Application of a Kalman Filter to a Multispecies Stock Complex

Application of a Kalman Filter to a Multispecies Stock Complex

Paul D. Spencer and James N. Ianelli

Application of a Kalman Filter to a Multispecies Stock ComplexThis is part of Fisheries Assessment and Management in Data-Limited Situations
PDF    
To download the free PDF [407.4 KB], please enter:
-or-

Description

Fish species are often managed as part of multispecies complexes when existing biological data and/or reported catch data are insufficient to conduct a more detailed single-species assessment. Catches of multispecies complexes may be reported for the aggregate complex, and an estimate of total catch by species can be made by applying species-specific catch proportions from fishery observer data (if available) to the total aggregated catch. However, this procedure would be expected to increase the variance of the resulting catch estimates, and possibly result in correlation in estimated catch between members of the complex. A Kalman filter is applied to such multispecies complexes to incorporate errors of this type, with the eastern Bering Sea/Aleutian Islands shortraker (Sebastes borealis) and rougheye (Sebastes aleutianus) rockfish used as an example. For this stock complex, consideration of sampling variability in observer data adds considerably to the assumed errors in catch. The Kalman filter also provides a methodology for treating the missing data that were common in the shortraker/rougheye data, and would be expected in most datapoor situations. By considering process errors, observation errors, and covariances between observations, the Kalman filter is a powerful tool for multispecies complexes that forces a consideration of sources and magnitude of the errors in the system. However, the shortraker/rough-eye example illustrates that this methodology cannot make informative parameter estimates from non-informative data, and emphasizes the importance of reducing the sampling variances of input data and developing suitable priors for model parameters.

Item details