Detection of Fishing Impacts

Better models to evaluate ocean harvest

fishing by vessels with ais map
Areas of heaviest activity by fishing vessels broadcasting their position on the automatic identification system are seen in yellow

Algorithms can help determine when and where vessels are fishing. When we apply them to vessel movement data, we can more accurately assess fishing activity and help decision makers manage global fisheries more equitably and sustainably.

A truer measure of fishing costs

The Food and Agriculture Organization of the United Nations estimates that over a billion people rely on fish as their primary source of protein. To truly understand the impact of fishing at this scale and better direct conservation efforts, we need to measure the presence and intensity of fishing activity around the globe. 

Data from the automatic identification system (AIS) provides detailed tracks of tens of thousands of fishing vessels. However, this data only reveals fishing vessel presence. For an accurate assessment of impact, we need to more completely characterize fishing activity and differentiate it from non-fishing activity. There are many knowledge gaps to fill, such as impacts by fishing gear type, seabird bycatch during line-setting and ocean floor habitat disturbance by bottom trawling vessels. To better understand the problems and potential solutions, it is important to more closely measure where, when and how much fishing occurs, particularly with high-impact gear types.

Connecting the dots on fishing activity 

To differentiate vessel activity, we developed a model using a convolutional neural network—an algorithm for arranging imagery and data—that classifies each AIS position as either fishing or not fishing for an accurate representation of global fishing activity. To address the need for detailed classification of fishing behavior, we also developed a specialized model that estimates where and when a longline vessel is starting and finishing its set, haul and transit activity. This information can be used to inform management measures and help monitor compliance. We are developing similar models for bottom-trawl fishing and pilot projects are in progress using alternative technology solutions to track small-scale fishing vessels that do not broadcast AIS.

Research Lead

Contributing Team



Recent Work
Scroll to Top