Résumé
Estimating bycatch, i. e. the incidental catch of non-target marine animals and undersized individuals of target species, by raising observer data to the whole fishery is routine practice. The annual bycatch of the European tropical tuna purse-seine fishery over the period 2003-2009 was estimated at 11 590 t [ 95% confidence interval: (8165-15 818 t)], corresponding to 4.7% of the tuna landings. An analysis of the variability in the precision of this estimate, based on generalized linear models and Monte Carlo simulations, showed that the current sampling coverage of the tropical tuna fishery observer programme, which is 4.6% of the fishing trips, resulted in large uncertainties in bycatch estimates by species, i. e. none of the estimates have a relative root mean square error smaller than 50%. Although the overall magnitude of bycatch of the fishery appeared to be small, the current sampling coverage was insufficient to give any reliable estimate for low-occurring species, such as marine turtles, some oceanic pelagic sharks, and some billfishes. Increasing the sampling coverage would likely improve bycatch estimates. Simulation outputs were produced to help define (i) trade-offs between the priority species to be monitored, (ii) the estimation precision, (iii) expected accuracy, and (iv) the associated sampling costs.