Predicting bycatch hotspots in tropical tuna purse seine fisheries at the basin scale
Also published as ITOC 2022 WPEB18-INF13.
Fisheries observer programs represent the most reliable way to collect data on fisheries bycatch. However, their limited coverage leads to important data gaps that preclude bycatch mitigation at the basin scale. Habitat models developed from available fisheries observer programs offer a potential solution to fill these gaps. We focus on tropical tuna purse seine fisheries (TTPSF) that span across the tropics and extensively rely on floating objects (FOBs) for catching tuna schools, leading to the bycatch of other species associated with these objects. Bycatch under floating objects is dominated by five species, including the vulnerable silky shark Carcharhinus falciformis and four bony fishes (oceanic triggerfish Canthidermis maculata, rainbow runner Elagatis bipinnulata, wahoo Acanthocybium solandri, and dolphinfish Coryphaena hippurus). Our objective was to predict possible bycatch hotspots associated with FOBs for these five species across two tropical oceans. We used bycatch data collected from observer programs onboard purse seiners in the Atlantic and Indian oceans. We developed a generalized additive model per species and per ocean relating bycatch to a set of environmental covariates (depth, chlorophyll-a concentration, sea surface temperature, mixed layer depth, surface salinity, total kinetic energy and the density of floating objects) and temporal covariates (year and month). We extrapolated modeled relationships across each ocean within the range of environmental covariates associated with the bycatch data and derived quarterly predictions. We then detected bycatch hotspots as the 90th percentiles of predictions. In the Atlantic Ocean, bycatch hotspots were predicted throughout tropical and subtropical waters with little overlap between species. By contrast in the Indian Ocean, major overlapping hotspots were predicted in the Arabian Sea throughout most of the year for four species, including the silky shark. Our modeling approach provides a new analytical way to fill data gaps in fisheries bycatch. Even with the lack of evaluation inherent to extrapolations, our modeling effort represents the first step to assist bycatch mitigation in TTPSF and is applicable beyond these fisheries.