Prediction of delayed mortality using vitality scores and reflexes, as well as catch, processing, and post-release conditions: evidence from discarded flatfish in the Western Baltic trawl fishery

Abstract

Abstract The EU discard ban and its high-survival exemption exposed our lack of scientific evidence on discard survival in the fisheries. Discard survival is known to be highly variable and influenced by numerous factors, including conditions during the catch, on-board the fishing vessels, and post-discard. Therefore, obtaining unambiguous results in discard survival experiments is challenging. We conducted the first systematic year-round discard survival study of flatfish in the Western Baltic Sea on-board a commercial stern trawler under realistic fishing conditions (13 monthly hauls from May 2015 to May 2016) to test whether delayed mortality can be predicted by vitality scores and reflex action mortality predictor (RAMP) scores in combination with variable conditions during catch, processing, and post-release. The factors vessel type, gear, haul duration, fishing ground, depth, handling time, and processing procedures were kept constant as much as possible. On-board, live individual flatfish were tested for vitality and the presence of reflexes for RAMP, then “discarded” and kept in cages on the bottom of the seafloor for about a week after which the delayed mortality was determined. The proportions of “discarded” plaice, flounder, and dab that were dead after being kept in the cages ranged from 5% to 100%, 0% to 96%, and 33% to 100%, respectively. Higher mortalities occurred in summer-autumn when air and water temperatures were higher, catches smaller, and catches contained smaller proportions of roundfish. Relationships between RAMP scores and mortality probabilities varied substantially across the monthly trials. Indeed, in addition to RAMP or vitality scores and individual reflexes, the factors air and water temperature and catch weight and catch composition were significant in logistic GLMs explaining delayed mortality. Cross-validations indicated that delayed mortality could be predicted by these models with a reasonable accuracy. Nevertheless, the presence of possible confounding effects calls for caution in inferring causality and extrapolating the conclusions on predictability.