Investigation of the paired-gear method in selectivity studies

We estimated selectivity parameters using simultaneously the paired-gear and covered codend method for two fish species and four different selection systems, for a total of eight study cases. The deviation (Δ) in L50 and SR between these sampling methods observed in a former simulation study was repeated throughout the eight cases in this investigation. When using the paired-gear method, the distribution of the estimated L50 and SR is wider; the distribution of the estimated split parameter has a higher variability than the true split; the estimated mean L50 and SR can be biased; the estimated between-haul variation is different from that estimated by the covered codend. ΔL50 and ΔSR decrease when the number of fish in the codend increases, but they do not necessarily progress towards zero. ΔL50 and ΔSR are positively correlated with the deviation between the split and the true split. We recommend that the methodology used to obtain selectivity estimates using the paired-gear method be reviewed.