Report a Shark Catch!

Electronic monitoring (or EM) technologies are evolving as an efficient and effective tool for fisheries monitoring. While data gathered from at-sea observers are an excellent source of information, the deployment of at-sea observers is not comprehensive. Observers are deployed at predetermined coverage rates based on the vessels' gear and size, with vessels <40ft LOA being excluded from coverage. This creates a significant lack of data, particularly on the catch taken by smaller vessels. EM technology can help address this by gathering data from vessels where either space or availability prevent at-sea observer presence.

Collecting such a vast amount of data does have its downsides, however. Review of EM video takes man-hours, which come at a cost. As a result, only about 30% of EM longline hauls are reviewed. Luckily, advances in machine learning systems have provided a potential solution to this problem. Our team, together with the University of Washington Information Processing Lab, has developed a program which we believe to be capable of automatically reviewing video for the presence of large sharks. Relative to target species, the capture of a large shark is rare- accurate catch estimates may therefore require reviewing a greater proportion of EM hauls.

For this objective, we are evaluating how well human video reviewers and the AI reviewers can detect sharks, in particular sharks that drop off the hooks at or near the surface of the water. This program is currently being tested, thanks in large part to the help of operators of Alaskan fishery vessels. When crews with EM equipment on their vessels report shark catches to us through the Catch Report Form, it provides us a known shark catch event to test our program against. If the program is found to be accurate at detecting these catches, it will provide a quick and accurate way to assess the number of sharks caught during EM video review. This, in turn, would improve the efficiency of the video review process and reduce the per haul cost of reviews, potentially allowing for greater video reviews.