Using habitat association models to predict Alewife and Blueback Herring marine distributions and overlap with Atlantic Herring and Atlantic Mackerel: can incidental catches be reduced?

Abstract

Concern over the impacts of incidental catches of Alewife, Alosa pseudoharengus and Blueback Herring, A. aestivalis (collectively managed as ‘river herring’) in the commercial Atlantic Herring (Clupea harengus) and Atlantic Mackerel (Scomber scombrus) fisheries has resulted in the recent implementation of river herring incidental catch limits. These incidental catches are highly variable in frequency and magnitude, and the environmental conditions associated with these catches are poorly understood. We used generalized additive models (GAMs) to describe habitat associations of Alewife, Blueback Herring, Atlantic Herring, and Atlantic Mackerel. Bottom temperature, bottom depth, bottom salinity, solar azimuth and elevation, and region of the Northeast U.S. continental shelf were all significant in the habitat models; GAMs explained 25.2, 16.9, 18.9, and 20.6% of the deviance observed for the presence/absence of Alewife, Blueback Herring, Atlantic Herring, and Atlantic Mackerel. A subset of the data was omitted from the model and the probability of presence was compared with observations; 66–77% of observations were correctly predicted. The individual probabilities of presence were used to quantify and evaluate the accuracy of modelled overlap of Alewife and Blueback Herring with Atlantic Herring (68–72% correct predictions) and Alewife and Blueback Herring with Atlantic Mackerel (57–69% correct predictions). Our findings indicate that environmental gradients influence the distributions and overlap of Alewife, Blueback Herring, Atlantic Herring, and Atlantic Mackerel, and with further testing and refinement these models could be developed into a tool to aid industry in reducing incidental catches of river herring.

Cooperative Research to Evaluate an Incidental Catch Distribution Forecast

“…Species' distribution forecasts were previously developed and evaluated with fishery-independent data (Turner et al, 2015(Turner et al, , 2017 for alewife, blueback herring, Atlantic herring, and Atlantic mackerel (not included in this study), encompassing most of the Northeast U.S. continental shelf (Figure 1). The spatial scale of this study (41.15-41.50 • N, − 71.86-− 71.40 • W; Figure 1) was substantially smaller than for initial model development and evaluation because the vessels used by the Atlantic herring small mesh bottom trawl in this study typically only take 1 day trips because of limited carrying capacity.…”

“…Generalized additive models for the habitat associations of alewife, blueback herring, and Atlantic herring, based on fisheryindependent data, were previously developed using the R package "mgcv" v. 1.8-6 (Hastie and Tibshirani, 1990;Wood, 2006;Politis et al, 2014;R Development Core Team, 2014;Turner et al, 2015). Species' presence was the response variable, using a binomial link function, and all models included smooth functions of bottom temperature, bottom salinity, and depth, a tensor product smooth of solar azimuth and solar elevation, and region as a factor variable (Turner et al, 2015).…”

“…Species' presence was the response variable, using a binomial link function, and all models included smooth functions of bottom temperature, bottom salinity, and depth, a tensor product smooth of solar azimuth and solar elevation, and region as a factor variable (Turner et al, 2015). Model probabilities of species presence using forecasted and observed conditions were evaluated for all three species using paired ttests.…”

“…Environmental forecast models have been used to predict species' distributions and habitat overlap with other commercially valuable species months in advance for southern bluefin tuna, Thunnus maccoyii (Hartog et al, 2011;Eveson et al, 2015). These models have longer temporal accuracy than those used here, but the spatial resolution is coarser; also, the study species' thermal associations are more distinct than observed for Atlantic herring, alewife, and blueback herring (Hartog et al, 2011;Turner et al, 2015). The difference between the spatial scales of the fishery-independent trawl survey (roughly 20 km or more between stations) that was used for model development and the Atlantic herring small-mesh bottom trawl fishery (total area of roughly 60 by 40 km) may further complicate our model evaluations.…”

“…We previously developed and evaluated species' distribution models using fishery-independent data (Turner et al, 2015), which we then coupled with an ocean forecast model (FVCOM; Chen et al, 2006) and evaluated its accuracy using newly collected fishery-independent data (Turner et al, 2017). Species' distribution forecasts correctly predicted presence for 68-69% of fishery-independent observations, but given the substantially different spatial scales of the trawl survey used for model development and initial testing and the fishery, fisherydependent evaluations were necessary.…”