New research on automated bycatch monitoring in Alaska pollock trawls concludes that developing accurate tracking methods is only an initial step toward potential applications.

Use of multi-object tracking (MOT) is currently limited by the challenges of collecting and automatically annotating footage of dynamic marine environments, according to research contracted by the Alaska Fisheries Science Center and National Marine Fisheries Service and published recently in Elsevier, a major Dutch academic journal.

Controversy over the number of Pacific salmon harvested incidentally in the Alaska pollock fishery has grown in recent years, driven by the decline of salmon available for commercial and subsistence fisheries, including along the Yukon River, where people have fished for thousands of years.

Alaska's pollock fishery, the largest in North America, was worth in excess of $500 million as of 2022, noted researchers Moses Lurbur, Katherine C. Wilson, and Noelle Yochum. Prohibited species catch limits for Pacific salmon are a significant concern for the fishery, which shuts down if those limits are reached.

Discard reduction efforts are an important tool for improving ecosystem health, fishery sustainability, and the vitality of fishing communities, they said.

Scientists and fishermen currently use recorded and live-feed video collected inside pollock trawls to monitor fishing progress and evaluate salmon excluder performance. The poorly lit, crowded, and highly dynamic environment makes detection and tracking of species uniquely difficult, even for human annotators, the report said.

The large volumes of quantitative data about fish movement that MOT automatically collects with minimal human annotation can inform the design of more selective fishing gear and enable real-time monitoring during fishing. But developing accurate tracking methods for the commercial fishing domain is only an initial step toward potential applications, the report said. More work is needed to develop standardized and efficient methods for creating and using MOT-generated data, including more accurate tracking algorithms, more marine datasets, and accessible tools for applying MOT data in fisheries research, the researchers concluded.