Time Series Analysis of Catch-At-Age Observations

Catches, stocks and fishing mortality rates are connected by given non-linear relationships. Multivariate time series models of fishing mortality rates are proposed and linear approximations to the Kalman filter algorithm are used to estimate the unobserved series of stocks and fishing mortality rates from observed series of catches at each age. The model parameters are estimated by the likelihood function of the catch prediction errors. Time series estimation does not require fishing effort measurements as well as the catch-at-age data, but it is also possible to include auxiliary information. Estimates of stocks and fishing mortality rates with acceptable accuracy can be obtained from exceptionally short series. They are not unduly sensitive to misspecifications and inaccurate estimation of model parameters.