Simulation-testing model-based and design-based bycatch estimators

Babcock EA, Harford WJ, Gedamke T, et al (2023) Simulation-testing model-based and design-based bycatch estimators. ICCAT Collect Vol Sci Papers 80:51–79

The BycatchEstimator tool developed by Babcock (2022) was used to estimate bycatch in fisheries simulated using the species distribution model and longline simulator (LLSIM) developed by Goodyear (2021). To compare the effectiveness of several design-based and model-based estimators that are used to estimate bycatch in a realistic context, an observer program similar to the US pelagic longline observer program (USPLL) was simulated. The estimates of total bycatch were precise and unbiased for all methods during recent years with high observer coverage. However, in the early years with lower observer coverage, the design-based methods (delta lognormal and ratio) performed worse than the delta lognormal model. The results were sensitive to how observers were allocated to trips. A geostatistical model showed that total bycatch estimates were more precise when spatial and/or spatiotemporal random effects were included. The BycatchEstimator tool was also applied to the real data from the USPLL. The tool was able to recreate the US Task 1 estimates in recent years, but when observer coverage was lower, estimates were sensitive to how strata with low sample sizes were pooled.