Abstract
Catch data from bottom trawl surveys are used in various ways (e.g. stock assessments, fisheries management, and ecosystem studies) to represent trends in fish populations across space, time, season, or size. Relative abundance indices assume constant capture efficiency, and area-swept abundance requires an estimate of capture efficiency. Therefore, it is important to develop a predictive understanding of the interaction between fish and survey gear. We conducted experiments to test two primary factors that influence the efficiency of survey trawls at capturing demersal groundfish: (1) footrope escapement-estimated by attaching a collection bag beneath the primary trawl, and (2) herding of the sweeps/doors-estimated by varying sweep length. Random forest models were used to disentangle the herding effect from patterns caused by environmental variables. Contrary to common assumptions, footrope efficiency was incomplete (< 100%) and herding was non-trivial (>0%), which introduces a bias in population metrics that rely on such assumptions. This bias varied by species and depended upon the relative strength of the counteracting effects of footrope escapement and herding. Our findings suggest that trawl efficiency should be estimated (not assumed) to derive area-swept abundance, and relative abundance indices should account for size-based efficiency and changing size compositions.