Proof of concept for a novel procedure to standardize multispecies catch and effort data

The effect of variability in targeting needs to be removed from catch-per-unit-effort (CPUE) data to estimate reliable abundance indices for multispecies fisheries. We test a Generalized Additive Model (GAM) that includes principal component scores (PCs) derived from the species composition in the catch, called the ‘Direct Principal Component’ (DPC) procedure, for its ability to remove the effect of variable targeting. Biomass trends are simulated for two multispecies, multi-habitat fishery scenarios: (i) four species distributed differentially across two habitats and (ii) ten species distributed differentially across four habitats. Tweedie distributed CPUE records are generated from the biomass trends for a fishery with constant targeting (control scenarios) and time-varying targeting (test scenarios). The DPC procedure is simulation-tested for its ability to estimate the underlying biomass trends for all species relative to the non-standardized CPUE index. The DPC procedure proved to be more accurate compared to nominal CPUE trends in the test scenarios. Even in the control scenarios, the DPC procedure offers greater accuracy for the estimated year effect by removing substantial variation from the data, with a small penalty on the accuracy of the underlying abundance trend. However, caution is advised if the DPC-derived index diverges noticeably from alternative models despite no indications for shifts in targeting. A selection procedure based on eigenvalues of the PCs is suitable to identifying the best-performing number of PCs to include in the GAM. The DPC procedure should be applicable for a variety of multispecies fisheries.