Shortfin mako hot sets – Defining high bycatch conditions as a basis for bycatch mitigation
Introduction
Shortfin mako sharks are highly migratory, long-lived sharks that are not known to aggregate. Shortfin makos tagged off the USA are distributed across the continental shelf and pelagic habitats following seasonal movements north and south that are correlated with seasonal patterns in ocean productivity (Queiroz et al., 2016, Vaudo et al., 2017). They prefer water temperatures 17–22 °C and appear to move inshore when the continental shelf waters warm, starting in April and May and move further east into the pelagic habitat in the fall and winter (Queiroz et al., 2016). Satellite tags also show that individual sharks exhibit high variability in their movements with some making long-distance migrations south into oligotrophic waters (Vaudo et al., 2017). Shortfin mako sharks’ distribution has high overlap with longline fleets (Queiroz et al., 2016, Queiroz et al., 2019).
In the Atlantic, shortfin makos are caught as bycatch in commercial longline fisheries targeting tunas and swordfish. They are assessed and managed by the International Commission for the Conservation of Atlantic Tunas (ICCAT) and historically have been one of the most commercially valuable sharks (Levesque, 2008). The population dynamics models used in the most recent stock assessment (Anonymous, 2017) and stock assessment update (Anonymous, 2019a) agree that the North Atlantic stock is overfished and experiencing overfishing. The ICCAT shark working group determined that the total shortfin mako catch needs to be reduced to less than 500 t from the level of 3115 t in 2017 to eliminate overfishing (Anonymous, 2019a). These findings contributed to the 2019 Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) listing the shortfin mako under Appendix II (Anonymous, 2019b). This listing will require that fishing nations demonstrate that fishing the shortfin mako would not threaten their chances for survival to obtain a CITES export permit for mako shark products (Anonymous, 1973). ICCAT recommendations and a listing under CITES acknowledge that shortfin mako sharks are not productive enough to rebuild without intervention.
This work looks specifically at the US portion of the pelagic longline fishery where shortfin mako are caught as bycatch but are not allowed to be retained. The US Pelagic Observer Program started in 1992 and monitors the US pelagic longline fleet operating from Newfoundland to Brazil, including the Caribbean and Gulf of Mexico (Beerkircher et al., 2004). The program is managed by NOAA’s Southeast Fisheries Science Center and has covered about 5% (1992–2001) to 8% (2002-present) of the vessels operating in this fishery. Observers on selected vessels record information about gear configurations and the species, size, sex, dead/alive status, and the ultimate fate (kept or discarded) of fish caught (Beerkircher et al., 2004). Bycatches of shortfin mako sharks recorded by observers in the US pelagic longline fishery are typically zero in most sets, so the shortfin mako bycatch per set data may be zero-inflated. When there is a positive shortfin mako bycatch, most of the time only 1 or 2 shortfin makos are caught per set. However, there are several events in the shortfin mako bycatch history that have caught upwards of 20 shortfin makos in one set. If we can identify the set of conditions that lead to sets with unusually high shortfin mako bycatch, hereafter referred to as “hot sets”, it may be possible to use this information to avoid high shortfin mako bycatch events and ultimately reduce the overall fishing mortality.
A generalized linear modeling (GLM) and generalized additive modeling (GAM) approach was used to predict mean shortfin mako bycatch per unit effort (CPUE), similar to the standardization methods used to generate abundance indices for past stock assessments (Cortes, 2007, Thorson et al., 2015). This study also determined conditions favoring particularly high CPUE, where much of the fishing mortality occurs. Quantile regression was used to focus on the upper tail of the CPUE distribution, rather than the mean, to reveal which conditions predict higher CPUE. Quantile regression is a method appropriate for abundance data with non-linear, non-symmetric, heterogenous scatter in response to a gradient (Koenker and Bassett, 1978, Cade and Noon, 2003, Anderson, 2008, Fukunaga et al., 2016). It addresses the scatter by allowing the coefficients to differ for the different parts of the abundance distribution (Koenker and Bassett, 1978) and has been used in ecology particularly to test whether environmental predictor variables predict areas of high or low abundance (Cade and Noon, 2003, Anderson, 2008, Fornaroli et al., 2016, Fukunaga et al., 2016). Application of quantile regression to the upper extreme allows the independent environmental variables to be viewed more easily as boundaries to suitable habitat (Fornaroli et al., 2016). Development of both model types involved a model selection process that included exploring categorical and numerical methods as well as GAMs with smoothers on the environmental variables to improve model fit and performance.
The traditional delta-lognormal approach (predicted probability of positive shortfin mako bycatch multiplied by the predicted mean of the positive shortfin mako bycatch to get the overall predicted mean CPUE) is hypothesized to perform poorly compared to the quantile regression fit to the upper extreme to predict hot sets. Quantile regression eliminates the need for a delta-lognormal approach and the difficulties in using zero-inflated data because the higher quantiles do not include zeros. For the purposes of determining conditions that favor a hot set, as opposed to standardized shortfin mako bycatch rates over time, quantile regression may be a better method. There is also more flexibility in model fit because there is no assumption of a normal distribution and the error distribution does not need to be specified. This method may provide the tools necessary to best identify potential hot sets and design a bycatch mitigation strategy.
This paper will present several ways to identify environmental conditions, regions and fishing methods that favor shortfin mako bycatch based on the outputs of the delta-lognormal model and quantile regression of the upper quantiles. The analysis will then determine which definition of a hot set and model could be used as a dynamic “no fish” algorithm to avoid shortfin mako bycatch and lower fishing mortality. The ultimate objective is to determine whether identifying conditions that favor high shortfin mako bycatch has the potential to decrease shortfin mako fishing mortality substantially while maintaining catches of target species.
Section snippets
Data preparation
Catch and effort data were obtained from the US pelagic longline observer program (1992–2016, 18,893 sets) (Southeast Fisheries Science Center, 2016), while weekly sea surface temperature composites on a 4 km grid (2003–2016) (GRIDDAP, 2018), daily sea surface height on a 0.25 degree grid (1992–2012) (Ducet et al., 2000), and daily chlorophyll-a concentrations (mg/m3) on a 4 km grid (2003–2021) (GRIDDAP, 2021) were downloaded from the NOAA CoastWatch satellite database and bathymetry on a 1
Results
In the US pelagic longline observer data, shortfin mako bycatch per set is generally low with a median of 0.0 and a mean of 0.68 ± 2.23 shortfin mako sharks. Most sets catch no shortfin mako sharks, or at most one or two (Fig. 1a). However, sets that catch one shortfin mako shark account for less than 18% of the total shortfin mako bycatch and sets that catch two or fewer shortfin mako account for less than 34% of the total shortfin mako bycatch (Fig. 1b). Sets with shortfin mako bycatches
Discussion
We presented several ways to identify environmental conditions, regions and fishing methods that favor high shortfin mako bycatch based on the outputs of the delta-lognormal model and quantile regression of the upper quantiles. We found that using the binomial portion of the delta-lognormal model, the probability of positive bycatch, was the best way to define a hot set basis for a “no fish” algorithm. Three potential mitigation strategies were designed based on the identification of hot sets
Conclusion
The shortfin mako catch in the pelagic longline fishery is highly clumped and patchy, with many sets catching none or one shark while some sets catch over 50 individuals at once. If most sets are not contributing to the bycatch rates, a method that focuses only on sets contributing to high bycatch rates would in theory be better, despite the added complexity. However, this was not the case; quantile regression, a more complicated method which can focus on high catch rates, did not perform
CRediT authorship contribution statement
Halie O’Farrell: Conceptualization, Methodology, Software, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Visualization, Elizabeth Babcock: Conceptualization, Methodology, Software, Resources, Writing – review & editing, Supervision, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
Thank you to Dr. David Die, Dr. Enric Cortes, Dr. Donald Olson, and Dr. John McManus for your scientific guidance and writing assistance. Thank you to the NOAA Observer Program and all observers, past and present, for collecting all the catch data used in this study. Thank you to NOAA Coastwatch, AVISO, and NASA's Goddard Space Flight Center for providing all the satellite data.
This project was funded by the NOAA Educational Partnership Program through the Living Marine Resources Cooperative
References (57)
Animal-sediment relationships re-visited: characterising species’ distributions along an environmental gradient using canonical analysis and quantile regression splines
J. Exp. Mar. Biol. Ecol.
(2008)- et al.
Revealing pelagic habitat use: the tagging of Pacific pelagics program
Oceanol. Acta
(2002) - et al.
Evaluating resource selection functions
Ecol. Model.
(2002) - et al.
Optimal flow for brown trout: habitat – prey optimization
Sci. Total Environ.
(2016) - et al.
Using delta generalized additive models to produce distribution maps for spatially explicit ecosystem models
Fish. Res.
(2014) - et al.
Evaluation of the impacts of different treatments of spatio-temporal variation in catch-per-unit-effort standardization models
Fish. Res.
(2019) - et al.
Generalized linear and generalized additive models in studies of species distributions: setting the scene
Ecol. Model.
(2002) International fisheries agreement: review of the International Commission for the Conservation of Atlantic Tunas case study – shark management
Mar. Policy
(2008)- et al.
Standardizing catch and effort data: a review of recent approaches
Fish. Res.
(2004) - et al.
Alternative error distribution models for standardization of catch rates of non-target species from a pelagic longline fishery: billfish species in the Venezuelan tuna longline fishery
Fish. Res.
(2004)