Monitoring drift gillnet vessels from space: leveraging low-cost methods to enhance understanding of bycatch in a data-poor fishery

Drift gillnets are commonly used in the Indian Ocean to catch tuna, where bycatch is estimated to be in the hundreds of thousands each year. In this region, drift gillnets are typically set by poorly documented vessels estimated to be responsible for very high levels of cetacean bycatch, despite the United Nations banning this gear on the high seas in 1992. Effort in this fishery is not documented and both catch and bycatch are largely underreported to the Indian Ocean Tuna Commission (IOTC), the fisheries management organization that manages tuna fisheries in this region. Here we tested several novel approaches to better document, monitor, and understand the size of drift gillnet fleets and, ultimately, bycatch, through satellite imagery and machine learning techniques. We focused on Pakistan’s drift gillnet fleet as a case study, which is one of the top-five gillnet fleets in the IOTC in terms of landed catch. Using Google-Earth Pro and World-View3 satellite imagery, machine learning, imagine annotation, and port-based interviews, we tested different methods to quantify the Pakistani tuna drift gillnet fleet from 2021-2022. Several low-cost image annotation tools (BIIGLE and FiftyOne Teams) combined with deep learning proved to be powerful tools to generate information from a fleet where other means of monitoring are not possible. We estimate that the Pakistani tuna drift gillnet fleet is comprised of 601 (+/- 41.87) gillnet vessels, with the highest number in Karachi. Vessel size ranged from 15.38 to 26.89 meters. Similarities between trawl vessels and gillnet vessels, and image availability complicated image detection. Based on the interviews and satellite boat estimate, we found that roughly 26,000 cetaceans are caught as bycatch in Pakistan tuna gillnet fisheries. We hope that this work will serve as a baseline for future analyses in data-poor fisheries to inform fishery characteristics and bycatch, as well as inform compliance with the 1992 UN driftnet ban.