Fitting fisheries models to standardised CPUE abundance indices

Abstract

The observation error method of fitting fisheries models to abundance indices is described. The effects of differing within year coefficients of variation between individual abundance indices are investigated. Likelihoods which are specifically designed to be used with standardised abundance indices estimated from catch and effort data when CV’s differ between years are presented. Monte Carlo simulation is used to investigate the proposed methods. Four main conclusions are drawn from the analysis: (1) Additional bias and reduced precision will be introduced into the population parameter estimates if the CV’s differ between abundance indices and if this difference is not incorporated into the fitting procedure. These are potentially large enough to cause concern. (2) Two procedures are proposed which can incorporate differing yearly CV’s. One method, the “Ad hoc” method appeared to perform the best. Both methods gave less biased estimates and lower average absolute error than the usual method of assuming a single CV for all index values. (3) The procedures proposed will give population parameter estimates which are better determined than if a single CV is assumed for all index values when there is a strong outlier index value with an associated high CV. (4) It is suggested that, when fitting population models to abundance indices generated by catch and effort data, both procedures (single and multiple CV’s) be applied to test the sensitivity of the population parameter estimates to this potential source of bias.