We used Gaussian Process Empirical Dynamic Models (GP-EDM) to forecast sockeye run sizes. We considered a variety of predictors, including total returns, returns of each age class, and several environmental variables (NPGO, PDO, PNI, and CUTI). All models were fit in R using the GPEDM package (https://tanyalrogers.github.io/GPEDM). Data from prior to 2016 were used to train each model, and then one-year-ahead predictions (with sequential model updating) were made on the 2016-2025 data. Of the models that achieved similar median APEs on the ten year test set, a single model was selected for most rivers. The exception was the Igushik and Kvichak; for these rivers, different model classes performed similarly but produced notably different forecasts, so model averaging was performed between two selected models.

The best models frequently included lags of age-specific returns and/or environmental variables. Age-structured return data were used in the selected models for six out of eight rivers, and environmental data were used in the selected models for four out of eight rivers.