Pacific herring, Clupea pallasi, recruitment in the Bering Sea and north-east Pacific Ocean, II: relationships to environmental variables and implications for forecasting

Previous studies have shown that Pacific herring populations in the Bering Sea and north‐east Pacific Ocean can be grouped based on similar recruitment time series. The scale of these groups suggests large‐scale influence on recruitment fluctuations from the environment. Recruitment time series from 14 populations were analysed to determine links to various environmental variables and to develop recruitment forecasting models using a Ricker‐type environmentally dependent spawner–recruit model. The environmental variables used for this investigation included monthly time series of the following: southern oscillation index, North Pacific pressure index, sea surface temperatures, air temperatures, coastal upwelling indices, Bering Sea wind, Bering Sea ice cover, and Bering Sea bottom temperatures. Exploratory correlation analysis was used for focusing the time period examined for each environmental variable. Candidate models for forecasting herring recruitment were selected by the ordinary and recent cross‐validation prediction errors. Results indicated that forecasting models using air and sea surface temperature data lagged to the year of spawning generally produced the best forecasting models. Multiple environmental variables showed marked improvements in prediction over single‐environmental‐variable models.