Deep Learning to uncover Watershed Specific Characteristics in Salmon Otolith Patterns to aid in Management of Pacific Salmon

Project Lead: Ben Makhlouf, UW Aquatic and Fisheries Sciences

Data Science Lead: Valentina Staneva

Pacific salmon in western Alaska support a multi-million-dollar commercial fishing industry and are a vital source of subsistence for upstream communities. In recent years, catastrophic declines in Chinook and Chum salmon populations have severely impacted subsistence opportunities, causing significant hardship across the region. As the crisis unfolds, efforts to identify the drivers of declining salmon abundance and assess the distribution of impacts across western Alaska have intensified. Several plausible hypotheses attribute these declines to oceanic mortality before salmon return to freshwater to spawn. However, testing these hypotheses is hindered by the lack of reliable methods to determine the river of origin for salmon caught at sea. Without this information, regulators face difficult decisions, including emergency management measures that could further restrict subsistence fishing or partially close some of the world’s largest commercial fisheries. Avoiding overly broad preventative actions requires a novel approach to assess the spatial distribution of bycatch impacts and enable targeted mitigation.

The analysis of isotopic signatures in otoliths, or fish ear stones, has emerged as a promising method for estimating the origins of migratory fish species. Otoliths incorporate elemental and isotopic ratios reflective of their environment as the fish grows, providing a natural record of geographic origin.Otoliths grow continuously, storing chemical information that can be extracted as sequential data and analyzed as a time series from rearing to death. While individual patterns may appear subtle or inconclusive, emerging data science techniques can reveal meaningful relationships and life history insights in the isotopic signatures. By analyzing known-origin otolith samples from western Alaska, we aim to develop a model capable of reliably estimating the geographic origin of marine-caught salmon. This tool could play a critical role in understanding the impacts of climate change and human activities on Pacific salmon populations in western Alaska.