Experimental work
Uncovering links to IUU vessels
Vessels involved in Illegal, Unreported and Unregulated (IUU) fishing are publicly listed to discourage engagement in such activities. Once a vessel is listed, it often receives further scrutiny, and is sometimes denied access to fishing grounds.
However, fishing vessels rarely operate in isolation. Bunkers, reefers, and other fishing vessels form a network in which catch, fuel, money, and people are exchanged. Often offshore, far from regulators' eyes. To improve IUU risk analysis, we're uncovering the associations between such vessels.
This is highly experimental work. We're publishing it to collaborate with others and improve. Expect this to break and change rapidly.
Building a network of offshore encounters
Vessel associations are often obfuscated by complex ownership and operating structures. Vessel encounters can help disentangle these associations. For instance, vessels that have encounters offshore may be part of the same fleet, share ownership/operation structure, or have commercial links with reefer or bunkering companies.
As a result, vessels that frequently encounter known IUU offenders might be more likely to be involved in IUU fishing themselves. Identifying these linked vessels is valuable for analysts and fisheries management organisations. However, inspecting the encounter history of relevant vessels can be laborious.
At Starboard, we're sifting through millions of vessel encounters and building a network of vessel associations so you don't have to do so.
A known problem in maritime monitoring is that vessels can turn off their AIS transponders. Because we use the vessel's AIS-encoded position to detect encounters between vessels, it's not always possible to identify those encounters directly.
We address that problem by looking not just at a vessel's neighbours (those which it encounters directly or direct links) but also at the neighbours' neighbours (indirect links). This approach allows us to find associations between vessels even when they might never meet each other directly. For instance, two fishing vessels might be determined to be associated with each other if they repeatedly transship catch to the same reefers, refuel from the same tankers, or share longlining buoys.
To identify IUU vessels, we use a list compiled by TMT, a not-for-profit organisation providing intelligence for improving fisheries governance.
Below, you can explore the network of encounters of all vessels in the TMT list for which we were able to construct associations. For clarity, we display each IUU vessel neighbourhood separately and exclude buoys from the visualisation. Zoom and hover over network elements for more information.
Overall, we found direct and indirect links between 29 IUU vessels and 3,995 other vessels (2,347 fishing vessels, 657 cargo vessels, 468 tankers, and 523 vessels of other types). This is a small number relative to the hundreds of thousands of vessels currently being tracked byStarboard and the tens of thousands related to fishing. But it is a remarkably large number given that it includes links from only 29 vessels included in the TMT list.
The vessels linked in this way represent 2—8% of the fishing-related vessels monitored by Starboard, helping to narrow the number of vessels that may deserve extra attention during inspections or fisheries operations.
Prioritising vessels
Identifying these links can be valuable for analysts in fisheries organisations tasked with assessing potentially-offending vessels. However, links between vessels are not all the same because some associations may be much stronger than others. We want to be able to focus on those vessels with the strongest associations.
Therefore, we scored each of the linked vessels based on the overall strength of their association with IUU vessels (see Methods for details). In short, the score is inversely related to the total “network distance” between a particular vessel and IUU vessels. The “closer” a pair of vessels are in a network, the stronger their association.
We designed this score such that it is larger for vessels with many links to IUU vessels (with direct links having a greater effect on the score than indirect ones). Nevertheless, vessels with the same number of links to IUU vessels can have vastly different scores, and this is because thehe IUU association score also accounts for the frequency, duration and age of the encounters between vessels. Longer, repeated and more recent encounters with IUU vessels (or their neighbours) indicate stronger associations.
The graph below shows the relationship and variability of the IUU association score with the number of links to IUU vessels. We show this for both direct and indirect links.
Most vessels have a relatively low association score, meaning the encounters with IUU vessels or their neighbours are short, old, or sporadic. However, the scores of a few vessels are over ten times the median score of 0.9.
The table below shows the top ten vessels with the largest score for different vessel types. These vessels are part of a large network that links them, directly or indirectly, to multiple listed IUU vessels through long, recent and repeated encounters.
You can also Download ↓ data for the top 100 vessel scores as a CSV file.
So what's next? Most importantly, we want to hear back from you.
We are also aware of multiple ways we can improve or extend this approach. An obvious one is that it's not (yet) possible to distinguish between vessels that score low because they have low or no association and those that avoid AIS monitoring altogether.
About Starboard Maritime Intelligence
Starboard tackles complex maritime domain awareness challenges, ranging from detecting illegal fishing and risk assessing arriving vessels to validating vessel movements and uncovering non-reporting dark vessels.
Learn more →Methods
We built encounter networks by inspecting vessel encounters between May 2020 and February 2022.
Encounters mean two vessels (or sometimes a vessel and a buoy) have met one of the following criteria: either they’re less than 50 metres from each other for at least 10 minutes, or less than 200 metres from each other for at least 30 minutes. This movement is consistent with the behaviour observed when two vessels meet at sea to exchange goods — fuel, fish, people or cargo, for example.
We weighted each encounter such that longer and more recent encounters have a larger weight, and therefore are judged more relevant than shorter or older ones: the length of an encounter was weighted as the exponential decay of its duration in hours. We set the half-life of encounters to one year, meaning that an encounter that occurred a year ago has half as much influence as an encounter of the same duration taking place now. A one-year half-life ensures that older encounters still provide information about current associations without overshadowing the relevance of more recent encounters.
After weighting each encounter individually, we quantified the overall weight S of the link between a pair of connected vessels by adding the weights of their individual encounters;hen we used these cumulative weights to quantify the score of associations between listed IUU vessels and all other vessels in the network (excluding buoys).
The score \(s_i\) for a vessel of interest \(i\) is the sum of the reciprocal distance between \(i\) and all IUU vessels \(j\) within its order-two neighbourhood (its neighbours and its neighbours' neighbours). In mathematical terms, \[s_{i} = \sum \frac{1}{d_{ij}}\] where \(d_{ij}\) is the weighted shortest-path distance between \(i\) and \(j\), which is defined as the minimum sum of reciprocal weights across all the paths connecting these two vessels. In simpler terms, the larger the weight of the links in the path connecting two vessels, the shorter the distance between them and the larger its association score.
For those familiar with graph theory, the score for a vessel of interest corresponds to its unnormalised harmonic closeness centrality but only in respect of listed IUU vessels in its neighbourhood.
