Combining information from commercial catches and research surveys to estimate recruitment: a comparison of methods

Abstract

Three basic methods for estimating year-class strength given several research surveys or commercial catch indices of recruitment are described. Two are regression methods - calibration regression and predictive regression. The third method is factor analysis, in which the covariance between the indices is modelled as a function of the relationship to the underlying true, but unobservable, recruitment. All three of the methods estimate recruitment as an inverse variance weighted average of the estimates from each of the index series.

In many cases, the commercial catch information will yield an estimate of recruitment for years 1 … T through some procedure such as virtual population analysis (VPA). This gives an estimate of absolute abundance (AA), while research surveys can give estimates of relative abundance (RA) or absolute abundance of recruits. The regression methods make specific assumptions concerning the relative magnitude of the errors in the estimates from different sources, e.g. that (AA) is most precise (calibration) or of no greater precision than the other indices (prediction). The difficulty arises in choosing between these methods with real data where the relative sizes of the errors in the available estimates of year-class strength are unknown.

Simulation tests were constructed to test the three basic methods as well as a calibration regression where the resulting weighted average includes a term for “shrinking” the estimate towards the mean of the AA series. The tests indicate that factor analysis and calibration with shrinkage perform best overall. Calibration can be quite sensitive to missing data, however, and may break down if the most recent year's recruitment is far from the mean of the AA series. Under these conditions, factor analysis performs better in simulation trials.

Author notes

present address: National Marine Fisheries Service, Woods Hole, Massachusetts 02543, USA.