Datasets and Code: Fishing vessels

Datasets and Code: Fishing vessels 2020-03-20T16:57:01-04:00

Over the course of a year, over 300,000 unique vessels broadcast their location via AIS. Tens of thousands of these vessels are industrial fishing vessels, and we identify them using three methods, listed below. Note that these lists are not the exact same as the vessels that currently appear on our interactive map. A fraction of vessels on these lists are highly inactive or engage in behaviors that make their AIS data unreliable, such as by offsetting their position by some fixed amount, or by simultaneously broadcasting the same MMSI as another vessel (“spoofing”). The Global Fishing Watch map removes these vessels in order to remove noisy or misleading information. Identifying and correcting for these behaviors is one of our ongoing prerogatives.

The three ways we identify fishing vessels include the following:

1. Self-Reported Fishing Vessels

AIS messages include a field shiptype, which is a two digit number corresponding to the vessel’s activity. The full list of these possible activities is listed on Marine Traffic. About seventy thousand vessels per year report that they are fishing. This information is mostly accurate, but because the user of the AIS device has to manually enter this information, there is potential for human error, and in some cases the shiptype is entered incorrectly. Also some reported fishing vessels are not actually fishing vessels, and some fishing vessels don’t report as such. We call vessels that self report as fishing likely fishing vessels.

2. Known Fishing Vessels

To identify fishing vessels we also match mmsi numbers to vessel registries, such as the European Union’s vessel registry, or the Consolidated List of Authorized Vessels. Many of these vessels also self-report as fishing. Matching self-reported fishing vessels with vessel registries gives us a higher degree of confidence, and we call these vessels known fishing vessels.

3. Inferred Fishing Vessels

The third method involves using machine learning techniques to identify vessels that behave like fishing vessels. This method is still under development. When vessels exhibiting fishing behavior are not also listed in registries or do not self-report, we call these inferred fishing vessels.

Vessels