An Analysis of Errors in Catch Projections for Canadian Atlantic Fish Stocks

Two techniques were used to identify and evaluate the components of error inherent in the catch projection method used in the assessment of Canadian Atlantic fish stocks. Firstly, a retrospective analysis was used to evaluate, for the projections made in 1979 and 1981, the impact of the actual error made in input variables such as catch-at-age, mean weight-at-age, recruitment, and stock size. Secondly, the variance of projections was estimated from the variance of input variables. From the retrospective analysis, the principal causes of error in catch projections were identified as the estimation of stock size, as well as the estimation and forecast of fishing mortalities for the projection years. The difference between actual catch and assumed catch for the first year of a 2-yr projection did not appear to be a major source of bias for catch projections. The coefficients of variation of the catch projected at F0.1 were estimated as follows: 15–20% for the projections of Atlantic cod (Gadus morhua) catches; 25–50% for those of haddock (Melanogrammus aeglefinus); 28% for those of pollock (Pollachius virens); 16% for those of redfish (Sebastes sp.); and 35–42% for those of herring (Clupea harengus). The uncertainties associated with abundance estimates for prerecruited age-groups and estimates of stock size emerged as key factors in the projections. Also important was the effect of uncertainties associated with the forecast of reference fishing mortalities (partial recruitment coefficients × F0.1) for the projection period, e.g. uncertainties resulting from changes in fish catchability, in fleet behavior, and in fishing patterns.