worth publishing.

Rankings

Rankings are computed within a ship type and a single reporting year, using EEOI. A vessel qualifies only if it reported the distance and transport work the ratio needs. Vessels that did not are shown as unranked with the reason given, rather than being silently placed last.

Vessel particulars

Tonnage, build year, builder, dimensions and photographs come from Wikidata, which publishes its statements under CC0. Coverage is partial and skewed towards larger and more notable ships. A missing build year means nobody has recorded it in an open source, not that the vessel is undated.

Flag state is not a reported MRV field. It is taken from the reported port of registry where that identifies a state unambiguously, and otherwise derived from the vessel's MMSI prefix. Each vessel page states which route was used, because a derived flag reflects the registry at the time of reporting and ships change registry.

Photographs

Images come from Wikimedia Commons. A photograph is shown only where the licence has been read from the file's own metadata and permits reuse, with the author and licence credited beside it. Files under licences that require the full licence text to travel with the image are not displayed at all. The underlying Wikidata statement being CC0 does not make the file CC0, and treating it as if it did is the most common way sites of this kind end up infringing.

Known limitations

Time at sea figures for the 2018 and 2019 periods sit in a column whose name changes between editions of the source file, and a small number of returns state impossible values, in one case over 600 million hours. Values above the number of hours in a year are rejected during import rather than displayed.

Emission categories in the MRV export overlap by design: emissions at berth are also counted within the voyage legs, so category shares on a vessel page do not sum to 100%.

Port call and voyage data is held in the database but is not published. Its geographic coverage is uneven enough that publishing it would imply a completeness that is not there.

Reading a vessel page

These notes applied to every vessel page and were repeated on each of them until 28 October 2026. They are collected here instead.

EEOI above 1,000 is withheld. Those values come from returns where reported transport work is near zero, which makes the ratio meaningless rather than extreme. Median across the dataset is around 12.

Distance is derived, intensity is reported. The MRV return publishes CO₂ per nautical mile but not the distance itself. The distance column is filled from the ranking step, which reconstructs it, and is blank for years where a vessel was not ranked. That is why a row can show an intensity without a distance.

Percentile is within type and year. It is the vessel's position among qualifying ships of the same type that year, where lower is more efficient. A blank means the vessel did not qualify for ranking that year.

Comparisons with the type use the median of all vessels of the same ship type on this register that filed a return for the same year, and are only made when at least five such vessels exist. Build year, tonnage and flag are compared with every vessel of the type that has the field recorded.

Sister ships share a builder, a type and a build year within two years of the vessel. That holds hull and machinery roughly constant and leaves operation as the variable, which is why they are listed before nearest-by-tonnage.

A change between years can be geography. EU MRV only covers voyages to, from and within the EU/EEA, so a change can reflect where the vessel sailed as much as how much it emitted. A return with zero CO₂ and zero time at sea means no EU/EEA voyages that year, not a cut in emissions.

Laden voyages separate productive sailing from ballast legs. For tankers, bulkers and ro-ro vessels that split is the operational question.

Methane and nitrous oxide start in 2024. They were not reported before that period, so earlier years cannot be compared against them.

Technical efficiency is a design figure fixed at construction. A vessel can hold a good index and still emit heavily by sailing fast, waiting at anchor or carrying little.

Movements come from the Norwegian Coastal Administration's open AIS extract and cover Norwegian waters only. Absence of positions does not mean the ship stopped sailing, and a gap is not evidence that a transponder was switched off.

Inspection records cover United States waters only. An empty record is not evidence of a clean history elsewhere.

Verifier changes are shown because they matter for audit and data continuity. A change is not in itself a sign of anything being wrong.

Corpus

The register currently holds 107,216 annual returns covering 2018 to 2025, across 17 ship types and 116 flag states.