Standardising 2020 fishing effort by flag state
2026-09-03
China operates the largest industrial fishing fleet in the world. That much is well established. But โlargestโ can mean several different things, and the answer changes depending on what you divide by.
Using Global Fishing Watchโs 2020 apparent fishing effort, aggregated by flag state, this note works through several ways of comparing national fleets and asks which of them tells you something you did not already know.
Throughout, โfleetโ means the industrial fleet visible to AIS. That is a specific and fairly small slice of global fishing, and the caveats at the end set out what it leaves out.
China records 14,295,173 fishing hours, 31.83% of the global total and four times the next fleet. This is the ranking everyone expects. Everything below asks what happens when you divide it by something.
Dividing by national population gives a footprint-per-citizen framing.
China disappears. The top of the ranking is small island and Nordic states, and the result is driven almost entirely by the denominator: the Falkland Islands have around 3,500 residents, the Faroe Islands around 53,000.
It is worth being blunt about what this metric does and does not measure. It does not measure fishing intensity, and it does not measure pressure on stocks. Several of the fleets at the top are foreign-owned vessels operating under local flags through licensing or joint-venture arrangements, so attributing that effort to the resident population is misleading. Population has no causal relationship to fishing effort. The chart is a good illustration of how a normalisation can produce a dramatic ranking while saying very little.
Hours per vessel measures how intensively a fleet is worked, and unlike population it is a property of the fishing operation itself.
The 150-vessel threshold matters here. Below it the metric stops measuring fleet behaviour and starts measuring AIS coverage: in countries whose fishing is predominantly small-scale, the handful of vessels visible to AIS are the industrial ones, often foreign-owned under a national flag. Senegalโs 58 detected vessels average 3,473 hours each, which says more about which vessels carry AIS than about how Senegal fishes.
Point size is total fishing hours; both axes are logarithmic. Hover for the numbers.
The plane splits cleanly. Toward the bottom right sit fleets of many vessels each fishing relatively few hours: China, with 26,170 vessels averaging 546 hours each, and Norway, which has more AIS-visible vessels than Taiwan but roughly a quarter of the hours per boat. Toward the top left sit fleets of few vessels fishing continuously: Russia at 2,628 hours per vessel from 850 boats, and Taiwan at 1,857 hours from 1,918.
Chinaโs dominance is therefore a matter of fleet size rather than intensity per boat. Its vessels are worked less hard, on average, than those of almost every distant-water fleet in the dataset.
Total hours say nothing about how effort is distributed in space, and from a conservation perspective the distribution is what matters. The same annual effort concentrated into a few hundred grid cells is a very different proposition from the same effort spread across several thousand.
Two things stand out.
Taiwanโs footprint is larger than Chinaโs. Taiwan touches 27,764 cells against Chinaโs 21,659, with a quarter of the effort, and records the lowest intensity per cell of any major fleet at 128 hours. That is the distant-water longline signature: a couple of thousand vessels spread thinly across three oceans. No fleet in the dataset reaches more of the worldโs ocean.
The Mediterranean fleets sit far above everyone else. That deserves its own section.
Ranking fleets by mean fishing hours per occupied cell puts the same handful of countries at the top, and they are all working the same small sea.
Italy applies 3279 hours to the average cell it fishes. China, across all the worldโs oceans, applies 660. Taiwan applies 128.
This is not a quirk of the metric. The Mediterranean is a small, semi-enclosed sea bordered by a dozen industrialised fishing nations with nowhere else to go. Effort that would be spread across an ocean basin is compressed into roughly two and a half million square kilometres, and it shows up as the highest fishing intensity per unit area anywhere in the dataset. It is consistent with what stock assessments in the region have been saying for years.
A fleet can post a modest mean while fishing a handful of cells extremely hard. The question is how many cells account for half a fleetโs annual effort.
China is the striking case: half of its 14,295,173 fishing hours fall inside 57 cells, out of the 21,659 it touches over the year. Vast reach and extreme concentration in the same fleet.
The obvious next step is to express D50 as a share of cells occupied, or to use an evenness index. Both turn out to be dominated by footprint size rather than by fishing behaviour.
Roughly a third of the variance in evenness is explained by footprint size alone, which makes cross-fleet comparison of the index close to meaningless. The absolute D50 count above is reported instead: it answers the same question without the mechanical dependence.
Three results in this dataset look more like artefacts than findings, and are flagged here rather than interpreted.
Malaysia records half its annual effort in a single cell. Portugal concentrates half of nearly 400,000 hours into eleven cells out of more than two thousand. Vietnamโs 178 vessels occupy 38 cells in total. All three patterns are consistent with vessels transmitting from port or anchorage being classified as fishing, and none should be read as a fishing signal until the underlying cells have been inspected.
This is the industrial fleet, not the fishing fleet. Effort here is inferred from AIS transmissions. AIS carriage is mandatory for large vessels and largely absent below roughly 25 metres, so what this dataset sees is the industrial segment: a small fraction of the worldโs fishing vessels, though one responsible for a disproportionate share of catch and of effort in offshore waters. Countries whose fishing is predominantly small-scale are systematically under-represented in every metric here. Indonesiaโs 399 detected vessels are not Indonesiaโs fishing fleet.
Not every state shares its data. AIS transmission can be switched off, and some countries do not make AIS or VMS records available. Absence of detected effort is not evidence of absence of fishing.
Apparent fishing effort is a model output, not observed fishing. Global Fishing Watch classifies vessel behaviour to infer when a vessel is fishing rather than steaming, and that classification carries its own error, as the section above illustrates.
Flag state is registration, not ownership. Open registries such as Panama, Vanuatu and the Cook Islands are excluded from the per-capita, per-vessel and spatial rankings for that reason, but the problem does not stop at those nine countries. Foreign-owned vessels operating under national flags through licensing arrangements are common, and this dataset cannot see through them.
Flag state is derived from the MMSI prefix, not from vessel registry metadata. About 10% of global fishing hours carry a prefix outside the assigned range and are excluded rather than guessed at.
Cells are not equal in area. A 0.5 degree cell at 60 degrees north covers roughly half the area of one at the equator. Cell counts in the charts are unweighted; an area-weighted count is included in the full table.
Population figures come from the World Bank for 2020. Taiwan, the Faroe Islands, the Falkland Islands and Curaรงao are not in that dataset and use approximate figures entered by hand.
Code and data pipeline: github.com/TheophileMt92/gfw-fleet-comparison. Effort data derives from the pipeline in gfw-seasonality. Fishing effort data: Global Fishing Watch, CC BY 4.0.