Use of the Kalman filter to reconstruct historical trends in productivity of Bristol Bay sockeye salmon (Oncorhynchus nerka)

Fisheries scientists and managers are concerned about potential long-term, persistent changes in productivity of fish stocks that might result from future climatic changes or other alterations in aquatic systems. However, because of large natural variability and measurement error in fisheries data, such changes are usually difficult to detect until long after they occur. Previous research using numerous Monte Carlo simulation trials showed that a Kalman filter performed better than standard estimation techniques in detecting such trends in a timely manner. Therefore, we used historical data along with a Kalman filter that included a time-varying Ricker a parameter to reconstruct changes in productivity (recruits per spawner at a given spawner abundance) of eight Bristol Bay, Alaska, sockeye salmon (Oncorhynchus nerka) stocks over the past 40 years. Productivity generally increased for most stocks but varied widely for others and dramatically decreased in another. Such large changes in productivity are important for management. They greatly affected optimal spawner abundances and optimal exploitation rates, suggesting that in the future, scientists should consider using models with time-varying productivity parameters.