Neural Network to Point Russian Fishers to New Fishing Grounds

Digitalising the Catch

For efficient commercial fishing, accurate forecasting of fishing conditions is critical. Such forecasts make it possible to predict, with justification, where and when target species are likely to concentrate, their migration routes, effective fishing depths, species composition, and expected catch volumes. These insights help fishing vessel captains steer fleets toward the most promising fishing areas. Equally important, fishing condition forecasts underpin decisions on strategy and tactics for the rational use of fish stocks, tracking trends and defining allowable catch levels. In Russia, this process is regulated by Federal Law No. 166-FZ of 20 December 2004 on fisheries and the conservation of aquatic biological resources.

A system for forecasting fishing conditions has been developed by the Fisheries Monitoring and Communications System Centre (CSMS), which operates under the authority of Rosrybolovstvo, in partnership with Russian research institutes and specialised IT organisations. The solution is based on mathematical modelling combined with artificial intelligence technologies.

Testing the system in real operating conditions has demonstrated its effectiveness. Specialists were able to locate Pacific sardine (iwashi) stocks that were previously believed to have left their traditional fishing grounds.

“The software analysed potential concentration zones for Pacific sardine. Fish were found outside Russia’s exclusive economic zone, the fishing fleet was redirected there, and results followed. We plan to continue using this system, including for forecasting the movement of other aquatic biological resources. The model will be further refined and improved,” said Ilya Shestakov, head of Rosrybolovstvo.

Faster and More Efficient Analysis

The system is built on a dynamic habitat model trained on large volumes of historical data. To generate accurate forecasts of fish concentrations, the neural network processes and analyses 11 parameters. These include oceanographic indicators such as temperature, salinity, sea level, and currents; biological factors such as zooplankton and chlorophyll concentration; atmospheric data including wind speed and direction and cloud cover; astronomical factors such as lunar phase; and seismic data, reflecting how earthquakes may influence fish behaviour.

We are validating the results through field research. At present, the AI-based forecasting horizon is three days. Improving performance requires large volumes of data. This is a large-scale effort that can significantly increase forecast efficiency and deliver strong economic returns. Right now, we face the task of rapidly locating Pacific sardine stocks, and we are going to demonstrate that innovative technologies are more effective than traditional survey methods

Andrei Yakovlev

Deputy Head of Rosrybolovstvo

Using the trained model, probability maps of fishing zones are generated and sent daily to fishing vessel captains to support operational planning. These materials can be cross-checked against onboard observation data.

A key advantage of AI is the significant reduction in decision-making time. For example, analysing 11 hours of infrared video takes the system around five hours, whereas a human analyst would need at least seven days to complete the same task.

Another defining feature of the system is its flexibility and capacity for learning. For each new species, the model undergoes a dedicated training phase using historical data. In addition to Pacific sardine, the system is being adapted for pollock and mackerel, with forecasts for saury included in future plans.

Harvesting While Preserving Stocks

Russia’s fisheries sector is moving to a new level of digitalisation, both in the operation of the fishing fleet and in the provision of state services related to aquatic biological resource management.

“Our long-term objective is to integrate the knowledge of ecologists, ichthyologists, and biologists with the expertise of programmers and roboticists to automate the full aquaculture cycle,” said Andrei Ronzhin, Director of the St Petersburg Federal Research Centre of the Russian Academy of Sciences.

AI-based analytics improve the efficiency of biological resource management and increase forecast accuracy for the fishing industry. They help cut the cost of locating fish, optimise logistics, and boost catch volumes without excessive pressure on ecosystems. Regulatory authorities will also gain access to high-quality analytical reports to support effective decision-making and sector oversight. Following testing in Russia, the solution could become an export product for fishing companies and research centres in friendly countries.