Combining Local Traditional Knowledge and Machine Learning to Predict the Future Safety of Alaskan Shellfish Harvests in a Changing Climate

Marine shellfish are an important diet and cultural resource for many Alaskans. Harmful algal blooms can produce toxins which accumulate in shellfish and can cause disease and death in consumers. Climate change is predicted to expand the timing and strength of harmful algal blooms which may affect food security for many Alaska Native Tribes and communities. Predicting when the shellfish are safe to consume is a critical component of establishing food security and adapting to climate change.  

Local traditional knowledge from Tlingit elders indicates that herring spawning events, which typically occur in April or May in Southeast Alaska, provide a warning signal for when to stop harvesting shellfish. The environmental drivers of toxin producing harmful algal blooms and the herring spawn are not well understood, but it is likely that temperature, particularly sea surface temperature, plays a role in each of those events. This project will incorporate advanced algorithms developed by the lead researcher, Dr. Harley, as well as local traditional knowledge from Tlingit elders, in order to build a predictive model for shellfish toxins. This model will incorporate modelled future conditions and environmental variables, including temperature, precipitation, freshwater discharge, and wind speeds to forecast harmful algal blooms and shellfish safety in a changing climate through 2050 for Southeast Alaska.  

The results of this project will facilitate a greater understanding of the future of harmful algal blooms and shellfish harvest in Southeast Alaska, allowing the researchers to describe changes to the potential bloom season and the strength of blooms over time. Tribal partners, recreational, and commercial shellfish harvesters will be able to use this information for resilience planning purposes, and Tribes and other regional networks can use this framework as a template to conduct assessments for their own monitoring networks. By incorporating local knowledge and novel computational analysis, this project will aid in addressing broader questions of the role of local traditional knowledge in planning for future climate conditions as well as the coupling and decoupling of seasonal events in a changing climate.