Using a generative adversarial network-based model to simulate fishing behavior in Antarctic krill fishery.

The existing implementation of a management policy for Antarctic krill fishery has faced challenges due to the diverse and variable management strategies. Understanding the fishing behavior of krill fishery is crucial for developing sustainable policies. In this study, krill fishery data collected on the Antarctic Peninsula, where is the key fishing ground, was used to model krill fishing behavior using generative adversarial networks (GANs). The GANs successfully captured fishing behavior, particularly important features such as temporal characteristics and the Lévy flight, and the performance was better than the previous approaches. Overall, this modeling approach shows promise as a tool for monitoring fishing behavior and management of krill fishery in the Southern Ocean. • Generative adversarial network-based model was applied to simulate fishing behavior in the Antarctic krill fishery. • Lévy flight property of krill fishing behavior was detected • Fishing behavior has not changed after the introduction of voluntary restriction zone. • Regularization model was deemed superior to baseline model for producing high-quality spatiotemporal data. • This study provides valuable insights into fishing behavior in the Antarctic krill fishery and can contribute to the conservation of Antarctic marine living resources. [ABSTRACT FROM AUTHOR]