Projects

Vessel Identity

Creating comprehensive vessel information spanning our global waters

fishing boat with mountain behind

To understand fishing activity, we need a complete picture of the identities of individual fishing vessels—including name, type, characteristics, owners and authorizations. Global Fishing Watch consolidates identity information by drawing on a large volume of vessel tracking data, machine learning and various public registry sources. Combined with vessel GPS data, this information is used to produce our technology applications, analysis reports and research papers. We work closely with researchers, fisheries authorities, practitioners and industry partners to continuously improve our vessel identity data and drive more transparent information about activity at sea. 

Changing names and flags

Finding a vessel’s true identity is not easy. Identity information is often fragmented, incomplete or conflicts directly with other information across multiple sources—and sometimes it is not publicly available at all. Other challenges include a large number of vessel identities. There are millions of vessels operating around the world, and their dynamics and identities change over time. Moreover, vessels that engage in illegal activity are more likely to frequently change their name or flag to avoid oversight. 

Unlocking the identity puzzle

We regularly collect information on vessel identity from more than 40 vessel registries, either available in the public domain or obtained from authorities and researchers. We train a convolutional neural network—a cutting-edge form of machine learning—to identify vessel characteristics. This model assigns one of our 40 vessel classes to every active vessel broadcasting automatic identification system (AIS) signal and predicts vessel characteristics like length, tonnage and engine power. We develop a system to consolidate data from all sources and minimize errors. After the integration, we select the best identity attributes for every vessel in the AIS data. To provide accurate and up-to-date information, we establish a data pipeline that automatically processes this data— covering more than 600,000 fishing and non-fishing vessels—and publish regular updates.

Recent Work
Scroll to Top