Redoubt lake is the largest source of subsistence sockeye salmon in southeast alaska

This federal Project Grant award from the U.S. Fish and Wildlife Service's Tribal Wildlife Grants (CFDA 15.639) program provides $200,000 to the Sitka Tribe of Alaska to install AI-enabled video chutes to automate real-time monitoring of sockeye salmon at Redoubt Lake in Southeast Alaska. The goal is to generate accurate in-season salmon counts at a lower cost than traditional manual monitoring, enabling the tribe to manage subsistence and commercial harvest regulations more effectively. Sitka Tribe will partner with the Wild Salmon Center and U.S. Forest Service to develop and train a deep learning model to identify and count salmon in the video footage, reducing the need for manual monitoring over the long term. This project aims to support the sustainable management of the culturally and economically important Redoubt Lake sockeye salmon fishery.

Harvest regulations are governed by the award-winning redoubt lake sockeye salmon management plan, which depends upon precise in-season salmon counts to liberalize subsistence harvest regulations, allow a community harvest permit to distribute sockeye to tribal elders, or open directed commercial fisheries. Currently, monitoring is conducted by manual counts, a very expensive process that management agencies have indicated is not sustainable into the future. Without in-season monitoring, harvest regulations cannot be liberalized and redoubt sockeye harvest will be drastically reduced. We propose partnering with wild salmon center and the US forest service to install artificial intelligence (AI)-enabled video chutes to produce automated real-time counts of redoubt salmon for a fraction of the cost of traditional weirs. Sitka tribe of alaska and US forest service staff will use a web APP to review video clips and train a deep learning model to accurately identify and count redoubt salmon. Once the model meets accuracy targets, IT will substantially reduce long-term monitoring costs and be able to generate accurate in-season salmon counts in real-time with minimal human input.