Drift gillnet vessels from space: leveraging low-cost methodologies for enhanced understanding of a data-poor fishery

The Indian Ocean produces the second-highest tuna catch across the world’s oceans. Here, the prevalence of drift gillnets – used to catch about one-third of tuna and tuna-like harvest – is unique compared to other global tuna fisheries, more commonly dominated by longlines and purse seines. Most drift gillnet fleets in the Indian Ocean are comprised of relatively small vessels under 24 meters in length overall. These vessels are poorly documented, fishing effort is opaque, and catch/bycatch is underreported. This is in contrast with purse seine and pelagic longline fleets operating in this region, for which fishing effort and catch are better understood and typically subject to more reporting requirements under the Indian Ocean Tuna Commission (IOTC), the regional body for managing tuna and tuna-like fisheries. Given existing data gaps and the lack of mandatory reporting to list these vessels on the IOTC Record of Authorized Vessels, this study set out to trial different approaches to better document, monitor, and understand drift gillnet fleets and, ultimately, bycatch, through satellite imagery. This study focuses on Pakistan’s drift gillnet fleet as a case study. Using image annotation, deep learning on satellite images, and port-based interviews in Pakistan, we tested different methods to quantify and describe the Pakistani tuna drift gillnet fleet and bycatch. We found that several low-cost image annotation methods and deep learning are powerful tools to illuminate information on a fleet where other monitoring and surveillance is missing. However, additional supporting information from local expertise, ground-truthing, and other considerations are necessary for robust estimates of fleet size. This paper describes 1) existing information on catch and bycatch in the Pakistani drift gillnet fleet, 2) the potential of satellite imagery analysis and deep learning towards fisheries management, and 3) the different methods, challenges, and lessons learned. This paper serves as a baseline for future similar analyses in the Indian Ocean and other regions toward a better understanding of data-poor fisheries.