ICES Journal of Marine Science: Journal du Conseil

van Gemert, Rob; Andersen, Ken H

doi: 10.1093/icesjms/fsx236pmid: N/A

Abstract Currently applied fisheries models and stock assessments rely on the assumption that density-dependent regulation only affects processes early in life, as described by stock–recruitment relationships. However, many fish stocks also experience density-dependent processes late in life, such as density-dependent adult growth. Theoretical studies have found that, for stocks which experience strong late-in-life density dependence, maximum sustainable yield (MSY) is obtained with a small fishery size-at-entry that also targets juveniles. This goes against common fisheries advice, which dictates that primarily adults should be fished. This study aims to examine whether the strength of density-dependent growth in actual fish stocks is sufficiently strong to reduce optimal fishery size-at-entry to below size-at-maturity. A size-structured model is fitted to three stocks that have shown indications of late-in-life density-dependent growth: North Sea plaice (Pleuronectes platessa), Northeast Atlantic (NEA) mackerel (Scomber scombrus), and Baltic sprat (Sprattus sprattus balticus). For all stocks, the model predicts exploitation at MSY with a large size-at-entry into the fishery, indicating that late-in-life density dependence in fish stocks is generally not strong enough to warrant the targeting of juveniles. This result lends credibility to the practise of predominantly targeting adults in spite of the presence of late-in-life density-dependent growth. Introduction Density dependence is a key process in population ecology. Negative (or compensatory) density dependence takes place when an increase in population size results in a decrease in individual growth, reproduction, or survival, usually due to increased intraspecific competition or increased predation mortality. Density dependence due to intraspecific competition can, for instance, stem from competition for food (Hassell, 1975) or spawning sites (Reichard et al., 2004). Likewise, density dependence as a result of predation mortality can stem from cannibalism (Ricker, 1954), or from the predator switching to the most abundant prey (type III functional response; Holling, 1959). Because population density is changed by exploitation, it is essential to understand how density dependence operates within a population when predicting how that population may respond to exploitation. The strength of density dependence varies with the size of the individual. Here we distinguish between two mechanisms of density dependence: early-in-life density-dependent recruitment and late-in-life density-dependent growth. In many stocks, individuals experience strong density dependence during the larval and early-juvenile stage. In spite of the wide prevalence of this early-in-life density dependence, its causal mechanisms are usually poorly understood. Therefore, it is sometimes referred to as density-dependent recruitment, as it takes place before the individual enters the “recruited” component of the stock. Whereas density-dependent recruitment takes place early in life, density-dependent growth can be assumed to be strongest later in life, after an individual reaches size-at-maturity. This is because density-dependent growth emerges due to resource competition, at the adult size where biomass of a cohort (and therefore its consumption) is usually the largest (e.g. Munch et al., 2005; Jennings et al., 2007). However, in spite of its potentially significant role in population regulation, late-in-life density-dependent growth is rarely incorporated in the calculation of fisheries reference points. Instead, current fisheries advice is generally given under the assumption that all density dependence occurs early in life, in the form of density-dependent recruitment (e.g. Beverton and Holt, 1957; Myers and Cadigan, 1993). This early-in-life density dependence is either described as a constant recruitment (yield-per-recruit models), or through a stock–recruitment relationship. This assumption of only early-in-life density dependence is likely acceptable for fish stocks that experience heavy fishing pressure, where fishing mortality relieves the exploited population component from late-in-life density dependence. However, during the last decade, improved fisheries management has led to many fish stocks in the NEA gradually showing signs of recovery from overfishing (Fernandes and Cook, 2013). For some species, this recovery coincided with reduced individual growth of older juveniles and adults, possibly as a result of late-in-life density-dependent resource competition (e.g. Cormon et al., 2016; Olafsdottir et al., 2016). Therefore, it may be problematic that late-in-life density-dependent growth is rarely taken into account in fisheries management. Optimal management strategies could differ substantially for stocks that experience late-in-life density-dependent growth. For example, a model study by Andersen et al. (2017) showed that if density-dependent regulation mainly happens late in life, maximum sustainable yield (MSY) is obtained by fishing on juvenile fish. This relieves the remaining juveniles from density dependence, thereby increasing the productivity of the entire stock. This prediction challenges reigning fisheries management procedures, which enforce minimum landing size regulations to avoid excessive fishing mortality on juveniles. The study of Andersen et al. (2017) only compared scenarios for hypothetical stocks, where density dependence either occurred mainly early in life or mainly late in life. However, density-dependent population regulation need not necessarily occur at only a single bottleneck. Given the widespread nature of density-dependent processes, it is likely that many fish stocks experience some form of density-dependent regulation at multiple life stages. For instance, Dover sole (Solea solea) recruitment appears to follow a classic Beverton-Holt stock–recruitment relationship (Lorenzen, 2005). This is indicative of strong early-in-life density dependence, but the stock also shows significant density-dependent growth in the recruited phase (Lorenzen and Enberg, 2002). Another example is North Sea plaice (P. platessa), which shows strong early-in-life density dependence when larvae settle in their nursery grounds (Van der Veer, 1986), but has also shown significant late-in-life density-dependent growth (Rijnsdorp and Van Leeuwen, 1992). Based on the findings of Andersen et al. (2017), fishery size-at-entry at which MSY is obtained should gradually decrease when the strength of late-in-life density-dependent growth increases (relative to that of early-in-life density dependence). However, it is unknown whether the late-in-life density-dependent growth that is experienced by marine fish stocks is actually strong enough to trigger a reduction in optimal fishery size-at-entry. We aim to explore whether marine fish stocks can actually experience late-in-life density-dependent growth that is strong enough to reduce optimal fishery size-at-entry (i.e. size-at-entry at which MSY is obtained). To this end we estimated the relative strengths of early- and late-in-life density dependence in three fish stocks, by fitting a dynamic single-stock size-structured model to empirical stock data. The three examined fish stocks, North Sea plaice (P. platessa) (Rijnsdorp and Van Leeuwen, 1992), NEA mackerel (S. scombrus) (Olafsdottir et al., 2016), and Baltic sprat (S. s. balticus) (Eero, 2012), have shown indications of experiencing some late-in-life density-dependent resource competition and only show little cannibalism. We focus on density-dependent resource competition as the primary mechanism behind late-in-life density-dependent regulation, to avoid any confounding effects of cannibalism. In our model, early-in-life density dependence is described by a stock–recruitment relationship. Late-in-life density dependence is not described with a single equation, but emerges through feeding on a dynamic resource spectrum. Varying the relative strengths of early- and late-in-life density dependence was possible by varying the stock–recruitment relationship’s maximum recruitment relative to the resource spectrum’s carrying capacity. This allowed us to examine whether the strength of density-dependent growth experienced by the stock is high enough so that optimal fishery size-at-entry is below size-at-maturity. Methods We apply a standard size-spectrum model (Andersen et al., 2015), adapted to represent only a single stock (Andersen et al., 2017). The model describes the population dynamics of a single fish stock feeding on a dynamic resource spectrum and incorporates early-in-life density dependence through a Beverton-Holt stock–recruitment relationship, and late-in-life density dependence emerges through size-based resource competition. Here we describe the main assumptions and principles of the model. Detailed descriptions of the assumptions and equations used in size-spectrum models such as this one can also be found e.g. in Hartvig et al. (2011), Andersen and Beyer (2015), and Andersen et al. (2015). All model equations and parameters are listed in Tables 1 and 2, and the numerical implementation of our model is given in Supplementary Appendix A. Throughout, size refers to body weight, w. Table 1. Governing model equations. Consumption  Size preference for prey ϕ(wwR)= exp ⁡[−(ln⁡(wwRβ))2/(2σ2)] M1  Encountered food Ee(w)=γwq∫0∞ϕ(wwR)wRNR(wR)dwR M2  Feeding level f(w)=Ee(w)Ee(w)+hwn M3 Growth  Available energy Ea(w)=αf(w)hwn−krwn−kaw M4  Switching function H(x)=(1+x−10)−1 M5  Maturation ψ(w)=H(wηmW∞)1−ϵa(w/W∞)n−1−ϵa M6  Growth rate g(w)=(1−ψ(w))Ea(w) M7 Mortality  Background predation μ0(w)=αpwn−1 M8  Fishing, trawl selectivity μF(w)=FH(wwF) M9 Reproduction  Egg production Rp=∫weggW∞ψ(w)Ea(w)2weggN(w)dw M10  Recruitment R=RmaxϵrRpRmax+ϵrRp M11 Mortality  Background predation μ0(w)=αpwn−1 M12  Fishing, trawl selectivity μF(w)=FH(wwF) M13 Population structure  Abundance spectrum ∂N(w)∂t+∂g(w)N(w)∂w=−[μ0(w)+μF(w)]N(w) M14  Boundary condition g(wegg)N(wegg)=R M15  SSB BSSB=∫weggW∞H(wηmW∞)wN(w)dw M16 Fishery performance  Yield Y=∫weggW∞μF(w)wN(w)dw M17 Resource  Predation on resource μp(wR)=∫weggW∞ϕ(wwR)(1−f(w))γwqN(w)dw M18  Resource spectrum ∂NR(wR)∂t=r0wRn−1[κwRλ−NR(wR)]−μp(wR)NR(wR) M19 Consumption  Size preference for prey ϕ(wwR)= exp ⁡[−(ln⁡(wwRβ))2/(2σ2)] M1  Encountered food Ee(w)=γwq∫0∞ϕ(wwR)wRNR(wR)dwR M2  Feeding level f(w)=Ee(w)Ee(w)+hwn M3 Growth  Available energy Ea(w)=αf(w)hwn−krwn−kaw M4  Switching function H(x)=(1+x−10)−1 M5  Maturation ψ(w)=H(wηmW∞)1−ϵa(w/W∞)n−1−ϵa M6  Growth rate g(w)=(1−ψ(w))Ea(w) M7 Mortality  Background predation μ0(w)=αpwn−1 M8  Fishing, trawl selectivity μF(w)=FH(wwF) M9 Reproduction  Egg production Rp=∫weggW∞ψ(w)Ea(w)2weggN(w)dw M10  Recruitment R=RmaxϵrRpRmax+ϵrRp M11 Mortality  Background predation μ0(w)=αpwn−1 M12  Fishing, trawl selectivity μF(w)=FH(wwF) M13 Population structure  Abundance spectrum ∂N(w)∂t+∂g(w)N(w)∂w=−[μ0(w)+μF(w)]N(w) M14  Boundary condition g(wegg)N(wegg)=R M15  SSB BSSB=∫weggW∞H(wηmW∞)wN(w)dw M16 Fishery performance  Yield Y=∫weggW∞μF(w)wN(w)dw M17 Resource  Predation on resource μp(wR)=∫weggW∞ϕ(wwR)(1−f(w))γwqN(w)dw M18  Resource spectrum ∂NR(wR)∂t=r0wRn−1[κwRλ−NR(wR)]−μp(wR)NR(wR) M19 Open in new tab Table 1. Governing model equations. Consumption  Size preference for prey ϕ(wwR)= exp ⁡[−(ln⁡(wwRβ))2/(2σ2)] M1  Encountered food Ee(w)=γwq∫0∞ϕ(wwR)wRNR(wR)dwR M2  Feeding level f(w)=Ee(w)Ee(w)+hwn M3 Growth  Available energy Ea(w)=αf(w)hwn−krwn−kaw M4  Switching function H(x)=(1+x−10)−1 M5  Maturation ψ(w)=H(wηmW∞)1−ϵa(w/W∞)n−1−ϵa M6  Growth rate g(w)=(1−ψ(w))Ea(w) M7 Mortality  Background predation μ0(w)=αpwn−1 M8  Fishing, trawl selectivity μF(w)=FH(wwF) M9 Reproduction  Egg production Rp=∫weggW∞ψ(w)Ea(w)2weggN(w)dw M10  Recruitment R=RmaxϵrRpRmax+ϵrRp M11 Mortality  Background predation μ0(w)=αpwn−1 M12  Fishing, trawl selectivity μF(w)=FH(wwF) M13 Population structure  Abundance spectrum ∂N(w)∂t+∂g(w)N(w)∂w=−[μ0(w)+μF(w)]N(w) M14  Boundary condition g(wegg)N(wegg)=R M15  SSB BSSB=∫weggW∞H(wηmW∞)wN(w)dw M16 Fishery performance  Yield Y=∫weggW∞μF(w)wN(w)dw M17 Resource  Predation on resource μp(wR)=∫weggW∞ϕ(wwR)(1−f(w))γwqN(w)dw M18  Resource spectrum ∂NR(wR)∂t=r0wRn−1[κwRλ−NR(wR)]−μp(wR)NR(wR) M19 Consumption  Size preference for prey ϕ(wwR)= exp ⁡[−(ln⁡(wwRβ))2/(2σ2)] M1  Encountered food Ee(w)=γwq∫0∞ϕ(wwR)wRNR(wR)dwR M2  Feeding level f(w)=Ee(w)Ee(w)+hwn M3 Growth  Available energy Ea(w)=αf(w)hwn−krwn−kaw M4  Switching function H(x)=(1+x−10)−1 M5  Maturation ψ(w)=H(wηmW∞)1−ϵa(w/W∞)n−1−ϵa M6  Growth rate g(w)=(1−ψ(w))Ea(w) M7 Mortality  Background predation μ0(w)=αpwn−1 M8  Fishing, trawl selectivity μF(w)=FH(wwF) M9 Reproduction  Egg production Rp=∫weggW∞ψ(w)Ea(w)2weggN(w)dw M10  Recruitment R=RmaxϵrRpRmax+ϵrRp M11 Mortality  Background predation μ0(w)=αpwn−1 M12  Fishing, trawl selectivity μF(w)=FH(wwF) M13 Population structure  Abundance spectrum ∂N(w)∂t+∂g(w)N(w)∂w=−[μ0(w)+μF(w)]N(w) M14  Boundary condition g(wegg)N(wegg)=R M15  SSB BSSB=∫weggW∞H(wηmW∞)wN(w)dw M16 Fishery performance  Yield Y=∫weggW∞μF(w)wN(w)dw M17 Resource  Predation on resource μp(wR)=∫weggW∞ϕ(wwR)(1−f(w))γwqN(w)dw M18  Resource spectrum ∂NR(wR)∂t=r0wRn−1[κwRλ−NR(wR)]−μp(wR)NR(wR) M19 Open in new tab Table 2. Model parameters. Symbol . Description . Value . Unit . Footnote(s) . Body size   W∞ Asymptotic size (weight)* stock specific g   wegg Egg weight 0.001 g a Consumption  n Metabolic exponent 3/4 – b  β Preferred predator–prey mass ratio 100 – c  σ Range of preferred prey size 1.3 – d  q Clearance rate exponent 0.8 – e  γ Clearance rate coefficient 6.57/κ g−q year−1 f  f0 Standard feeding level 0.6 – f  fc Critical feeding level 0.2 – f  h Maximum consumption* ≈3KW∞1/3/[α(f0−fc)] g1−n year−1 g Growth  α Assimilation efficiency 0.6 – f  ηm Size at maturation rel. to W∞ 0.25 – h  ϵa Fraction of energy for activity 0.8 – i  kr Standard metabolism coefficient fcαh g1−n year−1 f  ka Activity coefficient ϵaαh(f0−fc)W∞n−1 year−1 i Mortality  αp Mortality level* ≈M(ηmW∞)1−n g1−n year–1 j Reproduction  Rmax Maximum recruitment* stock specific year–1  ϵr Recruitment efficiency* stock specific – Fishery performance  wF Mean size-at-entry into the fishery variable g  F Fishing mortality variable year−1 Resource  κ Carrying capacity magnitude* stock specific g−1−λ  λ Carrying capacity exponent −2−q+n – e  r0 Resource growth rate coefficient 4 g1−n year−1 f, k Symbol . Description . Value . Unit . Footnote(s) . Body size   W∞ Asymptotic size (weight)* stock specific g   wegg Egg weight 0.001 g a Consumption  n Metabolic exponent 3/4 – b  β Preferred predator–prey mass ratio 100 – c  σ Range of preferred prey size 1.3 – d  q Clearance rate exponent 0.8 – e  γ Clearance rate coefficient 6.57/κ g−q year−1 f  f0 Standard feeding level 0.6 – f  fc Critical feeding level 0.2 – f  h Maximum consumption* ≈3KW∞1/3/[α(f0−fc)] g1−n year−1 g Growth  α Assimilation efficiency 0.6 – f  ηm Size at maturation rel. to W∞ 0.25 – h  ϵa Fraction of energy for activity 0.8 – i  kr Standard metabolism coefficient fcαh g1−n year−1 f  ka Activity coefficient ϵaαh(f0−fc)W∞n−1 year−1 i Mortality  αp Mortality level* ≈M(ηmW∞)1−n g1−n year–1 j Reproduction  Rmax Maximum recruitment* stock specific year–1  ϵr Recruitment efficiency* stock specific – Fishery performance  wF Mean size-at-entry into the fishery variable g  F Fishing mortality variable year−1 Resource  κ Carrying capacity magnitude* stock specific g−1−λ  λ Carrying capacity exponent −2−q+n – e  r0 Resource growth rate coefficient 4 g1−n year−1 f, k Parameters marked with an asterisk are specific for each stock, and the relation to standard parameters (K, W∞ ⁠, and M) are provided. a Neuheimer et al. (2015). b West et al. (1997). c Jennings et al. (2002). d Andersen et al. (2017). e Andersen and Beyer (2006). f Hartvig et al. (2011). g Juvenile growth rate (g/year) for w≪W∞ from (M7) is ≈αh(f0−fc)wn ⁠. A von Bertalanffy growth equation gives the growth rate for w≪W∞ as 3KW∞−1/3w2/3 ⁠. Ignoring the small difference in exponents gives the approximation in the table. h Jensen (1996), Froese and Binohlan (2000), and He and Stewart (2001). i Andersen and Beyer (2015). j The adult mortality from (M8) is M=αp(ηmW∞)n−1 ⁠, from which αp follows as αp=M(ηmW∞)1−n ⁠. k Savage et al. (2004). Open in new tab Table 2. Model parameters. Symbol . Description . Value . Unit . Footnote(s) . Body size   W∞ Asymptotic size (weight)* stock specific g   wegg Egg weight 0.001 g a Consumption  n Metabolic exponent 3/4 – b  β Preferred predator–prey mass ratio 100 – c  σ Range of preferred prey size 1.3 – d  q Clearance rate exponent 0.8 – e  γ Clearance rate coefficient 6.57/κ g−q year−1 f  f0 Standard feeding level 0.6 – f  fc Critical feeding level 0.2 – f  h Maximum consumption* ≈3KW∞1/3/[α(f0−fc)] g1−n year−1 g Growth  α Assimilation efficiency 0.6 – f  ηm Size at maturation rel. to W∞ 0.25 – h  ϵa Fraction of energy for activity 0.8 – i  kr Standard metabolism coefficient fcαh g1−n year−1 f  ka Activity coefficient ϵaαh(f0−fc)W∞n−1 year−1 i Mortality  αp Mortality level* ≈M(ηmW∞)1−n g1−n year–1 j Reproduction  Rmax Maximum recruitment* stock specific year–1  ϵr Recruitment efficiency* stock specific – Fishery performance  wF Mean size-at-entry into the fishery variable g  F Fishing mortality variable year−1 Resource  κ Carrying capacity magnitude* stock specific g−1−λ  λ Carrying capacity exponent −2−q+n – e  r0 Resource growth rate coefficient 4 g1−n year−1 f, k Symbol . Description . Value . Unit . Footnote(s) . Body size   W∞ Asymptotic size (weight)* stock specific g   wegg Egg weight 0.001 g a Consumption  n Metabolic exponent 3/4 – b  β Preferred predator–prey mass ratio 100 – c  σ Range of preferred prey size 1.3 – d  q Clearance rate exponent 0.8 – e  γ Clearance rate coefficient 6.57/κ g−q year−1 f  f0 Standard feeding level 0.6 – f  fc Critical feeding level 0.2 – f  h Maximum consumption* ≈3KW∞1/3/[α(f0−fc)] g1−n year−1 g Growth  α Assimilation efficiency 0.6 – f  ηm Size at maturation rel. to W∞ 0.25 – h  ϵa Fraction of energy for activity 0.8 – i  kr Standard metabolism coefficient fcαh g1−n year−1 f  ka Activity coefficient ϵaαh(f0−fc)W∞n−1 year−1 i Mortality  αp Mortality level* ≈M(ηmW∞)1−n g1−n year–1 j Reproduction  Rmax Maximum recruitment* stock specific year–1  ϵr Recruitment efficiency* stock specific – Fishery performance  wF Mean size-at-entry into the fishery variable g  F Fishing mortality variable year−1 Resource  κ Carrying capacity magnitude* stock specific g−1−λ  λ Carrying capacity exponent −2−q+n – e  r0 Resource growth rate coefficient 4 g1−n year−1 f, k Parameters marked with an asterisk are specific for each stock, and the relation to standard parameters (K, W∞ ⁠, and M) are provided. a Neuheimer et al. (2015). b West et al. (1997). c Jennings et al. (2002). d Andersen et al. (2017). e Andersen and Beyer (2006). f Hartvig et al. (2011). g Juvenile growth rate (g/year) for w≪W∞ from (M7) is ≈αh(f0−fc)wn ⁠. A von Bertalanffy growth equation gives the growth rate for w≪W∞ as 3KW∞−1/3w2/3 ⁠. Ignoring the small difference in exponents gives the approximation in the table. h Jensen (1996), Froese and Binohlan (2000), and He and Stewart (2001). i Andersen and Beyer (2015). j The adult mortality from (M8) is M=αp(ηmW∞)n−1 ⁠, from which αp follows as αp=M(ηmW∞)1−n ⁠. k Savage et al. (2004). Open in new tab Growth, mortality, and demography We assume that individuals feed on a resource NR(wR) that represents food of all sizes in the ecosystem. Individuals prefer food a factor β = 100 smaller than themselves (Jennings et al., 2002). Multiplying an individual’s size-preference with the biomass of that resource size, and integrating over all resource sizes, gives the total amount of food available to the individual. When multiplied with clearance rate γwq ⁠, this then gives the food actually encountered by the individual Ee(w) (M2). Consumption is described by a functional response type II (M3), with maximum consumption hwn and n = 3/4, giving the feeding level f(w) as consumed food relative to maximum consumption. From consumption we calculate the energy available for somatic growth and reproduction from an energy budget (M4). Energy is assimilated from consumed food with efficiency α and costs of standard metabolism (⁠ krwn ⁠) and activity (⁠ kaw ⁠) are paid. The remaining available energy Ea(w) is divided between somatic growth and reproduction, with individual growth rate g(w): g(w)=[1−ψ(w)][αf(w)hwn−krwn−kaw](1) Here ψ(w) represents the fraction of available energy invested into reproduction. The remaining available energy, (1−ψ(w)) ⁠, is invested into growth. ψ(w) approaches 0 so long as individual size w remains well below size at 50% maturation ηmW∞ ⁠. The switch to maturity is described by a sigmoid function that smoothly varies between 0 and 1 around size at 50% maturation (M5). Mature individuals still invest energy in growth, but this investment decreases as their size approaches W∞ ⁠, until at size W∞ all energy is invested into egg production (M6). This procedure results in a von Bertalanffy-like weight-at-age curve if food is plentiful, f(w)=f0 ⁠, but reduces growth if the resource has become depleted, f(w)<f0 ⁠. The energy not used for growth is invested into egg production: ψ(w)Ea(w) ⁠. Total egg production of the stock Rp emerges by integrating egg production over all individual sizes, taking into account that only females produce eggs (M10). We assume that natural mortality rate μ0(w) is mainly due to predation by other species, and decreases with individual size according to μ0(w)=αpwn−1 (M8). We assume that mortality due to cannibalism is negligible. The fishing mortality rate μF(w) is the product of a level F and a size-specific gear selectivity (M9). We use a sigmoid function that smoothly switches from 0 to 1 around size-at-entry into the fishery wF to resemble a trawl selectivity curve. The density of individuals across all sizes within the population makes up the abundance size-spectrum N(w) as calculated by the McKendrick-von Foerster conservation equation: ∂N(w)∂t+∂g(w)N(w)∂w=−[μ0(w)+μF(w)]N(w)(2) Spawning stock biomass BSSB can be calculated from the abundance size-spectrum by integrating mature biomass over all sizes (M16). Yield from fishing can be calculated by multiplying stock biomass targeted by the fishing gear with fishing mortality, and integrating over all sizes (M17). Density dependence Density dependence emerges from two sources: a stock–recruitment relationship determines the strength of density dependence early in life, and competition for food from the resource spectrum determines the strength of density dependence late in life. The relative importance of the two processes is described by the ratio between the parameters that describe the carrying capacity of the early life environment and the late-life environment. Below we first describe both of these processes individually, and then explain how they interact. A standard Beverton-Holt stock–recruitment relationship (Beverton and Holt, 1957) is used to describe recruitment R: R=RmaxϵrRpRmax+ϵrRp(3) where Rmax is the maximum recruitment, Rp is the number of eggs produced by the spawning stock, and ϵr is the stock–specific recruitment efficiency which accounts for costs of reproduction and egg survival. The recruitment R is used as a boundary condition for the conservation Equation (2): g(wegg)N(wegg)=R ⁠. Which specific type of stock–recruitment relationship we use here is of lesser importance; the most important thing is that it describes the density dependence that takes place early in life. We have used a Beverton-Holt stock–recruitment relationship because it is both simple and well-known. We consider the recruitment efficiency ϵr as constant for each stock. The maximum recruitment Rmax represents the carrying capacity of the early life environment and we therefore use this parameter to determine the strength of early-in-life density dependence relative to the strength of late-in-life density dependence. The resource spectrum NR(wR) represents all individuals, of all sizes, that do not belong to the focal stock. The change in resource abundance is described with a semi-chemostat: dNR(wR)dt=r0wRn−1[κwRλ−NR(wR)]−μp(wR)NR(wR)(4) where r0wRn−1 is the size-specific resource regeneration rate and μp(wR) (M18) is the size-specific resource mortality due to predation by the focal stock. Food abundance is determined by the carrying capacity of the resource κwRλ ⁠. The value of the slope λ has been determined as −2−q+n (Andersen and Beyer, 2006), meaning that the resource carrying capacity follows a Sheldon spectrum (Sheldon et al., 1977), where biomass is approximately constant in logarithmically-spaced size groups. Food availability is therefore largely independent of size, with the overall level determined by κ. The value of κ then determines the resource availability, and thereby the level of density-dependent competition and growth. Intraspecific competition for resources emerges when the consumption of any given resource size exceeds the regeneration of that resource size, thereby reducing its abundance (Figure 1a, grey lines). In fish, cohort biomass usually increases until maturity (and fishing) sets in. As the biomass of the fish stock increases with size (Figure 1a, black lines), the highest competition will be for the resource sizes that are targeted by mature fish. Density-dependent competition for resources therefore mainly takes place late in life, amongst the mature and late juvenile portion of the stock. Figure 1. Open in new tabDownload slide Mechanisms of density dependence in the model, illustrated with three different Rmax to κ ratios: 0.01 (dotted), 1 (solid), and 100 (dashed) g1+λ/year. These describe scenarios of only early-in-life, a mix of early- and late-in-life, and only late-in-life density dependence, respectively. The thin dotted lines indicate size-at-maturity. Shown for a W∞ of 1000 g, and no fishing mortality. (a) Stock (black) and resource (grey) biomass as a function of size. Note that the dotted stock line intersects the y-axis outside of the plotted range. (b) Feeding levels (ratio between consumption and maximum consumption) associated with the different Rmax to κ ratios, indicating that resource competition peaks around maturation size. (c) Weights-at-age associated with the different Rmax to κ ratios, showing how different strengths of early- and late-in-life density dependence affect growth. In the model, density-dependent population regulation emerges from two sources: early-in-life stock–recruitment, as determined by Rmax, and density-dependent growth as determined by the resource carrying capacity κ. Their ratio, Rmax/κ ⁠, controls the relative importance of the two processes of density dependence: a low value of Rmax/κ leads to a dominance of early-in-life density-dependent recruitment, whereas a high value leads to a dominance of late-in-life density-dependent growth (Figure 1). Fitting the model to fish stocks To find realistic Rmax to κ ratios for the examined stocks, we fitted the model to empirical data of three fish stocks: North Sea plaice, NEA mackerel, and Baltic sprat. These stocks vary in asymptotic size, show little-to-no cannibalism, and all have shown indications of density-dependent growth beyond the juvenile stage. The dynamics of each stock depend on stock-specific physiological parameters describing: growth (h), asymptotic size (⁠ W∞ ⁠), recruitment (ϵr), and mortality (αp), and on parameters that influence density dependence: maximum recruitment (Rmax) and resource abundance (κ). The physiological parameters are determined from classical parameters, the von Bertalanffy growth coefficient K, and adult mortality M, with the procedure described in Andersen et al. (2009); see Table 2 for relations, and Table 3 for parameters for each stock. For plaice and sprat, W∞ was calculated from the stock’s observed length-at-maturity (Lm) (Supplementary Appendix B). For mackerel, W∞ was calculated from the L∞ that was associated with the used value for K. Values for K are taken from empirical studies, and values for M from ICES assessments. Recruitment efficiency (ϵr) was set so that the model’s FMSY matched the advised FMSY of the stock, having set the size at 50% fisheries selectivity, wF, according to fisheries data. A more detailed explanation of the parameterization process for each stock can be found in Supplementary Appendix B. Table 3. Stock-specific parameters that were used as input for the model, and the resulting SSB and yield predicted by the model. . . Baltic sprat . NEA mackerel . North Sea plaice . Parameters Asymptotic size W∞ (g) 21 890 1600 Von Bertalanffy growth constant K (year–1) 0.68 0.18 0.16 Recruitment efficiency ϵr (−) 0.0055 0.00060 0.10 Size-at-entry into fishery wF (g) 3.6 240 120 Maximum recruitment Rmax (year–1) 2.5·1013 4.5·1010 5.0·109 Resource carrying capacity coeff. κ (⁠ g−1−λ ⁠) 2.5·1012 1.5·1013 3.3·1012 Low-SSB scenario Natural mortality M (year–1) 0.50 0.15 0.10 Fishing mortality F (year–1) 0.23 0.46 0.60 High-SSB scenario Natural mortality M (year–1) 0.20 0.15 0.10 Fishing mortality F (year–1) 0.39 0.29 0.050 Results Low-SSB scenario SSB BSSB (Mt) 0.36 1.7 0.15 Annual yield Y (Mt/year) 0.094 0.77 0.28 High-SSB scenario SSB BSSB (Mt) 2.1 3.0 2.7 Annual yield Y (Mt/year) 1.0 0.86 0.18 . . Baltic sprat . NEA mackerel . North Sea plaice . Parameters Asymptotic size W∞ (g) 21 890 1600 Von Bertalanffy growth constant K (year–1) 0.68 0.18 0.16 Recruitment efficiency ϵr (−) 0.0055 0.00060 0.10 Size-at-entry into fishery wF (g) 3.6 240 120 Maximum recruitment Rmax (year–1) 2.5·1013 4.5·1010 5.0·109 Resource carrying capacity coeff. κ (⁠ g−1−λ ⁠) 2.5·1012 1.5·1013 3.3·1012 Low-SSB scenario Natural mortality M (year–1) 0.50 0.15 0.10 Fishing mortality F (year–1) 0.23 0.46 0.60 High-SSB scenario Natural mortality M (year–1) 0.20 0.15 0.10 Fishing mortality F (year–1) 0.39 0.29 0.050 Results Low-SSB scenario SSB BSSB (Mt) 0.36 1.7 0.15 Annual yield Y (Mt/year) 0.094 0.77 0.28 High-SSB scenario SSB BSSB (Mt) 2.1 3.0 2.7 Annual yield Y (Mt/year) 1.0 0.86 0.18 The input parameters include the Rmax and κ values that resulted in the best model fit for each stock. Sources for the parameter values are listed in Supplementary Appendix B. Open in new tab Table 3. Stock-specific parameters that were used as input for the model, and the resulting SSB and yield predicted by the model. . . Baltic sprat . NEA mackerel . North Sea plaice . Parameters Asymptotic size W∞ (g) 21 890 1600 Von Bertalanffy growth constant K (year–1) 0.68 0.18 0.16 Recruitment efficiency ϵr (−) 0.0055 0.00060 0.10 Size-at-entry into fishery wF (g) 3.6 240 120 Maximum recruitment Rmax (year–1) 2.5·1013 4.5·1010 5.0·109 Resource carrying capacity coeff. κ (⁠ g−1−λ ⁠) 2.5·1012 1.5·1013 3.3·1012 Low-SSB scenario Natural mortality M (year–1) 0.50 0.15 0.10 Fishing mortality F (year–1) 0.23 0.46 0.60 High-SSB scenario Natural mortality M (year–1) 0.20 0.15 0.10 Fishing mortality F (year–1) 0.39 0.29 0.050 Results Low-SSB scenario SSB BSSB (Mt) 0.36 1.7 0.15 Annual yield Y (Mt/year) 0.094 0.77 0.28 High-SSB scenario SSB BSSB (Mt) 2.1 3.0 2.7 Annual yield Y (Mt/year) 1.0 0.86 0.18 . . Baltic sprat . NEA mackerel . North Sea plaice . Parameters Asymptotic size W∞ (g) 21 890 1600 Von Bertalanffy growth constant K (year–1) 0.68 0.18 0.16 Recruitment efficiency ϵr (−) 0.0055 0.00060 0.10 Size-at-entry into fishery wF (g) 3.6 240 120 Maximum recruitment Rmax (year–1) 2.5·1013 4.5·1010 5.0·109 Resource carrying capacity coeff. κ (⁠ g−1−λ ⁠) 2.5·1012 1.5·1013 3.3·1012 Low-SSB scenario Natural mortality M (year–1) 0.50 0.15 0.10 Fishing mortality F (year–1) 0.23 0.46 0.60 High-SSB scenario Natural mortality M (year–1) 0.20 0.15 0.10 Fishing mortality F (year–1) 0.39 0.29 0.050 Results Low-SSB scenario SSB BSSB (Mt) 0.36 1.7 0.15 Annual yield Y (Mt/year) 0.094 0.77 0.28 High-SSB scenario SSB BSSB (Mt) 2.1 3.0 2.7 Annual yield Y (Mt/year) 1.0 0.86 0.18 The input parameters include the Rmax and κ values that resulted in the best model fit for each stock. Sources for the parameter values are listed in Supplementary Appendix B. Open in new tab After having parameterized the model with stock-specific parameters, realistic Rmax to κ ratios were determined for each stock. For this, the aim was to match simulated density-dependent changes in individual growth and SSB with observed changes in individual growth and SSB, while also matching modelled fishery yield with historical yield data. To observe density-dependent changes in growth, the model was fitted to two historical scenarios between which there were significant differences in both SSB and individual growth (one scenario with low SSB and fast individual growth, and a second scenario with high SSB and slow individual growth). For North Sea plaice, the two scenarios were before and at the end of the Second World War. No fishing during the war resulted in roughly a tripling of plaice SSB at the end of the war (Margetts and Holt, 1948), and coincided with a reduction in late-juvenile and adult growth (Rijnsdorp and Van Leeuwen, 1992). No actual SSB data is known from this time, with SSB changes instead having been inferred from changes in catch-per-unit-effort. To be able to fit our model to plaice data, we therefore assumed that plaice SSB and yield figures from the 1990s would have been similar to those of pre-WWII, as for both these times F was around 0.6 year−1 (Beverton and Holt, 1957; ICES, 2015b). The model was fitted to NEA mackerel using scenarios from 2003 and 2013. In 2003 NEA mackerel was heavily fished (F: 0.46 year−1, Y: 680 kt; ICES, 2015a), SSB was relatively low (1900 kt; ICES, 2015a) and individual growth was fast (Olafsdottir et al., 2016). In 2013 fishing mortality had been decreased to 0.29 year−1 (ICES, 2015a), though yield had increased to 930 kt/year (ICES, 2015a). At the same time, SSB almost doubled to around 3600 kt (ICES, 2015a), and individual growth had decreased (Olafsdottir et al., 2016). Last, the model was fitted to Baltic sprat using scenarios from 1988 and 1998. The main predator of Baltic sprat, Eastern Baltic cod (Gadus morhua), suffered a large decrease in abundance during the mid-1980s (Köster et al., 2003). The reduction in predators reduced Baltic sprat mortality, and after 1988 sprat SSB started to increase. Whereas Baltic sprat SSB was around 415 kt in 1988, SSB had more than tripled to around 1400 kt in 1998 (ICES, 2015c) with a concurrent decrease in late-juvenile and adult growth (Eero, 2012). Furthermore, whereas in 1988 Baltic sprat yield was around 80 kt/year with a fishing mortality of 0.23 year−1, in 1998 yield had increased to 417 kt/year with a fishing mortality of 0.39 year−1 (ICES, 2015c). Using the empirical data from the above scenarios, the model was fitted to each of the three stocks. A detailed description of this fitting procedure is given in Supplementary Appendix B. After fitting, the size-at-entry into the fishery wF which yielded MSY was determined by running the fitted model with a range of wF and F combinations. For each value of wF, this resulted in a different highest sustainable yield and a different value of F leading to that highest sustainable yield. The wF with the largest value for highest sustainable yield is the wF that yields MSY. Furthermore, a sensitivity analysis was performed of the fitted variables ϵr and Rmax/κ ⁠, by varying their values with a factor 2. Those values were subsequently used to recalculate FMSY and optimal fishery size-at-entry respectively. The sensitivity analysis and its results are presented in more detail in Supplementary Appendix C. Results Fitted parameters, including the Rmax and κ values, are shown in Table 3. The resulting weight-at-age curves for both the high- and low-SSB scenarios are shown in Figure 2. Figure 2. Open in new tabDownload slide Weight-at-age (a–c) and highest sustainable yield as a function of size-at-entry into the fishery (d–f), modelled for Baltic sprat (a, d), NEA mackerel (b, e), and North Sea plaice (c, f). Highest sustainable yield is calculated separately for each size-at-entry value and, for each scenario, is shown relative to its maximum value among all size-at-entry values (MSY). Size-at-entry into the fishery is shown relative to W∞ ⁠. Grey lines represent the stock’s low-SSB scenario, and black lines represent the stock’s high-SSB scenario (Table 3). These lines overlap in (e) and (f), because only fishing mortality changes between scenarios there. The solid lines show the model fit of each stock. The grey dotted lines show the hypothetical model fit of each stock if all density dependence would occur early in life (⁠ Rmax/κ=0.00001 g1+λ/year). They are only shown for each stock’s low-SSB scenario (Table 3), and act as a reference to that scenario’s fitted curve (solid). The black dashed lines show the hypothetical model fit of each stock if all density dependence would occur late in life (⁠ Rmax/κ=100 000 g1+λ/year). They are only shown for each stock’s high-SSB scenario (Table 3), and act as a reference to that scenario’s fitted curve (solid). The thin dotted lines show size-at-maturity. Historical weight-at-age data points are shown for the low-SSB (open points) and high-SSB (filled points) scenarios of sprat and mackerel. They are not shown for plaice, because Rijnsdorp and Van Leeuwen (1992) do not show changes in weight-at-age but in growth-increments of length groups. Supplementary Appendix B contains an overview of how the model fit for plaice overlaps with this data type. Baltic sprat The modelled growth for Baltic sprat approaches the empirical weight-at-age data points of the high- and low-SSB scenarios (Figure 2a). In the low-SSB scenario modelled growth is high, and closely follows the reference line for only early-in-life density dependence. In the high-SSB scenario modelled growth is reduced by strong late-in-life density-dependent growth, and closely follows the reference line for only late-in-life density dependence. Fishery size-at-entry for which MSY is obtained is close to asymptotic size in the low-SSB scenario (Figure 2d), and closely follows the reference line for only early-in-life density dependence. In the high-SSB scenario, fishery size-at-entry for which MSY is obtained is smaller, but still greater than size-at-maturity. Again, the fitted curve closely follows the reference line for only late-in-life density dependence. The sensitivity analysis shows that both weight-at-age and optimal fishery size-at-entry are relatively unaffected by changes in the Rmax to κ ratio (Supplementary Appendix C). This indicates that, for Baltic sprat, a change in natural mortality M has a far stronger impact on strength of density-dependent growth than a change in the Rmax to κ ratio. NEA mackerel The historical change in weight-at-age of NEA mackerel could not be replicated (Figure 2b). Changing F from 0.46 to 0.29 year−1 resulted in only a minor reduction in growth. However, the reference line for only late-in-life density dependence is close to the high-SSB scenario data points. According to these results, it is likely that the observed decrease in NEA mackerel weight-at-age is not solely the result of a decrease in fishing mortality. For the Rmax to κ ratio predicted for NEA mackerel, MSY exploitation occurs with a large size-at-entry into the fishery (Figure 2e). Furthermore, the reference line for only late-in-life density dependence also peaks at a large size-at-entry into the fishery. This suggests that even if NEA mackerel would experience strong late-in-life density-dependent growth, optimal fishery size-at-entry would still be large. North Sea plaice For North Sea plaice, the fitted model was able to replicate historical growth data (Figure 2c, Supplementary Appendix B). For a high fishing mortality, growth is fast and almost all density dependence takes place early in life. When fishing mortality drops to nearly zero, late-in-life density dependence becomes stronger due to increased SSB, and growth is decreased. The reference line for only late-in-life density dependence predicts a scenario of severely reduced growth: density-dependent growth would be so strong that an average individual would not be able to grow to 50% size-at-maturity. For the Rmax to κ ratio predicted for North Sea plaice, MSY exploitation occurs with a large size-at-entry into the fishery (Figure 2f). The sensitivity analysis (Supplementary Appendix C) shows that this would still be the case if the Rmax to κ ratio would be a factor 2 higher (stronger density-dependent growth). The dashed reference line shows that, when late-in-life density dependence is very strong, there is a wide range of fishery size-at-entries for which MSY is obtained: from very small to larger than size-at-maturity. This is because throughout this size-at-entry range, the stock remained in a state of severe growth reduction. In this state, the stock had almost no tolerance for fishing mortality, so the yield was very small and almost independent of fishery size-at-entry. Discussion For all three analysed stocks the model predicts that fishing at MSY occurs with a large fishery size-at-entry. The optimal fishery size-at-entry can decrease somewhat when the strength of late-in-life density-dependent growth is high, but for the fitted stocks it always remained above size-at-maturity. Therefore, for the examined stocks this study indicates that the current practice of setting size-at-entry such that predominantly adults are targeted is sound, in spite of the presence of strong late-in-life density-dependent growth. However, this does not mean that late-in-life density-dependent growth should be completely disregarded when calculating fisheries reference points. For other stocks, if late-in-life density dependence is very strong, the optimal size-at-entry could be smaller than size-at-maturity. Further, strong late-in-life density-dependent growth will influence a stock’s FMSY reference point. Fishing on a stock that experiences strong late-in-life density-dependent growth will increase individual growth rate by relieving the stock of density dependence, and will thereby increase stock productivity. If this is ignored when calculating fisheries reference points, it is likely that the calculated FMSY will be lower than the actual FMSY. This would cause the fishery to lose out on potential yield. Therefore, it is important to consider density-dependent growth when calculating fisheries reference points. Previous theoretical work has indicated that stocks with a larger asymptotic size should have a larger density-dependent buffer against population decline, or in other words, they should experience stronger density-dependent regulation (Andersen and Beyer, 2015). Consequently, the issue of density-dependent growth might be most important for stocks of large-bodied species. Our results for North Sea plaice (which in this study is the stock with the greatest asymptotic size) give some confirmation of this. The model predicts that North Sea plaice is at risk of “stunted growth” (Alm, 1946; Ylikarjula et al., 1999) when late-in-life density-dependent growth is very strong, with growth stopping before size-at-maturity. Cases of stunted growth are, however, rarely observed in marine fish populations, possibly due to the large spatial extent of the habitat for adults in marine systems (Andersen et al., 2017). Whether our model is correct in predicting that North Sea plaice could become subject to stunted growth is therefore not completely certain. Model limitations We were unable to replicate NEA mackerel’s observed reduction in growth by only changing fishing mortality. We therefore assume that the observed reduction in NEA mackerel individual growth (Olafsdottir et al., 2016) is not, or not solely, the result of a reduction in fishing mortality and a subsequent SSB increase. If the observed growth reduction did occur via intraspecific density dependence, some environmental change should then be the cause. A possibility would be increased sea surface temperatures in NEA waters, which have been thought to have extended NEA mackerel’s feeding range northwards to Svalbard (Berge et al., 2015), and to have shifted the egg production centre-of-gravity of NEA mackerel’s western spawning component northward (Hughes et al., 2014). However, it is also possible that the observed growth reduction of NEA mackerel individuals is rather due to interspecific competition instead of intraspecific competition, as suggested by Olafsdottir et al. (2016). They show that the increase in NEA mackerel SSB occurred simultaneously with an increase in SSB of its competitor: Norwegian spring-spawning herring. Thus, increased interspecific competition for resources could also have caused or contributed to the observed growth decrease in NEA mackerel. Our model assumes a homogeneously distributed resource spectrum with a carrying capacity that follows a Sheldon spectrum (Sheldon et al., 1977). As a result of this, late-in-life resource competition is automatically highest for individuals with a size that is near the size-spectrum’s biomass peak (Figure 1). In reality however, marine fish often move through different habitats and resources as they grow. These may differ in a multitude of aspects from each other, with each habitat or resource type being able to contribute to density-dependent effects. It is especially important to consider habitat size, as this can be a major factor in shaping density dependence (Casini et al., 2016), in particular if habitat size changes during ontogeny (Andersen et al., 2017). To truly consider density dependence taking place throughout life requires incorporating this heterogeneity into the resource spectrum. Since this heterogeneity will be highly stock-specific, that can only be properly done when sufficient knowledge is available about it. This is only rarely the case. In the absence of this knowledge, our model offers a simplified method for incorporating density dependence both early and late in life. We have assumed that decreased resource availability reduces the feeding level of the individual, and thereby its growth rate, without affecting size-specific mortality. A reduced growth rate does cause individuals to spend a longer time at a smaller size, where mortality rate is higher (Peterson and Wroblewski, 1984), which decreases their chances of survival. Nevertheless, this does not change the size-specific mortality rate. In an experiment on reef fish, Forrester (1990) shows that density-dependent growth can take place without an associated mortality increase. However, other studies show that decreased resource availability can increase mortality rate, resulting in density-dependent mortality as well as density-dependent growth. For instance, individuals can attempt to prevent their feeding level from decreasing too much by increasing their time spent searching for food (Wyatt, 1972; Walters and Juanes, 1993) or by taking greater risks during foraging (Damsgird and Dill, 1998). An increase in search rate or risk-taking puts the individual at a greater risk of predation (Walters and Juanes, 1993; Biro et al., 2003, 2004), leading to an increased mortality rate. Furthermore, many fish stocks experience a reduction in body condition due to an increased stock density (e.g. Winters and Wheeler, 1994; Schindler et al., 1997; Olafsdottir et al., 2016). A decline in body condition can increase mortality rate. It may for instance decrease an individual’s ability to avoid predation (Hoey and McCormick, 2004), or increase mortality risk after spawning (Lambert and Dutil, 2000). We have not incorporated such density-dependent mortality mechanisms in this study. If these above processes have influenced the data to which we have fitted our model, this could therefore influence the interpretation of our results. An additional presence of density-dependent mortality alongside the observed density-dependent growth would indicate that late-in-life density dependence was stronger than what we have found. Optimal size-at-entry for MSY exploitation would then likely be smaller than what we have found. Interspecific density dependence We have mainly focussed on intraspecific density dependence, providing a method for analysing density dependence in fish stocks from a single-stock management perspective. For a while now however, an increasing amount of fisheries research has been devoted to ecosystem-based management. When modelling density dependence from an ecosystem perspective it is important to incorporate that fish stocks do not only experience intraspecific density dependence, but also react to density changes of interspecific prey, competitors, and predators. The model type that we have used can be a useful tool for describing density dependence throughout life from an ecosystem perspective. We have already partly done so in this study, by linking Baltic sprat predation mortality to Eastern Baltic cod stock size. However, we did this in a simplified way, and not for NEA mackerel or North Sea plaice. Fully incorporating interspecific density dependence into the model will require the addition of dynamic prey, competitor, and predator stocks. Unfortunately, the interplay between interspecific and intraspecific density dependence is hard to extract from field observations, and therefore difficult to accurately model. Nevertheless, understanding both of these processes is important for making long-term stock predictions, especially from an ecosystem point-of-view. Conclusion It is unlikely that the stocks examined in this study experience late-in-life density-dependent growth strong enough to decrease optimal fishery size-at-entry to below size-at-maturity. However, this could still change on a case-by-case basis, especially now that increased sustainable exploitation is increasingly leading to stock recovery in the NEA. This will likely lead to more available data on late-in-life density-dependent growth, which may change this conclusion. Nevertheless, right now the practice of advising a large fishery size-at-entry seems to be valid for the examined stocks, even in the presence of strong late-in-life density-dependent growth. Supplementary data Supplementary material is available at the ICESJMS online version of the article. Acknowledgements This work was supported by the Centre for Ocean Life, a VKR Centre of Excellence supported by the Villum Foundation. This work has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 675997. References Alm G. 1946 . Reasons for the occurrence of stunted fish populations with special regard to the perch . Report from the Institute of Freshwater Research, Drottningholm , 25 : 1 – 146 . Google Scholar OpenURL Placeholder Text WorldCat Andersen K. H. , Beyer J. E. 2006 . Asymptotic size determines species abundance in the marine size spectrum . The American Naturalist , 168 : 54 – 61 . Google Scholar Crossref Search ADS PubMed WorldCat Andersen K. H. , Beyer J. E. 2015 . Size structure, not metabolic scaling rules, determines fisheries reference points . Fish and Fisheries , 16 : 1 – 22 . Google Scholar Crossref Search ADS WorldCat Andersen K. H. , Farnsworth K. D., Pedersen M., Gislason H., Beyer J. E. 2009 . How community ecology links natural mortality, growth, and production of fish populations . ICES Journal of Marine Science , 66 : 1978 – 1984 . Google Scholar Crossref Search ADS WorldCat Andersen K. H. , Jacobsen N. S., Farnsworth K. D. 2015 . The theoretical foundations for size spectrum models of fish communities . Canadian Journal of Fisheries and Aquatic Sciences , 73 : 575 – 588 . Google Scholar Crossref Search ADS WorldCat Andersen K. H. , Jacobsen N. S., Jansen T., Beyer J. E. 2017 . When in life does density dependence occur in fish populations? Fish and Fisheries , 18 : 656 – 667 . Google Scholar Crossref Search ADS WorldCat Berge J. , Heggland K., Lønne O. J., Cottier F., Hop H., Gabrielsen G. W., Nøttestad L. et al. 2015 . First records of Atlantic mackerel (Scomber scombrus) from the Svalvard Archipelago, Norway, with possible explanations for the extension of its distribution . Arctic , 68 : 54 – 61 . Google Scholar Crossref Search ADS WorldCat Beverton R. J. H. , Holt S. J. 1957 . On the Dynamics of Exploited Fish Populations. Her Majesty’s Stationery Office, London, 540 pp. Biro P. A. , Abrahams M. V., Post J. R., Parkinson E. A. 2004 . Predators select against high growth rates and risk-taking behaviour in domestic trout populations . Proceedings of the Royal Society of London B: Biological Sciences , 271 : 2233 – 2237 . Google Scholar Crossref Search ADS WorldCat Biro P. A. , Post J. R., Parkinson E. A. 2003 . Density-dependent mortality is mediated by foraging activity for prey fish in whole-lake experiments . Journal of Animal Ecology , 72 : 546 – 555 . Google Scholar Crossref Search ADS PubMed WorldCat Casini M. , Käll F., Hansson M., Plikshs M., Baranova T., Karlsson O., Lundström K., et al. 2016 . Hypoxic areas, density-dependence and food limitation drive the body condition of a heavily exploited marine fish predator . Royal Society Open Science , 3 : 160416 . Google Scholar Crossref Search ADS PubMed WorldCat Cormon X. , Ernande B., Kempf A., Vermard Y., Marchal P. 2016 . North Sea saithe Pollachius virens growth in relation to food availability, density dependence and temperature . Marine Ecology Progress Series , 542 : 141 – 151 . Google Scholar Crossref Search ADS WorldCat Damsgird B. , Dill L. M. 1998 . Risk-taking behavior in weight-compensating coho salmon, Oncorhynchus kisutch . Behavioral Ecology , 9 : 26 – 32 . Google Scholar Crossref Search ADS WorldCat Eero M. 2012 . Reconstructing the population dynamics of sprat (Sprattus sprattus balticus) in the Baltic Sea in the 20th century . ICES Journal of Marine Science , 69 : 1010 – 1018 . Google Scholar Crossref Search ADS WorldCat Fernandes P. G. , Cook R. M. 2013 . Reversal of fish stock decline in the Northeast Atlantic . Current Biology , 23 : 1432 – 1437 . Google Scholar Crossref Search ADS PubMed WorldCat Forrester G. E. 1990 . Factors influencing the juvenile demography of a coral reef fish . Ecology , 71 : 1666 – 1681 . Google Scholar Crossref Search ADS WorldCat Froese R. , Binohlan C. 2000 . Empirical relationships to estimate asymptotic length, length at first maturity and length at maximum yield per recruit in fishes, with a simple method to evaluate length frequency data . Journal of Fish Biology , 56 : 758 – 773 . Google Scholar Crossref Search ADS WorldCat Hartvig M. , Andersen K. H., Beyer J. E. 2011 . Food web framework for size-structured populations . Journal of Theoretical Biology , 272 : 113 – 122 . Google Scholar Crossref Search ADS PubMed WorldCat Hassell M. P. 1975 . Density-dependence in single-species populations . Journal of Animal Ecology , 44 : 283 – 295 . Google Scholar Crossref Search ADS WorldCat He J. X. , Stewart D. J. 2001 . Age and size at first reproduction of fishes: predictive models based only on growth trajectories . Ecology , 82 : 784 – 791 . Google Scholar Crossref Search ADS WorldCat Hoey A. S. , McCormick M. I. 2004 . Selective predation for low body condition at the larval-juvenile transition of a coral reef fish . Oecologia , 139 : 23 – 29 . Google Scholar Crossref Search ADS PubMed WorldCat Holling C. S. 1959 . The components of predation as revealed by a study of small-mammal predation of the European Pine Sawfly . The Canadian Entomologist , 91 : 293 – 320 . Google Scholar Crossref Search ADS WorldCat Hughes K. M. , Dransfeld L., Johnson M. P. 2014 . Changes in the spatial distribution of spawning activity by north-east Atlantic mackerel in warming seas: 1977-2010 . Marine Biology , 161 : 2563 – 2576 . Google Scholar Crossref Search ADS WorldCat ICES . 2015a . Mackerel (Scomber scombrus) in Subareas I-VII and XIV and Divisions VIIIa-e and IXa (Northeast Atlantic). In ICES Advice 2015, Book 9. ICES . 2015b . Report of the Working Group on Widely Distributed Stocks (WGWIDE). ICES CM 2015/ACOM: 15. ICES . 2015c . Sprat (Sprattus sprattus) in Subdivisions 22-32 (Baltic Sea). In ICES Advice 2015, Book 8. Jennings S. , Oliveira J. A. A. D. E., Warr K. J. 2007 . Measurement of body size and abundance in tests of macroecological and food web theory . Journal of Animal Ecology , 76 : 72 – 82 . Google Scholar Crossref Search ADS PubMed WorldCat Jennings S. , Warr K. J., Mackinson S. 2002 . Use of size-based production and stable isotope analyses to predict trophic transfer efficiencies and predator-prey body mass ratios in food webs . Marine Ecology Progress Series , 240 : 11 – 20 . Google Scholar Crossref Search ADS WorldCat Jensen A. L. 1996 . Beverton and Holt life history invariants result from optimal trade-off of reproduction and survival . Canadian Journal of Fisheries and Aquatic Sciences , 53 : 820 – 822 . Google Scholar Crossref Search ADS WorldCat Köster F. W. , Möllmann C., Neuenfeldt S., Vinther M., St John M. A., Tomkiewicz J., Voss R. et al. 2003 . Fish stock development in the central Baltic Sea (1974-1999) in relation to variability in the environment . ICES Marine Science Symposia , 219 : 294 – 306 . Google Scholar OpenURL Placeholder Text WorldCat Lambert Y. , Dutil J.-D. 2000 . Energetic consequences of reproduction in Atlantic cod (Gadus morhua) in relation to spawning level of somatic energy reserves . Canadian Journal of Fisheries and Aquatic Sciences , 57 : 815 – 825 . Google Scholar Crossref Search ADS WorldCat Lorenzen K. 2005 . Population dynamics and potential of fisheries stock enhancement: practical theory for assessment and policy analysis . Philosophical Transactions of the Royal Society of London B: Biological Sciences , 360 : 171 – 189 . Google Scholar Crossref Search ADS WorldCat Lorenzen K. , Enberg K. 2002 . Density-dependent growth as a key mechanism in the regulation of fish populations: evidence from among-population comparisons . Proceedings of the Royal Society of London B: Biological Sciences , 269 : 49 – 54 . Google Scholar Crossref Search ADS WorldCat Margetts A. R. , Holt S. J. 1948 . The effect of the 1939-1945 war on the English North Sea trawl fisheries . Rapports et Procès-Verbaux des Réunions: Conseil Permanent International pour l’exploration de la mer , 122 : 26 – 46 . Google Scholar OpenURL Placeholder Text WorldCat Munch S. B. , Snover M. L., Watters G. M., Mangel M. 2005 . A unified treatment of top-down and bottom-up control of reproduction in populations . Ecology Letters , 8 : 691 – 695 . Google Scholar Crossref Search ADS WorldCat Myers R. A. , Cadigan N. G. 1993 . Density-dependent juvenile mortality in marine demersal fish . Canadian Journal of Fisheries and Aquatic Sciences , 50 : 1576 – 1590 . Google Scholar Crossref Search ADS WorldCat Neuheimer A. B. , Hartvig M., Heuschele J., Hylander S., Kiørboe T., Olsson K. H., Sainmont J. et al. 2015 . Adult and offspring size in the ocean over 17 orders of magnitude follows two life history strategies . Ecology , 96 : 3303 – 3311 . Google Scholar Crossref Search ADS PubMed WorldCat Olafsdottir A. H. , Slotte A., Jacobsen J. A., Oskarsson G. J., Utne K. R., Nøttestad L. 2016 . Changes in weight-at-length and size-at-age of mature Northeast Atlantic mackerel (Scomber scombrus) from 1984 to 2013: effects of mackerel stock size and herring (Clupea harengus) stock size . ICES Journal of Marine Science , 73 : 1255 – 1265 . Google Scholar Crossref Search ADS WorldCat Peterson I. , Wroblewski J. S. 1984 . Mortality rate of fishes in the pelagic ecosystem . Canadian Journal of Fisheries and Aquatic Sciences , 41 : 1117 – 1120 . Google Scholar Crossref Search ADS WorldCat Reichard M. , Jurajda P., Smith C. 2004 . Male-male interference competition decreases spawning rate in the European bitterling (Rhodeus sericeus) . Behavioral Ecology and Socio-Biology , 56 : 34 – 41 . Google Scholar Crossref Search ADS WorldCat Ricker W. E. 1954 . Stock and Recruitment . Journal of the Fisheries Research Board of Canada , 11 : 559 – 623 . Google Scholar Crossref Search ADS WorldCat Rijnsdorp A. D. , Van Leeuwen P. I. 1992 . Density-dependent and independent changes in somatic growth of female North Sea plaice Pleuronectes platessa between 1930 and 1985 as revealed by back-calculation of otoliths . Marine Ecology Progress Series , 88 : 19 – 32 . Google Scholar Crossref Search ADS WorldCat Savage V. M. , Gillooly J. F., Brown J. H., West G. B., Charnov E. L. 2004 . Effects of body size and temperature on population growth . The American Naturalist , 163 : 429 – 441 . Google Scholar Crossref Search ADS PubMed WorldCat Schindler D. E. , Hodgson J. R., Kitchell J. F. 1997 . Density-dependent changes in individual foraging specialization of largemouth bass . Oecologia , 110 : 592 – 600 . Google Scholar Crossref Search ADS PubMed WorldCat Sheldon R. W. , Sutcliffe W. H. Jr, Paranjape M. A. 1977 . Structure of pelagic food chain and relationship between plankton and fish production . Journal of the Fisheries Board of Canada , 34 : 2344 – 2353 . Google Scholar Crossref Search ADS WorldCat Van der Veer H. W. 1986 . Immigration, settlement, and density-dependent mortality of a larval and early postlarval 0-group plaice (Pleuronectes platessa) population in the western Wadden Sea . Marine Ecology Progress Series , 29 : 223 – 236 . Google Scholar Crossref Search ADS WorldCat Walters C. J. , Juanes F. 1993 . Recruitment limitation as a consequence of natural selection for use of restricted feeding habitats and predation risk taking by juvenile fishes . Canadian Journal of Fisheries and Aquatic Sciences , 50 : 2058 – 2070 . Google Scholar Crossref Search ADS WorldCat West G. B. , Brown J. H., Enquist B. J. 1997 . A general model for the origin of allometric scaling laws in biology . Science , 276 : 122 – 126 . Google Scholar Crossref Search ADS PubMed WorldCat Winters G. H. , Wheeler J. P. 1994 . Length-specific weight as a measure of growth success of adult Atlantic herring (Clupea harengus) . Canadian Journal of Fisheries and Aquatic Sciences , 51 : 1169 – 1179 . Google Scholar Crossref Search ADS WorldCat Wyatt T. 1972 . Some effects of food density on the growth and behaviour of plaice larvae . Marine Biology , 14 : 210 – 216 . Google Scholar OpenURL Placeholder Text WorldCat Ylikarjula J. , Heino M., Dieckmann U. 1999 . Ecology and adaptation of stunted growth in fish . Evolutionary Ecology , 13 : 433 – 453 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2018. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © International Council for the Exploration of the Sea 2018.

Bauer, Barbara; Meier, H E Markus; Casini, Michele; Hoff, Ayoe; Margoński, Piotr; Orio, Alessandro; Saraiva, Sofia; Steenbeek, Jeroen; Tomczak, Maciej T

doi: 10.1093/icesjms/fsy003pmid: N/A

Abstract In this study we investigate if eutrophication management has the potential to substantially affect which areas are going to be most suitable for commercial fishing in the future. We use a spatial ecosystem model, forced by a coupled physical-biogeochemical model, to simulate the spatial distribution of functional groups within a marine ecosystem, which depends on their respective tolerances to abiotic factors, trophic interactions, and fishing. We simulate the future long-term spatial developments of the community composition and their potential implications for fisheries under three different nutrient management scenarios and changing climate. The three nutrient management scenarios result in contrasting developments of bottom oxygen concentrations and phytoplankton abundance, with substantial effects on fish production. Nutrient load reduction increases the spatial extent of the areas suitable for the commercially most valuable demersal fish predator and all types of fisheries. This suggests that strategic planning of fishery management strategies could benefit from considering future changes in species distributions due to changes in eutrophication. We show that combining approaches from climate research, physical oceanography, biogeochemistry, biogeography, and trophic ecology with economical information provides a strong foundation to produce scientific knowledge that can support a multisectoral management of ecosystems. Introduction Eutrophication-induced habitat degradation directly affects demersal and demerso-pelagic fish, and may affect the fisheries exploiting them as well (Stortini et al., 2017; Townhill et al., 2017). Such fish commonly function as key predators in aquatic ecosystems. Therefore, changes in their spatial distribution as a result of management actions modifying underwater habitat quality can have large effects on the spatial distribution of their prey and the whole community. These predators and their prey are in some cases targeted by different segments of the fishery. Thus, eutrophication reduction actions may actually have different effects across the fisheries sectors. To predict such effects it is important to reliably estimate species distribution changes, which necessitates to consider not only direct effects of changes in habitat quality on commercial fish but also their biotic interactions (Godsoe et al., 2017). Here, we use a modelling framework to investigate the causal chain between nutrient load management and the spatial distribution of fishing efforts: changing abiotic conditions affecting species distributions and fish production across space, and the latter influencing relative suitability of fishing grounds. The framework consists of a regional climate model, a coupled physical-biogeochemical model and an ecosystem model incorporating economic information, parametrized to describe the central Baltic Sea ecosystem. Eutrophication is one of the main pressures on the Baltic Sea ecosystem and the extent of hypoxic areas increased 10-fold during the past 100 years (Carstensen et al., 2014). Increased nutrient loading is proposed to have increased production of forage fish (Eero et al., 2016), but reduced the suitable habitat of the eastern Baltic cod (Gadus morhua) (Casini et al., 2016) causing a mismatch in the spatial overlap of cod and its main forage fish species, which might be one of the reasons of the failed recovery of this cod stock from overfishing (Eero et al., 2012). Even though there is a number of models focusing on different aspects of the Baltic Sea ecosystem, there is a lack of process-based understanding of the spatial effects of environmental drivers on the whole food web. Previous studies on species distributions and pressures in the Baltic Sea (Gogina and Zettler, 2010; Casini et al., 2011, 2014; Voss et al., 2012; Uusitalo et al., 2016; Bartolino et al., 2017) ignore dynamic feedbacks among ecosystem components. Similarly, spatial process-based models of eastern Baltic cod stock and fisheries (Röckmann et al., 2007, 2008; Kraus et al., 2008; Bastardie et al., 2010a, b,, 2017) have not taken trophic interactions into account so far. Radtke et al. (2013) model spatial distributions of Baltic fish based on plankton food availability, omitting direct effects of environmental drivers on fish and the benthic part of the food web. Models looking at combined effects of environmental drivers and fisheries while representing food web interactions (Hansson et al., 2007; Österblom et al., 2007; Niiranen et al., 2013) lack a spatial component, with the exception of the model developed by Lindegren et al. (2014), which modelled the central Baltic Sea as three interlinked sub-basins. Previous studies generally showed a link between high nutrient loads, pronounced eutrophication and an increase of sprat abundance, whereas low nutrient loads are generally thought to lead to decreased eutrophication and an increase in cod abundance. However, it is an open question how these effects are going to be realized in space and if there are areas within the Baltic Sea that are going to especially benefit from the positive effects of reduced eutrophication. To answer this question, we use a modelling approach that goes beyond previous studies by incorporating both information on abiotic drivers of species distributions, trophic interactions, and fisheries effects on the food web in space. Ecospace is the spatial-temporal module of the commonly used Ecopath with Ecosim (EwE) suite of models (Walters et al., 1999; Pauly et al., 2000). The newest addition to Ecospace, the habitat capacity model, combines the strength of Species Distribution Models (Peterson et al., 2011) with dynamics approaches by incorporating a dynamic niche model that considers the responses of functional groups to any number of (changing) environmental conditions (Christensen et al., 2014). In the present study we use the habitat capacity model of Ecospace to identify potential shifts in distributions of functional groups as a result of changing environmental conditions under three different nutrient management scenarios and changing climate. In addition, we are going to investigate to what extent the suitability of different areas for fishing may change under these scenarios. Material and methods Study system The area represented in our model is the central Baltic Sea, a large brackish water body in northern Europe. Weather-driven inflows from the North Sea and anthropogenic nutrient loads from land determine oxygen concentrations (Meier et al., 2006; Matthäus et al., 2008). During the last decades, hypoxic conditions on the sea bottom have become more widespread (Figure 1), with adverse effects on the reproductive potential and stock production of demersal spawning fish and on benthic macroinvertebrates (Karlson et al., 2002; Meier et al., 2012a; Carstensen et al., 2014; Casini et al., 2016). Figure 1. Open in new tabDownload slide Study area. Shades show mean depth of the spatial cells used in the Ecopath with Ecosim model (resolution: 0.25 × 0.25 degrees). Red thick lines show the extent of hypoxic areas (<2 ml/l bottom O2 concentrations) according to the RCO-SCOBI coupled physical-biogeochemical model outputs (average values 2004–2008, reference scenario, see Methods for details). Figure 1. Open in new tabDownload slide Study area. Shades show mean depth of the spatial cells used in the Ecopath with Ecosim model (resolution: 0.25 × 0.25 degrees). Red thick lines show the extent of hypoxic areas (<2 ml/l bottom O2 concentrations) according to the RCO-SCOBI coupled physical-biogeochemical model outputs (average values 2004–2008, reference scenario, see Methods for details). The offshore central Baltic Sea contains a highly productive but low diversity ecosystem with three main commercially important fish stocks, the Eastern Baltic cod, and two clupeid stocks, sprat (Sprattus sprattus) and central Baltic herring (Clupea harengus) (ICES, 2016a). Flounder (Platichthys flesus) is also a relatively abundant species and caught commercially as well. Even though the grey seal (Halychoerus grypus) population is steadily increasing, the number of grey seals is still low (Härkönen et al., 2013), thus, cod is the main piscivore. Cod, flounder, and to some extent herring, consume benthic preys while herring and sprat are the main planktivores. Mysids (mainly Mysis mixta, M. relicta, and Neomysis integer) consume both phyto-and zooplankton as well as benthic material, thus, they provide an important trophic link between the benthic and pelagic parts of the food web. Regional ocean climate model We use scenario simulation results of the regional ocean climate model RCO-SCOBI which consists of the physical Rossby Centre Ocean (RCO) model (Meier et al., 2003) and the Swedish Coastal and Ocean Biogeochemical (SCOBI) model (Eilola et al., 2009) performed within the project ECOSUPPORT 2009–2011 (Advanced modeling tool for scenarios of the Baltic Sea ECOsystem to SUPPORT decision making, see Meier et al., 2014). The ocean model is coupled to a Hibler-type sea ice model and the subgrid-scale mixing in the ocean is parametrized using a k-ɛ turbulence closure scheme with flux boundary conditions (Meier et al., 2003). A flux-corrected, monotonicity-preserving transport scheme is embedded without explicit horizontal diffusion. In the northern Kattegat open lateral boundary conditions are used, where in case of inflow temperature, salinity, and nutrient values are nudged toward observed climatological profiles. Horizontal and vertical resolutions amount to 3.7 km and 3 m, respectively. SCOBI describes the dynamics of nitrate, ammonium, phosphate, oxygen, and hydrogen sulphide concentrations (the latter as negative oxygen), three phytoplankton species, zooplankton and detritus (Eilola et al., 2009). The sediment contains nutrients in the form of benthic nitrogen and benthic phosphorus. Processes like assimilation, remineralization, nitrogen fixation, nitrification, denitrification, grazing, mortality, excretion, sedimentation, resuspension, and burial are considered. Resuspension of organic matter is calculated using a simplified wave model (Almroth-Rosell et al., 2011). Downscaling of projected climate change Atmospheric forcing fields of RCO-SCOBI were calculated applying a dynamical downscaling approach using a regional coupled atmosphere-ice-ocean model (Meier et al., 2012b) with lateral boundary data from a global climate model HadCM3 (Gordon et al., 2000). For the projections 2001–2098 the greenhouse gas emission scenario A1B was selected (Nakićenović et al., 2000). Bias correction of atmospheric forcing data for the ocean model was not applied, except that wind speed extremes were improved using simulated gustiness (Meier et al., 2011). River runoff was calculated from the net water budget over land (precipitation minus evaporation) using a statistical model (Meier et al., 2012b). Finally, nutrient loads were derived from the product of river flow and riverine nutrient concentrations. For details of the modeling approach and climate model results, the reader is referred to Meier et al. (2012b, c). Food web model We constructed a food web model describing the environmental drivers of the functional groups and their trophic interactions in the offshore central Baltic Sea using the EwE food web modelling approach (Walters et al., 1997; Christensen and Walters, 2004). The first component of the suite, Ecopath, describes the average trophic flows in an ecosystem during one year in our case. The Ecosim model is a set of differential equations describing the temporal behaviour of the ecosystem, using the Ecopath model as initial condition. More details on the EwE approach are included in the Supplementary material. The capabilities and limitations of the approach have been described by Christensen and Walters (2004), Plagányi and Butterworth (2004), and Plagányi (2007). Ecospace is the spatially explicit component of EwE (Pauly et al., 2000; Christensen et al., 2014; Romagnoni et al., 2015). Ecospace is represented by a set of water and land grid cells. Functional groups and fisheries interact with each other within the water cells according to modified versions of Ecosim equations (see Supplementary Appendix S3). The representation of life histories in Ecospace compared to Ecosim is modified (Walters et al., 2010) and an effect of habitat capacity on predator-prey interactions is introduced. Low habitat capacity for a consumer species is modelled as decreased vulnerability of its prey to predation (Christensen et al., 2014). Habitat capacity in a cell for a functional group depends on the values of environmental drivers in the cell and the group’s response function to these (Supplementary Appendix S3.1). To initialize Ecospace simulations, biomasses of functional groups are distributed based on their respective overall relative habitat capacity values. These biomass distributions change in the following time-steps due to food web interactions. These biomass distributions change in the following time-steps due to the interplay of food web interactions, fishing, and species dispersal until Ecospace reaches spatial equilibrium. Therefore it is necessary to have a spin-up period under stable conditions in Ecospace, before introducing spatio-temporal forcing. Spatial migration among cells is represented by redistributing the functional groups’ biomasses among cells with a speed depending on their basal migration rate. Overall relative habitat capacity is inversely proportional to the rate of migration out of grid cells, as organisms are assumed to be less likely to leave habitats with higher capacity and more likely to migrate out of habitats with lower capacity (Christensen et al., 2014). Fishing efforts of fleets are distributed among cells based on the attractiveness of each cell for the fleet (eq. 7, Supplementary Appendix S3). Fishing mortality caused by each fleet on its target species in each cell is proportional to its fishing effort in that cell. Model parameterization and calibration Our Ecopath model describes annual trophic flows in the Baltic Proper during the early 2000s between 21 functional groups (composed of developmental stages of fish groups, species or groups of species) and detritus (Figure 2). Consistency of Ecopath input parameters with basic ecological principles was checked using the Prebal procedure (Link, 2010), described in detail in ICES (2016b, Annex 3). Figure 2. Open in new tabDownload slide Trophic diagram of the Baltic Proper food-web, boxes representing modelled functional groups and edges main predator-prey relationships (based on Tomczak et al., 2012). For more details on the definition of functional groups see Supplementary Appendix S1. Figure 2. Open in new tabDownload slide Trophic diagram of the Baltic Proper food-web, boxes representing modelled functional groups and edges main predator-prey relationships (based on Tomczak et al., 2012). For more details on the definition of functional groups see Supplementary Appendix S1. The Ecopath model includes the effects of fisheries on the food web by defining 10 fishing fleets operating in the region and the fishing mortality caused by them, calculated based on their landings and discards (Figure 3). We implemented three types of fleets in the model: (1) active demersal (ACT; mostly otter trawls and demersal seine) in three size categories: <18 m, 18–24 m, 24–40 m; (2) passive demersal (PAS; gillnets, trammel nets, longlines, and pots) in three size categories: <12 m, 12–18 m, 18-40 m; and (3) pelagic (PEL; pelagic trawl and pelagic seine) in four size categories: <18 m, 18–24 m, 24–40 m, >40 m. To parameterize the fisheries we used data made available by the European Commission’s Joint Research Centre fisheries data collection website (https://datacollection.jrc.ec.europa.eu/, accessed 15 September 2016), evaluated by the European Commission’s Scientific, Technical and Economic Committee (STECF), and from ICES (2015, 2016a). Ecopath model parameters are included in Supplementary Appendix S1. Biomass of fish groups, landings and discards by the fishing fleets are representative of the year 2004. For other parameters, data from a period as close as possible to this year was used. Figure 3. Open in new tabDownload slide Landings compositions of the fleets used in Ecopath, based on data from 2004. ACT: active demersal, PAS: passive demersal, PEL: pelagic. The first and last two numbers in the fleet code indicate the lower and upper limits of the included vessel size, respectively (“40OO” indicating vessels >40 m). Figure 3. Open in new tabDownload slide Landings compositions of the fleets used in Ecopath, based on data from 2004. ACT: active demersal, PAS: passive demersal, PEL: pelagic. The first and last two numbers in the fleet code indicate the lower and upper limits of the included vessel size, respectively (“40OO” indicating vessels >40 m). For this study, the Ecosim model described in ICES (2016b, Annex 3) was refitted to a number of reference time series using environmental forcing functions derived from RCO-SCOBI outputs, corresponding to the time period 2004–2013 (please see Supplementary Appendix S2.2 for details of the fitting procedure). The period 2004–2013 was chosen for fitting as 2004 was the first year when fishing effort (kW days at sea) data from STECF became available and 2013 the last year when an analytical stock assessment for the Eastern Baltic cod was performed (at the time of this study). Both types of information were used during the model fitting procedure. The procedure was the same as described in ICES (2016b). During the model fitting process, first we assessed the sensitivity of the sum of squared deviations measure (SS) to the number of “vulnerability blocks” (v–s) fitted using the “Stepwise fitting” plug-in of Ecosim (Christensen et al., 2008). Second, we set the v values to those maximizing model fit to time-series using the “Fit to time series” plug-in (Supplementary Appendix S2.1). As suggested by Heymans et al. (2016), we did not simply use the v-s resulting in the best fit to observed time series data, but applied additional tests on stock-recruitment and fishing mortality-catch relationships (Heymans et al., 2016; Stäbler et al., 2016) and model stability (Mackinson and Daskalov, 2007) to test for ecologically credible model behaviour and modified a few v-s accordingly (Supplementary Appendix S2.1). To set up the Ecospace model, driver maps were generated for each environmental driver (Supplementary Table S5). All environmental driver maps we used are derived from the outputs of the RCO-SCOBI model, with the exception of the depth map. The latter is based on the Depth Relief Map published by the HELCOM Map and Data Service (www.helcom.fi). We use yearly average phytoplankton biomass as relative primary production map, similarly to Coll et al., (2016). To parametrize environmental response functions (ERF) in the Ecospace model, we collected information about the responses of functional groups and species biomasses to abiotic factors from the species distribution modelling literature (Supplementary Table S5). We assumed three types of ERFs, “left-shoulder” (Supplementary Figure S8a), “trapezoid” (Supplementary Figure S8b), and “right-shoulder” (Supplementary Figure S8c) shapes. The choice of shape for a particular group-environmental driver pair does not reflect some general ecological characteristic of that group, rather it shows whether the environmental driver in the Baltic Sea have been described to encompass the entirety of the groups’ preferred range and values above and below that (trapezoid shape) or whether the group is only possibly limited by that driver because of too high (left-shoulder) or too low (right-shoulder) values in that ecosystem. In contrast to the Ecosim module, which we fit to non-spatial time series data, we assessed a fit of the Ecospace output to spatially explicit but temporally static empirical data (maps). There is no automated fitting procedure available for Ecospace. In the lack of temporal forcing, our model describes ecosystem behavior approximately of the year 2004. However, to make the model validation less sensitive to potential noise in the data and inherent natural variability in the system, we compared averaged observed stock biomass, catch and fleet effort distributions from the period 2004–2008 to model outputs. The Ecospace model validation process is described in more detail in Supplementary Appendix S3.2. Sensitivity analysis We tested the sensitivity of our biomass simulations to key ecological assumptions. First, we iteratively tested how excluding ERFs from the model influenced the correlation with data. This way we could identify those ERFs that were crucial to reproduce key patterns in observational data (Supplementary Appendix S3.3.1). We also investigated the sensitivity of model fit to two parameters related to fisheries (Supplementary Appendix S3.3.2): port placement, which influences spatial distribution of fleets via the fishing cost map (Supplementary Figure S9), and Effective Power (⁠ 1/σ in eq. 7, Supplementary Appendix S3). We reran the model using the same settings as for the validation run, with five variations of randomly placed ports and with values for Effective Power = 0.5, 1, 5 and 10. Scenario simulations First, we simulated three scenarios driven by differing nutrient loads using the RCO-SCOBI model. We then used environmental driver maps and temporal forcing derived from that model to drive distributions of functional groups, and, consequently, fishing efforts in Ecospace. The three scenarios of nutrient concentrations were selected to reflect rather contrasting socio-economic developments in the Baltic Sea catchment area (Meier et al., 2012b): (1) land nutrient loads reduced according to the Helsinki Commission’s Baltic Sea Action Plan (BSAP, see HELCOM, 2007) and 50% reduced atmospheric deposition; (2) Reference (REF) with current nutrient concentrations in rivers and atmospheric deposition (Eilola et al., 2009); and (3) Business-As-Usual (BAU) with an assumed exponential growth in agriculture and current atmospheric deposition. Model runs of RCO-SCOBI representing the present climate period 1961–2007 used average riverine nutrient concentrations that were calculated from observed loads. Then simulations 2008–2098 were run under the three above-mentioned scenarios based upon nutrient concentration changes relative to the period 1995–2002 (for details of the applied ramp function, see Meier et al., 2012b). Ecospace simulations were run over the period 2004–2098 after a spin-up period of 75 time steps (years) under static conditions corresponding to those of 2004. We used annually averaged maps in EwE as drivers as we focus on the effects of long-term changes in environmental conditions and not on the seasonal cycle or extreme events like salt water inflows. The driver maps were inserted into the running Ecospace model through the spatial-temporal data framework (Steenbeek et al, 2013). We considered the same warming climate and increasing seal population in all scenarios, to be able to compare eutrophication effects in a realistic environmental context. We kept the total level of fishing efforts per fleet over the whole modelled area constant at 2013 levels. However, the spatial distribution of efforts within the area was changing every time-step as a consequence of changes in spatial distributions of the targeted fish. This means that total fishing mortality caused by each fleet on the species they catch remained constant over time, but varied in space according to the simulated effort distribution. Temporal forcing used in the scenarios is described in Supplementary Appendix S2.3. Results First we compare the ecosystem response among the three modelled nutrient scenarios BSAP, REF, and BAU. Second, we present the main outcomes of the sensitivity analysis. Spatial ecosystem structure In our EwE projections, species or groups sensitive to O2 concentrations close to the seafloor generally benefit from reduced nutrient loads. Cod, flounder (Figure 4), and mysids (Figure 5) as well as all macrobenthos groups (Supplementary Figure S26) have a larger distribution range under the BSAP scenario due to higher bottom and below 60 m oxygen concentrations compared to the other two scenarios (Supplementary Figure S23). Under REF and BAU, hypoxia-tolerant meiobenthos is profiting from the absence of macrobenthic fauna and its biomass density increases in the deep basins (Supplementary Figure S26). Cod and flounder biomass density is low in the direct proximity of the coast and south-east from the island of Öland due to low bottom salinity in all scenarios, and west of Gotland due to low oxygen content especially in the REF and BAU scenarios. In the BAU and REF scenarios both species are mostly concentrated in the southern, and, in the case of flounder, eastern parts of the Baltic. They (especially flounder) reach high densities along the coasts, just beyond the shallowest areas, in these two scenarios. Changes in demersal fish distributions substantially affect the spatial distributions of some of their prey and top predator species. Besides clupeids, both juvenile and adult cod and adult flounder are important prey for grey seals, and therefore seal concentration is predicted to shift southwards under both REF and BAU scenarios compared to BSAP (Supplementary Figure S25). Sprat is present in all of the modelled area in all scenarios, but under BSAP it is rather concentrated toward shallower areas (Figure 4). Under REF and especially BAU sprat has a very high density across the whole area although it is relatively more concentrated in deep areas. In both cases, the distribution of sprat is negatively related to the distribution of cod, most probably due to strong cod predation on sprat. Compared to other fish, the spatial distribution of herring is less affected by the nutrient load scenarios, even though also for this species there is a general increase in density across the whole area in the REF and BAU scenarios. This is probably due to various factors affecting its distribution simultaneously (predation by cod and seal, benthic food availability, competition with sprat for zooplankton). Figure 4. Open in new tabDownload slide Projected density (t/km2) of adult fish (average values 2088–2098) under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Juvenile fish distributions are very similar to those of adult ones and therefore not shown. Average distributions 2004–2008 are shown in Supplementary Figures S11 and S12. Figure 4. Open in new tabDownload slide Projected density (t/km2) of adult fish (average values 2088–2098) under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Juvenile fish distributions are very similar to those of adult ones and therefore not shown. Average distributions 2004–2008 are shown in Supplementary Figures S11 and S12. Figure 5. Open in new tabDownload slide Projected density (t/km2) of selected lower trophic level functional groups (average values 2088–2098) under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Average distributions 2004–2008 are shown in Supplementary Figure S18. Figure 5. Open in new tabDownload slide Projected density (t/km2) of selected lower trophic level functional groups (average values 2088–2098) under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Average distributions 2004–2008 are shown in Supplementary Figure S18. Spatial distributions of the intermediate trophic level predators (the clupeids) affects the distributions of lower trophic level groups. Sprat and herring are the most important predators of the Pseudocalanus spp. and “other zooplankton”’ functional groups in the model, which therefore benefit from the relatively low densities of clupeids in the deep sea east of Gotland under the BSAP scenario (Figure 5). Even though the smaller Acartia spp. and Temora sp. are also consumed by clupeids, they are significantly predated upon by mysids as well. This is probably the reason why they do not show substantial differences in their distributions among the scenarios (Supplementary Figure S25). Differences in the spatial distribution of the primary producer group among the scenarios are the result of differences in zooplankton predation and nutrient loads. Phytoplankton density overall is increasingly higher when comparing BSAP, REF, and BAU scenarios due to an increasing level of nutrients available for primary production (Figure 5). While in the BSAP scenario phytoplankton in the deep offshore area east of Gotland is consumed by zooplankton, the low densities of Pseudocalanus spp. and the “other zooplankton” groups under REF and especially BAU result in an accumulation of phytoplankton biomass in the area. Distribution of fishing effort Figure 6 shows the distribution of fishing efforts of three selected fleet segments (one vessel size per each gear type) under three scenarios. Note that socio-economic drivers, such as port placement, fleet composition, and structure are assumed to be constant in time. Thus, modelled differences in fishing effort distributions across scenarios reflect differences in their target species distributions, higher priced fish having a larger influence. Thus, effort distributions indicate the relative suitability of fishing grounds under the three scenarios. Effort distributions of fleets using active and passive gears strongly reflect the biomass distribution of cod. Consequently, under the BSAP scenario their efforts are more evenly distributed over a larger area than in BAU and REF. This means that while under BSAP there are many similarly suitable fishing grounds in the model, increasing nutrient loads lead to intense fishing in small areas. Comparison of weighted center points of fishing effort distributions in 2004 to those from 2088 to 2098 shows that under BSAP fishing efforts of the demersal active and passive fisheries shift in a north-east direction, especially in the case of passive fleets. Under the REF and BAU scenarios weighted center points do not shift in space compared to 2004. Fleet effort distributions are projected to be very similar among fleet segments using differently sized vessels. Thus, the effort distributions shown in Figure 6 for mid-sized demersal trawlers and small vessels using passive gears are representative for all modelled vessel size categories of demersal trawlers and vessels using passive gears, respectively. Figure 6. Open in new tabDownload slide Projected fishing effort (average values 2088–2098) of selected fishing fleets, up to down: mid-sized demersal trawlers, small vessels using passive gears, and mid-sized pelagic trawlers, under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Values express fishing efforts relative to each fleet’s average effort over the entire modelled area in the initial year, 2004. Darker shades represent higher values. Brown triangles indicate the locations of the modelled weighted center points of the effort distributions in each of the last 11 simulated years (2088–2098). Orange circles show the same from 2004 (initial model state after spin-up period). Figure 6. Open in new tabDownload slide Projected fishing effort (average values 2088–2098) of selected fishing fleets, up to down: mid-sized demersal trawlers, small vessels using passive gears, and mid-sized pelagic trawlers, under three nutrient management scenarios: Baltic Sea Action Plan (left), Reference (middle) and Business-As-Usual (right column), in the modelled area (see Figure 1). Values express fishing efforts relative to each fleet’s average effort over the entire modelled area in the initial year, 2004. Darker shades represent higher values. Brown triangles indicate the locations of the modelled weighted center points of the effort distributions in each of the last 11 simulated years (2088–2098). Orange circles show the same from 2004 (initial model state after spin-up period). The fishing effort distribution of the pelagic fleet segment (Figure 6) reflects herring and sprat distribution in the BSAP scenario (Figure 4). Although this fleet segment mostly targets clupeids, it catches cod as well. This explains our projections which indicate that under the REF and BAU scenarios the location of the most suitable fishing grounds mirror the changes in clupeids’ distribution at the broad scale and the cod distribution at a finer scale. For all fleet segments, but especially for the pelagic fleets, the weighted center points during 2088–2098 are concentrated in a small area in the BSAP scenario relative to the other two scenarios, where they are more scattered. This means that under BSAP the year-to-year variability in effort distributions is smaller, indicating less change in the location of the most suitable fishing grounds between subsequent years. In contrast to the demersal fleets, effort distribution varied with vessel size in case of the pelagic fleet. In our model, vessels <24 m have a higher share of cod in their landings and therefore their distributions mostly reflect that of cod in all scenarios, similarly to demersal fleet segments. In contrast, landings of vessels >40 m consist almost entirely of clupeids and therefore their distributions follow that of the clupeids in all scenarios (Supplementary Figure S28). Sensitivity analysis The correlation between the modelled functional groups and the fleet effort distributions to empirical data and its sensitivity to model assumptions are described in Appendices S3.2 and S3.3, respectively. In general, model fit to observations measured by correlation is similar among biomasses, catches, and efforts (Supplementary Figure S10, Appendix S3.2). Most variables show a Kendall’s correlation coefficient of 0.2–0.4, with the exception of lower correlation coefficients in the case of juvenile and adult herring biomass of about ∼0.05–0.1. The correlation was not very sensitive to the choice of ERFs included in the model because correlation coefficients obtained by including only a subset of ERFs were similar (Supplementary Figure S20, Appendix S3.3.1). In contrast, port placement and Effective Power influenced the model fit (Supplementary Figures S21 and S22, Appendix S3.3.2). Discussion and conclusions In this study we present a mechanistic framework to assess how future nutrient management measures potentially alter the capacity of different areas of the central Baltic Sea to support commercial fisheries under climate change. We show that the implementation of a strong nutrient reduction policy, such as the BSAP, would strongly increase the spatial extent of areas that can support all types of fisheries. On the other hand, a smaller part of the Baltic Sea may experience increased densities of fish under scenarios assuming constant or further increasing nutrient loads. Such increased densities may cause population pressures and responses that are not included in our modelling framework, such as increased parasite infection rates and decreased individual growth (Eero et al., 2015; Casini et al., 2016). We found large differences among three modelled nutrient management scenarios in terms of spatial community composition and, consequently, fishing effort distributions across the whole modelled area. Although one region, the southeastern Baltic Sea, remained an important fishing ground in all scenarios, its relative importance compared to other areas changed dramatically. While in the highest nutrient load scenario it was the only area which could sustain both demersal and pelagic fisheries, in the lowest nutrient load scenario other, more northern areas also became suitable. Therefore, the relative location of most suitable fishing grounds for demersal fisheries shifted northwards, especially for the segments using passive gears. An extended potential range of operations may be particularly important for this segment as it is considered to be the most vulnerable within the Baltic fishery (Strehlow, 2010). Not only the spatial distribution of suitable fishing grounds, but also their interannual variability, differed among the scenarios. Under the low nutrient loading scenario, larger areas were suitable for fisheries and their location tended also to be more stable among years. This sort of spatial reliability of fish production may facilitate the inclusion of fisheries into marine spatial planning in the future. One of the most important outcomes of our study is that the differences in species distributions among modelled scenarios were the result of cumulative impacts of several environmental factors, in agreement with Stortini et al. (2017). While for individual groups one or two factors could be pinpointed as important drivers, changes in the spatial structure of the community as a whole were the result of the combined effects of changes in oxygen, salinity, primary productivity, and food web interactions. In the Baltic Sea, currently cod is the most important top predator and changes in its abundance potentially cause multilevel trophic cascades (Casini et al., 2008). Hypoxia-induced habitat compression of cod and its consequences for the spatial distribution of intermediate trophic level forage fish in the Baltic Sea are well documented (Casini et al., 2011, 2016). Our model results indicate that the habitat compression of cod may be reversed if nutrient load reduction policies are implemented. While constant and high nutrient load scenarios had adverse effects on benthic and demersal groups, abundances of phytoplankton and pelagic fish were predicted to increase. Similarly to other seas, there has been a positive link between increased nutrient loads and (especially forage) fish production in the Baltic (Chassot et al., 2007; Eero et al., 2016) also supported by our model results. However, it is questionable whether this relationship will hold in the future. Some evidence suggests that further increases in the eutrophication levels compared to today, especially under higher temperatures and lower abundances of higher trophic level predators, could contribute to shifts in primary producer composition to an unfavourable state for consumers. Such shifts include an increased proportion of smaller-sized organisms (Suikkanen et al., 2013), a more frequent occurrence of toxic cyanobacterial blooms (Lehtiniemi et al., 2002; Neumann et al., 2012), and an increased dominance of filamentous algae in coastal habitats (Borg et al., 1997). Our results also point out the environmental dependency of suitable areas for fisheries and possibly all human activities based on ecosystem functioning. This means that long-term, adaptive marine spatial planning needs to take into account changing abiotic conditions (Miller et al., 2013). Our modelling study suggests that the provision of wild-captured fish food, one of the important ecosystem services, may have a more even spatial distribution across the central Baltic Sea when nutrient loads are reduced. This could have important economic consequences for the fishing industry as spatial relation to the most productive fishing grounds is an important determinant of fleet efficiency (Hutniczak et al., 2015; Bastardie et al., 2017). When fish distribution consists of small pockets of high densities in space, as predicted under increasing nutrient loads, the risk for overexploitation is higher. Discard issues may also increase if species which are targeted and those that are caught as bycatch have similar requirements and their distributions become restricted to overlapping areas, such as cod and flounder in our model. Additionally, fisheries have to share the marine space with other human activities (Tidd et al., 2015; Yates et al., 2015). For example, in one area within the Sound (part of the Baltic Sea) a trawling ban has been in place since 1932 due to intense shipping traffic in the area (Lindegren et al., 2013). When the extent of areas suitable for fishing operations is decreased, together with the extent of areas supporting ecosystem functions, managers may face more difficult trade-offs in allocating marine areas for exploitation, conservation and other uses. In their recent study, Zurell et al. (2016) have shown that mechanistic modelling approaches, such as dispersal or population dynamics models and Bayesian process-based dynamic range models, outperform correlative species distribution models in predicting species range dynamics under climate change. We argue that the approach presented here is a useful complement to those evaluated by Zurell et al. (2016), as it simultaneously provides projections of all functional groups in an ecosystem without necessarily needing spatio-temporal data on abundances of all groups. For our ecosystem, the model was also not sensitive to the number of ERFs included and major patterns in the data could be reproduced by including a few key functions only (see Supplementary Appendix 3.3.1). Still, there is a need for the development of a consistent methodology for the parameterization of ERFs that express the responses of functional groups to abiotic factors. The empirical measurement of such responses is a highly active research area (e.g. Birnie-Gauvin et al., 2017). In addition, developing standard methodology to reliably assess the skill of spatial ecosystem models such as Ecospace is important to have an insight about the uncertainty of their predictions. Ideally, such a methodology would be based on a combination of metrics including correlation as used here, but also neighbourhood-based methods as described by Rose et al., (2009) and Stow et al., (2009) for oceanographic models. As Ecospace model parameterization is not based on automated statistical fitting but on expert judgement and literature values, it is especially important to explore the sensitivity of the results to assumptions made during model parameterization. Romagnoni et al. (2015) conducted an extensive sensitivity analysis of their Ecospace model for the North Sea. We have tested our model’s sensitivity to some of the same parameters they have found to be important. Both studies found a reasonably good agreement between modelled population distributions and spatial data from scientific surveys and a large effect of the parameter “Effective Power,” which affects the level of dispersion of modelled fleet efforts around profitable fishing areas. The agreement between modelled fishing efforts and spatial data from commercial fisheries was better in our Baltic model. In contrast to Romagnoni et al. (2015), we found an effect of port placement on fishing fleet distributions. The placement of fishing ports affects the calculation of a fishing cost map that reflects distance from ports. The fishing cost map is then used to distribute fishing effort, evaluating fleet- and cell-specific fishing costs based on the fleet-specific ratio of sailing- related costs to fixed fishing costs. The latter ratios were much higher in the case of the Baltic model which explains the higher sensitivity of our modelled fleets’ to port placement. Notably, some species distributions were also sensitive to port placement (see Supplementary Figure S21). The reason for this is that port placement influenced how fishing mortality was distributed in space via making areas far away from ports relatively less attractive for fishing fleets. This underlines the importance of considering both economic and environmental factors when making predictions about future species distributions. Compared to other modelled populations, our approach proved to be less successful in reproducing the distribution of one commercially important group, herring. Some earlier studies have shown that modelled distributions of species with distinct environmental preferences, such as cod, generally fit better to data than those of species with wide tolerances, such as herring (Somodi et al., 2017). In addition, pelagic species have more variable distributions than demersal ones which is harder to reproduce by models (Thorson et al., 2016). These results suggest that spatial management of such groups inevitably involves more uncertainty. Changes in habitat quality due to human impacts are increasingly common across the globe. As species shift their distributions to adapt to altered environmental conditions, the spatial provision of ecosystem services changes as well. Here we presented the projected effects of various nutrient management policies on various environmental variables and the cumulative effects of those factors across the marine food web and on commercial fisheries in the example of the Baltic Sea. Where data are available, the same approach could be used to evaluate the potential consequences of various environmental policies in other systems. In the Baltic Sea, it may provide inspiration for studies more focused on certain functional groups or areas. Our results indicate the effectivity of nutrient load reduction policies in recovering ecosystem function across large areas of the Baltic Sea, which may motivate environmental managers to further pursue such policies. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements BB, HEMM, and SS were funded by the BONUS BalticAPP (Well-being from the Baltic Sea—applications combining natural science and economics) and MC and AO by the BONUS INSPIRE project which have received funding from BONUS, the joint Baltic Sea research and development programme (Art 185), funded jointly from the European Union’s Seventh Programme for research, technological development and demonstration and from the Swedish Research Council (FORMAS). HEMM’s research is part of the Baltic Earth program (Earth System Science for the Baltic Sea region, http://www.baltic.earth). BB and PM have received funding from the MareFrame project (Co-creating Ecosystem-based Fisheries Management Solutions)—EU 7th FP under grant agreement no. 613571. BB is employed by the Baltic Sea Center at Stockholm University Baltic Nest Institute, which is supported by the Swedish Agency for Marine and Water Management through their grant 1: 11—Measures for marine and water environment. MTT is employed by the Baltic Sea Center Stockholm University Baltic Eye, a strategic partnership between Stockholm University and the BalticSea2020 fundation. We are grateful for two anonymous reviewers for their constructive comments and we thank M. Geibel for technical help. References Almroth-Rosell E. , Eilola K., Hordoir R., Meier H. E. M., Hall P. O. J. 2011 . Transport of fresh and resuspended particulate organic material in the Baltic Sea - a model study . Journal of Marine Systems , 87 : 1 – 12 . Google Scholar Crossref Search ADS WorldCat Bartolino V. , Tian H., Bergström U., Jounela P., Aro E., Dieterich C., Markus Meier H. E. et al. 2017 . Spatio-temporal dynamics of a fish predator: density-dependent and hydrographic effects on Baltic Sea cod population . PLoS One , 12 : e0172004. Google Scholar Crossref Search ADS PubMed WorldCat Bastardie F. , Vinther M., Nielsen J. R., Ulrich C., Paulsen M. S. 2010a . Stock-based vs. fleet-based evaluation of the multi-annual management plan for the cod stocks in the Baltic Sea . Fisheries Research , 101 : 188 – 202 . Google Scholar Crossref Search ADS WorldCat Bastardie F. , Nielsen J. R., Eero M., Fuga F., Rindorf A. 2017 . Effects of changes in stock productivity and mixing on sustainable fishing and economic viability . ICES Journal of Marine Science , 74 : 535 – 551 . Google Scholar Crossref Search ADS WorldCat Bastardie F. , Nielsen J. R., Kraus G. 2010b . The eastern Baltic cod fishery: A fleet-based management strategy evaluation framework to assess the cod recovery plan of 2008 . ICES Journal of Marine Science , 67 : 71 – 86 . Google Scholar Crossref Search ADS WorldCat Birnie-Gauvin K. , Costantini D., Cooke S. J., Willmore W. G. 2017 . A comparative and evolutionary approach to oxidative stress in fish: a review . Fish and Fisheries , 18 : 928 – 942 . Google Scholar Crossref Search ADS WorldCat Borg A. , Pihl L., Wennhage H. 1997 . Habitat choice by juvenile cod (Gadus morhua L.) on sandy soft bottoms with different vegetation types . Helgolander Meeresuntersuchungen , 51 : 197 – 212 . Google Scholar Crossref Search ADS WorldCat Carstensen J. , Andersen J. H., Gustafsson B. G., Conley D. J. 2014 . Deoxygenation of the Baltic Sea during the last century . Proceedings of the National Academy of Sciences , 111 : 5628 – 5633 . Google Scholar Crossref Search ADS WorldCat Casini M. , Kornilovs G., Cardinale M., Möllmann C., Grygiel W., Jonsson P., Raid T. et al. 2011 . Spatial and temporal density dependence regulates the condition of central Baltic Sea clupeids : compelling evidence using an extensive international acoustic survey . Population Ecology , 53 : 511 – 523 . Google Scholar Crossref Search ADS WorldCat Casini M. , Lövgren J., Hjelm J., Cardinale M., Molinero J-C., Kornilovs G. 2008 . Multi-level trophic cascades in a heavily exploited open marine ecosystem . Proceedings of the Royal Society B: Biological Sciences , 275 : 1793 – 1801 . Google Scholar Crossref Search ADS WorldCat Casini M. , Rouyer T., Bartolino V., Larson N., Grygiel W. 2014 . Density-dependence in space and time: opposite synchronous variations in population distribution and body condition in the Baltic Sea sprat (Sprattus sprattus) over three decades . PLoS One , 9 : e92278. Google Scholar Crossref Search ADS PubMed WorldCat Casini M. , Käll F., Hansson M., Plikshs M., Baranova T., Karlsson O., Lundström K. et al. 2016 . Hypoxic areas, density-dependence and food limitation drive the body condition of a heavily exploited marine fish predator . Royal Society Open Science , 3 : 160416. Google Scholar Crossref Search ADS PubMed WorldCat Chassot E. , Mélin F., Le Pape O., Gascuel D. 2007 . Bottom-up control regulates fisheries production at the scale of eco-regions in European seas . Marine Ecology Progress Series , 343 : 45 – 55 . Google Scholar Crossref Search ADS WorldCat Christensen V. , Coll M., Steenbeek J., Buszowski J., Chagaris D., Walters C. J. 2014 . Representing variable habitat quality in a spatial food web model . Ecosystems , 17 : 1397. Google Scholar Crossref Search ADS WorldCat Christensen V. , Walters C. J. 2004 . Ecopath with Ecosim: methods, capabilities and limitations . Ecological Modelling , 172 : 109 – 139 . Google Scholar Crossref Search ADS WorldCat Christensen V. , Walters C. J., Pauly D., Forrest R. 2008 . Ecopath with Ecosim Version 6. User Guide. Fisheries Centre, University of British Columbia, Vancouver, Canada. Coll M. , Steenbeek J., Sole J., Palomera I., Christensen V. 2016 . Modelling the cumulative spatial–temporal effects of environmental drivers and fishing in a NW Mediterranean marine ecosystem . Ecological Modelling , 331 : 100 – 114 . Elsevier B. V. Google Scholar Crossref Search ADS WorldCat Eero M. , Andersson H. C., Almroth-Rosell E., MacKenzie B. R. 2016 . Has eutrophication promoted forage fish production in the Baltic Sea? Ambio , 45 : 649 – 660 . Google Scholar Crossref Search ADS PubMed WorldCat Eero M. , Hjelm J., Behrens J., Buchmann K., Cardinale M., Casini M., Gasyukov P. et al. 2015 . Eastern Baltic cod in distress: biological changes and challenges for stock assessment . ICES Journal of Marine Science , 72 : 2180 – 2186 . Google Scholar Crossref Search ADS WorldCat Eero M. , Vinther M., Haslob H., Huwer B., Casini M., Storr-Paulsen M., Köster F. W. 2012 . Spatial management of marine resources can enhance the recovery of predators and avoid local depletion of forage fish . Conservation Letters , 5 : 486 – 492 . Google Scholar Crossref Search ADS WorldCat Eilola K. , Meier H. E. M., Almroth E., Almroth E. 2009 . On the dynamics of oxygen, phosphorus and cyanobacteria in the Baltic Sea; a model study . Journal of Marine Systems , 75 : 163 – 184 . Google Scholar Crossref Search ADS WorldCat Godsoe W. , Jankowski J., Holt R. D., Gravel D. 2017 . Integrating biogeography with contemporary niche theory . Trends in Ecology and Evolution , 32 : 488 – 499 . Google Scholar Crossref Search ADS PubMed WorldCat Gogina M. , Zettler M. L. 2010 . Diversity and distribution of benthic macrofauna in the Baltic Sea. Data inventory and its use for species distribution modelling and prediction . Journal of Sea Research , 64 : 313 – 321 . Google Scholar Crossref Search ADS WorldCat Gordon C. , Cooper C., Senior C. A., Banks H., Gregory J. M., Johns T. C., Mitchell J. F. B. et al. 2000 . The simulation of SST, sea ice extent and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments . Climate Dynamics , 16 : 147 – 166 . Google Scholar Crossref Search ADS WorldCat Hansson S. , Hjerne O., Harvey C., Kitchell J. F., Cox S. P., Essington T. E. 2007 . Managing Baltic Sea fisheries under contrasting production and predation regimes: ecosystem model analyses . Ambio , 36 : 265 – 271 . Google Scholar Crossref Search ADS PubMed WorldCat Härkönen T. , Galatius A., Bräger S., Karlsson O., Ahola M. 2013 . Population growth rate, abundance and distribution of marine mammals. HELCOM Core Indicator Report. Online: 1–34. HELCOM . 2007 . Toward a Baltic Sea unaffected by eutrophication. Background document to Helcom Ministerial Meeting, Krakow, Poland, Tech. rep., Helsinki Commission, Helsinki, Finland. Heymans J. J. , Coll M., Link J. S., Mackinson S., Steenbeek J., Walters C., Christensen V. 2016 . Best practice in developing, balancing, fitting and using Ecopath with Ecosim food-web models for ecosystem-based management . Ecological Modelling , 331 : 173 – 184 . Google Scholar Crossref Search ADS WorldCat Hutniczak B. , Nieminen E., Hoffmann J., Yletyinen J. 2015 . Input-efficiency of fishing cod in the Baltic Sea – comparing Major EU trawler fleets. University of Helsinki Dep. of Economics and Management Discussion Papers, 68. ICES . 2015 . Report of the Baltic Fisheries Assessment Working Group (WGBFAS), 14–21 April 2015, ICES HQ, Copenhagen, Denmark. ICES CM 2015/ACOM: 10. 806 pp. ICES . 2016a . Report of the Baltic Fisheries Assessment Working Group (WGBFAS), 12–19 April 2016, ICES HQ, Copenhagen, Denmark. ICES CM 2016/ACOM: 11. 593 pp. ICES . 2016b . Report of the Working Group on Multispecies Assessment Methods (WGSAM), 10–14 October 2016, Reykjavik, Iceland. ICES CM 2016/SSGEPI: 21. 94 pp. Karlson K. , Rosenberg R., Bonsdorff E. 2002 . Temporal and spatial large-scale effects of eutrophication and oxygen deficiency on benthic fauna in Scandinavian and Baltic Waters - a review . Oceanography and Marine Biology: An Annual Review , 40 : 427 – 489 . Google Scholar OpenURL Placeholder Text WorldCat Kraus G. , Pelletier D., Dubreuil J., Mollmann C., Hinrichsen H-H., Bastardie F., Vermard Y. et al. 2008 . A model-based evaluation of marine protected areas: the example of eastern Baltic cod (Gadus morhua callarias L.) . ICES Journal of Marine Science , 66 : 109 – 121 . Google Scholar Crossref Search ADS WorldCat Lehtiniemi M. , Engstrom-Ost J., Karjalainen M., Kozlowsky-Suzuki B., Viitasalo M. 2002 . Fate of cyanobacterial toxins in the pelagic food web: transfer to copepods or to faecel pellets? Marine Ecology Progress Series , 241 : 13 – 21 . Google Scholar Crossref Search ADS WorldCat Link J. S. 2010 . Adding rigor to ecological network models by evaluating a set of pre-balance diagnostics: A plea for PREBAL . Ecological Modelling , 221 : 1580 – 1591 . Google Scholar Crossref Search ADS WorldCat Lindegren M. , Andersen K. H., Casini M., Neuenfeldt S. 2014 . A metacommunity perspective on source—sink dynamics and management : the Baltic Sea as a case study . Ecological Applications , 24 : 1820 – 1832 . Google Scholar Crossref Search ADS PubMed WorldCat Lindegren M. , Waldo S., Nilsson P. A., Svedäng H., Persson A. 2013 . Towards sustainable fisheries of the Öresund cod (Gadus morhua) through sub-stock-specific assessment and management recommendations . ICES Journal of Marine Science , 70 : 1140 – 1150 . Google Scholar Crossref Search ADS WorldCat Mackinson S. , Daskalov G. 2007 . An ecosystem model of the North Sea to support an ecosystem approach to fisheries management: description and parameterisation: 196. Matthäus W. , Nehring D., Feistel R., Nausch G., Mohrholz V., Lass H. U. 2008 . The inflow of highly saline water into the Baltic Sea. In State and Evolution of the Baltic Sea, 1952–2005: A Detailed 50-Year Survey of Meteorology and Climate, Physics, Chemistry, Biology, and Marine Environment , pp. 265 – 309 . John Wiley & Sons, Inc ., Hoboken, NJ . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Meier H. E. M. , Andersson H. C., Arheimer B., Blenckner T., Chubarenko B., Donnelly C., Eilola K. et al. 2012a . Comparing reconstructed past variations and future projections of the Baltic Sea ecosystem—first results from multi-model ensemble simulations . Environmental Research Letters , 7 : 34005 . Google Scholar Crossref Search ADS WorldCat Meier H. E. M. , Andersson H. C., Arheimer B., Donnelly C., Eilola K., Gustafsson B. G., Kotwicki L. et al. 2014 . Ensemble modeling of the Baltic Sea ecosystem to provide scenarios for management . Ambio , 43 : 37 – 48 . Google Scholar Crossref Search ADS PubMed WorldCat Meier H. E. M. , Döscher R., Faxén T. 2003 . A multiprocessor coupled ice-ocean model for the Baltic Sea: application to salt inflow . Journal of Geophysical Research , 108 : 3273. Google Scholar Crossref Search ADS WorldCat Meier H. E. M. , Eilola K., Gustafsson B. G., Kuznetsov I., Neumann T., Savchuk O. P. 2012b . Uncertainty assessment of projected ecological quality indicators in future climate . Rapport Oceanograpfi No. , 112 , SMHI, Norrköping, Sweden. Google Scholar OpenURL Placeholder Text WorldCat Meier H. E. M. , Feistel R., Piechura J., Arneborg L., Burchard H., Fiekas V., Golenko N. et al. 2006 . Ventilation of the Baltic Sea deep water: a brief review of present knowledge from observations and models . Oceanologia , 48 : 133 – 164 . Google Scholar OpenURL Placeholder Text WorldCat Meier H. E. M. , Höglund A., Döscher R., Andersson H., Löptien U., Kjellström E. 2011 . Quality assessment of atmospheric surface fields over the Baltic Sea of an ensemble of regional climate model simulations with respect to ocean dynamics . Oceanologia , 53 : 193 – 227 . Google Scholar Crossref Search ADS WorldCat Meier H. E. M. , Müller-Karulis B., Andersson H. C., Dieterich C., Eilola K., Gustafsson B. G., Höglund A. et al. 2012c . Impact of climate change on ecological quality indicators and biogeochemical fluxes in the Baltic Sea - a multi-model ensemble study . Ambio , 41 : 558 – 573 . Google Scholar Crossref Search ADS WorldCat Miller K. A. , Munro G. R., Sumaila U. R., Cheung W. W. L. 2013 . Governing marine fisheries in a changing climate: a game-theoretic perspective . Canadian Journal of Agricultural Economics , 61 : 309 – 334 . Google Scholar Crossref Search ADS WorldCat Nakićenović N. , Alcamo J., Davis G., de Vries B. 2000 . Emission scenarios. A special report of working group III of the Intergovernmental Panel on Climate Change. Cambridge. Neumann T. , Eilola K., Gustafsson B., Müller-Karulis B., Kuznetsov I., Meier H. E. M., Savchuk O. P. 2012 . Extremes of temperature, oxygen and blooms in the Baltic Sea in a changing climate . Ambio , 41 : 574 – 585 . Google Scholar Crossref Search ADS PubMed WorldCat Niiranen S. , Yletyinen J., Tomczak M. T. 2013 . Combined effects of global climate change and regional ecosystem drivers on an exploited marine food web . Global Change Biology , 19 : 3327 – 3342 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Österblom H. , Hansson S., Larsson U., Hjerne O., Wulff F., Elmgren R., Folke C. 2007 . Human-induced trophic cascades and ecological regime shifts in the Baltic Sea . Ecosystems , 10 : 877 – 889 . Google Scholar Crossref Search ADS WorldCat Pauly D. , Christensen V., Walters C. 2000 . Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries . ICES Journal of Marine Science , 57 : 697 – 706 . Google Scholar Crossref Search ADS WorldCat Peterson A. T. , Soberón J., Pearson R. G., Anderson R. P., Martínez-Meyer E., Nakamura M., Araújo M. B. 2011 . Ecological Niches and Geographic Distributions (MPB-49). Princeton University Press , Princeton, New Jersey . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Plagányi É. E. 2007 . Models for an ecosystem approach to fisheries. FAO Fisheries Technical Paper, No. 477. FAO, Rome. 108 pp. Plagányi É. E. , Butterworth D. S. 2004 . A critical look at the potential of Ecopath with ecosim to assist in practical fisheries management . African Journal of Marine Science , 26 : 261 – 287 . Google Scholar Crossref Search ADS WorldCat Radtke H. , Neumann T., Fennel W. 2013 . A Eulerian nutrient to fish model of the Baltic Sea - a feasibility-study . Journal of Marine Systems , 125 : 61 – 76 . Google Scholar Crossref Search ADS WorldCat Röckmann C. , St. John M. A., Schneider U. A., Tol R. S. J. 2007 . Testing the implications of a permanent or seasonal marine reserve on the population dynamics of Eastern Baltic cod under varying environmental conditions . Fisheries Research , 85 : 1 – 13 . Google Scholar Crossref Search ADS WorldCat Röckmann C. , Tol R. S. J., Schneider U. A., John M. A. S. 2008 . Rebuilding the eastern Baltic cod stock under environmental change (Part II): taking into account the costs of a marine protected area . Natural Resource Modeling , 20 : 223 – 225 . Google Scholar Crossref Search ADS WorldCat Romagnoni G. , Mackinson S., Hong J., Eikeset A. M. 2015 . The Ecospace model applied to the North Sea: evaluating spatial predictions with fish biomass and fishing effort data . Ecological Modelling , 300 : 50 – 60 . Google Scholar Crossref Search ADS WorldCat Rose K. A. , Roth B. M., Smith E. P. 2009 . Skill assessment of spatial maps for oceanographic modeling . Journal of Marine Systems , 76 : 34 – 48 . Google Scholar Crossref Search ADS WorldCat Somodi I. , Lepesi N., Botta-Dukát Z. 2017 . Prevalence dependence in model goodness measures with special emphasis on true skill statistics . Ecology and Evolution , 7 : 863 – 872 . Google Scholar Crossref Search ADS PubMed WorldCat Stäbler M. , Kempf A., Mackinson S., Poos J. J., Garcia C., Temming A. 2016 . Combining efforts to make maximum sustainable yields and good environmental status match in a food-web model of the southern North Sea . Ecological Modelling , 331 : 17 – 30 . Google Scholar Crossref Search ADS WorldCat Steenbeek J. , Coll M., Gurney L., Mélin F., Hoepffner N., Buszowski J., Christensen V. 2013 . Bridging the gap between ecosystem modelling tools using geographic information systems: driving a food-web model with spatial-temporal primary production data . Ecological Modelling , 263 : 139 – 151 . Google Scholar Crossref Search ADS WorldCat Stortini C. H. , Chabot D., Shackell N. L. 2017 . Marine species in ambient low-oxygen regions subject to double jeopardy impacts of climate change . Global Change Biology , 23 : 2284 – 2296 . Google Scholar Crossref Search ADS PubMed WorldCat Stow C. A. , Jolliff J., McGillicuddy D. J., Doney S. C., Allen J. I., Friedrichs M. A. M., Rose K. A. et al. 2009 . Skill assessment for coupled biological/physical models of marine systems . Journal of Marine Systems , 76 : 4 – 15 . Google Scholar Crossref Search ADS PubMed WorldCat Strehlow H. V. 2010 . The multiannual management plan for cod in the Baltic Sea: reactions and sentiments in two German fishing communities . ICES Journal of Marine Science , 67 : 1963 – 1971 . Google Scholar Crossref Search ADS WorldCat Suikkanen S. , Pulina S., Engström-Öst J., Lehtiniemi M., Lehtinen S., Brutemark A., Lin S. 2013 . Climate change and eutrophication induced shifts in northern summer plankton communities . PLoS One , 8 : e66475. Google Scholar Crossref Search ADS PubMed WorldCat Thorson J. T. , Pinsky M. L., Ward E. J., Gimenez O. 2016 . Model-based inference for estimating shifts in species distribution, area occupied and centre of gravity . Methods in Ecology and Evolution , 7 : 990 – 1002 . Google Scholar Crossref Search ADS WorldCat Tidd A. N. , Vermard Y., Marchal P., Pinnegar J., Blanchard J. L., Milner-Gulland E. J., Hiddink J. G. 2015 . Fishing for space: fine-scale multi-sector maritime activities influence fisher location choice . PLoS One , 10 : e0116335. Google Scholar Crossref Search ADS PubMed WorldCat Tomczak M. T. , Niiranen S., Hjerne O., Blenckner T. 2012 . Ecosystem flow dynamics in the Baltic Proper-Using a multi-trophic dataset as a basis for food-web modelling . Ecological Modelling , 230 : 123 – 147 . Google Scholar Crossref Search ADS WorldCat Townhill B. L. , Pinnegar J. K., Righton D. A., Metcalfe J. D. 2017 . Fisheries, low oxygen and climate change: how much do we really know? Journal of Fish Biology , 90 : 723 – 750 . Google Scholar Crossref Search ADS PubMed WorldCat Uusitalo L. , Korpinen S., Jesper H., Niiranen S., Valanko S., Heiskanen A., Dickey-Collas M. 2016 . Exploring methods for predicting multiple pressures on ecosystem recovery: a case study on marine eutrophication and fisheries . Continental Shelf Research , 121 : 48 – 60 . Google Scholar Crossref Search ADS WorldCat Voss R. , Petereit C., Schmidt J. O., Lehmann A., Makarchouk A., Hinrichsen H. 2012 . The spatial dimension of climate-driven temperature change in the Baltic Sea and its implication for cod and sprat early life stage survival . Journal of Marine Systems , 100-101 : 1 – 8 . Google Scholar Crossref Search ADS WorldCat Walters C. J. , Christensen V., Pauly D. 1997 . Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments . Reviews in Fish Biology and Fisheries , 7 : 139 – 172 . Google Scholar Crossref Search ADS WorldCat Walters C. , Christensen V., Walters W., Rose K. 2010 . Representation of multistanza life histories in Ecospace models for spatial organization . Bulletin of Marine Science , 86 : 439 – 459 . Google Scholar OpenURL Placeholder Text WorldCat Walters C. , Pauly D., Christensen V. 1999 . ECOSPACE: prediction of mesoscale spatial patterns in trophic relationships of exploited ecosystems, with particular reference to impacts of marine protected areas . Ecosystems , 2 : 539 – 554 . Google Scholar Crossref Search ADS WorldCat Yates K. L. , Schoeman D. S., Klein C. J. 2015 . Ocean zoning for conservation, fisheries and marine renewable energy: assessing trade-offs and co-location opportunities . Journal of Environmental Management , 152 : 201 – 209 . Google Scholar Crossref Search ADS PubMed WorldCat Zurell D. , Thuiller W., Pagel J., Cabral J. S., Münkemüller T., Gravel D., Dullinger S. et al. 2016 . Benchmarking novel approaches for modelling species range dynamics . Global Change Biology , 22 : 2651 – 2664 . Google Scholar Crossref Search ADS PubMed WorldCat © International Council for the Exploration of the Sea 2018. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © International Council for the Exploration of the Sea 2018.

Trijoulet, Vanessa; Dobby, Helen; Holmes, Steven J; Cook, Robin M

doi: 10.1093/icesjms/fsx235pmid: N/A

Abstract The role grey seals have played in the performance of fisheries is controversial and a cause of much debate between fishers and conservationists. Most studies focus on the effects of seal damage to gears or fish and on prey population abundance but little attention is given to the consequences of the latter for the fisheries. We develop a model that quantifies the economic impact of grey seal predation on the West of Scotland demersal fisheries that traditionally targeted cod, haddock and whiting. Three contrasting fishing strategy scenarios are examined to assess impacts on equilibrium fleet revenues under different levels of seal predation. These include status quo fishing mortality (SQF, steady state with constant fishing mortality), open access fishing (bioeconomic equilibrium, BE) and the maximum economic yield (MEY). In all scenarios, cod emerges as the key stock. Large whitefish trawlers are most sensitive to seal predation due to their higher cod revenues but seal impacts are minor at the aggregate fishery level. Scenarios that consider dynamic fleet behaviour also show the greatest effects of seal predation. Results are sensitive to the choice of seal foraging model where a type II functional response increases sensitivity to seal predation. The cost to the fishery for each seal is estimated. Introduction There has long been controversy concerning the potential impact seals have on commercial fisheries (Lambert, 2001; Lavigne, 2003; Read, 2008), especially those where traditionally cod (Gadus morhua) formed a large portion of catches or revenues. The precipitous decline of cod stocks in the Northwest Atlantic (Hutchings and Myers, 1994) and the poor state of many cod stocks in the Northeast Atlantic (Fernandes and Cook, 2013) has fuelled arguments that seals have had a detrimental effect on these stocks (Butler et al., 2011; Gruber, 2014). A number of studies have evaluated the predation mortality rate of seal populations on cod both off the Canadian coast (Mohn and Bowen, 1996; Trzcinski et al., 2006; O’Boyle and Sinclair, 2012) and in European waters (Alexander et al., 2015; Cook et al., 2015). These studies primarily consider the dynamics of the resource and the role seal predation may have played in the decline of cod stocks or their failure to recover. Most analyses have concluded that fishing has been the principal cause for stock decline but that seal predation may be an important factor in limiting their recovery. Regardless of any role seal predation has had on the decline in fish stocks, there is a widely held perception that seals represent direct competition with commercial fisheries and are therefore detrimental to both total revenues and profitability even if the fish stocks themselves are in a sustainable state. An important question that arises is the extent to which fish consumed by seals affects commercial fisheries not only in terms of resource abundance but also on the economic performance of the fisheries. Studies quantifying the economics of depredation, the direct seal-induced damage, on fisheries are numerous but focus on losses due to damage to gears or fish (Bosetti and Pearce, 2003; Cronin et al., 2014; Holma et al., 2014). The economic impacts of grey seal predation on fisheries have rarely been fully examined. Here we focus on the economic impact on the fisheries as a result of changes to the resource dynamics driven by seal predation rather than the issue of the possible role of seals in stock decline or lack of recovery. The West of Scotland area, which corresponds to ICES (International Council for the Exploration of the Sea) Division 6a (Figure 1), offers an opportunity to investigate the economic impact of grey seal predation using data from seal diet studies carried out in 1985 and 2002 (Hammond et al., 2006; Harris, 2007). These studies have documented the importance of a number of commercially important demersal species in grey seal diets including cod, haddock (Melanogrammus aeglefinus) and whiting (Merlangius merlangus), which are the traditional target species in the mixed demersal fishery. Since the 1980s, the grey seal population has increased in the West of Scotland but has stabilized in recent years at around 30 thousand individuals (Thomas, 2015). Grey seal predation mortality on cod has been estimated for this area (Holmes, 2008; Holmes and Fryer, 2011; Cook et al., 2015; Cook and Trijoulet, 2016) and more recently also on haddock and whiting (Trijoulet et al., 2017). However, these studies only consider the biological impacts of seal predation. Figure 1. Open in new tabDownload slide Map showing ICES Division 6a; the study area. Bathymetry data taken from Amante and Eakins (2009). In this study, we consider the bioeconomic impact of grey seal predation on the West of Scotland demersal trawl fishery, and in particular UK vessels, as these are responsible for the majority of the whitefish catch in this area taking on average 75% of the combined cod, haddock, and whiting landings between 2008 and 2012 (ICES, 2013). There are two principal components to the fisheries: one directed at whitefish with haddock as the main target species and a second directed at Norway lobster, Nephrops norvegicus, which takes a bycatch of cod, haddock and whiting (ICES, 2016a). We use an age-structured mixed species multifleet model to evaluate the potential impacts of seal predation on fishing revenues and net profits under various levels of seal predation. Three equilibrium scenarios are considered that enable a comparison of grey seal impacts under alternative fishing strategies or regulations. Material and methods The simulation model The principal equations governing the resource dynamics and the costs and revenues in the model are presented in Table 1. For stocks with sufficient data, the populations are modelled using conventional age-structured methods (Hilborn and Walters, 1992). Each cohort is subject to a mortality comprising the sum of the fishing (⁠ F ⁠), natural (⁠ M ⁠) and seal predation (⁠ P ⁠) mortalities [Equations (T1.1) and (T1.2)]. New recruits to the stock are given by a Ricker stock recruitment function (Ricker, 1954) and subject to stochastic process error [Equation (T1.3)]. Fishing mortality is decomposed into an age effect representing selectivity (⁠ s ⁠) and a year/effort effect (⁠ E ⁠) (Pope and Shepherd, 1982) and is further partitioned by fleet (⁠ k ⁠) [Equation (T1.4)]. Following Cook et al. (2015), seal predation mortality is assumed to be the product of seal selectivity for each age class (⁠ sel ⁠), seal predation rate (ability of seals to catch fish, q ⁠), and the total number of seals (⁠ G ⁠) [Equation (T1.5)]. Table 1. Equations used in the simulation model. Number . Name . Equation . Comments . (T1.1) Fish abundance at age a and year y for species i Na,y,i=Na-1,y-1,ie-Za-1,i Exponential decay for cod, haddock, whiting and saithe (T1.2) Total mortality Za,i=Ma,i+Fa,i+Pa,i M is the natural mortality. P=0 for saithe (T1.3) Recruitment at age 1 N1,y,i=αiSSBy-1,ie-βiSSBy-1,ieεi Ricker curve with lognormal process errors, εi∼Normal(0,σ2) ⁠. The SSB is given by SSBy,i=∑aNa,y,ima,iwa,i, where m is the proportion of mature fish and w the fish weight. (T1.4) Fishing mortality for fleet k Fa,i,k=sa,i,kEk Product of fleet selectivity s and effort index E (T1.5) Seal predation mortality Pa,i=sela,iqiG Product of seal selectivity sel ⁠, seal predation rate q and seal number G (T1.6) Biomass for the other fish species By+1,i=By,i+4msyiKiBy,i1-By,iKi-Ly,i Schaefer model where msy is the maximum sustainable yield and K the carrying capacity (T1.7) Fishing catches Ca,y,i,k=Fa,i,kZa,iNa,y,i(1-e-Za,i) Baranov equation. Catches by seals are calculated by replacing F by P in T1.7 (T1.8) Landings for age-structured stocks Ly,i,k=∑aλa,i,kCa,y,i,k λ is the proportion of landings in the total catch (T1.9) Landings for the other species Ly,i,k=(1-e-Fi,k)By,i Baranov equation for biomass assuming F=Z (T1.10) Fishing revenues Ry,k=∑i(piLy,i,k) Product of fish landings and price p (T1.11) Fleet total cost c*k=v(cvk+cfk) Sum of the variable costs cv and the fixed costs cf per vessel multiplied by the number of vessels v ⁠. The variable costs are proportional to fleet effort using a constant ρ such as cvk=ρkEk (T1.12) Fleet net profit πy,k=Ry,k-c*k Number . Name . Equation . Comments . (T1.1) Fish abundance at age a and year y for species i Na,y,i=Na-1,y-1,ie-Za-1,i Exponential decay for cod, haddock, whiting and saithe (T1.2) Total mortality Za,i=Ma,i+Fa,i+Pa,i M is the natural mortality. P=0 for saithe (T1.3) Recruitment at age 1 N1,y,i=αiSSBy-1,ie-βiSSBy-1,ieεi Ricker curve with lognormal process errors, εi∼Normal(0,σ2) ⁠. The SSB is given by SSBy,i=∑aNa,y,ima,iwa,i, where m is the proportion of mature fish and w the fish weight. (T1.4) Fishing mortality for fleet k Fa,i,k=sa,i,kEk Product of fleet selectivity s and effort index E (T1.5) Seal predation mortality Pa,i=sela,iqiG Product of seal selectivity sel ⁠, seal predation rate q and seal number G (T1.6) Biomass for the other fish species By+1,i=By,i+4msyiKiBy,i1-By,iKi-Ly,i Schaefer model where msy is the maximum sustainable yield and K the carrying capacity (T1.7) Fishing catches Ca,y,i,k=Fa,i,kZa,iNa,y,i(1-e-Za,i) Baranov equation. Catches by seals are calculated by replacing F by P in T1.7 (T1.8) Landings for age-structured stocks Ly,i,k=∑aλa,i,kCa,y,i,k λ is the proportion of landings in the total catch (T1.9) Landings for the other species Ly,i,k=(1-e-Fi,k)By,i Baranov equation for biomass assuming F=Z (T1.10) Fishing revenues Ry,k=∑i(piLy,i,k) Product of fish landings and price p (T1.11) Fleet total cost c*k=v(cvk+cfk) Sum of the variable costs cv and the fixed costs cf per vessel multiplied by the number of vessels v ⁠. The variable costs are proportional to fleet effort using a constant ρ such as cvk=ρkEk (T1.12) Fleet net profit πy,k=Ry,k-c*k Open in new tab Table 1. Equations used in the simulation model. Number . Name . Equation . Comments . (T1.1) Fish abundance at age a and year y for species i Na,y,i=Na-1,y-1,ie-Za-1,i Exponential decay for cod, haddock, whiting and saithe (T1.2) Total mortality Za,i=Ma,i+Fa,i+Pa,i M is the natural mortality. P=0 for saithe (T1.3) Recruitment at age 1 N1,y,i=αiSSBy-1,ie-βiSSBy-1,ieεi Ricker curve with lognormal process errors, εi∼Normal(0,σ2) ⁠. The SSB is given by SSBy,i=∑aNa,y,ima,iwa,i, where m is the proportion of mature fish and w the fish weight. (T1.4) Fishing mortality for fleet k Fa,i,k=sa,i,kEk Product of fleet selectivity s and effort index E (T1.5) Seal predation mortality Pa,i=sela,iqiG Product of seal selectivity sel ⁠, seal predation rate q and seal number G (T1.6) Biomass for the other fish species By+1,i=By,i+4msyiKiBy,i1-By,iKi-Ly,i Schaefer model where msy is the maximum sustainable yield and K the carrying capacity (T1.7) Fishing catches Ca,y,i,k=Fa,i,kZa,iNa,y,i(1-e-Za,i) Baranov equation. Catches by seals are calculated by replacing F by P in T1.7 (T1.8) Landings for age-structured stocks Ly,i,k=∑aλa,i,kCa,y,i,k λ is the proportion of landings in the total catch (T1.9) Landings for the other species Ly,i,k=(1-e-Fi,k)By,i Baranov equation for biomass assuming F=Z (T1.10) Fishing revenues Ry,k=∑i(piLy,i,k) Product of fish landings and price p (T1.11) Fleet total cost c*k=v(cvk+cfk) Sum of the variable costs cv and the fixed costs cf per vessel multiplied by the number of vessels v ⁠. The variable costs are proportional to fleet effort using a constant ρ such as cvk=ρkEk (T1.12) Fleet net profit πy,k=Ry,k-c*k Number . Name . Equation . Comments . (T1.1) Fish abundance at age a and year y for species i Na,y,i=Na-1,y-1,ie-Za-1,i Exponential decay for cod, haddock, whiting and saithe (T1.2) Total mortality Za,i=Ma,i+Fa,i+Pa,i M is the natural mortality. P=0 for saithe (T1.3) Recruitment at age 1 N1,y,i=αiSSBy-1,ie-βiSSBy-1,ieεi Ricker curve with lognormal process errors, εi∼Normal(0,σ2) ⁠. The SSB is given by SSBy,i=∑aNa,y,ima,iwa,i, where m is the proportion of mature fish and w the fish weight. (T1.4) Fishing mortality for fleet k Fa,i,k=sa,i,kEk Product of fleet selectivity s and effort index E (T1.5) Seal predation mortality Pa,i=sela,iqiG Product of seal selectivity sel ⁠, seal predation rate q and seal number G (T1.6) Biomass for the other fish species By+1,i=By,i+4msyiKiBy,i1-By,iKi-Ly,i Schaefer model where msy is the maximum sustainable yield and K the carrying capacity (T1.7) Fishing catches Ca,y,i,k=Fa,i,kZa,iNa,y,i(1-e-Za,i) Baranov equation. Catches by seals are calculated by replacing F by P in T1.7 (T1.8) Landings for age-structured stocks Ly,i,k=∑aλa,i,kCa,y,i,k λ is the proportion of landings in the total catch (T1.9) Landings for the other species Ly,i,k=(1-e-Fi,k)By,i Baranov equation for biomass assuming F=Z (T1.10) Fishing revenues Ry,k=∑i(piLy,i,k) Product of fish landings and price p (T1.11) Fleet total cost c*k=v(cvk+cfk) Sum of the variable costs cv and the fixed costs cf per vessel multiplied by the number of vessels v ⁠. The variable costs are proportional to fleet effort using a constant ρ such as cvk=ρkEk (T1.12) Fleet net profit πy,k=Ry,k-c*k Open in new tab For the other fish species with no age-structured data available, a Schaefer surplus production function is used (Schaefer, 1954) following the formulation of Fletcher (1978) [Equation (T1.6)]. This describes the stock biomass dynamics in terms of carrying capacity (⁠ K ⁠) and maximum sustainable yield (⁠ msy ⁠). Catches for age-structured stocks are calculated from the Baranov (1918) Equation (T1.7) and partitioned into landings and discards (T1.8) while, for other species, landings are approximated directly from the biomass using Equation (T1.9). This equation corresponds to the Baranov catch equation for biomass assuming F=Z and provides an adequate approximation when F is large compared with M ⁠. For these other species, only the landings are modelled because the discard rates are low (Heath et al., 2015). Fleet revenues are obtained by multiplying landings by fish price (T1.10). Fleet costs are estimated following a cost function (T1.11). Variable costs are assumed proportional to fishing effort. Both the variable costs per vessel (cv) and the fixed costs (⁠ cf ⁠) are held constant in the model. The fleet net profits are calculated by taking the difference between fleet revenues and costs (T1.12). Modelled species and fleets For simplicity, species, in rank order by value that, along with cod, haddock and whiting, represent over 95% of the revenues of the UK demersal trawlers fishing in Division 6a (STECF, 2016a) were considered in the simulation model. These are saithe (Pollachius virens), anglerfish (Lophius sp.), megrim (Lepidorhombus spp.), European hake (Merluccius merluccius), ling (Molva molva), and Nephrops. Of these species, cod, haddock, whiting, ling, and saithe account for the greatest proportion of the grey seal diet (Harris, 2007). However, although the saithe biomass consumed by seals is of a comparable scale to whiting, it is a very small fraction of the saithe stock biomass (ICES, 2015b), while ling accounts for a very small part of the UK commercial catch (ICES, 2016b). Hence seal predation is considered only for cod, haddock and whiting. No trophic interaction is considered between fish species. Five fleets were selected based on definitions used by ICES (2015a) and are shown in Table 2. The fleets are identified by mesh size and by vessel length class. The “Others” fleet corresponds to all other gears used in UK fisheries in Division 6a and all foreign vessels catching cod, haddock, and whiting. Table 2. Fleets considered in the simulation model and their characteristics. Fleet code . Definition . Vessel length (m) . Net mesh size (mm) . Target species . Number of vessels . Variable costs (£’000) . Fixed costs (£’000) . TR1_10–24 Small UK whitefish trawlers 10–24 ≥120 Demersal whitefish 9 430.5 213.0 TR1>24 Large UK whitefish trawlers ≥24 ≥120 Demersal whitefish 10 1250.8 467.3 TR2<10 Small UK Nephrops trawlers <10 70–99 Nephrops 31 47.6 27.0 TR2_10–24 Large UK Nephrops trawlers 10–24 70–99 Nephrops 151 137.7 73.0 Others Other gear and foreign vessels All All Demersal whitefish, Nephrops 19 1236.3 618.1 Fleet code . Definition . Vessel length (m) . Net mesh size (mm) . Target species . Number of vessels . Variable costs (£’000) . Fixed costs (£’000) . TR1_10–24 Small UK whitefish trawlers 10–24 ≥120 Demersal whitefish 9 430.5 213.0 TR1>24 Large UK whitefish trawlers ≥24 ≥120 Demersal whitefish 10 1250.8 467.3 TR2<10 Small UK Nephrops trawlers <10 70–99 Nephrops 31 47.6 27.0 TR2_10–24 Large UK Nephrops trawlers 10–24 70–99 Nephrops 151 137.7 73.0 Others Other gear and foreign vessels All All Demersal whitefish, Nephrops 19 1236.3 618.1 The number of vessels and their associated annual costs per vessel are mean values for the years 2007–2011 obtained from Seafish. Open in new tab Table 2. Fleets considered in the simulation model and their characteristics. Fleet code . Definition . Vessel length (m) . Net mesh size (mm) . Target species . Number of vessels . Variable costs (£’000) . Fixed costs (£’000) . TR1_10–24 Small UK whitefish trawlers 10–24 ≥120 Demersal whitefish 9 430.5 213.0 TR1>24 Large UK whitefish trawlers ≥24 ≥120 Demersal whitefish 10 1250.8 467.3 TR2<10 Small UK Nephrops trawlers <10 70–99 Nephrops 31 47.6 27.0 TR2_10–24 Large UK Nephrops trawlers 10–24 70–99 Nephrops 151 137.7 73.0 Others Other gear and foreign vessels All All Demersal whitefish, Nephrops 19 1236.3 618.1 Fleet code . Definition . Vessel length (m) . Net mesh size (mm) . Target species . Number of vessels . Variable costs (£’000) . Fixed costs (£’000) . TR1_10–24 Small UK whitefish trawlers 10–24 ≥120 Demersal whitefish 9 430.5 213.0 TR1>24 Large UK whitefish trawlers ≥24 ≥120 Demersal whitefish 10 1250.8 467.3 TR2<10 Small UK Nephrops trawlers <10 70–99 Nephrops 31 47.6 27.0 TR2_10–24 Large UK Nephrops trawlers 10–24 70–99 Nephrops 151 137.7 73.0 Others Other gear and foreign vessels All All Demersal whitefish, Nephrops 19 1236.3 618.1 The number of vessels and their associated annual costs per vessel are mean values for the years 2007–2011 obtained from Seafish. Open in new tab Parameterization Age-structured stock dynamics For cod, haddock, and whiting, we used the age-structured stock assessment model described by Trijoulet et al. (2017) to provide estimates of the main input parameters. The model was fitted to the ICES stock assessment data (ICES, 2013) augmented with age compositions in seal diet derived from Harris (2007) and seal population size from Thomas (2013). Outputs from these analyses include a time series of fishing mortality, natural mortality, seal selectivity, seal predation rate, recruitment, and spawning stock biomass (SSB) that are provided in Supplementary Material. For saithe, the input values were taken from ICES (2013). Other species dynamics For the other species, those without a full age-based assessment, the Schaefer surplus production model was fitted by least squares to the biomass data from ICES reports (ICES, 2013, 2014) to obtain values for msy and K [Equation (T1.6)]. The landings were treated as known, error free, values. The status quo fishing mortality for these species was estimated using the average biomass and landings between 2007 and 2011 using equation (T1.9). No biomass estimates are available for ling and the landings were almost constant over the past 10 years. For simplicity, we assumed that ling landings scaled linearly with effort. Average landings between 2007 and 2011 were partitioned by fleet and assumed to correspond to an effort index of E=1 ⁠. Input values for the other species are given in Supplementary Material. Fishing selectivity by fleet Fleet specific catch data were used to partition the fishing mortality at age by fleet for the age-structured stocks. Total fishing mortality for the other species was partitioned down to fleet level by using the proportion of the fleet catch in the total catch. This is described in more detail in the Supplementary Material. Economic parameters Cost and revenue data for the years 2007–2011 for the four UK fleets were made available by the UK agency Seafish, and were corrected for inflation using the gross domestic product deflator with 2012 as the reference year. Economic data are usually aggregated for the North Sea and the West of Scotland (Anderson et al., 2013), so for this study, the West of Scotland data have been extracted by identifying the vessels that spend the majority of their time in Division 6a. Here, it is assumed that costs incurred due to fuel, crew share and other fishing costs are variable and that total vessel outlay, depreciation, interest and other financing expenses are fixed costs. Variable and fixed costs values used in the simulation model were averages over 2007–2011 to be consistent with the reference period used for the fish stock values. No cost data are available for the “Others” fleet. We assumed that this fleet was operating at the break-even point during the reference period 2007–2011 and used the revenues to estimate the costs. Within the UK fleets, average fixed costs per vessel are typically around half of the average variable costs. The total aggregated costs for “Others” was scaled to the number of vessels (all assumed foreign vessels), and partitioned using this ratio. The costs and the number of vessels for all fleets are summarized in Table 2. The price of fish in the West of Scotland is dictated by the European market (Scottish Fishermen’s Organization, 2016), which means a change in the quantity of local landings has little effect on fish prices. As a result, the fish prices are assumed to be constant for each species in the simulation model. They correspond to fixed average real prices between 2007 and 2011 taken from Marine Management Organization (2012) and are shown in Table 3. Table 3. Average fish price (⁠ p ⁠) per tonne (2007–2011) for the nine fish species considered in the simulation model and proportion of the total catch made by the UK vessels for indication. Species . p(£’000) . % of total catch by UK vessels . Cod 2.1 53 Haddock 1.2 76 Whiting 1.1 74 Saithe 0.8 43 Anglerfish 3.2 33 Megrim 3.0 54 Hake 1.9 26 Ling 1.4 32 Nephrops 2.9 99 Species . p(£’000) . % of total catch by UK vessels . Cod 2.1 53 Haddock 1.2 76 Whiting 1.1 74 Saithe 0.8 43 Anglerfish 3.2 33 Megrim 3.0 54 Hake 1.9 26 Ling 1.4 32 Nephrops 2.9 99 Open in new tab Table 3. Average fish price (⁠ p ⁠) per tonne (2007–2011) for the nine fish species considered in the simulation model and proportion of the total catch made by the UK vessels for indication. Species . p(£’000) . % of total catch by UK vessels . Cod 2.1 53 Haddock 1.2 76 Whiting 1.1 74 Saithe 0.8 43 Anglerfish 3.2 33 Megrim 3.0 54 Hake 1.9 26 Ling 1.4 32 Nephrops 2.9 99 Species . p(£’000) . % of total catch by UK vessels . Cod 2.1 53 Haddock 1.2 76 Whiting 1.1 74 Saithe 0.8 43 Anglerfish 3.2 33 Megrim 3.0 54 Hake 1.9 26 Ling 1.4 32 Nephrops 2.9 99 Open in new tab Equilibrium fishing scenarios Modelling regulations and fisher choices in the West of Scotland is complex. For simplicity, we chose to run the simulation model under equilibrium scenarios, which correspond to three different fishing or regulation strategies. This allows the comparison of grey seal impacts in contrasting scenarios to test the sensitivity of the results. The three scenarios “status quo F (SQF)”, “bioeconomic equilibrium (BE)”, and “maximum economic yield (MEY)” are outlined below. All the scenarios consider the impact of seal predation on fishing revenues and profitability under biological equilibrium conditions when the nine species considered show no change in mean SSB. The results presented are averages from the process error around recruitment over 50 years when SSB is at equilibrium. The SQF scenario keeps the fishing mortality at the base level constant (i.e. E=1 ⁠). It results in a biological equilibrium that assumes fleet behaviour does not respond to economic incentives. This scenario serves as a reference case for comparison with the other scenarios where fleet behaviour is dynamic and varies with the fleet net profit. The BE scenario assesses the impact of seal predation in the extreme open-access case where no regulation exists and vessels can enter or exit the fishery freely. Classical economic theory shows that, in this environment, fishers act independently and try to maximise their individual profit so that, in the long-term, the fishery tends to the bioeconomic equilibrium where total revenues equal total costs (Knowler, 2002). In this scenario, each UK fleet can invest or disinvest in effort or number of vessels following the value of its net profit. Given the value of the fleet net profit at the initial biological equilibrium [Equation (T1.12)], fishing effort is adjusted and the model run to the new biological equilibrium. This process is then repeated until the BE is reached. It is assumed that higher net profit will lead to larger investment in the number of vessels and effort per fleet. The MEY scenario represents the economic equilibrium assuming the fishery is closed to new entrants and the fleet composition is fixed. The fleets are assumed to collaborate to obtain a sustainable fishery where the aggregated fishery net profit is maximized at the equilibrium (Guillen et al., 2013). The goal is to determine the level of effort per fleet, which maximises the total fishery net profit. Because the cost function for the “Others” fleet is uncertain due to the lack of economic data for this fleet, its effort is kept constant in both the BE and MEY models so the fleet cannot modify its fishing behaviour with its net profit. Additional information on equilibrium scenarios is given in Supplementary Material. Seal predation scenarios Fleet revenues were compared at different levels of seal predation mortality (⁠ P ⁠). Scaling factors of 0.7–1.3 in steps of 0.1 were applied to the equation for P [Equation (T1.5)] in the three equilibrium scenarios. The scale range is limited to ±30% to avoid unrealistic departures from the current state. Assuming seal selectivity (⁠ sel ⁠) and predation rate (⁠ q ⁠) are more or less constant, applying a scaling factor to P corresponds to a change in seal population (⁠ G ⁠). In this study, the predation rate is assumed constant by default for all scenarios. However, q may be time varying especially if it is related to prey abundance such as in a functional response (Holling, 1959) and this is considered in the sensitivity analysis described below. In order to quantify the impact of a single seal on the fishery and on the fleet most affected by seal predation, we calculated the change in revenue per seal and the change in revenue per vessel when seal predation is changed by 10%. The change in revenue per seal is calculated as the difference between fishing revenues at the baseline number of seals and at increased/decreased seal predation, divided by the number of seals that represents 10% of the population. In order to allow comparison with fleet revenues, the weight of fish consumed by seals was converted to equivalent “revenues” by multiplying it with fish prices. Consistency check and sensitivity analysis The main parameters of the model are derived from the average state of the fishery between 2007 and 2011. As a check for consistency, the landings for this period were estimated by the model using mean population sizes from stock assessments for the same period. The estimated landings were then compared with observed values and shown to be consistent (Supplementary Material). Sensitivity to the different assumptions in the simulation model was tested as follows: The model was run for two other commonly used stock-recruitment relationships to test robustness to the choice of curve. These were Beverton and Holt (1957) and the smooth hockey-stick (Froese, 2008). The parameter estimates of the Schaefer surplus production function msy and K [Equation (T1.6)] were increased separately by 10% for all species to investigate estimation errors. A type II functional response of seals to cod biomass was applied as an alternative foraging model to the constant predation rate assumption. This was based on the cod partial biomass as described in Cook and Trijoulet (2016). This response is not considered for haddock and whiting due to difficulties fitting a type II functional response (Trijoulet, 2016). The BE and MEY scenarios are run allowing the fleet “Others” to vary its effort at each iteration with its net profit to test the assumption of constant effort. A SQF scenario was run in the absence of cod to examine the sensitivity of the results to the species composition in the fishery in the event of a cod stock collapse (Cook and Trijoulet, 2016). The sensitivity of the simulation model to seal predation was analysed by calculating the difference in seal impacts on fishing revenues when the seal predation is increased by 10%, between the initial model set up and when the sensitivity tests 1–5 are applied. For simplicity, results for sensitivity tests 1–4 are shown for the fleet most affected by seal predation only. Results Bioeconomic results Changes to SSB in the three scenarios resulting from different levels of seal predation are shown in Figure 2. Cod is the most sensitive to a change in grey seal numbers followed by whiting. The estimated equilibrium haddock SSB is little changed in all three scenarios even for large changes in seal population. Figure 2. Open in new tabDownload slide Change in mean equilibrium SSB (%) for cod, haddock, and whiting in the three different scenarios for small (±10%) and large (±30%) changes in seal population. The change in revenues and net profit at different levels of seal population is shown in Figure 3. Larger whitefish vessels (TR1 > 24) are most affected by a change in grey seal population in all scenarios. For this fleet, in the dynamic scenarios (BE and MEY), the percentage change in revenues is much larger than the change in seal population. The smaller whitefish fleet (TR1_10–24) and the “Others” fleet are less affected. As expected, the Nephrops trawlers show little change since cod, haddock, and whiting represent a very low proportion of their revenues. Although individual fleets show large changes in revenues, when the whole fishery is considered, changes in seal predation of ±30% result in about 5% changes in revenue. This arises because Nephrops have a high value relative to other stocks and are unaffected by seal predation in the model. Figure 3. Open in new tabDownload slide Change in mean equilibrium revenues (%) or net profit (MEY scenario only) by fleet and for the entire fishery in the three different equilibrium scenarios for small (±10%) and large (±30%) changes in seal population. The MEY equilibrium is the only scenario where profits respond to seal predation. Here, the changes in net profit with seal predation are similar to the changes in revenues for all fleets except TR1 > 24, where the impact on the net profit is less than on the revenues (Figure 3). The value of the quantity of fish eaten by seals was compared with fleet revenues for the current number of seals in the Division 6a (Table 4). When revenues from cod, haddock, and whiting are compared (Table 4a), seal” revenues” only represent a small proportion (<0.5%) of the total revenues and this proportion is considerably smaller than the proportion for the whitefish fleets. Note that seal revenues of cod, haddock, and whiting can be larger than those of the TR2 < 10 fleet, but this arises because the fleet catches mainly Nephrops (Supplementary Figure S2). When seal revenues are compared with fleet revenues for all fish species combined (Table 4b), the value of seal predation is negligible since it represents <2% of each fleet revenue. Table 4. Comparison of fleet and seal revenues from cod, haddock, and whiting with that for seals under the three scenarios and at the baseline number of seals. (a) Revenue of cod, haddock, and whiting by fleet expressed as a proportion (%) of the total cod, haddock, and whiting revenue from all fleets including revenue from consumption by seals. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . Seals . SQF 12.90 54.81 0.07 5.23 26.70 0.29 BE 50.24 26.78 0.91 0.87 20.99 0.21 MEY 20.99 23.60 0.10 7.07 47.79 0.45 (a) Revenue of cod, haddock, and whiting by fleet expressed as a proportion (%) of the total cod, haddock, and whiting revenue from all fleets including revenue from consumption by seals. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . Seals . SQF 12.90 54.81 0.07 5.23 26.70 0.29 BE 50.24 26.78 0.91 0.87 20.99 0.21 MEY 20.99 23.60 0.10 7.07 47.79 0.45 (b) Revenue of cod, haddock, and whiting taken by seals expressed as a proportion (%) of the total fleet revenue including all species. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . SQF 0.46 0.19 1.22 0.10 0.10 BE 0.12 0.29 0.08 0.55 0.10 MEY 0.56 0.80 1.72 0.15 0.13 (b) Revenue of cod, haddock, and whiting taken by seals expressed as a proportion (%) of the total fleet revenue including all species. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . SQF 0.46 0.19 1.22 0.10 0.10 BE 0.12 0.29 0.08 0.55 0.10 MEY 0.56 0.80 1.72 0.15 0.13 The weight of fish consumed by seals is converted to seal “revenue” using fish price. Open in new tab Table 4. Comparison of fleet and seal revenues from cod, haddock, and whiting with that for seals under the three scenarios and at the baseline number of seals. (a) Revenue of cod, haddock, and whiting by fleet expressed as a proportion (%) of the total cod, haddock, and whiting revenue from all fleets including revenue from consumption by seals. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . Seals . SQF 12.90 54.81 0.07 5.23 26.70 0.29 BE 50.24 26.78 0.91 0.87 20.99 0.21 MEY 20.99 23.60 0.10 7.07 47.79 0.45 (a) Revenue of cod, haddock, and whiting by fleet expressed as a proportion (%) of the total cod, haddock, and whiting revenue from all fleets including revenue from consumption by seals. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . Seals . SQF 12.90 54.81 0.07 5.23 26.70 0.29 BE 50.24 26.78 0.91 0.87 20.99 0.21 MEY 20.99 23.60 0.10 7.07 47.79 0.45 (b) Revenue of cod, haddock, and whiting taken by seals expressed as a proportion (%) of the total fleet revenue including all species. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . SQF 0.46 0.19 1.22 0.10 0.10 BE 0.12 0.29 0.08 0.55 0.10 MEY 0.56 0.80 1.72 0.15 0.13 (b) Revenue of cod, haddock, and whiting taken by seals expressed as a proportion (%) of the total fleet revenue including all species. . Scenario . TR1_10–24 . TR1 > 24 . TR2 < 10 . TR2_10–24 . Others . SQF 0.46 0.19 1.22 0.10 0.10 BE 0.12 0.29 0.08 0.55 0.10 MEY 0.56 0.80 1.72 0.15 0.13 The weight of fish consumed by seals is converted to seal “revenue” using fish price. Open in new tab Table 5 shows the change in annual fishing revenues for a 10% change in seal population for the entire fishery and the TR1 > 24 fleet. Also shown is the “cost” per seal to the fishery or fleet. The results are of the same order of magnitude for all scenarios. For the TR1 > 24 fleet, the cost per seal is less than that for the fishery in all but one case but the cost per vessel is large as the losses are distributed among few vessels. For the whole fishery, the costs per vessel are lowest in the BE scenario because the Nephrops fleets expand to dissipate the profits. In contrast, for TR1 > 24, the costs per vessel are highest under this scenario (BE) because some vessels exit the fishery. Table 5. Change in annual fishing revenues (£’000) for the fishery and for TR1 > 24 following an increase or decrease in seal population of 10% (3204 individuals). Seal scenario . Equilibrium scenario . Fishery . TR1 > 24 . Whole . Per vessel . Per seal . Whole . Per vessel . Per seal . +10% SQF −1350 −6.13 −0.421 −715 −71.54 −0.223 BE −1618 −2.69 −0.505 −1289 −257.83 −0.402 MEY −1405 −6.39 −0.439 −903 −90.25 −0.282 −10% SQF 1414 6.43 0.441 763 76.32 0.238 BE 1456 2.41 0.454 1541 220.21 0.481 MEY 1601 7.28 0.500 1165 116.46 0.363 Seal scenario . Equilibrium scenario . Fishery . TR1 > 24 . Whole . Per vessel . Per seal . Whole . Per vessel . Per seal . +10% SQF −1350 −6.13 −0.421 −715 −71.54 −0.223 BE −1618 −2.69 −0.505 −1289 −257.83 −0.402 MEY −1405 −6.39 −0.439 −903 −90.25 −0.282 −10% SQF 1414 6.43 0.441 763 76.32 0.238 BE 1456 2.41 0.454 1541 220.21 0.481 MEY 1601 7.28 0.500 1165 116.46 0.363 The change is given at the level of the whole fishery or fleet, per vessel and per seal. Open in new tab Table 5. Change in annual fishing revenues (£’000) for the fishery and for TR1 > 24 following an increase or decrease in seal population of 10% (3204 individuals). Seal scenario . Equilibrium scenario . Fishery . TR1 > 24 . Whole . Per vessel . Per seal . Whole . Per vessel . Per seal . +10% SQF −1350 −6.13 −0.421 −715 −71.54 −0.223 BE −1618 −2.69 −0.505 −1289 −257.83 −0.402 MEY −1405 −6.39 −0.439 −903 −90.25 −0.282 −10% SQF 1414 6.43 0.441 763 76.32 0.238 BE 1456 2.41 0.454 1541 220.21 0.481 MEY 1601 7.28 0.500 1165 116.46 0.363 Seal scenario . Equilibrium scenario . Fishery . TR1 > 24 . Whole . Per vessel . Per seal . Whole . Per vessel . Per seal . +10% SQF −1350 −6.13 −0.421 −715 −71.54 −0.223 BE −1618 −2.69 −0.505 −1289 −257.83 −0.402 MEY −1405 −6.39 −0.439 −903 −90.25 −0.282 −10% SQF 1414 6.43 0.441 763 76.32 0.238 BE 1456 2.41 0.454 1541 220.21 0.481 MEY 1601 7.28 0.500 1165 116.46 0.363 The change is given at the level of the whole fishery or fleet, per vessel and per seal. Open in new tab Sensitivity analysis Table 6 shows the changes in grey seal impacts on TR1 > 24 for the different sensitivity scenarios. The three fishery scenarios show little change for all sensitivity tests except for the seal foraging model. Here a type II functional response for cod has a large effect. Overall, the dynamic scenarios show greater sensitivity than the SQF scenario. Table 6. Sensitivity of the three scenarios expressed as the change in seal impacts on TR1 > 24 revenues (%) for an increase in seal population of 10%. Sensitivity test . Sensitivity to the . Change considered . SQF . BE . MEY . 1 Ricker stock-recruitment model Beverton–Holt 0.0 4.1 0.0 Hockey-stick −0.1 2.5 3.5 2 Schaefer parameters msy+10% −0.2 −0.1 −6.2 K+10% 0.0 5.0 0.6 3 Constant seal predation rate Type II seal functional response to cod biomass 10.7 23.7 10.7 4 Constant effort for “Others” Effort can vary with fleet net profit None −0.6 −2.5 Sensitivity test . Sensitivity to the . Change considered . SQF . BE . MEY . 1 Ricker stock-recruitment model Beverton–Holt 0.0 4.1 0.0 Hockey-stick −0.1 2.5 3.5 2 Schaefer parameters msy+10% −0.2 −0.1 −6.2 K+10% 0.0 5.0 0.6 3 Constant seal predation rate Type II seal functional response to cod biomass 10.7 23.7 10.7 4 Constant effort for “Others” Effort can vary with fleet net profit None −0.6 −2.5 The change in impacts is calculated by taking the difference between changes in revenues for the initial simulation results and changes in revenues for the sensitivity test results. For instance, a value of 4.1 (BE scenario, sensitivity test 1) means that seal impacts on the fleet revenues are increased by 4.1% when a Beverton–Holt stock recruitment relationship is used compared with a Ricker relationship. Open in new tab Table 6. Sensitivity of the three scenarios expressed as the change in seal impacts on TR1 > 24 revenues (%) for an increase in seal population of 10%. Sensitivity test . Sensitivity to the . Change considered . SQF . BE . MEY . 1 Ricker stock-recruitment model Beverton–Holt 0.0 4.1 0.0 Hockey-stick −0.1 2.5 3.5 2 Schaefer parameters msy+10% −0.2 −0.1 −6.2 K+10% 0.0 5.0 0.6 3 Constant seal predation rate Type II seal functional response to cod biomass 10.7 23.7 10.7 4 Constant effort for “Others” Effort can vary with fleet net profit None −0.6 −2.5 Sensitivity test . Sensitivity to the . Change considered . SQF . BE . MEY . 1 Ricker stock-recruitment model Beverton–Holt 0.0 4.1 0.0 Hockey-stick −0.1 2.5 3.5 2 Schaefer parameters msy+10% −0.2 −0.1 −6.2 K+10% 0.0 5.0 0.6 3 Constant seal predation rate Type II seal functional response to cod biomass 10.7 23.7 10.7 4 Constant effort for “Others” Effort can vary with fleet net profit None −0.6 −2.5 The change in impacts is calculated by taking the difference between changes in revenues for the initial simulation results and changes in revenues for the sensitivity test results. For instance, a value of 4.1 (BE scenario, sensitivity test 1) means that seal impacts on the fleet revenues are increased by 4.1% when a Beverton–Holt stock recruitment relationship is used compared with a Ricker relationship. Open in new tab The impact of grey seals on all fleet revenues, and therefore, the whole fishery is substantially reduced if the cod stock collapses (Figure 4). Even reducing the seal population by 30% only increases the revenues of TR1 > 24, the most affected fleet, by <3%. Figure 4. Open in new tabDownload slide Change in revenues (%) by fleet and for the entire fishery for a small (±10%) and large (±30%) change in seal population in the initial SQF scenario and for the SQF scenario in the absence of cod. Discussion In the model, an increase in grey seal predation resulted in a clear decrease in the cod and whiting stocks. However, even large changes in grey seal predation have little impact on the haddock biomass. This is partly because the predation mortality on haddock is low compared with fishing mortality and also because seals show very low selectivity on the younger ages, which contribute most to the stock biomass. This study suggests that the impact of seal predation on the haddock stock is likely to be low. Cod is the key stock in evaluating the impacts of seal predation on the demersal fishery. Seal predation mortalities are much greater on cod than haddock and whiting (Trijoulet et al., 2017) so seal predation effects are more substantial for this stock. In addition, the price per tonne of cod is roughly twice that of haddock and whiting, so cod make a proportionately larger contribution to the revenues. The three scenarios, SQF, BE, and MEY, represent very different fishing strategies but a clear pattern emerges that the larger whitefish trawlers (TR1 > 24) are most sensitive to the effects of seal predation (mainly on revenues, less so on profits) and that this is largely due to revenues accruing from cod. In the scenario where the cod stock has collapsed, although the TR1 > 24 fleet still shows the greatest effects of seal predation, the impact is substantially reduced. For the TR1_10–24 fleet, whitefish are a principal target, yet Nephrops makes a significant contribution to the catches. As Nephrops is nearly twice as valuable as cod, the revenues of this fleet are less sensitive to cod biomass and any seal predation on it. Not surprisingly, the TR2 fleets that target Nephrops are little affected by seal predation. Overall, the value of fish caught by seals is low in comparison to the fleet revenues and seal predation impacts are relatively small at the level of the whole fishery because Nephrops dominates the value of the total landings. We chose a number of fishing scenarios to explore whether seal predation effects were sensitive to contrasting fleet behaviour. While none represent the current fishery accurately they show similar effects that may characterize, qualitatively, what may occur in reality. The SQF scenario shows the smallest effects of predation while both the BE and MEY scenarios show substantially greater sensitivity to seals. Both of these scenarios allow vessels to adapt their fishing strategy in response to economic incentives and such behaviour appears to magnify the effects of seal predation. Current estimates of the economic performance of the fleets suggest that they are operating close to BE (Lawrence et al., 2016), a scenario which heightens sensitivity to seal predation compared with SQF and reduces it compared with MEY. However, the magnitude of the change in revenues due to increased seal predation is much more sensitive to the population model assumptions (stock recruitment function, seal functional response, etc.) in the dynamic fishing scenarios. The results of the BE and MEY scenarios should therefore be treated as more uncertain than when fishing at SQF. For all scenarios, a small change in grey seal population of ±10% did not show substantial variations in fleet revenues and the results appear relatively robust to most model assumptions, with the possible exception of seal functional response to cod biomass. The type II functional response results show that an alternative seal foraging model may alter the results significantly. The effect of the response is to accelerate decline when stocks are already declining and similarly accelerate increase when stock are increasing. Inevitably this will contribute to greater sensitivity to seal predation as the effect is inversely density dependent. This highlights the need for a more realistic seal foraging model. Depredation and seal-induced infections are a different source of impact that would need to be added to predation effects to get a more complete estimate of the economic effects of seals. There have been a number of studies estimating the cost of seal-induced infections and depredation. These give an annual cost between £300 and £4800 per fisher or processor (Bjørge et al., 1981; Bosetti and Pearce, 2003; Butler et al., 2011) and a corresponding cost per seal between £15 and £290. Given the estimates of cost of seal predation in the West of Scotland from this study, it would suggest the costs including depredation could be as high as £700 per seal. Although seals may represent a cost to the fishery, they may support positive benefits to the economy from activities such as ecotourism. Grey seals are the third most popular wildlife attraction in Scotland after cetaceans and seabirds (Woods-Ballard et al., 2003). In the West of Scotland, tourism gains from whale and seal-watching have been estimated at around £1.8 million in 2001 and the indirect income from other tourism attractions during the visitor stay can reach £7.8 million per year (Warburton et al., 2001). Consequently, it can be argued that even if grey seals represent only a portion of these gains, grey seal presence may be more beneficial than harmful to the Scottish economy. However, these gains do not benefit the fishers that suffer the costs. Our model does not consider predatory interactions other than that of seals on three major species. Seabirds and cetaceans are also responsible for removal of large quantities of commercial fish (Overholtz and Link, 2007) and the largest predation on demersal fish comes from predatory fish themselves (Sparholt, 1994; Engelhard et al., 2014). Incorporating trophic interactions is likely to have a minor effect on the estimated direction of change seen from the model given that this study investigates the sensitivity to seal predation under average conditions. The results describe the relative impacts of seal predation on the different fleets under various exploitation scenarios rather than predict actual revenues and profit in the long-term. There are a number of additional reasons for treating the results presented here with caution. Seal predation mortality was estimated using only 2 years of seal diet data (Harris, 2007) that are themselves highly uncertain. This should not have a major impact on the qualitative impact of seals on the different fleets and fish stocks but may cause uncertainty in its magnitude. In addition, this study also makes the assumption that the fish population is homogeneous and equally available to seals and fishers, which are in direct competition with each other. Currently, the majority of cod landings are taken in the far north of Division 6.a and along the continental shelf edge (STECF, 2016b) while seal foraging mostly occurs on the continental shelf (Jones et al., 2015) including areas considered unsuitable for trawl fishing (Marine Environmental Mapping Programme, 2015). Seals may therefore predate on fish, which are not directly available to fishers and although the absence of overlap between fishing and foraging zones does not mean the absence of competition, the interaction between seals and fishers is likely to be more complex than assumed here. This has potential to bias resulting model estimates and is an issue that requires further investigation. Conclusion Overall, seal predation effects on revenues are small at the whole fishery scale. The TR1 > 24 fleet is the most sensitive to seal predation, and this is primarily due to the importance of cod in its catch. It seems, therefore that the importance of the seal-fishery interaction in the West of Scotland is limited to one major fleet and stock. However, assessing the significance of this interaction is heavily dependent on the assumption of the seal foraging model and is an area in need of further research. Supplementary material Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements This work was supported by funds from the University of Strathclyde, Marine Scotland and MASTS through the Scottish Funding Council (grant 388 HR09011). We thank Alex Dickson for his suggestions on the economic part of the model. We also thank Steve Lawrence and John Anderson for extracting the economic data used in this study. References Alexander K. A. , Heymans J. J., Magill S., Tomczak M. T., Holmes S. J., Wilding T. A. 2015 . Investigating the recent decline in gadoid stocks in the west of Scotland shelf ecosystem using a foodweb model . ICES Journal of Marine Science , 72 : 436 – 449 . Google Scholar Crossref Search ADS WorldCat Amante C. , Eakins B. W. 2009 . ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24. National Geophysical Data Center, NOAA. Anderson J. , Stewart A.-M., Curtis H., Dawe R. 2013 . 2011 Economic Survey of the UK Fishing Fleet, Key Features . Seafish Economics. Edinburgh, pp. 1–26. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Baranov F. I. 1918 . On the question of the biological basis of fisheries . Nauchnge Issledovaniya Ikhtiologicheskii Instituta Izvestiya , 1 : 81 – 128 . Google Scholar OpenURL Placeholder Text WorldCat Beverton R. J. H. , Holt S. J. 1957 . On the dynamics of exploited fish populations , Springer Netherlands, Dordrecht . 538 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bjørge A. , Christensen I., Øritsland T. 1981 . Current Problems and Research Related to Interactions Between Marine Mammals and Fisheries in Norwegian Coastal and Adjacent Waters. ICES Document 1981/N: 18. 10 pp. Bosetti V. , Pearce D. 2003 . A study of environmental conflict: the economic value of Grey Seals in southwest England . Biodiversity and Conservation , 12 : 2361 – 2392 . Google Scholar Crossref Search ADS WorldCat Butler J. R. A. , Middlemas S. J., Graham I. M., Harris R. N. 2011 . Perceptions and costs of seal impacts on Atlantic salmon fisheries in the Moray Firth, Scotland: Implications for the adaptive co-management of seal-fishery conflict . Marine Policy , 35 : 317 – 323 . Google Scholar Crossref Search ADS WorldCat Cook R. M. , Holmes S. J., Fryer R. J. 2015 . Grey seal predation impairs recovery of an over-exploited fish stock . Journal of Applied Ecology , 52 : 969 – 979 . Google Scholar Crossref Search ADS WorldCat Cook R. M. , Trijoulet V. 2016 . The effects of grey seal predation and commercial fishing on the recovery of a depleted cod stock . Canadian Journal of Fisheries and Aquatic Sciences , 73 : 1319 – 1329 . Google Scholar Crossref Search ADS WorldCat Cronin M. , Jessopp M., Houle J., Reid D. 2014 . Fishery-seal interactions in Irish waters: current perspectives and future research priorities . Marine Policy , 44 : 120 – 130 . Google Scholar Crossref Search ADS WorldCat Engelhard G. H. , Peck M. A., Rindorf A., C. Smout S., van Deurs M., Raab K., Andersen K. H., et al. 2014 . Forage fish, their fisheries, and their predators: who drives whom? ICES Journal of Marine Science , 71 : 90 – 104 . Google Scholar Crossref Search ADS WorldCat Fernandes P. G. , Cook R. M. 2013 . Reversal of fish stock decline in the Northeast Atlantic . Current Biology , 23 : 1432 – 1437 . Google Scholar Crossref Search ADS PubMed WorldCat Fletcher R. I. 1978 . Time-dependent solutions and efficient parameters for stock-production models . Fishery Bulletin , 76 : 377 – 388 . Google Scholar OpenURL Placeholder Text WorldCat Froese R. 2008 . The continuous smooth hockey stick: a newly proposed spawner‐recruitment model . Journal of Applied Ichthyology , 24 : 703 – 704 . Google Scholar Crossref Search ADS WorldCat Gruber C. P. 2014 . Social, Economic, and Spatial Perceptions of Gray Seal (Halichoerus grypus) Interactions with Commercial Fisheries in Cape Cod. Master's project, Duke University, MA. 68 pp. Guillen J. , Macher C., Merzéréaud M., Bertignac M., Fifas S., Guyader O. 2013 . Estimating MSY and MEY in multi-species and multi-fleet fisheries, consequences and limits: an application to the Bay of Biscay mixed fishery . Marine Policy , 40 : 64 – 74 . Google Scholar Crossref Search ADS WorldCat Hammond P. S. , Grellier R., Harris R. N. 2006 . Grey seal diet composition and prey consumption in the North Sea and west of Scotland. SCOS Briefing Paper 06/06. pp. 79–81. Harris R. N. 2007 . Assessing grey seal (Halichoerus grypus) diet in western Scotland, Thesis. NERC Sea Mammal Research Unit (SMRU) Theses. University of St Andrews, St Andrews. Heath M. R. , Cook R. M., Lee S. Y. 2015 . Hind-casting the quantity and composition of discards by mixed demersal fisheries in the North Sea . PLoS One , 10 : e0117078 . Google Scholar Crossref Search ADS PubMed WorldCat Hilborn R. , Walters C. J. 1992 . Quantitative Fisheries Stock Assessment: Choice, Dynamics and Uncertainty . Chapman and Hall, New York . 570 pp. Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Holling C. S. 1959 . The components of predation as revealed by a study of small-mammal predation of the European pine sawfly . The Canadian Entomologist , 91 : 293 – 320 . Google Scholar Crossref Search ADS WorldCat Holma M. , Lindroos M., Oinonen S. 2014 . The economics of conflicting interests: northern Baltic salmon fishery adaption to gray seal abundance . Natural Resource Modeling , 27 : 275 – 299 . Google Scholar Crossref Search ADS WorldCat Holmes S. J. 2008 . Seal Predation on VIa Cod and its Effect on the Assessment of the VIa Cod Stock. ICES WD 6: Working Documents of the Working Group on the Assessment of Northern Shelf Demersal Stocks (WGNSDS) 2008. pp. 49 – 77 . Holmes S. J. , Fryer R. J. 2011 . Significance of seal feeding on cod west of Scotland: results from a state space stock assessment model. ICES CM 2011/I:22 Hutchings J. A. , Myers R. A. 1994 . What can be learned from the collapse of a renewable resource? Atlantic cod, Gadus morhua, of Newfoundland and Labrador . Canadian Journal of Fisheries and Aquatic Sciences , 51 : 2126 – 2146 . Google Scholar Crossref Search ADS WorldCat ICES . 2013 . Report of the Working Group for Celtic Seas Ecoregion (WGCSE). 8–17 May 2013, Copenhagen, Denmark. ICES CM 2013/ACOM: 12. 1986 pp. ICES . 2014 . Report of the Working Group for the Bay of Biscay and the Iberian waters Ecoregion (WGBIE), 7–13 May 2014, Lisbon, Portugal. ICES CM 2014/ACOM: 11. 714 pp. ICES . 2015a . Report of the Working Group on Mixed Fisheries Advice (WGMIXFISH-ADVICE). 25–29 May 2015, Copenhagen, Denmark. ICES CM 2015/ACOM: 21. 171 pp. ICES . 2015b . Report of the Working Group on the Assessment of Demersal Stocks in the North Sea and Skagerrak (WGNSSK). 28 April-7 May, Copenhagen, Denmark. ICES CM 2015/ACOM: 13. 1031 pp. ICES . 2016a . Report of the Working Group for the Celtic Seas Ecoregion (WGCSE). 12–21 May 2015, Copenhagen, Denmark. ICES CM 2015/ACOM: 12. 1432 pp. ICES . 2016b . Report of the Working Group on the Biology and Assessment of Deep-sea Fisheries Resources (WGDEEP). 692 pp. 20–27 April 2016, Copenhagen, Denmark. ICES CM 2016/ACOM:18. 692 pp. Jones E. L. , Smout S., McConnell B. J. 2015 . Determine environmental covariates for usage preference around the UK. Sea Mammal Research Unit, University of St Andrews, Report to Scottish Government, St Andrews. MR 5.1. 37–49 pp. Knowler D. 2002 . A review of selected bioeconomic models with environmental influences in fisheries . Journal of Bioeconomics , 4 : 163 – 181 . Google Scholar Crossref Search ADS WorldCat Lambert R. A. 2001 . Grey seals to cull or not to cull? History Today , 51 : 30 – 32 . Google Scholar OpenURL Placeholder Text WorldCat Lavigne D. 2003 . Marine Mammals and Fisheries: The Role of Science in the Culling Debate . CSIRO Publishing , Victoria . pp. 31 – 47 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lawrence S. , Motova A., Russell J. 2016 . Fleet economic performance Dataset 2008–2015. Seafish Economics , Edinburgh . pp. 1 – 68 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Marine Environmental Mapping Programme. 2015 . BGS SeaBed Sediment (250K) Map. http://www.maremap.ac.uk/view/search/searchMaps.html (last accessed 03 November 2015). Marine Management Organization . 2012 . UK Sea Fisheries Statistics 2011. Marine Management Organisation, London , pp. 1–140. Mohn R. , Bowen W. 1996 . Grey seal predation on the eastern Scotian Shelf: modelling the impact on Atlantic cod . Canadian Journal of Fisheries and Aquatic Sciences , 53 : 2722 – 2738 . Google Scholar Crossref Search ADS WorldCat O’Boyle R. , Sinclair M. 2012 . Seal–cod interactions on the Eastern Scotian Shelf: Reconsideration of modelling assumptions . Fisheries Research , 115–116 : 1 – 13 . Google Scholar Crossref Search ADS WorldCat Overholtz W. J. , Link J. S. 2007 . Consumption impacts by marine mammals, fish, and seabirds on the Gulf of Maine–Georges Bank Atlantic herring (Clupea harengus) complex during the years 1977–2002 . ICES Journal of Marine Science , 64 : 83 – 96 . Google Scholar Crossref Search ADS WorldCat Pope J. G. , Shepherd J. G. 1982 . A simple method for the consistent interpretation of catch-at-age data . Journal Du Conseil , 40 : 176 – 184 . Google Scholar Crossref Search ADS WorldCat Read A. J. 2008 . The looming crisis: interactions between marine mammals and fisheries . Journal of Mammalogy , 89 : 541 – 548 . Google Scholar Crossref Search ADS WorldCat Ricker W. E. 1954 . Stock and recruitment . Journal of the Fisheries Board of Canada , 11 : 559 – 623 . Google Scholar Crossref Search ADS WorldCat Schaefer M. B. 1954 . Some Aspects of the Dynamics of Populations Important to the Management of the Commercial Marine Fisheries. Inter-American Tropical Tuna Commission , La Jolla, California . pp. 27 – 56 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Scottish Fishermen’s Organization . 2016 . Production and marketing plan 2015. pp. 1 – 18 . Sparholt H. 1994 . Fish species interactions in the Baltic Sea . Dana , 10 : 131 – 162 . Google Scholar OpenURL Placeholder Text WorldCat STECF . 2016a . The 2016 Annual Economic Report on the EU Fishing Fleet (STECF 16-11) . Publications Office of the European Union , Luxembourg . pp. 1 – 470 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC STECF . 2016b . Fisheries Dependent Information (STECF-16-20) . Publications Office of the European Union . Luxembourg . pp. 1 – 858 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Thomas L. 2013 . Estimating the size of the UK grey seal population between 1984 and 2012, using established and draft revised priors. SCOS Briefing Papers 13/02. pp. 89 – 109 . Thomas L. 2015 . Estimating the size of the UK grey seal population between 1984 and 2014. SCOS Briefing Papers 15/02. pp. 64 – 83 . Trijoulet V. 2016 . Bioeconomic modelling of seal impacts on West of Scotland fisheries, Doctoral thesis. University of Strathclyde, Glasgow. 338 pp. Trijoulet V. , Holmes S. J., Cook R. M. 2017 . Grey seal predation mortality on three depleted stocks in the West of Scotland: What are the implications for stock assessments? Canadian Journal of Fisheries and Aquatic Sciences . doi: 10.1139/cjfas-2016-0521. Google Scholar OpenURL Placeholder Text WorldCat Trzcinski M. K. , Mohn R., Bowen W. D. 2006 . Continued decline of an Atlantic cod population: how important is gray seal predation? . Ecological Applications , 16 : 2276 – 2292 . Google Scholar Crossref Search ADS PubMed WorldCat Warburton C. , Parsons E., Woods-Ballard A., Hughes A., Johnston P. 2001 . Whale-watching in West Scotland: Report for the Department for Environment , Food and Rural Affairs , London . pp. 1–81 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Woods-Ballard A. , Parsons E., Hughes A., Velander K., Ladle R., Warburton C. 2003 . The sustainability of whale-watching in Scotland . Journal of Sustainable Tourism , 11 : 40 – 55 . Google Scholar Crossref Search ADS WorldCat Author notes Present address: Integrated Statistics, under contract to Northeast Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 166 Water St., Woods Hole, MA 02543, USA. © International Council for the Exploration of the Sea 2018. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © International Council for the Exploration of the Sea 2018.

Roach, Michael; Cohen, Mike; Forster, Rodney; Revill, Andrew S; Johnson, Magnus

doi: 10.1093/icesjms/fsy006pmid: N/A

Abstract Offshore wind farms (OWF) form an important part of many countries strategy for responding to the threat of climate change, their development can conflict with other offshore activities. Static gear fisheries targeting sedentary benthic species are particularly affected by spatial management that involves exclusion of fishers. Here we investigate the ecological effect of a short-term closure of a European lobster (Homarus gammarus (L.)) fishing ground, facilitated by the development of the Westermost Rough OWF located on the north-east coast of the United Kingdom. We also investigate the effects on the population when the site is reopened on completion of the construction. We find that temporary closure offers some respite for adult animals and leads to increases in abundance and size of the target species in that area. Reopening of the site to fishing exploitation saw a decrease in catch rates and size structure, this did not reach levels below that of the surrounding area. Opening the site to exploitation allows the fishery to recuperate some of the economic loss during the closure. We suggest that our results may indicate that temporary closures of selected areas may be beneficial and offer a management option for lobster fisheries. Introduction Globally there has been an increase in energy provided from the wind industry, surpassing 63 GW in 2015, an 18% increase since 2014 (Global Wind Energy Council, 2015). Wind energy developments are often the most used tool by national governments to meet their energy demands from renewable sources, seeing an increase in offshore wind developments in recent years. Offshore wind developments are often located to exploit the optimum wind energy and be able to transmit the energy to shore. For example, the United States is estimated to have a 4000 GW capacity for wind energy (US Department of Energy 2012). The development of offshore wind farms (OWF) can cause spatial conflicts with other sea users. For example, the eastern sea board of the United States is prime location for both offshore wind energy and nationally important crustacean fisheries (Breton and Moe, 2009; Brehme et al., 2015). Co-location of marine users and spatial management of resources is being observed in the United Kingdom; one of the world leaders in offshore wind exploitation (Hooper and Austen, 2014; Kota et al., 2015). UK government has a target of 15% of its energy from renewable sources by 2020 (European Commission, 2016). There are currently 25 OWF operational or under construction within UK waters currently providing approximately 5% of the UK demand with a further 16 with development consent (The Crown Estates, 2017). There has been a steady increase in research into the impacts of OWF on the marine ecosystem. This increase in literature is largely review based with the few empirical studies available, not being able to give a reliable assessment of the cumulative impact of offshore wind development (Lindeboom et al., 2015). The current empirical studies have largely focussed on the impact to seabird interactions (15 out of 78 publications reviewed by Hooper et al. [2017]), marine mammals (Madsen et al., 2006; Thomsen et al., 2008; Bailey et al., 2010; Brandt et al., 2011), substrate and infauna disturbance (Coates et al., 2014; De Backer et al., 2014; Vandendriessche et al., 2015) and fish populations (Wahlberg and Westerberg, 2005; De Troch et al., 2013; Bergman et al., 2014; Stenberg et al., 2015). Most empirical studies investigating effects on macrobenthic crustaceans form part of an environmental impact assessment or statutory monitoring programmes. To date the majority of OWF constructed in European waters are in shallow water (typically less than 30 m) on sand-based substrates. The introduction of individual turbines and associated stone protection (used to protect monopole bases from sand scour), can introduce a new hard substrate habitat to an area. This can increase shelter and hard substrate habitat in areas that it may not have existed previously. This introduced habitat has been found to increase biodiversity and biomass of associated fauna in some areas (Lindeboom et al., 2011; De Backer et al., 2014; Stenberg et al., 2015). Krone et al. (2017) observed over 5000 Cancer pagurus on individual monopoles with scour stone protection, which was more than double that found on monopoles without scour stone protection. However, this was observed in areas characterized by sandy substrate, the effect of scour stone protection on areas characterized by rock and cobble is yet to be understood. Using studies from sites that are not comparable to each other to understand effects of OWF installations can lead to misunderstanding of the processes involved (Lindeboom et al., 2015). OWF and individual turbines can act as fish aggregation devices, providing a refuge for fish species from predation and exploitation, although the effects can be spatially limited to the OWF (Griffin et al., 2016). An OWF can act as a quasi-marine protected area (MPA) or no take zone (NTZ). This can be due to exclusion during construction or operation, to all fishing vessels or the physical presence of the turbines excluding certain gear types such as mobile gear (Bergman et al., 2014; Krone et al., 2017). There is potential for co-location of fisheries and OWF developments, however, these are predominantly static fisheries (Christie et al., 2014; Hooper and Austen 2014; Stelzenmüller et al., 2016). The effects of OWF on mobile benthic megafauna that are targeted by static gear fisheries are little understood (Hooper and Austen 2014; Lindeboom et al., 2015). The potential of spill-over effects of MPA/NTZ can be difficult to ascertain (Moland et al., 2013; Smyth et al., 2015; Vandendriessche et al., 2015), the temporal scale of studies can often not be of sufficient duration to observe the spill over. However Goñi et al. (2010) and Hoskin et al. (2011) observed spill-over effects of different lobster populations within a closed area over a period of 10 and 4 years, respectively. Homarus gammarus have been shown to have strong site fidelity and defined home ranges (Bannister and Addison, 1998; Smith et al., 1998; Moland et al., 2011) although there is a seasonal migration to deeper water during the colder months. Any spill-over effects of closed areas are likely to be only observed locally or during immigration/emigration from the site. The implementation of an MPA/NTZ has often been met with resistance by commercial fisheries. The potential for positive ecological and possible economic effects of closed areas are often met with scepticism from the fishing industry. This is due to the implementation of surveys not reflecting the way fishermen operate and the fact that data are not in the public domain (Hooper and Austen 2014; Hooper et al., 2015). The spill-over effect can lead to the process of “fishing the line,” where fishing intensity is increased on the boundaries of a closed area (Kellner et al., 2007). Spatial displacement of effort into another area can increase pressure on fisheries and lead to increased competition among fishers. This is especially the case in static gear fisheries where individual fishers can have a strong fidelity to specific sites (Hart et al., 2002; Turner et al., 2013). The implementation of closed areas can often be considered by industry and some in the scientific community to be conducted for political purposes as opposed to ecological. The use of MPA’s as a fisheries management tool should be treated as a rigorously designed experiment with accurate cost/benefit analysis (Kaiser 2005; Caveen et al., 2014). During construction of OWF, the fishing industry are often excluded from the area for safety reasons; this can have a potential short-term positive effect on the local population due to the removal of fishing mortality. Here we investigate the short-term effects of construction of an OWF on a commercially exploited European lobster (H. gammarus, Homaridae (L.) subsequently referred to as lobster) population. We also highlight the effects of reopening the site to exploitation on completion of the OWF construction and their use as a potential management tool. The study also highlighted the potential positive effects of the fishing industry engaging in research of OWF effects. Methods Site description The Holderness coast lobster fishery is the largest lobster fishery in the United Kingdom, representing approximately 20% of national lobster landings. Landings of European lobster, into the regions’ main port of Bridlington in 2015 were 405 tonnes with an estimated first sale value of £4.2 million (Marine Management Organisation, 2015). The fishery in the area targets lobster almost exclusively using static creels generally baited with mackerel. Creels are immersed for varying periods depending on the fisher, but generally 2–3 days. The Westermost Rough wind farm, constructed in 2014/2015 at a cost of £800 million, and is located within the Holderness fishery, situated within one of the fisheries main target areas. The site was one of the first to be located on a rock and cobble substrate. The Westermost Rough OWF extends from 7.7 km off the coast to 13.3 km offshore and is approximately 35 km2 in area (Figure 1). It consists of 35, 6 MW turbines and associated substation, located in a depth of water ranging from 15 to 23 m. The substrate is predominantly rock and cobble with sand patches, the area was subjected to boulder removal prior to the construction phase. Figure 1. Open in new tabDownload slide Location of the Westermost Rough OWF, the individual turbine locations marked and the locations of the treatment strings within the turbine array and the control strings to the North of the site. Figure 1. Open in new tabDownload slide Location of the Westermost Rough OWF, the individual turbine locations marked and the locations of the treatment strings within the turbine array and the control strings to the North of the site. The study was conducted using a fishing industry managed research vessel, the R.V. Huntress. The study was a collaboration between the local fishery; The Holderness Fishing Industry Group (www.hfig.org.uk) and the OWF developer, DONG Energy. Sampling methods There were two sites chosen to assess the effects of the construction of the Westermost Rough OWF, one site in the OWF (treatment [subsequently referred to as the wind farm]) and a site to the north of the OWF (control) (Figure 1). The sites were restricted in their spatial distribution within the OWF due to the process of the construction of the OWF. The site was agreed with the developer as the area that could be surveyed without disruption to the sampling protocol. The control site was located 1 km to the north of the OWF. The prevailing current drifts north/south, any effects of the construction should not have been observed. This site was also selected due to the substrate reflecting that within the OWF. There were further spatial restrictions of the control site due to displacement of fishing gear from the OWF to the surrounding area, care was taken to avoid gear conflict. Sampling strings consisting of 30 creels were deployed both within the wind farm and the control. The strings consisted of 25 standard commercial creels with a 70 mm mesh and 96.5 cm base; and 5 creels with a 30 mm mesh and a 76.2 cm base. All creels were exempt from local byelaws ordering the use of escape gaps. The smaller mesh creels were used to sample catch that may escape the larger mesh creels. On every haul, each creel was baited with two mackerel “frames,” which are commonly used in the region to target lobsters. Each string was secured at either end with a 20 kg anchor and marked with a surface marker buoy. The gear configuration mirrored that of the commercial fishing strings in the area. A baseline survey was carried out prior to the wind farm construction, taking into account the spatial restrictions that were predicted once the construction began. The survey was timed to target the lobster fishery between June and September of 2013, maintaining a mean immersion period of 3.0 days (SD ±1.34 days) and all creels from both the control and treatment were processed on every survey day (n = 24 hauls each site). Following the before/after, control/impact (BACI) approach (Carstensen et al., 2006; Hoskin et al., 2011; Moland et al., 2013; Vandendriessche et al., 2015) sampling was mirrored in June–September of 2015 for the first-year post build of the wind farm. The immersion period of the creels in 2015 was 3.9 days (SD ±2.1 days) and all creels from both the control and treatment were processed on every survey day (n = 23 hauls each site). Variation in immersion periods to the baseline survey was due to inclement weather. There was no survey during 2014 as the site was under construction. During the baseline survey of 2013 both the wind farm and control were subjected to fishing exploitation for the entire period. During construction, the wind farm was closed to fishing exploitation for a period of 20 months during 2014/2015, until the middle of August 2015 (13/08/2015). This was part way through the 2015 sampling period, with 13 sample days when the site was closed and 11 sample days when the site was open to exploitation. For the entire survey periods in both 2013 and 2015 there were no restrictions to fishing exploitation in the control site and the main management of effort in the area was based on minimum landing size (87 mm carapace length [CL]) of the catch. Abundance of lobster was recorded from each creel. Sex, condition, ovigerous status and size (CL) were recorded for the aggregated catch within each string. All animals were returned to sea after recording. The survey was timed and designed to assess the effects of the wind farm construction on the region’s most valuable fishery, this study reports lobster status only. Data analysis Analysis was conducted on the overall differences in size and catch rates between the baseline (2013) and the first-year post build (2015). Because the previously closed site was reopened to fishing during the 2015 sampling regime, analysis was also conducted on the status of the wind farm (open/closed to fishing exploitation). All analysis was conducting using R statistical software (R Core Team, 2017). Size distribution Differences in size frequency for both between years and between statuses of the wind farm (open and closed) were analysed using the two-sample Kolmogorov–Smirnov (K-S) test. Empirical cumulative distribution function (ECDF) plots were generated to demonstrate the proportion of lobsters between the two sites that are less than each observed length (Ogle, 2016). Generalized Linear Mixed Models (GLMM) are used when the data are not normally distributed and when there is the potential for pseudo-replication (Zuur and Ieno, 2016). Due to the limitations of survey sites, the size data not conforming to normality (K-S, p < 0.05) and the variability in the number of lobsters sampled on each day (range = 13–137 [2013], 44–179 [2015]), GLMM was deemed a more suitable analysis. We therefore applied a GLMM in which the relative catch probability of the lobsters entering the pots within each site/year was the response variable, carapace length (length) of lobster as the fixed effect and haul (survey day) was used as a crossed random intercept. A binomial error was applied due to the response variable being the relative catch probability of lobsters entering pots within each site/year. Sex, ovigerous status and condition of the lobsters were investigated as factors within the model but were rejected due to either insignificance (p > 0.05 [sex and condition]) or unsuitable factors to include (ovigerous status). Soak time was investigated to see whether it should be included as an offset within the model. There was a poor relationship (R2 < 0.1 on all occasions) between the daily abundance of lobsters within each string and the soak time they were subjected to. Soak time was also negated for the between sites comparison within the survey design, as both sites were subjected to the same soak time on all occasions. It was decided that soak time was not required as an offset within the GLMM. Therefore, the simplest model was the best description of the relative catch probability of lobsters of each size entering the strings/pots between the two sites/status of the wind farm (open/closed); Pr⁡{TestTest + Control}=1/(1+e-haul+β1×length+β2×length2) GLMM was applied using the lme4 package in R statistical software (Bates et al., 2015). This follows the methodology described by Holst and Revill (2009), analysing differences in catch composition at length between tests and controls (Van Marlen et al., 2014; Vogel et al., 2017). Test was determined as the strings sampled in 2015 for the between years comparison with the strings sampled in 2013 (baseline) as the control. The analysis of the wind farm status in 2015 (open/closed), test was always the wind farm site and the control was always the control site to the north of the wind farm. Validation of each GLMM model was conducted by checking that the normality of the standardized residuals conformed to a normal distribution (Shapiro–Wilkes, p > 0.05) (Thomas et al., 2015) and also comparing the GLMM results to the two-sample K-S analysis. GLMM results were also presented graphically, allowing for inference as to where within the distribution the significance lay. Catch comparison Catch per unit of effort (CPUE) was determined as the total number of lobsters caught in a string (Davies et al., 2015). Landings per unit of effort (LPUE) was determined as the total number of lobsters per string that were above the minimum landing size (87 mm CL) and of a good enough quality (i.e. not missing limbs and no visible signs of disease) to be landed to market. The CPUE and LPUE data conformed to a normal distribution (K-S, p > 0.05) but the variances could not be considered equal (F-test, p < 0.05). A Welch’s t-test assuming unequal variances was applied to the CPUE and LPUE to analyse the differences in site, year and wind farm status (open/closed). Results A total of 1440 creels (720 at each site) were hauled during the baseline data collection in 2013 (n = 24 survey days) recording 6051 lobsters. During the 2015 post-build survey (n = 23 survey days), 1380 creels (690 at each site) were hauled and 8734 lobsters were recorded. Size distribution The size frequency distributions of lobsters differed significantly between the two years (K-S, D = 0.10, p < 0.001). The windfarm in 2015 showed a larger proportion of lobsters at a larger size (>100 mm CL) than sampled in 2013 (Figure 2a and b), there was a greater proportion of lobsters from the MLS (87 mm)—96 mm CL sampled in 2013. There was a broader size range, 39–126 mm CL in 2015 as opposed to 56–114 mm CL in 2013. The ECDF plot (Figure 2c) demonstrates that the greatest difference in distributions were between 75 and 92 mm CL. This was supported by the GLMM plot (Table 1, Figure 2d), which demonstrates that there was a greater proportion of lobsters sampled over 70 mm CL in 2015 than in 2013. Table 1. GLMM parameters for both the comparison between years and the comparison between the control and wind farm, in relation to the status of the wind farm being subjected to fishing exploitation. Treatment . Response . Intercept variance . Parameter . Estimate . Standard error . Wind farm Between 2013 and 2015 0.755 β0 0.215 0.347 β1 −0.009 0.004 Wind farm closed Wind farm and control 0.031 β0 6.678 0.385 β1 −0.081 0.005 Wind farm open Wind farm and control 0.036 β0 2.045 0.464 β1 −0.020 0.006 Treatment . Response . Intercept variance . Parameter . Estimate . Standard error . Wind farm Between 2013 and 2015 0.755 β0 0.215 0.347 β1 −0.009 0.004 Wind farm closed Wind farm and control 0.031 β0 6.678 0.385 β1 −0.081 0.005 Wind farm open Wind farm and control 0.036 β0 2.045 0.464 β1 −0.020 0.006 Open in new tab Table 1. GLMM parameters for both the comparison between years and the comparison between the control and wind farm, in relation to the status of the wind farm being subjected to fishing exploitation. Treatment . Response . Intercept variance . Parameter . Estimate . Standard error . Wind farm Between 2013 and 2015 0.755 β0 0.215 0.347 β1 −0.009 0.004 Wind farm closed Wind farm and control 0.031 β0 6.678 0.385 β1 −0.081 0.005 Wind farm open Wind farm and control 0.036 β0 2.045 0.464 β1 −0.020 0.006 Treatment . Response . Intercept variance . Parameter . Estimate . Standard error . Wind farm Between 2013 and 2015 0.755 β0 0.215 0.347 β1 −0.009 0.004 Wind farm closed Wind farm and control 0.031 β0 6.678 0.385 β1 −0.081 0.005 Wind farm open Wind farm and control 0.036 β0 2.045 0.464 β1 −0.020 0.006 Open in new tab Figure 2. Open in new tabDownload slide Size distributions of lobsters sampled within the Westermost Rough OWF for the baseline survey in 2013 (a) and the first-year post-build survey in 2015 (b), both plots fitted with the density curve of the distribution and the bins set to 2 mm carapace length. (c) ECDF for the sampled lobsters for the wind farm and control site in 2013 (red and black) and the wind farm and control site in 2015 (blue and grey). (d) Plot derived from GLMM modelling of the proportion of the lobsters sampled at each size in 2013 (top box) and 2015 (bottom box). The grey-shaded areas represent the 95% confidence intervals and the bold black line the mean value. The central horizontal line represents the 0.5 (50%) value, points overlapping this line indicate that there was no significant difference in the proportion of that sized animal between the two years. A value of 0.75 indicates that 75% of the lobsters sampled at that size were sampled in 2013 and the other 25% were sampled in 2015. This applies to all subsequent plots derived from GLMM analysis. The vertical line on all plots represents the minimum landing size of lobsters in the fishery, which is 87 mm carapace length. This applies to all subsequent plots reported. Figure 2. Open in new tabDownload slide Size distributions of lobsters sampled within the Westermost Rough OWF for the baseline survey in 2013 (a) and the first-year post-build survey in 2015 (b), both plots fitted with the density curve of the distribution and the bins set to 2 mm carapace length. (c) ECDF for the sampled lobsters for the wind farm and control site in 2013 (red and black) and the wind farm and control site in 2015 (blue and grey). (d) Plot derived from GLMM modelling of the proportion of the lobsters sampled at each size in 2013 (top box) and 2015 (bottom box). The grey-shaded areas represent the 95% confidence intervals and the bold black line the mean value. The central horizontal line represents the 0.5 (50%) value, points overlapping this line indicate that there was no significant difference in the proportion of that sized animal between the two years. A value of 0.75 indicates that 75% of the lobsters sampled at that size were sampled in 2013 and the other 25% were sampled in 2015. This applies to all subsequent plots derived from GLMM analysis. The vertical line on all plots represents the minimum landing size of lobsters in the fishery, which is 87 mm carapace length. This applies to all subsequent plots reported. During the wind farm closure in 2015 (prior to 13/08/2015), the size distribution of lobsters in the control site (Figure 3b) had a narrower distribution (39–117 mm CL) than within the wind farm (40–126 mm CL) and there was generally a greater proportion of lobsters within the wind farm than within the control site (Figure 3a). The size distribution of lobsters within the wind farm was significantly different to both the control site (K-S, D = 0.32, p = < 0.0001) and the baseline data (K-S, D = 0.14, p < 0.0001) (Figure 3a and b). Although Figure 3c shows that the size of lobsters within the wind farm (red) differed from the baseline (black and grey) and the control (blue) between 60–107 mm CL, Figure 3d shows that the distribution was split approximately at the MLS (vertical line). The graphical representation of the GLMM (Table 1) shows that there was a greater proportion of lobsters below the MLS in the control site and the inverse in the wind farm. Figure 3. Open in new tabDownload slide Size distributions of lobsters sampled at the Westermost Rough OWF for both the wind farm site (a), which was closed to fishing for the period and the control (b), which was subjected to fishing throughout the period. ECDF plot for the period before the wind farm site was opened to fishing showing the wind farm (red), control (blue) and baseline for the two sites: black (wind farm), grey (control) (c) and the plot derived from GLMM analysis for both the control and wind farm site (d) for the period before the wind farm was opened to fishing. Figure 3. Open in new tabDownload slide Size distributions of lobsters sampled at the Westermost Rough OWF for both the wind farm site (a), which was closed to fishing for the period and the control (b), which was subjected to fishing throughout the period. ECDF plot for the period before the wind farm site was opened to fishing showing the wind farm (red), control (blue) and baseline for the two sites: black (wind farm), grey (control) (c) and the plot derived from GLMM analysis for both the control and wind farm site (d) for the period before the wind farm was opened to fishing. There was a decline in the proportion of lobsters above MLS in the control site after the wind farm had been opened to fishing (Figure 4a) in comparison to the period when the wind farm was closed (Figure 3a) (K-S, D = 0.07, P < 0.05). This was also reflected within the wind farm site (Figures 3b and 4b) (K-S, D = 0.28, p < 0.0001). The sampling period post opening of the wind farm to fishing demonstrated a greater proportion of lobsters within the wind farm in comparison to the control site (K-S, D = 0.11, p < 0.0001). Although there was a difference in the cumulative distribution between the wind farm and the control site between 70 and100 mm CL, both sites also showed a difference from the baseline data (Figure 4c). GLMM analysis (Table 1) shows that after opening of the site to fishing there was a greater proportion of lobsters below MLS in the control site as opposed to the wind farm (Figure 4d). There was no significant difference in the proportion of lobsters above MLS between the two sites post opening of the site. Figure 4. Open in new tabDownload slide Size distributions of lobsters sampled at the Westermost Rough OWF for both the wind farm site (a) after the site was opened to fishing and the control (b), which was subjected to fishing throughout the period. ECDF plot for the period after the wind farm site was opened to fishing showing the wind farm (red), control (blue) and baseline for the two sites: black (wind farm), grey (control) (c) and plot derived from GLMM analysis for both the control and wind farm site (d) for the period after the wind farm was opened to fishing. Figure 4. Open in new tabDownload slide Size distributions of lobsters sampled at the Westermost Rough OWF for both the wind farm site (a) after the site was opened to fishing and the control (b), which was subjected to fishing throughout the period. ECDF plot for the period after the wind farm site was opened to fishing showing the wind farm (red), control (blue) and baseline for the two sites: black (wind farm), grey (control) (c) and plot derived from GLMM analysis for both the control and wind farm site (d) for the period after the wind farm was opened to fishing. Catch and landings per unit of effort between years Mean CPUE (Table 2) was significantly greater in 2015 for both sites than in 2013 (p < 0.01, Table 3), however did not differ significantly between control and wind farm within the same year (p > 0.05, Table 3). Mean LPUE (Table 2) was also significantly greater in the wind farm in 2015 than in 2013 and it was also significantly greater in the wind farm than the control site in the year 2015 (p < 0.01, Table 3). The control site showed no significant difference in mean LPUE between sample years (p > 0.05, Figure 5 and Table 3). The greatest ratio between CPUE and LPUE (0.25) was within the wind farm during the year 2015, this was when the wind farm was closed for a period during the sampling regime (Figure 5). Table 2. Descriptive statistics of CPUE and LPUE of lobsters sampled at both sites of the Westermost Rough OWF during the 2013 and 2015 surveys. Year . Site . Effort . Mean . SD . 2013 Wind farm CPUE 63.14 34.68 2013 Control CPUE 74.27 45.48 2015 Wind farm CPUE 93.30 32.14 2015 Control CPUE 107.30 29.46 2013 Wind farm LPUE 11.51 6.75 2013 Control LPUE 11.28 5.71 2015 Wind farm LPUE 23.39 16.68 2015 Control LPUE 10.26 4.67 Status Site Effort Mean SD Closed Wind farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Year . Site . Effort . Mean . SD . 2013 Wind farm CPUE 63.14 34.68 2013 Control CPUE 74.27 45.48 2015 Wind farm CPUE 93.30 32.14 2015 Control CPUE 107.30 29.46 2013 Wind farm LPUE 11.51 6.75 2013 Control LPUE 11.28 5.71 2015 Wind farm LPUE 23.39 16.68 2015 Control LPUE 10.26 4.67 Status Site Effort Mean SD Closed Wind farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Open in new tab Table 2. Descriptive statistics of CPUE and LPUE of lobsters sampled at both sites of the Westermost Rough OWF during the 2013 and 2015 surveys. Year . Site . Effort . Mean . SD . 2013 Wind farm CPUE 63.14 34.68 2013 Control CPUE 74.27 45.48 2015 Wind farm CPUE 93.30 32.14 2015 Control CPUE 107.30 29.46 2013 Wind farm LPUE 11.51 6.75 2013 Control LPUE 11.28 5.71 2015 Wind farm LPUE 23.39 16.68 2015 Control LPUE 10.26 4.67 Status Site Effort Mean SD Closed Wind farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Year . Site . Effort . Mean . SD . 2013 Wind farm CPUE 63.14 34.68 2013 Control CPUE 74.27 45.48 2015 Wind farm CPUE 93.30 32.14 2015 Control CPUE 107.30 29.46 2013 Wind farm LPUE 11.51 6.75 2013 Control LPUE 11.28 5.71 2015 Wind farm LPUE 23.39 16.68 2015 Control LPUE 10.26 4.67 Status Site Effort Mean SD Closed Wind farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Open in new tab Table 3. Results from Welch’s two-sample t-test for the mean CPUE/LPUE data analysed between the control and treatment sites of the Westermost Rough OWF and between the baseline and post-build surveys. Factors analysed . Response . p . t . DF . Treatment between years CPUE <0.01 −3.02 29.27 Control between years CPUE <0.01 −2.88 35.75 Treatment between years LPUE <0.01 −3.16 29.27 Control between years LPUE n.s. 0.65 40.62 Treatment vs. control in 2013 CPUE n.s. 0.91 39.25 Treatment vs. control in 2015 CPUE n.s. 1.54 43.67 Treatment vs. control in 2013 LPUE n.s. −0.12 40.88 Treatment vs. control in 2015 LPUE <0.01 −3.64 25.43 Factors analysed . Response . p . t . DF . Treatment between years CPUE <0.01 −3.02 29.27 Control between years CPUE <0.01 −2.88 35.75 Treatment between years LPUE <0.01 −3.16 29.27 Control between years LPUE n.s. 0.65 40.62 Treatment vs. control in 2013 CPUE n.s. 0.91 39.25 Treatment vs. control in 2015 CPUE n.s. 1.54 43.67 Treatment vs. control in 2013 LPUE n.s. −0.12 40.88 Treatment vs. control in 2015 LPUE <0.01 −3.64 25.43 The significant results are displayed in bold. Open in new tab Table 3. Results from Welch’s two-sample t-test for the mean CPUE/LPUE data analysed between the control and treatment sites of the Westermost Rough OWF and between the baseline and post-build surveys. Factors analysed . Response . p . t . DF . Treatment between years CPUE <0.01 −3.02 29.27 Control between years CPUE <0.01 −2.88 35.75 Treatment between years LPUE <0.01 −3.16 29.27 Control between years LPUE n.s. 0.65 40.62 Treatment vs. control in 2013 CPUE n.s. 0.91 39.25 Treatment vs. control in 2015 CPUE n.s. 1.54 43.67 Treatment vs. control in 2013 LPUE n.s. −0.12 40.88 Treatment vs. control in 2015 LPUE <0.01 −3.64 25.43 Factors analysed . Response . p . t . DF . Treatment between years CPUE <0.01 −3.02 29.27 Control between years CPUE <0.01 −2.88 35.75 Treatment between years LPUE <0.01 −3.16 29.27 Control between years LPUE n.s. 0.65 40.62 Treatment vs. control in 2013 CPUE n.s. 0.91 39.25 Treatment vs. control in 2015 CPUE n.s. 1.54 43.67 Treatment vs. control in 2013 LPUE n.s. −0.12 40.88 Treatment vs. control in 2015 LPUE <0.01 −3.64 25.43 The significant results are displayed in bold. Open in new tab Figure 5. Open in new tabDownload slide Mean catch per unit effort (a) and landings per unit effort (b) for the wind farm (dark grey) and the control site (light grey) for the baseline survey (2013) and the first-year post-build survey (2015). The top of the bars represent the mean CPUE/LPUE and the error bars the standard deviation of the data. The values above the LPUE bars represent the ratio between CPUE and LPUE. The letters above the bars indicate the factors that showed a significant difference. This applies to all subsequent bar plots reported. Figure 5. Open in new tabDownload slide Mean catch per unit effort (a) and landings per unit effort (b) for the wind farm (dark grey) and the control site (light grey) for the baseline survey (2013) and the first-year post-build survey (2015). The top of the bars represent the mean CPUE/LPUE and the error bars the standard deviation of the data. The values above the LPUE bars represent the ratio between CPUE and LPUE. The letters above the bars indicate the factors that showed a significant difference. This applies to all subsequent bar plots reported. Influence of wind farm opening After the wind farm opened to fishing exploitation, mean CPUE (Table 4) within the wind farm reduced significantly (p < 0.001, Table 5), this was not the case in the control site (p > 0.05, Table 5). Mean CPUE (Table 4) was also significantly greater prior to the wind farm being opened to fishing exploitation (p < 0.001, Figure 6 and Table 5). Mean LPUE (Table 4) was significantly greater in the wind farm when it was closed than the control site during the same period and once the wind farm was opened to fishing (p < 0.001, Table 5). Mean LPUE (Table 4) was significantly greater in the control site when the wind farm was closed than the period when the area was open to fishing exploitation (p < 0.05, Figure 6 and Table 5). The greatest ratio of LPUE against CPUE (0.33) was during the period when the wind farm was closed to fishing, indicating a higher proportion of high quality lobsters that were not being exploited (Figure 6). Table 4. Descriptive statistics of CPUE and LPUE of lobsters sampled at both sites of the Westermost Rough OWF before and after the wind farm was opened to fishing exploitation. Status . Site . Effort . Mean . SD . Closed Wind Farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind Farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind Farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind Farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Status . Site . Effort . Mean . SD . Closed Wind Farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind Farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind Farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind Farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Open in new tab Table 4. Descriptive statistics of CPUE and LPUE of lobsters sampled at both sites of the Westermost Rough OWF before and after the wind farm was opened to fishing exploitation. Status . Site . Effort . Mean . SD . Closed Wind Farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind Farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind Farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind Farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Status . Site . Effort . Mean . SD . Closed Wind Farm CPUE 113.08 29.31 Closed Control CPUE 107.08 35.44 Open Wind Farm CPUE 71.73 18.59 Open Control CPUE 107.55 22.98 Closed Wind Farm LPUE 36.83 10.43 Closed Control LPUE 12.08 4.23 Open Wind Farm LPUE 8.73 6.25 Open Control LPUE 8.27 4.47 Open in new tab Table 5. Results from Welch’s two-sample t-test for the mean CPUE/LPUE data analysed between the status of the Westermost Rough OWF in 2015, i.e. open or closed to fishing. Factors analysed . Response . p . t . DF . Treatment between status of OWF CPUE <0.001 4.08 18.79 Control between status of OWF CPUE n.s. −0.04 19.01 Treatment between status of OWF LPUE <0.0001 7.92 18.23 Control between status of OWF LPUE <0.05 2.10 20.56 Treatment when OWF was open CPUE <0.001 4.02 19.16 Control when OWF was closed CPUE n.s. −0.45 21.25 Treatment when OWF was open LPUE n.s. −0.20 18.12 Control when OWF was closed LPUE <0.0001 −7.62 14.53 Factors analysed . Response . p . t . DF . Treatment between status of OWF CPUE <0.001 4.08 18.79 Control between status of OWF CPUE n.s. −0.04 19.01 Treatment between status of OWF LPUE <0.0001 7.92 18.23 Control between status of OWF LPUE <0.05 2.10 20.56 Treatment when OWF was open CPUE <0.001 4.02 19.16 Control when OWF was closed CPUE n.s. −0.45 21.25 Treatment when OWF was open LPUE n.s. −0.20 18.12 Control when OWF was closed LPUE <0.0001 −7.62 14.53 The significant results are displayed in bold. Open in new tab Table 5. Results from Welch’s two-sample t-test for the mean CPUE/LPUE data analysed between the status of the Westermost Rough OWF in 2015, i.e. open or closed to fishing. Factors analysed . Response . p . t . DF . Treatment between status of OWF CPUE <0.001 4.08 18.79 Control between status of OWF CPUE n.s. −0.04 19.01 Treatment between status of OWF LPUE <0.0001 7.92 18.23 Control between status of OWF LPUE <0.05 2.10 20.56 Treatment when OWF was open CPUE <0.001 4.02 19.16 Control when OWF was closed CPUE n.s. −0.45 21.25 Treatment when OWF was open LPUE n.s. −0.20 18.12 Control when OWF was closed LPUE <0.0001 −7.62 14.53 Factors analysed . Response . p . t . DF . Treatment between status of OWF CPUE <0.001 4.08 18.79 Control between status of OWF CPUE n.s. −0.04 19.01 Treatment between status of OWF LPUE <0.0001 7.92 18.23 Control between status of OWF LPUE <0.05 2.10 20.56 Treatment when OWF was open CPUE <0.001 4.02 19.16 Control when OWF was closed CPUE n.s. −0.45 21.25 Treatment when OWF was open LPUE n.s. −0.20 18.12 Control when OWF was closed LPUE <0.0001 −7.62 14.53 The significant results are displayed in bold. Open in new tab Figure 6. Open in new tabDownload slide Mean CPUE (a) and LPUE from the wind farm (dark grey) and the control site (light grey) before and after the wind farm was opened to fishing exploitation. Figure 6. Open in new tabDownload slide Mean CPUE (a) and LPUE from the wind farm (dark grey) and the control site (light grey) before and after the wind farm was opened to fishing exploitation. Discussion Size distribution The exclusion of fishing effort within the OWF was found to have an effect on the size distribution of lobsters within the area. There was a greater total number of lobsters observed during the post-build survey than during the baseline survey (n = 2683 difference). Within the wind farm, there was a greater proportion of lobsters, over MLS observed in 2015 than in 2013 (Figure 2). The absence of fishing exploitation within the wind farm during construction acted as a NTZ, protecting lobsters greater than the MLS. There is potential for spill-over effects of an MPA/NTZ (McClanahan and Mangi 2000; Smyth et al., 2015), this can be observed locally for species with reduced movement patterns. (Moland et al., 2011). When a NTZ is created it has been reported to initially show an increase in lobster abundance and biomass (Hoskin et al., 2011; Wootton et al., 2012; Davies et al., 2015). The increase in the proportion of larger lobster (> 100 mm CL) reported in Figure 2 and the overall higher number of lobsters observed was expected due to the closure of the site when compared to the baseline data. Prior to the wind farm being opened, the size distribution of lobsters within the wind farm was significantly different to the control site and the wind farm baseline distribution (Figure 3). The density of lobsters can be influenced by the availability of shelters within a habitat (Ball et al., 2001; Steneck 2006). The size of lobsters within a population has also been demonstrated to be limited by the size and number of shelters available (Bushmann and Atema 1997; Debuse et al., 1999). The addition of scour stone protection to the base of each monopole could potentially increase the available habitat and shelters for lobsters. The Westermost Rough OWF site was subjected to boulder removal prior to construction so the additional habitat creation may have been negated by the boulder removal. As the difference in size within the wind farm was described by lobsters over 75 mm CL (Figure 3d), the absence of fishing effort in the site is most likely to have greater influence than the habitat change. Opening of the site to fishing exploitation led to a rapid, short-term increase in landings from the wind farm in comparison to the surrounding area. After the wind farm was reopened to fishing, the previously unfished population of larger lobsters was reduced by intensive fishing over a short period. This reduction, however, did not drop below that reflected by the control site, which was subjected to exploitation for the entire period. The proportion of lobsters above MLS did not differ between the wind farm and the control site in the period after opening to fishing (Figure 4d). However, there was a greater proportion of smaller lobsters observed in the control site than within the wind farm (Figure 4d). The presence of a greater abundance of larger lobsters may have deterred the smaller lobsters from the wind farm (Steneck 2006; Émond et al., 2010). Their immigration into the site once lobsters above MLS were again being exploited may not have occurred during the timeframe of the survey. The smaller lobsters may have also been displaced into areas surrounding the wind farm due to inter-specific competition for resources (Wahle et al., 2013), which was reflected in the control site (Figures 3b and 4b). This indicates potential overspill effects, however, of the pre-recruits rather than recruits into a fishery. There are also influences in catch dynamics of a creel, smaller lobsters may use creels as shelter from predation and as feeding stations (Grabowski et al., 2010). The greater abundance of larger lobsters in the area that were subsequently caught in the survey creels, may have deterred the smaller lobsters from entering the creels (Jury et al., 2001). This interaction could have skewed the data to present a population biased in favour of larger lobsters. Alternatively the construction of the OWF and associated disturbance may have had a greater effect on the smaller, less robust lobsters (Rodmell and Johnson, 2002). As the fishing pressure returns to a stable state, again removing lobsters above MLS, it’s likely that smaller lobsters will again be observed within the area. Catch and landings per unit of effort The increase in lobster abundance observed in 2015 was reflected by the CPUE, showing a significant increase in catch rate of lobsters for both sites in 2015. There was no significant difference in catch rates of lobsters between the wind farm and control site in 2015 (Figure 5). Indicating that the difference between the years was due to natural variation and not just the closure of the wind farm. The LPUE, i.e. the number of lobsters of good quality per string that were above the MLS of 87 mm CL also showed a significant increase within the wind farm between years and with the control site in 2015. Wootton et al. (2012) and Davies et al. (2015) both observed a greater prevalence of injury and disease in lobsters within an NTZ. The increased LPUE observed in this study, indicating a greater abundance of lobsters without injury above MLS, suggests that this was not the case here. This could be attributed to the period of closure, as this site was only closed for 20 months in comparison to 2 and 5 years in their respective studies. The area may not have been closed long enough for true competition of resources that can result in increased occurrences of injuries. After the wind farm was opened to fishing the CPUE reduced significantly within the wind farm when compared to when the site was closed. It also differed significantly to the control site after opening (Figure 6 and Table 5). This demonstrates the effect of opening the area to exploitation after a period of closure. The mean LPUE however, after opening of the wind farm did not differ significantly to the control site (Figure 6 and Table 5). This indicates that although effort was high, within a relatively short period (6 weeks survey period after opening), the landings fishermen were getting within the site reflected the surrounding area. It has been demonstrated that periodic (Murawski et al., 2000) or permanent (Bergman et al., 2014) closure of areas to exploitation can enhance commercial fisheries. Closure of areas can allow the larger, more fecund lobsters to contribute to the spawning stock without fishing pressure (Moland et al., 2010; Leal et al., 2012). Periodically closing and reopening of the site has the potential to offset the possible detrimental effects of a permanent NTZ as observed by Wootton et al. (2012) and Davies et al. (2015). Economic loss to the fishery of a closed area may be offset by the increased earning potential once the site has been opened. Figure 6 demonstrates a 22% increase in LPUE in comparison to the control site. This, however, was only a short-term effect as the LPUE returned to background levels within the 6-week period of the survey. There is the potential for OWFs with their easily identifiable delineation to be used as a stock management tool for lobster fisheries. Combined with other suitable and easily identifiable sites, rotational closures could protect spawning stocks whilst offsetting economic loss and detrimental effects of permanently closed areas (Cohen and Foale, 2013). Conclusion This study has demonstrated the short-term effects on size distribution, CPUE and LPUE of OWF construction on a commercially important lobster fishery. The construction of the OWF created a temporary NTZ during the construction period, which resulted in an increase in larger, good quality lobsters in comparison to both the baseline data and control sites. The opening of the wind farm during the sampling period has highlighted that exploitation levels immediately following reopening of a site are high but quickly return to reflect surrounding areas. This study, whilst spatially limited has also presented a BACI approach to monitoring effects of OWF construction. Presenting a high number of individuals sampled, that represented the main fishing season for lobsters in the area. The collaboration between industry and developers has led to a study using industry data collection, that enables a high number of lobsters sampled, to aid in addressing a current gap in the literature. Subsequent monitoring of the site will highlight any longer-term effects of the OWF construction and its operation on the local lobster stocks when fishing exploitation is stable. Opening of the site during the sampling period has also highlighted the potential for OWF sites to be used as a stock management tool for periodic closures. Acknowledgements We would like to thank DONG Energy for funding the research and encouraging industry/developer collaboration. The crew of the R.V. Huntress for extensive sea time and advice on local fisheries and Rebecca Skirrow for assistance with data recording. The authors would also like to thank the reviewers for their advice in improving the manuscript. References Bailey H. , Senior B., Simmons D., Rusin J., Picken G., Thompson P. M. 2010 . Assessing underwater noise levels during pile-driving at an offshore windfarm and its potential effects on marine mammals . Marine Pollution Bulletin , 60 : 888 – 897 . Google Scholar Crossref Search ADS PubMed WorldCat Ball B. , Linnane A., Munday B., Browne R., Mercer J. P. 2001 . The effect of cover on in situ predation in early benthic phase European lobster Homarus gammarus . Journal of Marine Biological Association , 81 : 639 – 642 . Google Scholar Crossref Search ADS WorldCat Bannister R. C. A. , Addison J. T. 1998 . Enhancing lobster stocks: a review of recent European methods, results, and future prospects . Bulletin of Marine Science , 62 : 369 – 387 . Google Scholar OpenURL Placeholder Text WorldCat Bates D. , Mächler M., Bolker B. M., Walker S. C. 2015 . Fitting linear mixed-effects models using lme4 . Journal of Statistical Software , 67 : 1 – 48 . Google Scholar Crossref Search ADS WorldCat Bergman M. J. N. , Ubels S. M., Duineveld G. C. A., Meesters E. W. G. 2014 . Effects of a 5-year trawling ban on the local benthic community in a wind farm in the Dutch coastal zone . ICES Journal of Marine Science , 71 : 1088 – 1099 . Google Scholar Crossref Search ADS WorldCat Brandt M. J. , Diederichs A., Betke K., Nehls G. 2011 . Responses of harbour porpoises to pile driving at the Horns Rev II offshore wind farm in the Danish North Sea . Marine Ecology Progress Series , 421 : 205 – 216 . Google Scholar Crossref Search ADS WorldCat Brehme C. E. , McCarron P., Tetreault H. 2015 . A dasymetric map of Maine lobster trap distribution using local knowledge . The Professional Geographer , 67 : 98 – 109 . Google Scholar Crossref Search ADS WorldCat Breton S. P. , Moe G. 2009 . Status, plans and technologies for offshore wind turbines in Europe and North America . Renewable Energy , 34 : 646 – 654 . Google Scholar Crossref Search ADS WorldCat Bushmann P. J. , Atema J. 1997 . Shelter sharing and chemical courtship signals in the lobster Homarus americanus . Canadian Journal of Fisheries & Aquatic Sciences , 54 : 647 – 654 . Google Scholar Crossref Search ADS WorldCat Carstensen J. , Henriksen O. D., Teilmann J. 2006 . Impacts of offshore wind farm construction on harbour porpoises: acoustic monitoring of echo-location activity using porpoise detectors (T-PODs) . Marine Ecology Progress Series , 321 : 295 – 308 . Google Scholar Crossref Search ADS WorldCat Caveen A. J. , Fitzsimmons C., Pieraccini M., Dunn E., Sweeting C. J., Johnson M. L., Bloomfield H., et al. 2014 . Diverging strategies to planning an ecologically coherent network of MPAs in the North Sea: the roles of advocacy, evidence and pragmatism in the face of uncertainty . Advances in Marine Biology , 69 : 325 – 370 . Google Scholar Crossref Search ADS PubMed WorldCat Christie N. , Smyth K., Barnes R., Elliott M. 2014 . Co-location of activities and designations: a means of solving or creating problems in marine spatial planning? Marine Policy , 43 : 254 – 261 . Google Scholar Crossref Search ADS WorldCat Coates D. a. , Deschutter Y., Vincx M., Vanaverbeke J. 2014 . Enrichment and shifts in macrobenthic assemblages in an offshore wind farm area in the Belgian part of the North Sea . Marine Environmental Research , 95 : 1 – 12 . Google Scholar Crossref Search ADS PubMed WorldCat Cohen P. J. , Foale S. J. 2013 . Sustaining small-scale fisheries with periodically harvested marine reserves . Marine Policy , 37 : 278 – 287 . Google Scholar Crossref Search ADS WorldCat Davies C. E. , Johnson A. F., Wootton E. C., Greenwood S. J., Clark K. F., Vogan C. L., Rowley A. F. 2015 . Effects of population density and body size on disease ecology of the European lobster in a temperate marine conservation zone . ICES Journal of Marine Science , 72 : 128 – 138 . Google Scholar Crossref Search ADS WorldCat De Backer A. , Van Hoey G., Coates D., Vanaverbeke J., Hostens K. 2014 . Similar diversity-disturbance responses to different physical impacts: three cases of small-scale biodiversity increase in the Belgian part of the North Sea . Marine Pollution Bulletin , 84 : 251 – 262 . Google Scholar Crossref Search ADS PubMed WorldCat De Troch M. , Reubens J. T., Heirman E., Degraer S., Vincx M. 2013 . Energy profiling of demersal fish: a case-study in wind farm artificial reefs . Marine Environmental Research , 92 : 224 – 233 . Google Scholar Crossref Search ADS PubMed WorldCat Debuse V. J. , Addison J. T., Reynolds J. D. 1999 . Effects of breeding site density on competition and sexual selection in the European lobster . Behavioral Ecology , 14 : 396 – 402 . Google Scholar Crossref Search ADS WorldCat Émond K. , Sainte-Marie B., Gendron L. 2010 . Relative growth, life-history phases, and sexual maturity of American lobster (Homarus americanus) . Canadian Journal of Zoology , 88 : 347 – 358 . Google Scholar Crossref Search ADS WorldCat European Commission . 2016 . Proposal for a Directive of the European Parliament and of the Council on the promotion of the use of energy from renewable sources. Global Wind Energy Council . 2015 . Global Wind Report 2015. 75 pp. http://www.gwec.net/global-figures/wind-energy-global-status/ (last accessed 9 December 2016). Goñi R. , Hilborn R., Díaz D., Mallol S., Adlerstein S. 2010 . Net contribution of spillover from a marine reserve to fishery catches . Marine Ecology Progress Series , 400 : 233 – 244 . Google Scholar Crossref Search ADS WorldCat Grabowski J. H. , Clesceri E. J., Baukus A. J., Gaudette J., Weber M., Yund P. O. 2010 . Use of herring bait to farm lobsters in the Gulf of Maine . PLoS One , 5 : 1 – 11 . Google Scholar Crossref Search ADS WorldCat Griffin R. A. , Robinson G. J., West A., Gloyne-Phillips I. T., Unsworth R. K. F. 2016 . Assessing fish and motile fauna around offshore windfarms using stereo baited video . PLoS One , 11 : 1 – 14 . Google Scholar OpenURL Placeholder Text WorldCat Hart P. J. , Blyth R. B., Kaiser M. J., Edwards Jones G. 2002 . Sustainable exploitation with minimal conflict: is it possible? In Who Owns the Sea? pp. 11 – 24 . The University of Hull (ISDN 978-1-4092-5410-2). Holst R. , Revill A. 2009 . A simple statistical method for catch comparison studies . Fisheries Research , 95 : 254 – 259 . Google Scholar Crossref Search ADS WorldCat Hooper T. , Austen M. 2014 . The co-location of offshore windfarms and decapod fisheries in the UK: constraints and opportunities . Marine Policy , 43 : 295 – 300 . Google Scholar Crossref Search ADS WorldCat Hooper T. , Ashley M., Austen M. 2015 . Perceptions of fishers and developers on the co-location of offshore wind farms and decapod fisheries in the UK . Marine Policy , 61 : 16 – 22 . Google Scholar Crossref Search ADS WorldCat Hooper T. , Beaumont N., Hattam C. 2017 . The implications of energy systems for ecosystem services: a detailed case study of offshore wind . Renewable and Sustainable Energy Reviews , 70 : 230 – 241 . Google Scholar Crossref Search ADS WorldCat Hoskin M. G. , Coleman R. a., von Carlshausen E., Davis C. M. 2011 . Variable population responses by large decapod crustaceans to the establishment of a temperate marine no-take zone . Canadian Journal of Fisheries and Aquatic Sciences , 68 : 185 – 200 . Google Scholar Crossref Search ADS WorldCat Jury S. H. , Howell H., O'Grady D. F., Watson W. H. III, 2001 . Lobster trap video: in situ video surveillance of the behaviour of Homarus americanus in and around traps . Marine and Freshwater Research , 52 : 1125 – 1132 . Google Scholar Crossref Search ADS WorldCat Kaiser M. J. 2005 . Are marine protected areas a red herring or fisheries panacea? Canadian Journal of Fisheries and Aquatic Sciences , 62 : 1194 – 1199 . Google Scholar Crossref Search ADS WorldCat Kellner J. B. , Tetreault I., Gaines S. D., Nisbet R. M. 2007 . Fishing the line near marine reserves in single and multispecies fisheries . Ecological Applications , 17 : 1039 – 1054 . Google Scholar Crossref Search ADS PubMed WorldCat Kota S. , Bayne S. B., Nimmagadda S. 2015 . Offshore wind energy: a comparative analysis of UK, USA and India . Renewable and Sustainable Energy Reviews , 41 : 685 – 694 . Google Scholar Crossref Search ADS WorldCat Krone R. , Dederer G., Kanstinger P., Krämer P., Schneider C., Schmalenbach I. 2017 . Mobile demersal megafauna at common offshore wind turbine foundations in the German Bight (North Sea) two years after deployment – increased production rate of Cancer pagurus . Marine Environmental Research , 123 : 53 – 61 . Google Scholar Crossref Search ADS PubMed WorldCat Leal M. C. , Pochelon P. N., Silva T. L., Reis A., Rosa R., Calado R. 2012 . Variable within-brood maternal provisioning in newly extruded embryos of Homarus gammarus . Marine Biology , 160 : 763 – 772 . Google Scholar Crossref Search ADS WorldCat Lindeboom H. , Degraer S., Dannheim J., Gill A. B., Wilhelmsson D. 2015 . Offshore wind park monitoring programmes, lessons learned and recommendations for the future . Hydrobiologia , 756 : 169 – 180 . Google Scholar Crossref Search ADS WorldCat Lindeboom H. J. , Kouwenhoven H. J., Bergman M. J. N., Bouma S., Brasseur S., Daan R., Fijn R. C., et al. 2011 . Short-term ecological effects of an offshore wind farm in the Dutch coastal zone; a compilation . Environmental Research Letters , 6 : 35101 . Google Scholar Crossref Search ADS WorldCat Madsen P. T. , Wahlberg M., Tougaard J., Tyack P. 2006 . Wind turbine underwater noise and marine mammals: implications of current knowledge and data needs . Marine Ecology Progress Series , 309 : 279 – 295 . Google Scholar Crossref Search ADS WorldCat Marine Management Organisation . 2015 . UK sea fisheries annual statistics report 2014. https://www.gov.uk/government/statistics/uk-sea-fisheries-annual-statistics-report-2014 (last accessed 16 June 2015). McClanahan T. R. , Mangi S. 2000 . Spillover of exploitable fishes from a marine park and its effect on the adjacent fishery author . Ecological Applications , 10 : 1792 – 1805 . Google Scholar Crossref Search ADS WorldCat Moland E. , Olsen E. M., Stenseth N. C. 2010 . Maternal influences on offspring size variation and viability in wild European lobster Homarus gammarus . Marine Ecology Progress Series , 400 : 165 – 173 . Google Scholar Crossref Search ADS WorldCat Moland E. , Olsen E. M., Andvord K., Knutsen J. A., Stenseth N. C. 2011 . Home range of European lobster (Homarus gammarus) in a marine reserve: implications for future reserve design . Canadian Journal of Fisheries and Aquatic Sciences , 1210 : 1197 – 1210 . Google Scholar Crossref Search ADS WorldCat Moland E. , Olsen E. M., Knutsen H., Garrigou P., Espeland S. H., Kleiven A. R., André C., et al. 2013 . Lobster and cod benefit from small-scale northern marine protected areas: inference from an empirical before-after control-impact study . Proceedings. Biological Sciences/the Royal Society , 280 : 20122679 . Google Scholar Crossref Search ADS WorldCat Murawski S. A. , Brown R., Lai H. L., Rago P. J., Hendrickson L. 2000 . Large-scale closed areas as a fishery tool in temperate marine systems : the Georges Bank experience . Bulletin of Marine Science , 66 : 775 – 798 . Google Scholar OpenURL Placeholder Text WorldCat Ogle, D. 2016. Cumulative Frequencies, pp. 111–112. In Introductory Fisheries Analysis with R. Ed. by Ogle, D. Taylor and Francis Group, Boca Raton. R Core Team 2017 . R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (last accessed 15 January 2017). Rodmell D. , Johnson M. 2002 . The development of marine based wind energy generation and inshore fisheries in UK waters: are they compatible? In Who owns the sea? pp. 76 – 103 . The University of Hull (ISDN 978-1-4092-5410-2). Smith I. P. , Collins K. J., Jensen a. C. 1998 . Movement and activity patterns of the European lobster, Homarus gammarus, revealed by electromagnetic telemetry . Marine Biology , 132 : 611 – 623 . Google Scholar Crossref Search ADS WorldCat Smyth K. , Christie N., Burdon D., Atkins J. P., Barnes R., Elliott M. 2015 . Renewables-to-reefs? – Decommissioning options for the offshore wind power industry . Marine Pollution Bulletin , 90 : 247 – 258 . Google Scholar Crossref Search ADS PubMed WorldCat Stelzenmüller V. , Diekmann R., Bastardie F., Schulze T., Berkenhagen J., Kloppmann M., Krause G., et al. 2016 . Co-location of passive gear fisheries in offshore wind farms in the German EEZ of the North Sea: a first socio-economic scoping . Journal of Environmental Management , 183 : 794 – 805 . Google Scholar Crossref Search ADS PubMed WorldCat Stenberg C. , Støttrup J., van Deurs M., Berg C., Dinesen G., Mosegaard H., Grome T., et al. 2015 . Long-term effects of an offshore wind farm in the North Sea on fish communities . Marine Ecology Progress Series , 528 : 257 – 265 . Google Scholar Crossref Search ADS WorldCat Steneck R. S. 2006 . Possible demographic consequences of intraspecific shelter competition among American lobsters . Journal of Crustacean Biology , 26 : 628 – 638 . Google Scholar Crossref Search ADS WorldCat The Crown Estate . 2017 . Offshore wind energy; Project details. London, UK. https://www.thecrownestate.co.uk/energy-minerals-and-infrastructure/offshore-wind-energy/our-portfolio/project-details (last accessed 12 February 2017). Thomas R. , Lello J., Medeiros R., Pollard A., Seward A., Smith J., Vafidis J., et al. 2015 . Data Analysis with R Statistical Software A Guidebook for Scientists, pp. 103–112. In Statistical Models: Generalised Linear Mixed Models. Ed. by Thomas, R. Eco-explore, Cardiff, UK. Thomsen F. , Lüdemann K., Kafemann R., Piper W. 2008 . Effects of offshore wind farm noise on marine mammals and fish . Bioacoustics , 17 : 221 – 223 . Google Scholar Crossref Search ADS WorldCat Turner R. a. , Gray T., Polunin N. V. C., Stead S. M. 2013 . Territoriality as a driver of fishers’ spatial behavior in the Northumberland lobster fishery . Society & Natural Resources , 26 : 491 – 505 . Google Scholar Crossref Search ADS WorldCat US Department of Energy . 2012 . Energy Efficiency & Renewable Energy; wind and water program. https://www.nrel.gov/docs/fy12osti/49222.pdf (last accessed 12 February 2017). Van Marlen B. , Wiegerinck J. A. M., van Os-Koomen E., van Barneveld E. 2014 . Catch comparison of flatfish pulse trawls and a tickler chain beam trawl . Fisheries Research , 151 : 57 – 69 . Google Scholar Crossref Search ADS WorldCat Vandendriessche S. , Derweduwen J., Hostens K. 2015 . Equivocal effects of offshore wind farms in Belgium on soft substrate epibenthos and fish assemblages . Hydrobiologia , 756 : 19 – 35 . Google Scholar Crossref Search ADS WorldCat Vogel C. , Kopp D., Méhault S. 2017 . From discard ban to exemption: how can gear technology help reduce catches of undersized Nephrops and hake in the Bay of Biscay trawling fleet? Journal of Environmental Management , 186 : 96 – 107 . Google Scholar Crossref Search ADS PubMed WorldCat Wahlberg M. , Westerberg H. 2005 . Hearing in fish and their reactions to sounds from offshore wind farms . Marine Ecology Progress Series , 288 : 295 – 309 . Google Scholar Crossref Search ADS WorldCat Wahle R. A. W. , Brown C., Hovel K. 2013 . The geography and body-size dependence of top-down forcing in New England's lobster-groundfish interaction . Bulletin of Marine Science , 89 : 189 – 212 . Google Scholar Crossref Search ADS WorldCat Wootton E. C. , Woolmer A. P., Vogan C. L., Pope E. C., Hamilton K. M., Rowley A. F. 2012 . Increased disease calls for a cost-benefits review of marine reserves . PLoS One , 7 : e51615 . Google Scholar Crossref Search ADS PubMed WorldCat Zuur A. F. , Ieno E. N. 2016 . A protocol for conducting and presenting results of regression-type analyses . Methods in Ecology and Evolution , 7 : 636 – 645 . Google Scholar Crossref Search ADS WorldCat © International Council for the Exploration of the Sea 2018. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © International Council for the Exploration of the Sea 2018. All rights reserved. For Permissions, please email: [email protected]

Lynch, Patrick D; Shertzer, Kyle W; Cortés, Enric; Latour, Robert J

doi: 10.1093/icesjms/fsy008pmid: N/A

Abstract Relative abundance trends of highly migratory species (HMS) have played a central role in debates over the health of global fisheries. However, such trends have mostly been inferred from fishery catch rates, which can provide misleading signals of relative abundance. While many biases are accounted for through traditional catch rate standardization, pelagic habitat fished is rarely directly considered. Using a method that explicitly accounts for temperature regimes, we analysed data from the US pelagic longline fishery to estimate relative abundance trends for 34 HMS in the Atlantic Ocean from 1987 through 2013. This represents one of the largest studies of HMS abundance trends. Model selection emphasized the importance of accounting for pelagic habitat fished with water column temperature being included in nearly every species’ model, and in extreme cases, a temperature variable explained 50–60% of the total deviance. Our estimated trends represent observations from one fishery only, and a more integrated stock assessment should form the basis for conclusions about stock status overall. Nonetheless, our trends serve as indicators of stock abundance and they suggest that a majority of HMS (71% of analysed species) are either declining in relative abundance or declined initially with no evidence of rebuilding. Conversely, 29% of the species exhibited stable, increasing, or recovering trends; however, these trends were more prevalent among tunas than either billfishes or sharks. By estimating the effects of pelagic habitat on fishery catch rates, our results can be used in combination with ocean temperature trends and forecasts to support bycatch avoidance and other time-area management decisions. Introduction Fish stock assessments provide the quantitative basis for sustainable fisheries management. Assessment models typically rely on information about changes in stock abundance over time, and because it is impossible to conduct a census of most marine organisms, indices of relative abundance are often used to characterize population trends (Quinn and Deriso, 1999; Maunder and Punt, 2004). Within assessment models, indices are often treated as “observed” measures of relative abundance, thereby giving them substantial influence over assessment results. Unfortunately, relative abundance trends of highly migratory species (HMS) are rarely obtained through comprehensive, scientifically designed, survey programs (due to the high cost of implementation), but rather from fishery-dependent catch and effort data (Maunder and Punt, 2004; Lynch et al., 2011) (HMS in this study include fishes only [tunas, billfish, and sharks]). This poses a considerable challenge to estimating an accurate index of relative abundance, because fisheries frequently change their fishing practices in response to various socio-economic drivers. When fishery catch rates, or catch per unit effort (CPUE), are assumed to be proportional to stock abundance, changes in fishing practices need to be accounted for because they can cause the proportionality assumption to be violated (Maunder and Punt, 2004). In the Atlantic Ocean, pelagic longline fisheries are responsible for the bulk of the fishing mortality experienced by many HMS. These fisheries have altered fishing practices over time by changing gear configurations, target species, and the spatio-temporal distribution of effort (Majkowski, 2007). Although contemporary statistical approaches to estimating HMS relative abundance trends do account for changes in fishing practices, ocean conditions are variable and pelagic habitats fished are related to both fishing practices and environmental conditions. While the distributions of HMS can be roughly characterized by depth and geography, temperature regimes are likely the main governing factor (Brill and Lutcavage, 2001; Bigelow and Maunder, 2007). Therefore, when estimating HMS relative abundance trends, it is important to consider pelagic habitats exploited (e.g. temperature regimes) in addition to fishing practices. Temperature information is not straightforward to incorporate analytically when estimating relative abundance trends from pelagic longline fisheries data, because estimates of fishing depth and environmental conditions at depth are required. Longline fishing depths are notoriously difficult to estimate with accuracy (Ward and Myers, 2006; Rice et al., 2007) and environmental conditions at a given depth, time, and location are often not recorded, and can only be estimated through analysis of a global ocean database. Therefore, HMS relative abundance trends are typically estimated without accounting for the pelagic habitats exploited by the fishery, which inevitably vary over time. Despite the challenges associated with accounting for pelagic habitat fished, Lynch et al. (2012) proposed a method for incorporating this information using a delta-generalized linear model (delta-GLM), and showed that it can improve the estimation accuracy of HMS relative abundance trends. The method is also relatively insensitive to errors in estimates of longline fishing depths, which is contrary to other methods that incorporate habitat, such as habitat-based standardization (HBS; Hinton and Nakano, 1996) and the statistical counterpart to HBS (statHBS; Maunder et al., 2006). The HBS and statHBS approaches have both demonstrated high sensitivity to model inputs, such as estimates of longline fishing depth (Goodyear, 2003; Lynch et al., 2012). For fisheries stock assessments of Atlantic HMS, we are unaware of any occasions where the relative abundance trends used in the assessment incorporated detailed pelagic habitat information. Here, we accounted for temperature regimes in the application of delta-GLMs (some of which included mixed effects; i.e. delta-GLMMs) to fisher logbook data from the US pelagic longline fishery (USLL). These analyses resulted in new abundance trends for 34 HMS (Table 1) in the Atlantic Ocean. For comparison, we also analysed data collected by scientific observers aboard pelagic longline fishing vessels (US Pelagic Longline Observer Program). In general, relative abundance trends for species caught in the USLL are estimated by US members of the Standing Committee on Research and Statistics (SCRS), a committee within the International Commission for the Conservation of Atlantic Tunas (ICCAT). All of the 34 species analysed fall under the management purview of ICCAT, either as directly managed species or as bycatch species. However, not all species managed by ICCAT have been formally assessed using modern stock assessment methods. To our knowledge, only 13 of the 34 species (38%) have been assessed (Table 1). Table 1. Species for which abundance trends were generated using fisher logbook and pelagic longline observer program data from the USLL. Speciesa . Logbook . Observer . Species . Logbook . Observer . Swordfish, Xiphias gladius 256643 (99.6%) 17496 (100.0%) Silky shark, Carcharhinus falciformis 145539 (56.5%) 15333 (87.6%) Yellowfin tuna, Thunnus albacares 255815 (99.3%) 17496 (100.0%) Bigeye thresher, Alopias superciliosus 141026 (54.8%) 17496 (100.0%) Dolphinfish, Coryphaena hippurus 253666 (98.4%) — Dusky shark, Carcharhinus obscurus 137124 (53.2%) 15333 (87.6%) Bigeye tuna, Thunnus obesus 243036 (94.4%) 17496 (100.0%) Blacktip shark, Carcharhinus limbatus 125346 (48.7%) 14460 (82.6%) Wahoo, Acanthocybium solandri 233435 (90.6%) — Spearfishes, Tetrapturus spp. 105661 (41.0%) — Blue marlin, Makaira nigricans 221178 (85.9%) — Sandbar shark, Carcharhinus plumbeus 108111 (42.0%) 14235 (81.4%) Albacore tuna, Thunnus alalunga 225525 (87.6%) 17496 (100.0%) Oceanic whitetip shark, Carcharhinus longimanus 95149 (36.9%) 15333 (87.6%) White marlin, Kajikia albida 220633 (85.7%) — Skipjack tuna, Katsuwonus pelamis 97107 (37.7%) 15333 (87.6%) Atlantic bluefin tuna, Thunnus thynnus 218430 (84.8%) 17292 (98.8%) Night shark, Carcharhinus signatus 71202 (27.6%) 14664 (83.8%) Longfin mako, Isurus paucus 203654 (79.1%) 15333 (87.6%) Scalloped hammerhead, Sphyrna lewini 54493 (21.2%) 15129 (86.5%) Blue shark, Prionace glauca 198479 (77.1%) 17496 (100.0%) Atlantic bonito, Sarda sarda 49258 (19.1%) — Tiger shark, Galeocerdo cuvier 193050 (74.9%) 17496 (100.0%) Smooth hammerhead, Sphyrna zygaena 28920 (11.2%) 4072 (23.3%) Hammerhead sharks, Sphyrna spp. 186753 (72.5%) 15333 (87.6%) White shark, Carcharodon carcharias 34393 (13.4%) — Shortfin mako, Isurus oxyrinchus 186905 (72.6%) 17496 (100.0%) Spinner shark, Carcharhinus brevipinna 34608 (13.4%) 11773 (67.3%) Blackfin tuna, Thunnus atlanticus 188078 (73.0%) — Porbeagle, Lamna nasus 16384 (6.4%) 5739 (32.8%) Oilfish, Gempylidae spp. 173749 (67.5%) — Bignose shark, Carcharhinus altimus 13527 (5.3%) — Sailfish, Istiophorus albicans 163142 (63.3%) — Common thresher, Alopias vulpinus 166262 (64.5%) 11232 (87.6%) Speciesa . Logbook . Observer . Species . Logbook . Observer . Swordfish, Xiphias gladius 256643 (99.6%) 17496 (100.0%) Silky shark, Carcharhinus falciformis 145539 (56.5%) 15333 (87.6%) Yellowfin tuna, Thunnus albacares 255815 (99.3%) 17496 (100.0%) Bigeye thresher, Alopias superciliosus 141026 (54.8%) 17496 (100.0%) Dolphinfish, Coryphaena hippurus 253666 (98.4%) — Dusky shark, Carcharhinus obscurus 137124 (53.2%) 15333 (87.6%) Bigeye tuna, Thunnus obesus 243036 (94.4%) 17496 (100.0%) Blacktip shark, Carcharhinus limbatus 125346 (48.7%) 14460 (82.6%) Wahoo, Acanthocybium solandri 233435 (90.6%) — Spearfishes, Tetrapturus spp. 105661 (41.0%) — Blue marlin, Makaira nigricans 221178 (85.9%) — Sandbar shark, Carcharhinus plumbeus 108111 (42.0%) 14235 (81.4%) Albacore tuna, Thunnus alalunga 225525 (87.6%) 17496 (100.0%) Oceanic whitetip shark, Carcharhinus longimanus 95149 (36.9%) 15333 (87.6%) White marlin, Kajikia albida 220633 (85.7%) — Skipjack tuna, Katsuwonus pelamis 97107 (37.7%) 15333 (87.6%) Atlantic bluefin tuna, Thunnus thynnus 218430 (84.8%) 17292 (98.8%) Night shark, Carcharhinus signatus 71202 (27.6%) 14664 (83.8%) Longfin mako, Isurus paucus 203654 (79.1%) 15333 (87.6%) Scalloped hammerhead, Sphyrna lewini 54493 (21.2%) 15129 (86.5%) Blue shark, Prionace glauca 198479 (77.1%) 17496 (100.0%) Atlantic bonito, Sarda sarda 49258 (19.1%) — Tiger shark, Galeocerdo cuvier 193050 (74.9%) 17496 (100.0%) Smooth hammerhead, Sphyrna zygaena 28920 (11.2%) 4072 (23.3%) Hammerhead sharks, Sphyrna spp. 186753 (72.5%) 15333 (87.6%) White shark, Carcharodon carcharias 34393 (13.4%) — Shortfin mako, Isurus oxyrinchus 186905 (72.6%) 17496 (100.0%) Spinner shark, Carcharhinus brevipinna 34608 (13.4%) 11773 (67.3%) Blackfin tuna, Thunnus atlanticus 188078 (73.0%) — Porbeagle, Lamna nasus 16384 (6.4%) 5739 (32.8%) Oilfish, Gempylidae spp. 173749 (67.5%) — Bignose shark, Carcharhinus altimus 13527 (5.3%) — Sailfish, Istiophorus albicans 163142 (63.3%) — Common thresher, Alopias vulpinus 166262 (64.5%) 11232 (87.6%) The number and percent of logbook and observer records analysed (of a potential 257581 logbook and 17496 observer records) after filtering the data to include only the regions and vessels with catch rates above predetermined thresholds. We did not have observer data for 11 of the species analysed. Species highlighted in bold text are those for which stock assessments are known to have been previously conducted. a In addition to individual species, there were three species groups (i.e. identified to the genus level) included in the analyses: oilfish (Gempylidae spp.), spearfishes (Tetrapturus spp.), and hammerhead sharks (Sphyrna spp.). We use “HMS” and “species” throughout to collectively refer to individual species and species groups. Open in new tab Table 1. Species for which abundance trends were generated using fisher logbook and pelagic longline observer program data from the USLL. Speciesa . Logbook . Observer . Species . Logbook . Observer . Swordfish, Xiphias gladius 256643 (99.6%) 17496 (100.0%) Silky shark, Carcharhinus falciformis 145539 (56.5%) 15333 (87.6%) Yellowfin tuna, Thunnus albacares 255815 (99.3%) 17496 (100.0%) Bigeye thresher, Alopias superciliosus 141026 (54.8%) 17496 (100.0%) Dolphinfish, Coryphaena hippurus 253666 (98.4%) — Dusky shark, Carcharhinus obscurus 137124 (53.2%) 15333 (87.6%) Bigeye tuna, Thunnus obesus 243036 (94.4%) 17496 (100.0%) Blacktip shark, Carcharhinus limbatus 125346 (48.7%) 14460 (82.6%) Wahoo, Acanthocybium solandri 233435 (90.6%) — Spearfishes, Tetrapturus spp. 105661 (41.0%) — Blue marlin, Makaira nigricans 221178 (85.9%) — Sandbar shark, Carcharhinus plumbeus 108111 (42.0%) 14235 (81.4%) Albacore tuna, Thunnus alalunga 225525 (87.6%) 17496 (100.0%) Oceanic whitetip shark, Carcharhinus longimanus 95149 (36.9%) 15333 (87.6%) White marlin, Kajikia albida 220633 (85.7%) — Skipjack tuna, Katsuwonus pelamis 97107 (37.7%) 15333 (87.6%) Atlantic bluefin tuna, Thunnus thynnus 218430 (84.8%) 17292 (98.8%) Night shark, Carcharhinus signatus 71202 (27.6%) 14664 (83.8%) Longfin mako, Isurus paucus 203654 (79.1%) 15333 (87.6%) Scalloped hammerhead, Sphyrna lewini 54493 (21.2%) 15129 (86.5%) Blue shark, Prionace glauca 198479 (77.1%) 17496 (100.0%) Atlantic bonito, Sarda sarda 49258 (19.1%) — Tiger shark, Galeocerdo cuvier 193050 (74.9%) 17496 (100.0%) Smooth hammerhead, Sphyrna zygaena 28920 (11.2%) 4072 (23.3%) Hammerhead sharks, Sphyrna spp. 186753 (72.5%) 15333 (87.6%) White shark, Carcharodon carcharias 34393 (13.4%) — Shortfin mako, Isurus oxyrinchus 186905 (72.6%) 17496 (100.0%) Spinner shark, Carcharhinus brevipinna 34608 (13.4%) 11773 (67.3%) Blackfin tuna, Thunnus atlanticus 188078 (73.0%) — Porbeagle, Lamna nasus 16384 (6.4%) 5739 (32.8%) Oilfish, Gempylidae spp. 173749 (67.5%) — Bignose shark, Carcharhinus altimus 13527 (5.3%) — Sailfish, Istiophorus albicans 163142 (63.3%) — Common thresher, Alopias vulpinus 166262 (64.5%) 11232 (87.6%) Speciesa . Logbook . Observer . Species . Logbook . Observer . Swordfish, Xiphias gladius 256643 (99.6%) 17496 (100.0%) Silky shark, Carcharhinus falciformis 145539 (56.5%) 15333 (87.6%) Yellowfin tuna, Thunnus albacares 255815 (99.3%) 17496 (100.0%) Bigeye thresher, Alopias superciliosus 141026 (54.8%) 17496 (100.0%) Dolphinfish, Coryphaena hippurus 253666 (98.4%) — Dusky shark, Carcharhinus obscurus 137124 (53.2%) 15333 (87.6%) Bigeye tuna, Thunnus obesus 243036 (94.4%) 17496 (100.0%) Blacktip shark, Carcharhinus limbatus 125346 (48.7%) 14460 (82.6%) Wahoo, Acanthocybium solandri 233435 (90.6%) — Spearfishes, Tetrapturus spp. 105661 (41.0%) — Blue marlin, Makaira nigricans 221178 (85.9%) — Sandbar shark, Carcharhinus plumbeus 108111 (42.0%) 14235 (81.4%) Albacore tuna, Thunnus alalunga 225525 (87.6%) 17496 (100.0%) Oceanic whitetip shark, Carcharhinus longimanus 95149 (36.9%) 15333 (87.6%) White marlin, Kajikia albida 220633 (85.7%) — Skipjack tuna, Katsuwonus pelamis 97107 (37.7%) 15333 (87.6%) Atlantic bluefin tuna, Thunnus thynnus 218430 (84.8%) 17292 (98.8%) Night shark, Carcharhinus signatus 71202 (27.6%) 14664 (83.8%) Longfin mako, Isurus paucus 203654 (79.1%) 15333 (87.6%) Scalloped hammerhead, Sphyrna lewini 54493 (21.2%) 15129 (86.5%) Blue shark, Prionace glauca 198479 (77.1%) 17496 (100.0%) Atlantic bonito, Sarda sarda 49258 (19.1%) — Tiger shark, Galeocerdo cuvier 193050 (74.9%) 17496 (100.0%) Smooth hammerhead, Sphyrna zygaena 28920 (11.2%) 4072 (23.3%) Hammerhead sharks, Sphyrna spp. 186753 (72.5%) 15333 (87.6%) White shark, Carcharodon carcharias 34393 (13.4%) — Shortfin mako, Isurus oxyrinchus 186905 (72.6%) 17496 (100.0%) Spinner shark, Carcharhinus brevipinna 34608 (13.4%) 11773 (67.3%) Blackfin tuna, Thunnus atlanticus 188078 (73.0%) — Porbeagle, Lamna nasus 16384 (6.4%) 5739 (32.8%) Oilfish, Gempylidae spp. 173749 (67.5%) — Bignose shark, Carcharhinus altimus 13527 (5.3%) — Sailfish, Istiophorus albicans 163142 (63.3%) — Common thresher, Alopias vulpinus 166262 (64.5%) 11232 (87.6%) The number and percent of logbook and observer records analysed (of a potential 257581 logbook and 17496 observer records) after filtering the data to include only the regions and vessels with catch rates above predetermined thresholds. We did not have observer data for 11 of the species analysed. Species highlighted in bold text are those for which stock assessments are known to have been previously conducted. a In addition to individual species, there were three species groups (i.e. identified to the genus level) included in the analyses: oilfish (Gempylidae spp.), spearfishes (Tetrapturus spp.), and hammerhead sharks (Sphyrna spp.). We use “HMS” and “species” throughout to collectively refer to individual species and species groups. Open in new tab With the exception of the incorporation of pelagic habitat fished, our relative abundance trends were estimated following an approach used for yellowfin tuna (Thunnus ablacares) by the SCRS (Walter, 2011). This framework represents the contemporary approach used by the SCRS, so our trends can be compared to those estimated by the SCRS with minimal concern over methodological differences. The independent variables included in our final models were objectively selected by considering the percent of total deviance explained by each variable. This allowed us to compare the importance of the temperature variables as related to the variables normally considered by the SCRS. Finally, we characterized general population trends by calculating instantaneous rates of change for each species. We used a flexible approach to detect measurable changes in relative abundance trends over time. Methods Fishery data Relative abundance trends were generated for 34 HMS routinely caught by the USLL (Table 1). Fisher logbook and observer data for the USLL were obtained from the National Marine Fisheries Service. The logbook data contain longline set-specific information, including catches (numbers of individuals), effort (number of hooks), gear configurations, dates, time, and spatial locations (Figure 1). The primary target species of the USLL include swordfish (Xiphias gladius), yellowfin tuna, and bigeye tuna (Thunnus obesus); however, bycatch rates in this fishery are relatively high, particularly for sharks (Mandelman et al. 2008). While the USLL covers a large portion of the distributions of most species analysed, fishing effort is largely focused along the US east coast. The USLL in the early through mid-1970s was considered an “underground” fishery, and initially used a gear configuration similar to Japanese and Norwegian shark longline fisheries (Hoey and Bertolino, 1988). Between 1978 and 1983, various gear modifications occurred as the fishery evolved to using lighter monofilament line with increased hook spacing and depth, and chemical lightsticks. Other features of this fishery have been described in detail by Hoey and Bertolino (1988). Figure 1. Open in new tabDownload slide Map of the distribution of longline sets (total number per cell) between 1987 and 2010 for the USLL in the northwest Atlantic Ocean. The geographical regions used for classifying the fishery include the Caribbean Sea (CAR), Gulf of Mexico (GOM), Florida east coast (FEC), south Atlantic bight (SAB), mid-Atlantic bight (MAB), north-east coastal (NEC), north-east distant waters (NED), Sargasso Sea (SAR), and offshore waters (OFS). Figure 1. Open in new tabDownload slide Map of the distribution of longline sets (total number per cell) between 1987 and 2010 for the USLL in the northwest Atlantic Ocean. The geographical regions used for classifying the fishery include the Caribbean Sea (CAR), Gulf of Mexico (GOM), Florida east coast (FEC), south Atlantic bight (SAB), mid-Atlantic bight (MAB), north-east coastal (NEC), north-east distant waters (NED), Sargasso Sea (SAR), and offshore waters (OFS). While fishers continually adjust their practices, the logbook and observer programs track this information on a set-by-set basis, allowing catch rates to be analysed and interpreted accordingly. The logbook program began in 1986, although data for that year are incomplete; thus, our analyses use data beginning in 1987. The major gear changes described by Hoey and Bertolino (1988) occurred before the start of the logbook program, so there is not a need to address those shifts in this study; however, we do account for the variety of fishing practices and time/area dynamics observed since 1987. There have been several time-area management measures imposed on the USLL, particularly since 2000 (Mandelman et al. 2008; Walter, 2011). Our treatment of the data, including data filtering is described in the Supplementary data. Oceanographic data Detailed oceanographic data were necessary for generating estimates of pelagic habitats fished. We designated temperature regimes as habitats; therefore, we assigned each longline set a fixed temperature-at-depth profile. Ocean temperature profiles were obtained from the National Oceanographic Data Center (www.nodc.noaa.gov) using the World Ocean Atlas (WOA) data series (Locarnini et al., 2010). These data were available as average monthly temperature profiles following 1° latitude by 1° longitude spatial resolution, covering a depth range of 0–1500 m over variable increments. The climatologies were derived from averaging decadal climatologies between 1955 and 2006 (Locarnini et al., 2010). For the rare instances where temperature profiles were not available for a given combination of geographical location and month, the longline set record was removed entirely (<2% of the logbook records). Pelagic habitat variables To incorporate pelagic habitat fished, estimates of longline fishing depths and corresponding estimates of temperature at depth were required (Lynch et al., 2012). See Supplementary data for a description of the methods used to calculate longline hook depths. Fishing depths for each longline set were related to temperature at depth for the corresponding month and geographical location of the set. Because temperatures were available at discrete depths, the temperature at the depth closest to estimated fishing depth was specified as the temperature fished for a given hook. Following Lynch et al. (2012), temperatures fished were converted to 1°C increments relative to surface temperature in the corresponding time/space. The maximum deviation from sea surface temperature (MaxΔT), or deepest, coldest pelagic habitat fished, was then assigned to each longline set as a single value (0°, …, 15°C) thereby characterizing the contrast in temperatures fished for that set. For example, if surface water temperature is 25°C for a given longline set, and the temperature associated with the deepest hook fished in that set is 15°C, then the MaxΔT factor would have a value of 10°C for that set. In addition to MaxΔT, we evaluated a variable that characterized each longline set as the minimum temperature fished (MinT) in that set. This variable was specified as categorical with 5° temperature bins from 1°C to 30°C. In the example stated above, the MinT variable would have a value of 15°C for that set. While MaxΔT directly accounts for the vertical distribution of the species being analysed, MinT accounts for the distribution of the species geographically, as well as vertically. The inclusion of temperature regimes fished is a non-trivial undertaking, but an important consideration. While temperature is likely related to depth, the correlation between these variables is not perfect due to dynamic ocean patterns. Furthermore, HMS distributions and behaviour are more a function of temperature than depth (Brill and Lutcavage, 2001). Thus, we concluded it was crucial to estimate temperature regimes fished, rather than depths, which would have been simpler. The inclusion of these pelagic habitat variables represents the primary difference between our study and prior estimates of relative abundance for Atlantic HMS. Making only one change in methodology facilitated the comparison of results to previous work; however, it is important to consider if these new variables were correlated with any of the traditional variables (see Other variables), which may confound the comparisons. Because these habitat variables are included to account for potential biases due to the temperature-driven vertical distribution of HMS in the location of fishing, we conclude that the patterns in these variables are not captured by any of the traditional variables. Other variables A suite of additional explanatory variables was also considered in the analyses. These variables were modelled as categorical factors, and included Year (year in which the set occurred), Region (nine geographical regions commonly used to classify the longline fishery: Figure 1), Season (calendar quarters: January–March, April–June, July–September, October–December), Lightstick (the ratio of lightsticks per hook categorized with four levels: 0,  >0–0.4,  >0.4–0.7,  >0.7), hooks between floats (HBF) categorized with seven levels (0–3, 4–6, 7–9, 10–15, 16–21, 22–29, 30+), Time (time at the beginning of the set: a.m., p.m., or unknown), and Bait (type of bait used: live, dead, mixture, unknown). These variables are all thought to potentially affect catch rates of various species encountered by the USLL (Walter, 2011). Modelling framework We used a two-stage delta-GLM approach for estimating relative abundance trends (e.g. Aitchison, 1955; Lo et al., 1992; Stefánsson, 1996; Maunder and Punt, 2004). A GLM is a linear model that can accommodate non-normal error structure using a link function to relate dependent and independent variables. The delta-GLM (also referred to as a hurdle model) accounts for zero-inflated data by combining two GLMs, one that models the probability of observing a zero catch as a function of predictor variables and a separate model of the non-zero catches. The delta-GLM is represented as: Pr(Y=y)={w y=0(1−w)f(y) otherwise(1) where w is the probability of observing a zero for the response (CPUE) and f(y) is a model of the mean of non-zero data (CPUE). Accordingly, our abundance trends were determined by combining two linear models, one of which modelled the presence/absence of a particular species as a linear function of explanatory variables, assuming a binomial error distribution (logit link function). The second modelled CPUE, calculated as numbers of individuals caught in a set per 1000 hooks. For this model, only the records with a positive catch rate (i.e. CPUE > 0) were included, and we assumed a lognormal error distribution by using log(CPUE) as the response variable (identity link function). For both models, explanatory variables and interaction terms were modelled as fixed effects, with the exception of interaction terms that included the Year variable, which were modelled as random effects to facilitate deriving abundance estimates using the year effects. Technically, when random effects were included, delta-GLMMs were applied, but we use the term “GLM” generally throughout to refer to our modelling framework. Annual estimates of relative abundance were obtained by multiplying the probability of a positive catch rate (1 − w) in a given year from the binomial GLM by the mean CPUE in that same year from the lognormal GLM. The probability of a positive catch was calculated as the back-transformed mean probabilities for each year, predicted when all factors other than Year were set to their mode level (Maunder and Punt, 2004). Mean CPUE for each year was calculated as back-transformed year means adjusted by an infinite series lognormal bias correction (Lo et al., 1992), and standard errors of the annual abundance estimates were calculated using the delta method (Seber, 1982; Lo et al., 1992). Model selection We based the selection of variables to include in our component GLMs on percent deviance explained with a threshold for inclusion of 5%. This mimics the approach commonly used when estimating relative abundance trends for HMS (Ortiz and Arocha, 2004; Walter, 2011; Supplementary data). By incorporating our temperature variables (MaxΔT, MinT) into the established approach to model selection, we evaluated the importance of these variables relative to other variables commonly considered in these analyses. We considered all first-order interaction terms in our model selection exercise, but observed increasing model instability when multiple interaction terms were included. Thus, our final models only incorporated the interaction term that explained the highest percent of the total deviance (if the percent explained exceeded at least 5%). General patterns We used linear regression as a simple approach to characterizing the general patterns observed in our relative abundance trends (e.g. increasing/decreasing). Each trend was scaled to have a mean of one, and the general direction over time was estimated by regressing scaled relative abundance on Year (treated as a continuous variable). In addition to standard linear regression, we modelled each trend using piecewise, or segmented, regression with one breakpoint. We then used Akaike’s Information Criterion, corrected for small sample size (Burnham and Anderson, 2002) to select between standard and segmented regression models. This provided an objective characterization of the general pattern in abundance as being either unidirectional over time, or one that exhibited a change in direction. There may have been cases where trends could have been characterized by more than two segments, but to avoid overparameterization, we did not fit these more complex models. The slope parameters from the regression models represent instantaneous rates of change, and these were extracted for making comparisons across species. There were either one or two slope parameters for each species, depending on whether the standard or segmented regression model was selected for describing the abundance trend. We characterized the populations as stable over time when the slopes were not significantly different from zero, but when significantly positive or negative, we considered the populations to be increasing or decreasing, respectively. All quantitative analyses were implemented using the statistical programming language R (R Core Team, 2016). Results The USLL spatial coverage in the Atlantic Ocean can be characterized as broad with areas of concentrated fishing effort (Figure 1). Due to our data filtering technique (Supplementary data), we analysed a different number of USLL logbook records for each of the 34 HMS included in this study (Table 1). Observer data were not available for all species (Table 1), but when analysed, the number of available observer records was filtered by region (not by historical catches per vessel as with logbook records—Supplementary data). Species with more catch records (after data filtering was applied) tended to have a higher frequency of occurrence in the fishery (Figure 2a), but with the exception of swordfish and yellowfin tuna, positive catches were less frequent than catches equal to zero. Thus, most species we analysed were rarely encountered by the fishery. While our models accounted for excessive zeros in the data, the ability to infer population trajectories for rarely encountered species may be limited. Figure 2. Open in new tabDownload slide Number of logbook records analysed (a), including proportion of positive catch records for species captured in the USLL. Also, the percent of the total deviance explained by the habitat factors MaxΔT (b), and MinT (c) for analysis of presence/absence of a given species (Binomial) or the positive catch records (Positive). The deviance threshold used for determining inclusion of the variable in the final model (5%) was provided for reference (black line). Figure 2. Open in new tabDownload slide Number of logbook records analysed (a), including proportion of positive catch records for species captured in the USLL. Also, the percent of the total deviance explained by the habitat factors MaxΔT (b), and MinT (c) for analysis of presence/absence of a given species (Binomial) or the positive catch records (Positive). The deviance threshold used for determining inclusion of the variable in the final model (5%) was provided for reference (black line). A wide variety of model structures was selected for the binomial and positive catch component models of the delta-GLMs (Supplementary Tables S1–S34). According to our selection criteria (at least 5% of total deviance explained by the variable), the MinT habitat variable was selected for the binomial and/or positive models for almost every species (Figure 2c, Supplementary Tables S1–S35). This suggests that MinT explained a substantial amount of the variability in the catch rates of target and incidentally captured species of the USLL. For several species, MinT explained 50–60% of the total deviance. In addition to MinT, we evaluated MaxΔT; however, this variable explained greater than 5% of the total deviance for only five species (wahoo, blackfin tuna, Atlantic bonito, white marlin, and night shark), and in these cases, the percent explained was only slightly above the threshold for inclusion (Figure 2b). Overall, at least one of our pelagic habitat variables was important to include when estimating abundance trends for all but five species (yellowfin tuna, swordfish, spinner shark, white shark, and bignose shark). Estimates of MinT explained substantial variability surrounding observed CPUE, and visualizing the influence of this variable on species-specific catch rates highlights behavioural patterns (Figure 3). Encounter rates (proportion of sets with positive CPUE) and median positive catch rates both exhibited variability across estimates of MinT. The highest encounter rates and median positive CPUE values were observed for swordfish and blue sharks when the coldest habitats were fished. In fact, the highest overall median CPUE corresponded with blue sharks at approximately 50 sharks per 1000 hooks. Other species with higher catch rates in cooler habitats include bluefin tuna, shortfin mako, hammerhead sharks, and porbeagle. The encounter rates of swordfish and yellowfin tuna (two important target species of this fishery) exhibited opposing gradients in response to MinT, with the highest rates for yellowfin tuna occurring when the warmest habitats were fished. Along with yellowfin tuna, wahoo, blackfin tuna, oilfish, skipjack tuna, dolphinfish, the billfishes (excluding swordfish), tiger shark, thresher sharks, and night shark had higher encounter and catch rates in the warmer habitats. Figure 3. Open in new tabDownload slide Catch rates (CPUE) by species from the USLL, presented as the proportion of positive catches (a) and the median of the positive catches (b) observed in 5°C temperature bins corresponding with the estimated minimum temperature fished per set. Figure 3. Open in new tabDownload slide Catch rates (CPUE) by species from the USLL, presented as the proportion of positive catches (a) and the median of the positive catches (b) observed in 5°C temperature bins corresponding with the estimated minimum temperature fished per set. The majority of our relative abundance trends declined over the time series (Figure 4, Supplementary Tables S1–S35); however, the magnitude of change was highly variable. For instance, the declines observed for the primary target species, swordfish and yellowfin tuna, were much less severe than those observed for many of the sharks. When compared with relative abundance trends estimated from observer program data (Supplementary Figure S3), observer trends were more variable than those estimated from logbook data. Logbook and observer trends exhibited significant positive correlations for 57% of species (13 of the 23 species for which observer data were analysed). We also compared relative abundance trends estimated from logbook data to those previously estimated by the SCRS (Supplementary Figure S4), and 79% of these trends were significantly positively correlated. Figure 4. Open in new tabDownload slide Abundance trends estimated for each species using fisher logbook data from the USLL (thick line), with linear trends fit to the abundance patterns (thin line). Each abundance trend was scaled to its mean value, and the corresponding median of the annual coefficients of variation was presented next to each species name in parentheses. Figure 4. Open in new tabDownload slide Abundance trends estimated for each species using fisher logbook data from the USLL (thick line), with linear trends fit to the abundance patterns (thin line). Each abundance trend was scaled to its mean value, and the corresponding median of the annual coefficients of variation was presented next to each species name in parentheses. General relative abundance patterns were characterized using either continuous or piecewise linear trends (Figure 4). Linear trends from the logbook analyses were compared with those estimated from observer data (Supplementary Figure S5), and in general, directionality was consistent across data sets, with obvious exceptions for blue shark, porbeagle, common thresher, scalloped hammerhead, smooth hammerhead, night shark, and spinner shark. As a measure of precision, the median of the annual coefficients of variation (MCV) was calculated for each relative abundance trend (Figure 4). According to MCV, eight (24%) of the trends were estimated with poor precision (i.e. MCV > 1), suggesting that the annual estimates of relative abundance for these particular trends should be interpreted with caution. We further characterized relative abundance trends using instantaneous rates of change estimated from the logbook (Figure 5) and observer (Supplementary Figure S6) analyses. Strongly negative rates were most prevalent early in the time series, particularly for sharks, but most species with steep initial declines in abundance have either stabilized or are experiencing less severe declines in recent years. Eight patterns in instantaneous rates of change emerged from the logbook analyses: (1) decreasing (negative) throughout, (2) decreasing then stable (not significantly different from zero), (3) decreasing then increasing (positive), (4) stable throughout, (5) stable then increasing, (6) increasing throughout, (7) increasing then stable, and (8) increasing then decreasing. A summary of these patterns (Table 2) indicated that approximately 71% of HMS analysed are either decreasing in recent years or have decreased without evidence of recovery (patterns 1, 2, and 5), while 29% exhibited other, more favourable trends (patterns 3, 4, and 6–8). These patterns were also summarized according to taxonomic grouping (Table 2), which emphasized that relative abundance trends are generally more favourable for tunas than for either billfishes or sharks. For tunas, 67% of the species fell into the favourable categories, whereas 20% of billfishes and 16% of shark species followed favourable patterns. Table 2. Patterns observed for instantaneous rates of change in abundance estimated from the logbook analyses, presented as the total number and percent of species analysed corresponding to each pattern. Pattern . All . Tunas . Billfish . Sharks . 1. Decreasing 9 (26.5%) 2 (22.2%) 1 (20.0%) 6 (31.6%) 2. Decreasing then stable 14 (41.2%) 1 (11.1%) 3 (60.0%) 10 (52.6%) 3. Decreasing then increasing 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 4. Stable 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 5. Stable then increasing 2 (5.9%) 0 (0.0%) 1 (20.0%) 1 (5.3%) 6. Increasing 1 (2.9%) 1 (11.1%) 0 (0.0%) 0 (0.0%) 7. Increasing then stable 3 (5.7%) 1 (11.1%) 0 (0.0%) 2 (10.5%) 8. Increasing then decreasing 1 (2.9%) 0 (0.0%) 0 (0.0%) 0 (0.0%) Pattern . All . Tunas . Billfish . Sharks . 1. Decreasing 9 (26.5%) 2 (22.2%) 1 (20.0%) 6 (31.6%) 2. Decreasing then stable 14 (41.2%) 1 (11.1%) 3 (60.0%) 10 (52.6%) 3. Decreasing then increasing 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 4. Stable 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 5. Stable then increasing 2 (5.9%) 0 (0.0%) 1 (20.0%) 1 (5.3%) 6. Increasing 1 (2.9%) 1 (11.1%) 0 (0.0%) 0 (0.0%) 7. Increasing then stable 3 (5.7%) 1 (11.1%) 0 (0.0%) 2 (10.5%) 8. Increasing then decreasing 1 (2.9%) 0 (0.0%) 0 (0.0%) 0 (0.0%) Patterns were summarized for all HMS analysed, tunas (Suborder: Scombroidei), billfish (Suborder: Xiphiodei), and sharks (Superorder: Euselachii). The single increasing then decreasing trend is associated with dolphinfish. Open in new tab Table 2. Patterns observed for instantaneous rates of change in abundance estimated from the logbook analyses, presented as the total number and percent of species analysed corresponding to each pattern. Pattern . All . Tunas . Billfish . Sharks . 1. Decreasing 9 (26.5%) 2 (22.2%) 1 (20.0%) 6 (31.6%) 2. Decreasing then stable 14 (41.2%) 1 (11.1%) 3 (60.0%) 10 (52.6%) 3. Decreasing then increasing 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 4. Stable 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 5. Stable then increasing 2 (5.9%) 0 (0.0%) 1 (20.0%) 1 (5.3%) 6. Increasing 1 (2.9%) 1 (11.1%) 0 (0.0%) 0 (0.0%) 7. Increasing then stable 3 (5.7%) 1 (11.1%) 0 (0.0%) 2 (10.5%) 8. Increasing then decreasing 1 (2.9%) 0 (0.0%) 0 (0.0%) 0 (0.0%) Pattern . All . Tunas . Billfish . Sharks . 1. Decreasing 9 (26.5%) 2 (22.2%) 1 (20.0%) 6 (31.6%) 2. Decreasing then stable 14 (41.2%) 1 (11.1%) 3 (60.0%) 10 (52.6%) 3. Decreasing then increasing 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 4. Stable 2 (5.9%) 2 (11.1%) 0 (0.0%) 0 (0.0%) 5. Stable then increasing 2 (5.9%) 0 (0.0%) 1 (20.0%) 1 (5.3%) 6. Increasing 1 (2.9%) 1 (11.1%) 0 (0.0%) 0 (0.0%) 7. Increasing then stable 3 (5.7%) 1 (11.1%) 0 (0.0%) 2 (10.5%) 8. Increasing then decreasing 1 (2.9%) 0 (0.0%) 0 (0.0%) 0 (0.0%) Patterns were summarized for all HMS analysed, tunas (Suborder: Scombroidei), billfish (Suborder: Xiphiodei), and sharks (Superorder: Euselachii). The single increasing then decreasing trend is associated with dolphinfish. Open in new tab Figure 5. Open in new tabDownload slide Instantaneous rates of change in relative abundance ±95% confidence intervals. A single or initial rate of change is presented for each species (▪), and a second, more recent rate of change is presented for species where piecewise regression outperformed simple linear regression (○). Figure 5. Open in new tabDownload slide Instantaneous rates of change in relative abundance ±95% confidence intervals. A single or initial rate of change is presented for each species (▪), and a second, more recent rate of change is presented for species where piecewise regression outperformed simple linear regression (○). Discussion In this study we estimated relative abundance trends (1987–2013) for 34 HMS in the western Atlantic Ocean using an approach that accounts for pelagic habitat fished. This represents one of the most comprehensive analyses of HMS to date, and the individual species trends offer a variety of potential benefits. For the species that have previously been assessed by ICCAT (Table 1), our trends are useful in a comparative sense, because where available, stock assessment results should serve as the primary basis for understanding stock status and trends in abundance. However, our methodology may result in more accurate indices of relative abundance from the USLL fleet, which may improve the stock assessments of these species if our trends are incorporated. For the species that are not regularly assessed, including dolphinfish, wahoo, blackfin tuna, oilfish, spearfishes, and several sharks, we provide first-ever, or updated abundance trends that may well represent the best current understanding of their abundance trends. Overall, USLL abundance trends indicate population declines of varying degrees without noticeable recovery for most HMS analysed (71% of the species). Declines in relative abundance of large predatory fishes have been cited as evidence of a global fisheries crisis (Jackson et al., 2001; Baum et al., 2003; Myers and Worm, 2003; Worm et al., 2006; Myers et al., 2007; Ferretti et al., 2008). While these studies have garnered considerable attention from the media, general public, and scientific community, many have been criticized for analytical flaws, some of which may have been critical to the conclusions (Walters, 2003; Burgess et al., 2005; Hampton et al., 2005; Polacheck, 2006; Wilberg and Miller, 2007). Examples of common criticisms include the use of aggregated CPUE (Walters, 2003), a failure to consider USLL observer data (Burgess et al., 2005), and ignoring habitat, vertical distributions, and other factors that can bias trends in fishery CPUE (Burgess et al., 2005; Hampton et al., 2005; Polacheck, 2006). In our study, we did not aggregate CPUE across species or spatial cells, we included an analysis of USLL observer data, and we considered a full suite of variables (including habitats fished) that have been hypothesized to potentially bias CPUE trends. We fully recognize the difficulty in inferring population trends from fishery data, but given that there are no scientific monitoring programs operating at the population scale, fisheries offer the best available information. Thus, we have been careful to address many of the concerns associated with estimating relative abundance trends using fishery data. Using USLL-derived indices of abundance, we observed substantial declines for many species; however, complete extirpation of all large predators does not appear imminent unless several abundance trends suddenly decline. Approximately ten species (29%) did not show a statistically significant negative trend in relative abundance over the past several years (albacore tuna, bluefin tuna, blackfin tuna, wahoo, oilfish, Atlantic bonito, spearfishes, tiger shark, shortfin mako, and porbeagle), and some stocks showed signs of growth or recovery. It should be noted that while not statistically significant, shortfin mako and porbeagle appear to be declining in relative abundance. In contrast, if recent increases in blue shark relative abundance continue, we anticipate that our analyses would identify a favourable change (i.e. significantly positive instantaneous rate of change) starting around 2005. While our results indicate that many HMS have declined in abundance over time, the species that exhibited favourable patterns suggest that either the purported demise of marine predators was overly pessimistic, or that some of these species began to rebuild since the earlier studies were conducted (we suspect both explanations to be true). The range of relative abundance patterns observed in this study support the conclusions of Worm et al. (2009), who, in a comprehensive analysis of global marine ecosystems, described a combination of overexploited and recovering fish stocks. Changes in fishing pressure, due to management actions or socio-economic dynamics, are likely a strong driver of HMS abundance, but across all 34 species analysed, it would be very challenging to disentangle fishing effects from other potential drivers, such as climate change, environmental variability, and predator-prey dynamics. The data used for our analyses comprise one of the best sources available for making inferences about HMS relative abundance in the Atlantic Ocean (Baum et al., 2003). Pelagic longline fisheries typically cover a wide geographic range, and they have been operating in the Atlantic Ocean since the 1950s (Majkowski, 2007). Longline fleets from nations with a long-term presence in the Atlantic (e.g. Japan and Taiwan) are also potentially valuable sources of data for evaluating HMS abundance; however, to account for changing fishery dynamics, information about fishing practices must be available. When recorded, this information is often considered proprietary, and therefore can be difficult to obtain. We analysed fisher logbook data from the USLL, which includes detailed set-specific information concerning fishery dynamics. We encourage similar studies using pelagic longline data from other nations, such as Japan, if reliable data on fishing practices are available. Analyzing data from fisheries with longer time series may be most beneficial, because the first complete year of USLL logbook records was 1987, and relative abundance in the first year of our time series may have already been reduced following years of intense fishing pressure. In general, stock assessments (Quinn and Deriso, 1999) that integrate multiple sources of information (including relative abundance trends) provide a more complete evaluation of fish stock dynamics than simple trend analyses. For the few species that have been assessed, management decisions should be (and are) based on assessment results rather than fishery-derived relative abundance trends; however, our trends have the novelty of adjusting for exploited habitats and may be useful in future stock assessments. Relative abundance trends previously estimated using logbook data from the USLL are available for species that have been assessed in a fishery stock assessment context or by individual research projects (e.g. Baum et al., 2003). Our relative abundance trends are not completely divergent from those previously estimated for stock assessments, and they extend the estimates beyond the final year of the earlier time series (Supplementary Figure S4). We observed that previous relative abundance trajectories have continued for many species, while the direction of others has reversed (mainly those that exhibited signs of population growth in recent years). The relative abundance trends we estimated for swordfish and skipjack tuna are in contrast with previous estimates used in stock assessments. We showed a declining, rather than stable swordfish relative abundance over time, and we did not observe a sudden increase in skipjack tuna relative abundance as previously shown. An analysis of USLL observer data by Baum and Blanchard (2010) estimated relative abundance trends for many of the same shark species we analysed. Although Baum and Blanchard (2010) aggregated several of the shark species and conducted analyses at the genus or species group level, our estimated trends (Supplementary Figure S3) were similar to theirs through 2005 (the final year of data analysed by Baum and Blanchard [2010]). When comparing and evaluating relative abundance trends for individual species, the population biology and fishery data collection for that species should be considered. For instance, estimates of relative abundance used in recent swordfish stock assessments relied on fishery weigh-out data to compute catches by age, and then aggregated catches over ages 3–10. We did not have weigh-out data available for our analyses, nor did we attempt to partition catches by age. Also, regulatory effects were considered when analysing the swordfish weigh-out data, and we did not explicitly consider species-specific regulations. These methodological differences between our analysis and the swordfish stock assessment may explain the divergent abundance trends. For billfishes, primarily white marlin, the recent validation of roundscale spearfish (Tetrapturus georgii) as a species (Shivji et al., 2006) may have affected catch reporting accuracy by shifting catches that were historically reported as “white marlin” and other billfishes to “spearfishes.” Abundance trends used in previous Atlantic bluefin tuna stock assessments were estimated using only records from the Gulf of Mexico during January–May (NMFS, 1993), yet we used data throughout the year. There are also important considerations concerning the use of USLL logbook data to make inferences about the relative abundance of sharks (although these concerns may not apply to blue and shortfin mako sharks). Burgess et al. (2005) discussed regulatory changes in 1993 that might have contributed to false declines in catch rates of some sharks; however, we note that many of the shark species we analysed exhibited declines before 1993. Additional issues noted by Burgess et al. (2005) that may contribute significant errors to the logbook database include misidentification, errors in reporting, and failure to record bycatch species. However, random errors in identification and data recording are much less problematic than an unaccounted sudden change or systematic pattern in data recording. Although, for some species, such as white shark (Carcharadon carcharias), errors in the data may be substantial enough to make our relative abundance trends uninformative (most recorded white shark catches are likely the result of misidentification; Burgess et al., 2005). Fishery observer data likely contain fewer issues related to misidentification or errors in reporting. Thus, positive correlations between abundance trends estimated from logbook data and those based on fishery observer data provide a degree of validation for 57% of the stocks with observer data (Supplementary Figure S3). For species with divergent logbook and observer trends, the trends based on logbook data should be interpreted with caution. Also, we recommend additional work to compare logbook and observer data collected on the same trip. Catches observed in relation to the MinT habitat variable (Figure 3) highlight the expected result that exploited pelagic habitats (which are a function of gear configuration, fishing location, and environmental conditions) largely govern the composition of species encountered. This conclusion provides strong support for including a temperature variable in models designed to estimate HMS relative abundance trends. Furthermore, the incorporation of pelagic habitat fished allows a post-hoc evaluation of the role of pelagic habitat on HMS catches. For instance, blue sharks exhibited a higher encounter rate when cooler habitats were fished. This is not necessarily surprising (Cortés et al., 2007); however, when the fishery exploited the absolute coldest habitat (1–5°C) and blue sharks were encountered, their catch rates were higher than those for any other species caught by the fishery. Because blue sharks are a bycatch species in the USLL fishery, managers could use this information to impose time-area restrictions on certain gear configurations to avoid fishing the coldest habitat and possibly reduce overall bycatch of blue sharks. Evaluating habitat-specific catch rates would not only be useful for blue sharks, but potentially for all species analysed, especially those with high catch rates in specific habitats (e.g. shortfin mako shark, hammerhead sharks, sandbar shark, spinner shark, porbeagle, and bignose shark). Many shark species are particularly vulnerable to overfishing due to their relatively low fecundity, slow growth rates, and late maturity (Musick et al. 2000), and in fact, various stocks of scalloped hammerhead sharks are listed as either threatened or endangered under the Endangered Species Act (http://www.nmfs.noaa.gov/pr/species/esa/listed.htm#fish). Thus, our habitat-specific catch rates may facilitate conservation of many sharks and other species that are vulnerable to overfishing. The pelagic habitat variables explained a relatively small amount of variance in catch rates of the primary target species, such as swordfish and yellowfin tuna (Figure 2). One explanation for this result is that, in order to maximize catch rates, fishermen purposefully deploy gear in the preferred habitats of their target species. Thus, variation in target species catch rates may be more related to changes in abundance and targeting practices than habitat-driven availability. For bycatch species, however, fishermen are not seeking to maximize their catch rates, and overlaps between fishing effort and their distributions are less frequent and likely more driven by incidentally fishing in their preferred habitats. The relative lack of importance of MaxΔT was unexpected considering the results of a simulation study conducted by Lynch et al. (2012); however, that study was based on the dynamics of the Japanese pelagic longline fishery. The Japanese fishery has substantially changed fishing practices over time, resulting in strong contrast in pelagic habitats exploited. The USLL has also exhibited systematic changes in fishing practices over the time period we analysed, but these changes did not occur on the temporal and spatial scales of the Japanese fishery. This does not suggest that relative temperature is not an important factor governing the population dynamics of HMS, but rather that the minimal contrast observed in MaxΔT precludes it from explaining considerable variability in USLL catch rates. We maintain that future efforts to estimate relative abundance trends from HMS fishery data consider both MinT and MaxΔT in model development. Several of our relative abundance trends were not estimated with high precision, and this uncertainty should be kept in mind when interpreting the patterns. In some cases, the inclusion of temperature variables may have increased uncertainty in relation to relative abundance trends previously estimated without these variables. However, increased uncertainty would be a poor justification for ignoring important dynamics, such as pelagic habitat fished, and in fact, our results suggest that pelagic habitat variables can explain substantial variability in HMS catch rates. Empirical evidence highlights the importance of temperature in governing HMS vertical distributions (Brill and Lutcavage, 2001), and our modelling exercises can be useful for understanding how HMS catch rates may respond to ocean dynamics. By including the temperature variables, our analyses may have placed a higher value on accuracy than precision, but we encourage that future studies seek to reduce uncertainty while maintaining the consideration of pelagic habitat. Also, to improve the characterization of habitats fished, we encourage enhanced sampling of oceanographic variables during fishing operations to be recorded in logbooks and by fishery observers. In addition to precision, several underlying model assumptions warrant attention. For instance, to estimate the temperature fished in each longline set, we assumed that all sections of the gear were distributed identically throughout the water column. This is unlikely, because longline fishing depth is governed by numerous dynamic processes, including wind, hydrodynamics, and the behaviour of hooked organisms (Bigelow et al., 2006; Ward and Myers, 2006; Rice et al., 2007). Also, by relating fishing depth to temperature using average ocean temperatures we ignored interannual variability in temperature at depth for a given time and location. However, one benefit of ignoring interannual variability is that our analyses were not confounded by potential changes in stock productivity related to changing ocean temperature; rather, our temperature variables accounted for changes in availability due to monthly ocean dynamics. In the broader context of improving relative abundance estimates, future analyses might consider additional environmental factors, such as the oxygen minimum zone (Prince et al., 2010), or other statistical treatments of spatio-temporal data (e.g. Thorson et al., 2015). Despite potential caveats, we believe this study advances the methodology for deriving fishery-dependent indices of abundance from HMS longline fisheries. Our habitat variables generally explained a substantial amount of deviation in catch rates. Thus, we recommend that these variables be considered in future stock assessments that incorporate estimates of relative abundance from longline catch rates. Further, the results of this study can help inform discussions about the health of global fisheries, particularly for species that are not regularly assessed. Overall, we observed a mixture of declining, stable, and increasing trends in relative abundance, which indicates that global fisheries are not likely following a unidirectional pattern. However, in general terms, declines observed for bycatch species were more severe than those for target species. This may suggest that bycatch species of HMS fisheries are more susceptible to overfishing than target species. With this challenge in mind, the habitat-specific catch rates we observed (Figure 3) may serve as a valuable management tool for reducing fishing pressure on bycatch species. Supplementary data Supplementary material is available at the ICESJMS online version of the manuscript. Acknowledgements We thank R. Ahrens, L. Beerkircher, T. Boyer, K. Erickson, T. Gedamke, D. Gloeckner, K. Keene, and K. Logan for help obtaining data; we thank R. Bell, C. Brown, J. Brubaker, A. Buchheister, C. Cotton, J. Graves, T. Miller, K. Parsons, J. Walter, and C. Wor for assistance in developing this manuscript; and we thank K. Andrews, M. Lauretta, and anonymous reviewers for comments on earlier versions. Funding was provided by the National Oceanic and Atmospheric Administration (NA09OAR4170119). This is Virginia Institute of Marine Science contribution number 3715. The views expressed are those of the authors, and do not necessarily represent findings or policy of any government agency. References Aitchison J. 1955 . On the distribution of a positive random variable having a discrete probability mass at the origin . Journal of the American Statistical Association , 50 : 901 – 908 . Google Scholar OpenURL Placeholder Text WorldCat Baum J. K. , Blanchard W. 2010 . Inferring shark population trends from generalized linear mixed models of pelagic longline catch and effort data . Fisheries Research , 102 : 229 – 239 . Google Scholar Crossref Search ADS WorldCat Baum J. K. , Myers R. A., Kehler D. G., Worm B., Harley S. J., Doherty P. A. 2003 . Collapse and conservation of shark populations in the northwest Atlantic . Science , 299 : 389 – 392 . Google Scholar Crossref Search ADS PubMed WorldCat Bigelow K. A. , Maunder M. N. 2007 . Does habitat or depth influence catch rates of pelagic species? Canadian Journal of Fisheries and Aquatic Sciences , 64 : 1581 – 1594 . Google Scholar Crossref Search ADS WorldCat Bigelow K. , Musyl M. K., Poisson F., Kleiber P. 2006 . Pelagic longline gear depth and shoaling . Fisheries Research , 77 : 173 – 183 . Google Scholar Crossref Search ADS WorldCat Brill R. , Lutcavage M. 2001 . Understanding environmental influences on movements and depth distributions of tunas and billfishes can significantly improve stock assessments. In Island in the Stream: Oceanography and Fisheries of the Charleston Bump, Pp. 179–198 . Ed. by Sedberry G. R.. American Fisheries Society , Bethesda, Maryland . 240 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Burgess G. H. , Beerkircher L. R., Cailliet G. M., Carlson J. K., Cortés E., Goldman K. J., Grubbs R. D. 2005 . Is the collapse of shark populations in the northwest Atlantic Ocean and Gulf of Mexico real? Fisheries , 30 : 19 – 26 . Google Scholar Crossref Search ADS WorldCat Burnham K. P. , Anderson D. R. 2002 . Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach , 2nd edn. Springer-Verlag , New York, NY . 488 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cortés E. , Brown C. A., Beerkircher L. R. 2007 . Relative abundance of pelagic sharks in the western North Atlantic Ocean, including the Gulf of Mexico and Caribbean Sea . Gulf and Caribbean Research , 19 : 37 – 52 . Google Scholar Crossref Search ADS WorldCat Ferretti F. , Myers R. A., Serena F., Lotze H. K. 2008 . Loss of large predatory sharks from the Mediterranean Sea . Conservation Biology , 22 : 952 – 964 . Google Scholar Crossref Search ADS PubMed WorldCat Goodyear C. P. 2003 . Tests of the robustness of habitat-standardized abundance indices using simulated blue marlin catch-effort data . Marine and Freshwater Research , 54 : 369 – 381 . Google Scholar Crossref Search ADS WorldCat Hampton J. , Sibert J. R., Kleiber P., Maunder M. N., Harley S. J. 2005 . Fisheries: decline of Pacific tuna populations exaggerated? Nature , 434 : E1 – E2 . Google Scholar Crossref Search ADS PubMed WorldCat Hinton M. G. , Nakano H. 1996 . Standardizing catch and effort statistics using physiological, ecological, or behavioral constraints and environmental data, with an application to blue marlin (Makaira nigricans) catch and effort data from Japanese longline fisheries in the Pacific . Inter-American Tropical Tuna Commission Bulletin , 21 : 171 – 200 . Google Scholar OpenURL Placeholder Text WorldCat Hoey J. J. , Bertolino A. 1988 . Review of the U.S. fishery for swordfish, 1978 to 1986 . Collective Volume of Scientific Papers ICCAT , 27 : 256 – 266 . Google Scholar OpenURL Placeholder Text WorldCat Jackson J. B. C. , Kirby M. X., Berger W. H., Bjorndal K. A., Botsford L. W., Bourque B. J., Bradbury R. H. 2001 . Historical overfishing and the recent collapse of coastal ecosystems . Science , 293 : 629 – 638 . Google Scholar Crossref Search ADS PubMed WorldCat Lo N. C. H. , Jacobson L. D., Squire J. L. 1992 . Indices of relative abundance from fish spotter data based on delta-lognormal models . Canadian Journal of Fisheries and Aquatic Sciences , 49 : 2515 – 2526 . Google Scholar Crossref Search ADS WorldCat Locarnini R. A. , Mishonov A. V., Antonov J. I., Boyer T. P., Garcia O. K., Baranova O. K., Zweng M. M. 2010 . World Ocean Atlas 2009, Volume 1: temperature. In NOAA Atlas NESDIS 68 . Ed. by Levitus S.. U.S. Government Printing Office , Washington, DC . 184 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lynch P. D. , Graves J. E., Latour R. J. 2011 . Challenges in the assessment and management of highly migratory bycatch species: a case study of the Atlantic marlins, pp. 197–225. In Sustainable Fisheries: Multi-Level Approaches to a Global Problem . Ed. by Taylor W. W., Lynch A. J., Schechter A. J.. American Fisheries Society , Bethesda, MD . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lynch P. D. , Shertzer K. W., Latour R. J. 2012 . Performance of methods used to estimate indices of abundance for highly migratory species . Fisheries Research , 125-126 : 27 – 39 . Google Scholar Crossref Search ADS WorldCat Majkowski J. 2007 . Global fishery resources of tuna and tuna-like species. FAO Fisheries Technical Paper 483. Food and Agriculture Organization of the United Nation, Rome. 54 pp. Mandelman J. W. , Cooper P. W., Werner T. B., Lagueux K. M. 2008 . Shark bycatch and depredation in the U.S. Atlantic pelagic longline fishery . Reviews in Fish Biology and Fisheries , 18 : 427 – 442 . Google Scholar Crossref Search ADS WorldCat Maunder M. N. , Hinton M. G., Bigelow K. A., Langley A. D. 2006 . Developing indices of abundance using habitat data in a statistical framework . Bulletin of Marine Science , 79 : 545 – 559 . Google Scholar OpenURL Placeholder Text WorldCat Maunder M. N. , Punt A. E. 2004 . Standardizing catch and effort data: a review of recent approaches . Fisheries Research , 70 : 141 – 159 . Google Scholar Crossref Search ADS WorldCat Musick J. A. , Burgess G. H., Camhi M., Cailliet G., Fordham S. 2000 . Management of sharks and their relatives (Elasmobranchii) . Fisheries , 25 : 9 – 13 . Google Scholar Crossref Search ADS WorldCat Myers R. A. , Baum J. K., Shepherd T. D., Powers S. P., Peterson C. H. 2007 . Cascading effects of the loss of apex predatory sharks from a coastal ocean . Science , 315 : 1846 – 1850 . Google Scholar Crossref Search ADS PubMed WorldCat Myers R. A. , Worm B. 2003 . Rapid worldwide depletion of predatory fish communities . Nature , 423 : 280 – 283 . Google Scholar Crossref Search ADS PubMed WorldCat NMFS (National Marine Fisheries Service). 1993 . Fishery management plan for sharks of the Atlantic Ocean. National Oceanic and Atmospheric Administration, Silver Spring, MD. Ortiz M. , Arocha F. 2004 . 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 . Fisheries Research , 70 : 275 – 297 . Google Scholar Crossref Search ADS WorldCat Polacheck T. 2006 . Tuna longline catch rates in the Indian Ocean: did industrial fishing result in a 90% rapid decline in the abundance of large predatory species? Marine Policy , 30 : 470 – 482 . Google Scholar Crossref Search ADS WorldCat Prince E. D. , Luo J., Goodyear C. P., Hoolihan J. P., Snodgrass D., Orbesen E. S., Serafy J. E., et al. 2010 . Ocean scale hypoxia-based habitat compression of Atlantic istiophorid billfishes . Fisheries Oceanography , 19 : 448 – 462 . Google Scholar Crossref Search ADS WorldCat Quinn T. J. , Deriso R. B. 1999 . Quantitative Fish Dynamics . Oxford University Press , Oxford . 212 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC R Core Team . 2016 . R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at http://www.R-project.org/. Rice P. H. , Goodyear C. P., Prince E. D., Snodgrass D., Serafy J. E. 2007 . Use of catenary geometry to estimate hook depth during near-surface pelagic longline fishing: theory versus practice . North American Journal of Fisheries Management , 27 : 1148 – 1161 . Google Scholar Crossref Search ADS WorldCat Seber G. A. F. 1982 . The Estimation of Animal Abundance and Related Parameters, 2nd edn . Oxford University Press , New York . 654 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Shivji M. S. , Magnussen J. E., Beerkircher L. R., Hinteregger G., Lee D. W., Serafy J. E., Prince E. D. 2006 . Validity, identification, and distribution of the roundscale spearfish, Tetrapturus georgii (Teleostei: istiophoridae): morphological and molecular evidence . Bulletin of Marine Science , 79 : 483 – 491 . Google Scholar OpenURL Placeholder Text WorldCat Stefánsson G. 1996 . Analysis of groundfish survey abundance data: combining the GLM and delta approaches . ICES Journal of Marine Science , 53 : 577 – 588 . Google Scholar Crossref Search ADS WorldCat Thorson J. T. , Shelton A. O., Ward E. J., Skaug H. J. 2015 . Geostatistical delta-generalized linear mixed models improve precision for estimated abundance indices for West Coast groundfishes . ICES Journal of Marine Science , 72 : 1297 – 1310 . Google Scholar Crossref Search ADS WorldCat Walter J. 2011 . Standardized catch rate in number and weight of yellowfin tuna (Thunnus albacares) from the United States pelagic longline fishery 1987–2010 . Collective Volume of Scientific Papers ICCAT , 68 : 915 – 952 . Google Scholar OpenURL Placeholder Text WorldCat Walters C. 2003 . Folly and fantasy in the analysis of spatial catch rate data . Canadian Journal of Fisheries and Aquatic Sciences , 60 : 1433 – 1436 . Google Scholar Crossref Search ADS WorldCat Ward P. J. , Myers R. A. 2006 . Do habitat models accurately predict the depth distribution of pelagic fishes? Fisheries Oceanography , 15 : 60 – 66 . Google Scholar Crossref Search ADS WorldCat Wilberg M. J. , Miller T. J. 2007 . Comment on “Impacts of biodiversity loss on ocean ecosystem services” . Science , 316 : 1285b. Google Scholar Crossref Search ADS WorldCat Worm B. , Barbier E. B., Beaumont N., Duffy J. E., Folke C., Halpern B. S., Jackson J. B. C., et al. 2006 . Impacts of biodiversity loss on ocean ecosystem services . Science , 314 : 787 – 790 . Google Scholar Crossref Search ADS PubMed WorldCat Worm B. , Hilborn R., Baum J. K., Branch T. A., Collie J. S., Costello C., Fogarty M. J., et al. 2009 . Rebuilding global fisheries . Science , 325 : 578 – 585 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes Present address: NOAA, NMFS, Office of Science and Technology, 1315 East West Highway, Silver Spring, MD 20910, USA. Published by International Council for the Exploration of the Sea 2018. This work is written by US Government employees and is in the public domain in the US. Published by International Council for the Exploration of the Sea 2018.

Deroba, Jonathan J

doi: 10.1093/icesjms/fsy013pmid: N/A

Abstract Spatial and temporal variation in stomach-contents data is often unquantified or combined in such a way (e.g. averaged among years) that true signal in diets may be lost. Using a delta approach, this paper fits generalized additive mixed models (GAMMs) to the amount of Atlantic herring (Clupea harengus) identified in predator stomachs using only data from stomachs in which herring occurred, and to the probability that a stomach contained herring. Both the amount of herring in stomachs and the probability of a stomach containing herring varied seasonally, spatially, and among years. Of the random effects in each GAMM, the effect of predator species had the largest variance. An index of herring abundance derived from the stomach-contents data was generally consistent with recent herring stock assessments. The temporal and spatial variation in the stomach-contents data suggested that the effect of averaging or combining stomach-contents data among years, seasons, or areas may lead to falsely precise or biased estimates from multispecies assessments or in estimates of consumption, and may restrain the relevance of static foodweb models. Introduction Fish diet and subsequent stomach-contents (gut) data vary temporally and spatially (Reum and Essington, 2008; Nunn et al., 2012), and the utility of stomach data has been argued to be subject to a range of unquantifiable errors and biases (Baker et al., 2014). For example, stomach data represent a short time-span of prey selection that may not represent seasonal or even daily predator preference (Reum and Essington, 2008; Baker et al., 2014). Furthermore, as a result of partial digestion, prey items in stomach contents are difficult to distinguish to species, and prey size is difficult to measure or estimate. Stomach-contents data, however, have a range of uses, including multispecies stock assessment models, foodweb models, and estimating total annual consumption of prey species (Overholtz et al., 2008; Tyrrell et al., 2008; Gaichas et al., 2011; Curti et al., 2013). The treatment of stomach-contents data to serve those purposes has different consequences. Multispecies virtual population analyses (MSVPA) and multispecies statistical catch-at-age (MSCAA) models attempt to capture some temporal (usually annual) variation among species interactions, but concerns about relatively high measurement error have forced analysts into combining or averaging stomach contents among seasons, years, and across broad geographic regions (Gislason and Helgason, 1985; Livingston and Jurado-Molina, 2000; Lewy and Vinther, 2004; Tyrrell et al., 2008; Kinzey and Punt, 2009; Curti et al., 2013). Combining or averaging stomach samples among space and time may result in artificially reducing true variation in the data, as opposed to just measurement error, and may induce bias if systematic differences among space or time are ignored. Subsequent estimates from multispecies assessments may then be falsely precise or biased. MSCAA models are also sensitive to the model-fitting weight given to stomach-contents data relative to other data sources (Curti et al., 2013; Van Kirk et al., 2015). Thus, not accounting for true variation in the data may result in false confidence in the data and biased model estimates. Static foodweb models provide insight into ecosystem function for a snapshot of time and have used stomach-contents data from relatively few years to represent longer time-frames, sometimes over broad geographic areas (Harvey et al., 2003; Gaichas et al., 2010). Systematic and significant variation among space and time in stomach contents that is not accounted for in the modelling will consequently affect where and for how long these models remain relevant for management. Dynamic foodweb models also provide insight into ecosystem function and attempt to inform the degree of temporal variation in the processes (Gaichas et al., 2011). Consequently, not accounting for variation in stomach-contents data among space and time may produce falsely precise or biased estimates, as in multispecies assessments. Stomach contents of fish predators have also been used to estimate total annual consumption of a prey fish, with the intent of using the consumption estimates to allow for estimation of natural mortality in single-species stock assessments of the prey (Overholtz and Link, 2007; Overholtz et al., 2008; Moustahfid et al., 2009). Not accounting for non-random spatial or seasonal variation in the stomach-contents data would likely result in biased estimates of consumption for a given year and, subsequently, biased estimates of natural mortality and other stock assessment quantities. Similarly, averaging stomach contents among years in these contexts would dampen true variation and may result in inaccurate estimates of annual natural mortality. Issues with the treatment of stomach-contents data in the estimation of consumption are further compounded if the measurement and estimation uncertainty that stems from the predator stock assessments is ignored when deriving consumption estimates (Brooks and Deroba, 2015). Understanding sources of variation in stomach-contents data can also inform possible consequences of climate change and the strength of predator–prey relationships. For example, water temperatures in the Northwest Atlantic have increased over years, and the strength of those changes varied by season (Thomas et al., 2017). Fish species in the region are expected to exhibit a range of responses to these temperature changes (Hare et al., 2016). Coupled with an understanding of temporal and spatial variation in stomach contents, the effects of species distribution shifts in response to climate change on predator–prey interactions might be better anticipated. Similarly, understanding how stomach contents vary among predator species can reveal the relative importance of predators in their ability to affect prey mortality rates. Thus, the effects of changes in predator abundance on prey mortality rates can be anticipated and possibly incorporated into management. Stomach contents can also be seen as biological samples of available prey, which permits predator diet data to be used to create indices of prey abundance (Link, 2004; Mills et al., 2007; Buchheister and Latour, 2016). Frequency of occurrence was used as an index of benthic prey abundance in the Northeast United States (Link, 2004). Likewise, proportion of juvenile rockfish (Sebastes spp.) in the diet of seabirds and number of rockfish in chinook salmon (Oncorhynchus tshawytscha) stomachs were combined to create an index of rockfish abundance in the California Current System (Mills et al., 2007). Such indices are useful as points of comparison with more commonly used indices of abundance, such as trawls, and with stock assessment estimates of abundance (Mills et al., 2007). Atlantic herring (C. harengus) (hereafter herring) in the Northwest Atlantic are preyed upon by fish, seabirds, and marine mammals and can account for 20–50% of the diet of these predators (Overholtz and Link, 2007; Smith and Link, 2010; Curti et al., 2013). Atlantic herring have also been the focal species in several multispecies modelling efforts that utilized stomach-contents data (Read and Brownstein, 2003; Overholtz and Link, 2007; Tyrrell et al., 2008). Thus, an increased understanding of the stomach contents of herring predators in the Northwest Atlantic would be especially relevant and impactful. Bottom-trawl surveys used in herring stock assessments are also relatively imprecise, and changes in trawl gear and vessel have caused temporal changes in catchability that increased assessment uncertainty (Miller et al., 2010; NEFSC, 2012; Miller, 2013; Jech and Sullivan, 2014). So, having an index of abundance based on predator stomach contents would be useful in the stock assessment process. The first objective of this manuscript was to evaluate sources of variation in the amount and occurrence of herring in the stomachs of piscivorous predators in the Northwest Atlantic during 1973–2014. This objective was addressed by fitting separate generalized additive mixed models (GAMMs) to: (i) the amount of herring observed in predator stomachs using only those stomachs in which herring were identified, and (ii) a model of the probability of a stomach containing herring using data from all sampled stomachs. This method is analogous to the delta approach that has been used to standardize catch-per-effort data and has been previously applied in fish diet studies (Stefánsson and Pálsson, 1997; Maunder and Punt, 2004; Buchheister and Latour, 2016). The second objective was to develop an index of herring abundance by treating the stomach contents as catch-per-effort observations, and combining the results of the GAMMs as in the delta approach. Methods Data Stomach-contents data were collected on National Marine Fisheries Service Northeast Fisheries Science Center spring and fall bottom-trawl surveys. Details about the methods for sampling stomach contents can be found in Link and Almeida (2000) and Smith and Link (2010). Details about bottom-trawl survey design can be found in Grosslein (1969), Azarovitz (1981), and Miller et al. (2010). A brief overview was provided here. Bottom-trawl survey sampling stations between Nova Scotia, Canada and Cape Hatteras, NC were selected using a stratified random design, with strata defined by depth and latitude. A total of 350–400 stations were sampled each year and season, which resulted in sampling being approximately proportional to stratum area. A minimum of two stations were sampled per stratum. Catch was sorted by species and weighed, individuals were measured for length, and a subset of species was sampled for food habits. Quantitative stomach contents have been sampled since 1973. Total stomach contents and individual prey mass were measured to the nearest 0.01 g. Prey was identified to the lowest possible taxonomic group. For this analysis, unidentified clupeid remains were combined with explicit herring observations to define the amount and occurrence of herring in a stomach (explicit herring observations accounted for 66% of the herring weight observed among all stomachs). Atlantic herring are the dominant clupeid prey in the region, and most of the unidentified clupeid remains are also likely Atlantic herring (Smith and Link, 2010; NEFSC, 2012). Analyses were restricted to those predators that had at least ten stomach observations that contained herring and at least 0.1% of all stomachs sampled among all years contained herring. Restricting the analyses to these 15 predators was similar to what has been done in recent herring stock assessments (NEFSC, 2012), but was also intended to help avoid model convergence problems that might occur by including predators with relatively low sample sizes. GAMM for amount of herring in stomachs with positive herring occurrence In GAMMs for the amount of herring in stomachs, the natural log of the weight (g) of herring in the stomach was always the dependent variable. All models were fit using package gamm4 in the R statistical software (version R-3.3.3; Wood and Scheipl, 2014; R Core Team, 2017). Fixed effects included factors for (i) geographic area (Georges Bank, Gulf of Maine, Mid-Atlantic Bight, Southern New England, and Scotian Shelf; Figure 1; a ⁠), (ii) season (spring or fall; s ⁠), or (iii) the product factor of area and season (⁠ αas ⁠; the product factor was never included in a model with either of the individual effects to avoid collinearity; see below). Area was considered a fixed effect because samples covered the entire range of the Atlantic herring stock and represented the entire spatial domain of interest. Season was considered a fixed effect because spring and fall do not represent subsamples from a larger population of interest, which would justify a random effect, but systematically chosen sampling times. Treating season as random would also require estimating a variance for a distribution using two observations (i.e. spring and fall), which would be inestimable or poorly determined at best. Smooths in the form of thin plate regression splines (Wood, 2003) were applied to predator length [f(li)] and the amount of herring catch in the tow from which a stomach was sampled [f(ci)]. Random intercepts were included for year [by∼ N0,σb2] ⁠, predator species [mr∼ N0,σm2] ⁠, and all two- and three-way interactions of year, predator, area, and season. Models with the four-way interaction did not converge. Random intercepts were assumed to be normally distributed with mean zero and variance estimated by the model. Random effects of year, predator species, and interaction of year and predator species each nested within the fixed effects of area and/or season were also evaluated, where the random effects were assumed distributed as multivariate normal, with each row of the variance/covariance matrix corresponding to a level of the given fixed effect (e.g. a random effect nested within area would have a separate variance estimated for each level of area; Bates et al., 2015). Year was considered a random effect because variance among years was of interest and the ability to make inference about years outside those sampled was desired. Similarly, the bottom-trawl surveys are not efficient samplers of all herring predators [e.g. striped bass (Morone saxatilis), blue shark (Prionace glauca), and bluefish (Pomatomus saltatrix)], but the ability to draw inference about predators that were poorly or not sampled was of interest, as was the variance among predators (see Results). The fully saturated model, excluding all of the random interactions for brevity, was lnhi=μ+αas+fli+fci+by+mr+qy|a+oy|s+kr|a+zr|s+ɛi,(1) where h was the observed weight of herring in stomach i ⁠, μ was the overall model intercept, qy|a was the random year effect nested within area and ∼MVN 0,σq2 ⁠, oy|s was the random year effect nested within season and ∼MVN 0,σo2 ⁠, kr|awas the random effect of predator species nested within area and ∼MVN 0,σk2 ⁠, zr|s was the random effect of predator species nested within season and ∼MVN 0,σz2 ⁠, and ɛ was residual error ∼ N0,σɛ2 ⁠. A random effect for tow that would account for the correlation among stomachs sampled from the same tow was considered, but ∼75% of tows only had one stomach that contained herring, which made estimation of a tow effect impractical. Figure 1. Open in new tabDownload slide Map of the geographic areas. Figure 1. Open in new tabDownload slide Map of the geographic areas. Model selection was conducted using Akaike’s Information Criterion (AIC; Burnham and Anderson, 2002) and a two-step procedure. In step 1, the random effects were evaluated while retaining all the fixed effects and smooths in the model (Ngo and Brand, 1997; Deroba and Bence, 2009), and with model fitting done using restricted maximum likelihood (REML). REML was used in step 1 because it is superior to maximum likelihood (ML) for estimating random effects (McCulloch and Searle, 2001). Random effects were evaluated before fixed effects so that the final model had the simplest error structure possible (i.e. retaining a fixed effect that explained a similar source of variation was preferred to including a random effect). The full factorial combination of random effects was evaluated. In step 2, the fixed effects were evaluated while using the set of random effects that had the lowest AIC in step 1, and with model fitting done with ML instead of REML. Models in step 2 were fit using ML instead of REML because comparisons with AIC based on fits using REML are not valid for models with different fixed effects (Deroba and Bence, 2009). The full factorial combination of fixed effects and smooths was evaluated. The product factor of area and season was never included in a model with either of the individual effects to avoid collinearity. Smooths were evaluated with the fixed effects because smooths in gamm4 are parameterized to be composed of a fixed effect and a random effect with eight levels (Wood, 2006; Wood and Scheipl, 2014). The variance estimate for the random effect portion dictates the degree of smoothness. In this way, the GAMM reduces to a generalized linear mixed model and no longer requires the use of penalized likelihood or the somewhat subjective determination of a basis dimension and effective number of parameters, as is typically required in a generalized additive model (Hastie and Tibshirani, 1990). The fixed effects and smooths from the model with the lowest AIC were retained in the final model, along with the set of random effects that had the lowest AIC in step 1. Results were reported for the final model fit using REML. Results for the fixed effects, smooths, and random effects in the final model were reported by exponentiating the sum of the model intercept and each coefficient: h^j=eμ+j,(2) where j was a generic representation of any coefficient from a fixed effect, smooth, or random effect. The method puts results in more intuitive units of grams of herring in stomachs h^ associated with the given coefficient. The method also isolates the results for a given effect while ignoring the other effects retained in the final model. GAMM for probability of a positive herring occurrence in a stomach The probability p of a positive herring occurrence was modelled as binomial using a GAMM with a logit link function. Initially, model selection was attempted using the same approach as for GAMMs of the amount of herring in stomachs, but nearly all models did not converge. Models in the binomial family, especially using data with many zeros, are known to have convergence issues (Collett, 2003; Wood, 2006; Buchheister and Latour, 2016). Instead, GAMMs were fit using the mgcv package in the R statistical software (Wood, 2004; R Core Team, 2017). The GAMMs fit in mgcv used penalized likelihood for model fitting; therefore, the more parsimonious parameterization of gamm4, especially as it pertains to the smooths and random effects, was lost. Changing the modelling approach to improve model convergence was preferred over further reducing the dataset to predators that have more frequent positive occurrences of herring in their stomachs because continued restriction of the dataset would reduce generality and comparability of the predator species that have been considered of interest for stock assessment (NEFSC, 2012). The fixed effects, smooths, and random intercepts that were considered were the same as in the GAMMs for the amount of herring in stomachs. Package mgcv does not, however, have the capability to nest random effects within other factors; therefore, those types of random effects were not evaluated. Treating year as a random effect (alone or as an interaction) resulted in non-convergence or a variance parameter on the bound of 0.0. Consequently, year was evaluated as a fixed effect because temporal trends in the probability of herring occurrence in stomachs was still of interest. Smooths for predator length and the amount of herring catch in the tow were still applied using thin plate regression splines (Wood, 2003). The fully saturated model, again excluding the random interactions for brevity, was lnpi1-pi=μ+βy+αas+fli+fci+mr;(3) where βy was the fixed effect of year, and all other symbols were defined as above. The variance terms of random effects were estimated, and the coefficients for each level of the random effects were estimated using an identity penalty matrix (i.e. a ridge penalty; Wood, 2008). The identity penalty is equivalent to assuming that the coefficients are independent and identically distributed as normal. Unlike fits using REML or ML, where the coefficients associated with each level of the random effects are integrated out of the likelihood and do not contribute to the number of parameters, the coefficients for random effects using penalized ML contributed to the effective number of parameters (Wood, 2008). The two-step model-selection procedure used for GAMMs for the amount of herring in stomachs, where random effects were evaluated using REML before fixed effects were evaluated using ML, was not needed here because the superiority of REML for random effects was lost by estimating the coefficients of the random effects using penalized ML. So, all models were fit using penalized ML in a full factorial design with model selection done using AIC. Results for the fixed effects, smooths, and random effects in the final model were reported by summing the model intercept and each coefficient, and then converting this logit scale value into a probability p^ ⁠: p^j=eμ+j1+eμ+j.(4) Developing an index of herring abundance An annual index of herring abundance Iy was developed using the year effect coefficients from the GAMM for the amount of herring in stomachs by ⁠, and the probability of a stomach containing a herring βy ⁠: h^y=eμ+by; p^y=eμ+βy1+eμ+βy;(5) Iy=h^y×p^y. Measures of uncertainty (e.g. confidence intervals) were not provided for the index of abundance because methods for combining uncertainty measures from the multistage sampling of the stomachs within the bottom-trawl survey and those from the separate GAMMs have not been developed. The trend among years in the index of abundance was qualitatively compared to the time-series of estimated total herring biomass from the 2015 stock assessment (Deroba, 2015). Results GAMM for amount of herring in stomachs with positive herring occurrence The model with the set of random effects that had the lowest AIC and was, therefore, considered “best” was 2.26 units better than the second-best model, and all other models had <0.01% probability of being the best (Table 1). Similarly, the model with the best set of fixed effects had an AIC that was 3.99 units better than the second-best model, and all other models had <0.01% probability of being the best (Table 1). The overall best model, for which results were reported, included a fixed effect for the product factor of area and season, a smooth for predator length, and random intercepts for year, predator species, the interaction of year and the product factor of area and season dy,as ⁠, and the interaction of year, predator species, and the product factor of area and season gy,r,as ⁠: lnhi=μ+αas+fli+by+mr+dy,as+gy,r,as+ɛi.(6) Table 1. Fit and model selection criteria for the five best models based on AIC from each of the GAMMs. No. of parameters . Log likelihood . AIC . AIC-best AIC . Model likelihood . Model probability . Model covariates . Random effects for GAMM for the amount of herring in stomachs 19 –3 514.60 7 067.19 0.00 1.00 0.76 αas+fli+fci+by+mr+dy,as+gy,r,as 18 –3 516.72 7 069.45 2.26 0.32 0.24 αas+fli+fci+mr+dy,as+gy,r,as 72 –3 470.07 7 084.13 16.94 0.00 0.00 αas+fli+fci+mr+qy|a+oy|s+gy,r,as 73 –3 470.08 7 086.16 18.97 0.00 0.00 αas+fli+fci+by+mr+qy|a+oy|s+gy,r,as 34 –3 509.21 7 086.42 19.23 0.00 0.00 αas+fli+fci+mr+qy|a+vy,r|s Fixed effects for GAMM for the amount of herring in stomachs 17 –3 502.36 7 038.73 0.00 1.00 0.88 αas+fli+by+mr+dy,as+gy,r,as 19 –3 502.36 7 042.72 3.99 0.14 0.12 αas+fli+fci+by+mr+dy,as+gy,r,as 8 –3 518.87 7 053.74 15.02 0.00 0.00 fli+by+mr+dy,as+gy,r,as 9 –3 518.35 7 054.70 15.97 0.00 0.00 τs+fli+by+mr+dy,as+gy,r,as 12 –3 516.03 7 056.07 17.34 0.00 0.00 θa+fli+by+mr+dy,as+gy,r,as GAMM for the probability that a stomach contains herring 132.75 –10 806.38 21 878.26 0.00 1.00 0.32 βy+αas+ fli+ fci+mr+nr,a,s 136.61 –10 804.18 21 881.58 3.31 0.19 0.06 βy+τs+ fli+ fci+mr+nr,a,s 138.31 –10 802.56 21 881.74 3.48 0.18 0.06 βy+θa+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+nr,a,s No. of parameters . Log likelihood . AIC . AIC-best AIC . Model likelihood . Model probability . Model covariates . Random effects for GAMM for the amount of herring in stomachs 19 –3 514.60 7 067.19 0.00 1.00 0.76 αas+fli+fci+by+mr+dy,as+gy,r,as 18 –3 516.72 7 069.45 2.26 0.32 0.24 αas+fli+fci+mr+dy,as+gy,r,as 72 –3 470.07 7 084.13 16.94 0.00 0.00 αas+fli+fci+mr+qy|a+oy|s+gy,r,as 73 –3 470.08 7 086.16 18.97 0.00 0.00 αas+fli+fci+by+mr+qy|a+oy|s+gy,r,as 34 –3 509.21 7 086.42 19.23 0.00 0.00 αas+fli+fci+mr+qy|a+vy,r|s Fixed effects for GAMM for the amount of herring in stomachs 17 –3 502.36 7 038.73 0.00 1.00 0.88 αas+fli+by+mr+dy,as+gy,r,as 19 –3 502.36 7 042.72 3.99 0.14 0.12 αas+fli+fci+by+mr+dy,as+gy,r,as 8 –3 518.87 7 053.74 15.02 0.00 0.00 fli+by+mr+dy,as+gy,r,as 9 –3 518.35 7 054.70 15.97 0.00 0.00 τs+fli+by+mr+dy,as+gy,r,as 12 –3 516.03 7 056.07 17.34 0.00 0.00 θa+fli+by+mr+dy,as+gy,r,as GAMM for the probability that a stomach contains herring 132.75 –10 806.38 21 878.26 0.00 1.00 0.32 βy+αas+ fli+ fci+mr+nr,a,s 136.61 –10 804.18 21 881.58 3.31 0.19 0.06 βy+τs+ fli+ fci+mr+nr,a,s 138.31 –10 802.56 21 881.74 3.48 0.18 0.06 βy+θa+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+nr,a,s Fixed effect product factor of area and season αas ⁠, smooth of predator length fli ⁠, smooth of herring in catch fci ⁠, random year effect by ⁠, random predator species effect mr ⁠, random interaction of year and product factor of area and season dy,as ⁠, random interaction of year, predator species, and product factor of area and season gy,r,as ⁠, random effect of year nested within area qy|a ⁠, random effect of year nested within season oy|s ⁠, random interaction of year and predator species nested within season vy,r|s ⁠, fixed effect of season τs ⁠, fixed effect of area θa ⁠, fixed effect of year βy ⁠, and random interaction of predator species, area, and season nr,a,s ⁠. Open in new tab Table 1. Fit and model selection criteria for the five best models based on AIC from each of the GAMMs. No. of parameters . Log likelihood . AIC . AIC-best AIC . Model likelihood . Model probability . Model covariates . Random effects for GAMM for the amount of herring in stomachs 19 –3 514.60 7 067.19 0.00 1.00 0.76 αas+fli+fci+by+mr+dy,as+gy,r,as 18 –3 516.72 7 069.45 2.26 0.32 0.24 αas+fli+fci+mr+dy,as+gy,r,as 72 –3 470.07 7 084.13 16.94 0.00 0.00 αas+fli+fci+mr+qy|a+oy|s+gy,r,as 73 –3 470.08 7 086.16 18.97 0.00 0.00 αas+fli+fci+by+mr+qy|a+oy|s+gy,r,as 34 –3 509.21 7 086.42 19.23 0.00 0.00 αas+fli+fci+mr+qy|a+vy,r|s Fixed effects for GAMM for the amount of herring in stomachs 17 –3 502.36 7 038.73 0.00 1.00 0.88 αas+fli+by+mr+dy,as+gy,r,as 19 –3 502.36 7 042.72 3.99 0.14 0.12 αas+fli+fci+by+mr+dy,as+gy,r,as 8 –3 518.87 7 053.74 15.02 0.00 0.00 fli+by+mr+dy,as+gy,r,as 9 –3 518.35 7 054.70 15.97 0.00 0.00 τs+fli+by+mr+dy,as+gy,r,as 12 –3 516.03 7 056.07 17.34 0.00 0.00 θa+fli+by+mr+dy,as+gy,r,as GAMM for the probability that a stomach contains herring 132.75 –10 806.38 21 878.26 0.00 1.00 0.32 βy+αas+ fli+ fci+mr+nr,a,s 136.61 –10 804.18 21 881.58 3.31 0.19 0.06 βy+τs+ fli+ fci+mr+nr,a,s 138.31 –10 802.56 21 881.74 3.48 0.18 0.06 βy+θa+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+nr,a,s No. of parameters . Log likelihood . AIC . AIC-best AIC . Model likelihood . Model probability . Model covariates . Random effects for GAMM for the amount of herring in stomachs 19 –3 514.60 7 067.19 0.00 1.00 0.76 αas+fli+fci+by+mr+dy,as+gy,r,as 18 –3 516.72 7 069.45 2.26 0.32 0.24 αas+fli+fci+mr+dy,as+gy,r,as 72 –3 470.07 7 084.13 16.94 0.00 0.00 αas+fli+fci+mr+qy|a+oy|s+gy,r,as 73 –3 470.08 7 086.16 18.97 0.00 0.00 αas+fli+fci+by+mr+qy|a+oy|s+gy,r,as 34 –3 509.21 7 086.42 19.23 0.00 0.00 αas+fli+fci+mr+qy|a+vy,r|s Fixed effects for GAMM for the amount of herring in stomachs 17 –3 502.36 7 038.73 0.00 1.00 0.88 αas+fli+by+mr+dy,as+gy,r,as 19 –3 502.36 7 042.72 3.99 0.14 0.12 αas+fli+fci+by+mr+dy,as+gy,r,as 8 –3 518.87 7 053.74 15.02 0.00 0.00 fli+by+mr+dy,as+gy,r,as 9 –3 518.35 7 054.70 15.97 0.00 0.00 τs+fli+by+mr+dy,as+gy,r,as 12 –3 516.03 7 056.07 17.34 0.00 0.00 θa+fli+by+mr+dy,as+gy,r,as GAMM for the probability that a stomach contains herring 132.75 –10 806.38 21 878.26 0.00 1.00 0.32 βy+αas+ fli+ fci+mr+nr,a,s 136.61 –10 804.18 21 881.58 3.31 0.19 0.06 βy+τs+ fli+ fci+mr+nr,a,s 138.31 –10 802.56 21 881.74 3.48 0.18 0.06 βy+θa+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+mr+nr,a,s 138.66 –10 802.34 21 881.98 3.72 0.16 0.05 βy+ fli+ fci+nr,a,s Fixed effect product factor of area and season αas ⁠, smooth of predator length fli ⁠, smooth of herring in catch fci ⁠, random year effect by ⁠, random predator species effect mr ⁠, random interaction of year and product factor of area and season dy,as ⁠, random interaction of year, predator species, and product factor of area and season gy,r,as ⁠, random effect of year nested within area qy|a ⁠, random effect of year nested within season oy|s ⁠, random interaction of year and predator species nested within season vy,r|s ⁠, fixed effect of season τs ⁠, fixed effect of area θa ⁠, fixed effect of year βy ⁠, and random interaction of predator species, area, and season nr,a,s ⁠. Open in new tab More northerly areas (Georges Bank and Gulf of Maine) tended to have higher amounts of herring in stomachs in fall than in spring, with the exception of the Scotian Shelf, which was similar between seasons, while the opposite was true for the more southerly areas (Mid-Atlantic Bight and Southern New England; Figure 2a). The amount of herring in stomachs generally increased with predator length (Figure 3a). Of the random effects, predator species had the highest variance estimate, year had the lowest, and the interactions were intermediate (Figure 4a). The random coefficients for the year effect generally did not have a consistent trend among years (Figure 5a). Flatfish (Pleuronectiformes) and skates (Rajidae) tended to have less herring in their stomachs than did other predators (Figure 6a). Figure 2. Open in new tabDownload slide Results for the product factor of area and season from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). The area and season combinations were: Georges Bank in fall (GBFALL), Georges Bank in spring (GBSPRING), Gulf of Maine in fall (GoMFALL), Gulf of Maine in spring (GoMSPRING), Mid-Atlantic Bight in fall (MABFALL), Mid-Atlantic Bight in spring (MABSPRING), Scotian Shelf in fall (ScSFALL), Scotian Shelf in spring (ScSSPRING), Southern New England in fall (SNEFALL), and Southern New England in spring (SNESPRING). Figure 2. Open in new tabDownload slide Results for the product factor of area and season from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). The area and season combinations were: Georges Bank in fall (GBFALL), Georges Bank in spring (GBSPRING), Gulf of Maine in fall (GoMFALL), Gulf of Maine in spring (GoMSPRING), Mid-Atlantic Bight in fall (MABFALL), Mid-Atlantic Bight in spring (MABSPRING), Scotian Shelf in fall (ScSFALL), Scotian Shelf in spring (ScSSPRING), Southern New England in fall (SNEFALL), and Southern New England in spring (SNESPRING). Figure 3. Open in new tabDownload slide Results for smooths of predator length from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). The grey-shaded areas are 95% confidence intervals, and the vertical bars along the x-axis are a “rug plot” that indexes the distribution of data at each predator length. Figure 3. Open in new tabDownload slide Results for smooths of predator length from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). The grey-shaded areas are 95% confidence intervals, and the vertical bars along the x-axis are a “rug plot” that indexes the distribution of data at each predator length. Figure 4. Open in new tabDownload slide Standard deviation estimates for the random effects from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). Figure 4. Open in new tabDownload slide Standard deviation estimates for the random effects from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). Figure 5. Open in new tabDownload slide Results for the random effect of year from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). Figure 5. Open in new tabDownload slide Results for the random effect of year from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). Figure 6. Open in new tabDownload slide Results for the random effect of predator species from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). Figure 6. Open in new tabDownload slide Results for the random effect of predator species from a GAMM for the amount of herring in stomachs (a) and the probability of a stomach containing herring (b). GAMM for probability of a positive herring occurrence in a stomach The best model had an AIC that was 3.31 units better than the second-best model, and all models other than the best had ≤6% probability of being the best (Table 1). The overall best model, for which results were reported, included fixed effects for year and the product factor of area and season, smooths for predator length and the amount of herring catch in the tow from which a stomach was sampled, and random effects for predator species, and the interaction of predator species and the product factor of area and season nr,as ⁠: lnpi1-pi=μ+βy+αas+ fli+ fci+mr+nr,as.(7) The probability of a stomach containing a herring generally increased from the 1970s to the late 1990s and has varied without trend since (Figure 5b). Similar to the amount of herring in stomachs, more northerly areas (Georges Bank and Gulf of Maine) had higher probabilities in fall than in spring, and the Scotian Shelf was similar between seasons (Figure 2b). In Southern New England, probabilities were higher in spring than in fall, but the opposite was true for the Mid-Atlantic Bight (Figure 2b). The probability of a stomach containing herring was dome shaped with predator length, with the probability increasing to a peak at ∼90 cm and declining thereafter, although uncertainty was relatively high at larger sizes (Figure 3b). The probability of a stomach containing herring increased with the amount of herring catch in a tow from 0.0 to ∼50 kg, and varied without trend at larger catches where changes in probability were also more likely due to low sample size than true effects (Figure 7). As with the amount of herring in stomachs, the random effect for predator species had the largest variance of the random effects, with standard deviation being nearly double that of the effect for the interaction of predator species and the product factor of area and season (Figure 4b). Also, similar to results for the amount of herring in stomachs, the probability of a stomach containing a herring was generally lower for flatfish and skates than for other species (Figure 6b). While sea raven (Hemitripterus americanus) had a below-average probability of a stomach containing a herring, it had the largest weight of herring observed in those stomachs that did have herring (Figure 6a and b). Figure 7. Open in new tabDownload slide Results for a smooth of the amount of herring caught in a tow from a GAMM for the probability of a stomach containing herring. The grey-shaded areas are 95% confidence intervals, and the vertical bars along the x-axis are a “rug plot” that indexes the distribution of data at each amount of herring catch. Figure 7. Open in new tabDownload slide Results for a smooth of the amount of herring caught in a tow from a GAMM for the probability of a stomach containing herring. The grey-shaded areas are 95% confidence intervals, and the vertical bars along the x-axis are a “rug plot” that indexes the distribution of data at each amount of herring catch. Developing an index of herring abundance The index of abundance generally increased from the 1970s to a peak in 2000, decreased for 3 years, and varied without trend through the end of the time-series (Figure 8). The index of abundance generally matched the trend among years in estimated total herring biomass (Figure 8). Figure 8. Open in new tabDownload slide An index of Atlantic herring abundance derived from GAMMs of stomach-contents data (solid line) and time-series estimates of total herring biomass from a stock assessment (dashed line; Deroba, 2015). Figure 8. Open in new tabDownload slide An index of Atlantic herring abundance derived from GAMMs of stomach-contents data (solid line) and time-series estimates of total herring biomass from a stock assessment (dashed line; Deroba, 2015). Discussion Stomach-contents data vary temporally and spatially for a variety of reasons, such as ontogeny, habitat variability, prey diversity, and temperature (Nunn et al., 2012), and knowledge of this variation can improve the use of stomach-contents data in multispecies assessments, foodweb models, and predator consumption estimates. Reum and Essington (2008) defined predator guilds in Puget Sound, WA using stomach-contents data and found that one-third of predators switched guilds among fall, winter, and summer seasons. Although in a different context, the results of this study are consistent with Reum and Essington (2008). More specifically, the product factor of area and season was included in the final models for the amount of herring in stomachs and the probability of a stomach containing a herring. Stomach-contents data are often not available from all seasons, however, so samples from one season have been assumed to reflect conditions in other seasons in order to define annual diet compositions or consumption. Stomach-contents data that varied in availability by quarter of the year (i.e. a proxy for season) were used to construct a MSVPA in the eastern Bering Sea, with the implicit assumption that in some cases samples from a single quarter represented annual diets (Livingston and Jurado-Molina, 2000). If diet varies among seasons, as in this study, however, the observations from one season would be a biased representation of the annual stomach contents and result in biased estimates from the MSVPA. Using similar datasets as in this study, annual predator consumption of Atlantic mackerel (Scomber scombrus) and herring were estimated by assuming stomach contents from the fall also represented the winter, and spring observations also represented the summer (Overholtz and Link, 2007; Overholtz et al., 2008; Moustahfid et al., 2009). While only variation between the spring and fall seasons were evaluated in the GAMMs here, the differences suggest that the stomach contents of herring predators during summer and winter may also systematically differ. If so, then the annual consumption estimates will be biased to some degree because accounting for only the two seasons is insufficient to represent annual habits, and consumption estimates calculated in this manner have been shown to be sensitive to diet composition (Overholtz and Link, 2007). Stomach-contents samples from the summer season in 2 years were used to construct a static foodweb model representative of the early 1990s in the Gulf of Alaska (Gaichas et al., 2010). Such methods produce a model indicative of that single season, which may restrict utility for management decisions needed in other seasons or on an annual basis. These results suggest a need for broader seasonal sampling of stomach contents. As with seasons, stomach-contents samples have also been combined across broad geographic regions, or samples from a limited spatial range have been assumed to represent a broader range, but knowledge of how stomach contents vary spatially can be used to evaluate these assumptions and make model improvements. An MSVPA for the Northeast US continental shelf combined stomach contents across the entire shelf and assumed 100% spatial overlap between predators and prey (Tyrrell et al., 2008). Similarly, annual predator consumption of Atlantic mackerel and herring were estimated by combining stomach contents across the entire Northwest Atlantic, and these consumption estimates were used to estimate natural mortality of Atlantic mackerel and herring (Overholtz and Link, 2007; Overholtz et al., 2008; Moustahfid et al., 2009). Both the amount of herring in stomachs and the probability of a stomach containing herring varied spatially in this study, which also focused on a similar study area (i.e. Northwest Atlantic). Results from the GAMMs could be used to inform how stomach-contents data should be combined in a spatially stratified way, which would improve measures of uncertainty in the stomach contents and consumption estimates. Improved measures of uncertainty could inform sensitivity analysis for the MSVPA and be used in weighting consumption estimates in subsequent statistical model fitting (e.g. statistical catch-at-age models). Some of the spatial variation in stomach contents is also likely driven by variation in overlap of herring and their predators; not acknowledging this variation might induce positive bias in the strength and scale of the predator–prey interactions from an MSVPA (Tyrrell et al., 2008) or positively bias estimates of consumption (Overholtz and Link, 2007; Overholtz et al., 2008; Moustahfid et al., 2009). Further studies on movement of predators and prey would be required to inform a more accurate specification of overlap. Stomach-contents data have also been combined among years to reduce noise caused by measurement error (Van Kirk et al., 2010; Curti et al., 2013), but results of this study suggest that some of the variation among years is likely true process variation. In a MSCAA of Georges Bank (Curti et al., 2013) and in estimating consumption of Atlantic mackerel in the Northwest Atlantic (Moustahfid et al., 2009), stomach contents were averaged over 5-year periods. The amount of variation among years (i.e. including random interactions with year) in both GAMMs in this study suggests that such averaging would likely miss true variation in herring predator diets, which may also be true for other prey items like mackerel. Averaging stomach contents among years and missing true signals in the data could induce bias of scale in a MSCAA and in estimating consumption by falsely increasing or decreasing the strength of interactions in some years. Similarly, such averaging could induce bias in trends by assigning observations to years that do not reflect conditions in those years. A trade-off is likely to exist between averaging enough years to reduce the effect of measurement error, but not averaging so many as to induce bias. In a static foodweb model of the Gulf of Alaska, 2 years of stomach-contents data were used to reflect conditions in the early 1990s (Gaichas et al., 2010), while stomach contents from single years for some predators were used to reflect conditions during 1974–2000 in a foodweb model of the Baltic Sea (Harvey et al., 2003). The degree of variation in stomach-contents data among years will dictate how long such models will remain indicative of the system, but such an evaluation (e.g. the GAMMs in this study) will require sampling stomach contents more consistently among years. Similarly, a dynamic foodweb model was applied in the Gulf of Alaska, but did not use stomach-contents data (Gaichas et al., 2011). Fits of the foodweb model to input biomass and catch time-series were generally poor, and this may have been due to not permitting enough variation in diets among years, which was supported by existing stomach-contents data from the Gulf of Alaska, but not used in model fitting (Gaichas et al., 2011). Gaichas et al. (2011) encouraged the expansion of the foodweb model to include stomach-contents data in fitting, which would also serve the purpose of reducing the number of plausible hypotheses related to predator and prey vulnerabilities, although foodweb models have been shown to be less sensitive to uncertainty in diet composition data than other inputs (e.g. input biomass; Essington, 2007). Results of GAMMs like those in this study could serve as the basis for an operating model that simulates predator–prey interactions and the collection of diet data with measurement error. The estimates from the GAMMs could define how the predator–prey interactions (i.e. amount and probability of prey occurring in a stomach) vary spatially and temporally, and the degree of measurement error in the simulated diet data could be consistent with the residual and random-effect variance estimates. The simulated diet data could then be combined across space, seasons, or years to evaluate the consequences of such data aggregations in subsequent multispecies modelling. This approach has been used extensively to evaluate the performance of single-species stock assessments (Deroba et al., 2015) and for some multispecies applications (Essington, 2007). Using the results of the GAMMs has the advantage of conditioning the operating model on actual data, as opposed to using uninformative uniform distributions to represent the uncertainty in diet composition, as has been done previously (Overholtz and Link, 2007). Conditioning operating models on fits to real data has been identified as a best practice for simulation tests of this sort (Deroba et al., 2015). Results of models such as GAMs that explore variation in stomach-contents data could also be used to more efficiently allocate sampling effort. Stefánsson and Pálsson (1997) used a delta approach and GAM models to analyse the spatial variation in Icelandic cod (Gadus morhua) stomach contents. The spatial variance of the stomach contents was suggested for use to allocate stomach sampling effort so that resources were not wasted trying to collect many stomachs from locations where prey were scarce and observations imprecise. The only species of interest in Stefánsson and Pálsson (1997), however, was Icelandic cod, whereas the surveys used to collect stomach-contents data in this study were designed for multispecies sampling. If the information lost by focusing sampling efforts on the stomachs of herring predators is considered acceptable, however, then this study could serve as a starting point to consider revisions to the spatial and temporal distribution of stomach sampling in the Northwest Atlantic because results inform how the amounts and probabilities of herring encounters vary among seasons and areas. However, follow-up studies directed at measuring the intraregion/season variance of stomach contents would still be needed. Results for the product factor of area and season in the GAMMs were generally consistent with knowledge about the migration patterns of Atlantic herring. Gulf of Maine and Georges Bank herring spawn in September–October and then migrate south for the winter months before returning to feed in the Gulf of Maine and Georges Bank areas in summer (Reid et al., 1999). Herring on the Scotian Shelf, however, spawn approximately at the same time, but migrate north along the east coast of Nova Scotia in winter, returning south in summer (Reid et al., 1999). These migration patterns are likely why the amount of herring and probability of herring were generally higher in predator stomachs in the Gulf of Maine and Georges Bank in fall than in spring, while the opposite was true for the Mid-Atlantic Bight and Southern New England, and results for the Scotian Shelf were less variable between seasons. Water temperatures in the Northwest Atlantic have increased over years in some seasons and areas, and these changes may have implications for predator–prey interactions and subsequent diet (Thomas et al., 2017). Thomas et al. (2017) found that the strongest increases in sea surface temperature occurred in the Scotian Shelf and Gulf of Maine, while the Mid-Atlantic Bight had weaker increasing trends. The trends were also strongest in late summer (July–September), while winter temperature trends were relatively constant or weak. Predators of herring, and herring themselves, are expected to have a range of responses to these temperature changes, including distribution shifts and changes in mortality and growth, with responses varying by functional group (Hare et al., 2016). Consequently, the role of herring as prey is likely to change among years, areas, and predators, and the results of the GAMMs support this conclusion. Year and the product factor of area and season were included in the final models of both GAMMs. Furthermore, the inclusion of random effects for the interaction of year and the product factor of area and season, and the interaction of year, predator species, and the product factor of area and season in the final GAMM for the amount of herring in stomachs suggests that significant temporal and spatial variation has already occurred and that the effects differ by predator. Continued monitoring and quantification of variation in predator diet would be beneficial and could be used to anticipate foodweb shifts as a result of climate change. The variance of the random effect for predator species was the largest of the random effects in both GAMMs, which suggests that a main source of variation is features of the predators themselves. Flatfish and skates were consistently below average in both GAMMs. One explanation may be that predators with subterminal or inferior mouths are less efficient predators of herring than species with terminal mouths, which seems logical given that herring is generally a pelagic species. Gape size may also offer an explanation. In the Northwest Atlantic, Scharf et al. (2000) found that relatively large-gaped predators like goosefish (Lophius americanus) consumed larger prey than relatively small-gaped predators like spiny dogfish (Squalus acanthias), which fed more on invertebrates than fish. Results of this study are consistent with the conclusions of Scharf et al. (2000), as the coefficients for the random effect of predator identified relatively large-gaped predators (e.g. goosefish) as containing more herring with greater probability than relatively smaller-gaped predators (e.g. spiny dogfish). The importance of predator effects in the GAMMs also implies that changes in the abundance of different predators will have different consequences for Atlantic herring. For example, given spatial and temporal overlap, changes in the abundance of consistently above-average predators will have a greater effect on the amount of herring consumed than changes in consistently below-average predators. The GAMM for the probability of a stomach containing a herring included smooths for predator length and the amount of herring catch in the tow. The smooth for predator length was dome shaped, which implies that predators’ preference for herring as prey is also dome shaped. One explanation may be that the range of prey available to a predator increases as the predator grows and their gape width increases, which expands the prey field to fish that generally grow longer than herring (Scharf et al., 2000). Predator length was also included in GAMM models examining variation in diet data in Chesapeake Bay, United States, but the shape of the relationship varied by predator and prey (Buchheister and Latour, 2016). An expanding prey field that increased with body size and gape width was also suggested as an explanation for some of the results in that study. Alternatively, digestion rates may differ among predator sizes such that prey identification is more difficult in stomachs from larger predators. Examining the diversity of prey in relationship to predator length or controlled experiments in prey identification would help disentangle these possibilities, and this might be done using the same stomach-contents data used in this study. The smooth for the amount of herring catch in a tow could be used to develop a threshold that indicates a level of herring biomass, as indexed by survey catches, below which the availability of herring to predators is reduced. The probability of a predator stomach containing a herring increased in tows with herring catches from 0 to ∼50 kg, and the number of observations in this range were relatively high such that the effect was likely real, compared to some of the variation above ∼50 kg where the number of observations were relatively few. The changes in the probability between 0 and ∼50 kg may be indicative of the availability of herring to predators when herring biomass varies in this range. Consequently, mean herring survey catches of ∼ 50 kg tow−1 might serve as a threshold below which the availability of herring to predators is reduced. The consequences of reduced availability of herring, and possible management responses, should be the focus of future research. The index of herring abundance developed from the stomach contents was generally consistent with recent Atlantic herring stock assessment estimates of total biomass (Figure 8; NEFSC, 2012; Deroba, 2015). Stomach contents are relatively time consuming to collect, however, so their use for the sole purpose as an index of abundance may be relatively inefficient when compared to directed surveys, unless a species of interest is not well sampled in surveys (Link, 2004). Nonetheless, treating predators as a survey platform can provide a useful point of comparison for stock assessments and other survey indices. A measure of precision has not been developed for the index of abundance derived from the stomach-contents data in this study. Such measures of precision are necessary, however, to evaluate the index for trends among years and for weighting this index relative to others. Given that the stomach contents were collected during standardized bottom-trawl surveys that also provide indices of abundance, the measures of precision from the bottom-trawl survey indices could serve as a lower bound for the indices derived from stomach contents. Accounting for the correlation in stomach contents sampled from the same tow was not possible in this analysis, but since ∼75% of tows only had one predator stomach that contained herring, the results and conclusions were likely robust. Stefánsson and Pálsson (1997) reported significant within-tow correlation in stomach-contents samples of Icelandic cod, but also discussed the difficulty in accounting for such correlation, including in the GAM models used in their analysis. Buchheister and Latour (2016) included a random effect for tow in delta-GAMMs of predator diets in Chesapeake Bay, United States and advocated for the continued use of random effects as a way to account for correlation among observations. Given the relative sparseness of stomach-contents datasets, accounting for such correlation may significantly increase measures of precision in models of predator–prey relationships (e.g. GAMM models as in this study), and software to include random effects has expanded (e.g. package gamm4), which makes the suggestions of Buchheister and Latour (2016) increasingly feasible. Acknowledgements I am grateful to the staff and crews of the National Marine Fisheries Service Northeast Fisheries Science Center bottom-trawl surveys. I thank Brian Smith for help with data preparation, answering my clarifying questions, comments on draft versions of this work, and making Figure 1. Gary Shepherd, Kiersten Curti, and Sarah Gaichas provided helpful comments and reviews of this research at various stages. Larry Jacobson, Dan Hennen, and Tim Miller were helpful in constructing and understanding the GAMM models. Friends and staff at Papa Jake’s in East Falmouth, MA ensured that I found time for leisure that ultimately improved the research product. Handling Editor Emory Anderson and two anonymous reviewers also provided constructive feedback. References Azarovitz T. R. 1981 . A brief historical review of the Woods Hole Laboratory trawl survey time series. In Bottom Trawl Surveys , pp. 62 – 67 . Ed. by Doubleday W.G., Rivard D.. Canadian Special Publication of Fisheries and Aquatic Sciences 58 . Government of Canada, Fisheries and Oceans, Hull, Quebec. 273 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Baker R. , Buckland A., Sheaves M. 2014 . Fish gut content analysis: robust measures of diet composition . Fish and Fisheries , 15 : 170 – 177 . Google Scholar Crossref Search ADS WorldCat Bates D. , Mächler M., Bolker B. M., Walker S. C. 2015 . Fitting linear mixed-effects models using lme4 . Journal of Statistical Software , 67 : 1 – 48 . Google Scholar Crossref Search ADS WorldCat Brooks E. N. , Deroba J. J. 2015 . When “data” are not data: the pitfalls of post hoc analyses that use stock assessment model output . Canadian Journal of Fisheries and Aquatic Sciences , 72 : 634 – 641 . Google Scholar Crossref Search ADS WorldCat Buchheister A. , Latour R. J. 2016 . Dynamic trophic linkages in a large estuarine system – support for supply-driven dietary changes using delta generalized additive mixed models . Canadian Journal of Fisheries and Aquatic Sciences , 73 : 5 – 17 . Google Scholar Crossref Search ADS WorldCat Burnham K. P. , Anderson D. R. 2002 . Model Selection and Multimodel Inference: A Practical Information Theoretic Approach , 2nd edn. Springer-Verlag , New York . 488 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Collett D. 2003 . Modeling Binary Data , 2nd edn. Chapman & Hall/CRC , Boca Raton, FL . 408 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Curti K. L. , Collie J. S., Legault C. M., Link J. S. 2013 . Evaluating the performance of a multispecies statistical catch-at-age model . Canadian Journal of Fisheries and Aquatic Sciences , 70 : 470 – 484 . Google Scholar Crossref Search ADS WorldCat Deroba J. J. 2015 . Atlantic herring operational assessment report 2015. Northeast Fisheries Science Center Reference Document 15-16. 30 pp. Deroba J. J. , Bence J. R. 2009 . Developing model-based indices of lake whitefish abundance using commercial fishery catch and effort data in Lakes Huron, Michigan, and Superior . North American Journal of Fisheries Management , 29 : 50 – 63 . Google Scholar Crossref Search ADS WorldCat Deroba J. J. , Butterworth D. S., Methot R. D. Jr, De Oliveira J. A. A., Fernandez C., Nielsen A., Cadrin S. X., et al. 2015 . Simulation testing the robustness of stock assessment models to error: some results from the ICES strategic initiative on stock assessment methods . ICES Journal of Marine Science , 72 : 19 – 30 . Google Scholar Crossref Search ADS WorldCat Essington T. E. 2007 . Evaluating the sensitivity of a trophic mass-balance model (Ecopath) to imprecise data input . Canadian Journal of Fisheries and Aquatic Sciences , 64 : 628 – 637 . Google Scholar Crossref Search ADS WorldCat Gaichas S. K. , Aydin K. Y., Francis R. C. 2010 . Using food web model results to inform stock assessment estimates of mortality and production for ecosystem-based fisheries management . Canadian Journal of Fisheries and Aquatic Sciences , 67 : 1490 – 1506 . Google Scholar Crossref Search ADS WorldCat Gaichas S. K. , Aydin K. Y., Francis R. C. 2011 . What drives dynamics in the Gulf of Alaska? Integrating hypotheses of species, fishing, and climate relationships using ecosystem modeling . Canadian Journal of Fisheries and Aquatic Sciences , 68 : 1553 – 1578 . Google Scholar Crossref Search ADS WorldCat Gislason H. , Helgason T. 1985 . Species interaction in assessment of fish stocks with special application to the North Sea . Dana , 5 : 1 – 44 . Google Scholar OpenURL Placeholder Text WorldCat Grosslein M. D. 1969 . Groundfish survey methods. Bureau of Commercial Fisheries, Woods Hole, MA Laboratory Reference Document 69-02. 34 pp. Hare J. A. , Morrison W. E., Nelson M. W., Stachura M. M., Teeters E. J., Griffis R. B., Alexander M. A., et al. 2016 . A vulnerability assessment of fish and invertebrates to climate change on the northeast US continental shelf . PLoS One , 11 : e0146756. Google Scholar Crossref Search ADS PubMed WorldCat Harvey C. J. , Cox S. P., Essington T. E., Hansson S., Kitchell J. F. 2003 . An ecosystem model of food web and fisheries interactions in the Baltic Sea . ICES Journal of Marine Science , 60 : 939 – 950 . Google Scholar Crossref Search ADS WorldCat Hastie T. J. , Tibshirani R. J. 1990 . Generalized Additive Models . Chapman & Hall , New York . 352 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jech J. M. , Sullivan P. J. 2014 . Distribution of Atlantic herring in the Gulf of Maine from 1998 to 2012 . Fisheries Research , 156 : 26 – 33 . Google Scholar Crossref Search ADS WorldCat Kinzey D. , Punt A. E. 2009 . Multispecies and single-species model of fish population dynamics: comparing parameter estimates . Natural Resource Modeling , 22 : 67 – 104 . Google Scholar Crossref Search ADS WorldCat Lewy P. , Vinther M. 2004 . A stochastic age-length structured multispecies model applied to North Sea stocks. ICES Document CM 2004/FF: 20. 33 pp. Link J. S. 2004 . Using fish stomachs as samplers of the benthos: integrating long-term and broad scales . Marine Ecology Progress Series , 269 : 265 – 275 . Google Scholar Crossref Search ADS WorldCat Link J. S. , Almeida F. P. 2000 . An overview and history of the food web dynamics program of the Northeast Fisheries Science Center, Woods Hole, MA. NOAA Technical Memorandum NMFS-NE-159. 60 pp. Livingston P. A. , Jurado-Molina J. 2000 . A multispecies virtual population analysis of the eastern Bering Sea . ICES Journal of Marine Science , 57 : 294 – 299 . Google Scholar Crossref Search ADS WorldCat Maunder M. N. , Punt A. E. 2004 . Standardizing catch and effort data: a review of recent approaches . Fisheries Research , 70 : 141 – 159 . Google Scholar Crossref Search ADS WorldCat McCulloch C. E. , Searle S. R. 2001 . Generalized, Linear, and Mixed Models . Wiley , New York . 325 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Miller T. J. 2013 . A comparison of hierarchical models for relative catch efficiency based on paired-gear data for US Northwest Atlantic fish stocks . Canadian Journal of Fisheries and Aquatic Sciences , 70 : 1306 – 1316 . Google Scholar Crossref Search ADS WorldCat Miller T. J. , Das C., Politis P. J., Miller A. S., Lucey S. M., Legault C. M., Brown R. W., et al. 2010 . Estimation of Albatross IV to Henry B. Bigelow calibration factors. Northeast Fisheries Science Center Reference Document 10-05. 233 pp. Mills K. L. , Laidig T., Ralston S., Sydeman W. J. 2007 . Diets of top predators indicate pelagic juvenile rockfish abundance in the California Current System . Fisheries Oceanography , 16 : 273 – 283 . Google Scholar Crossref Search ADS WorldCat Moustahfid H. , Link J. S., Overholtz W. J., Tyrrell M. C. 2009 . The advantage of explicitly incorporating predation mortality into age-structured stock assessment models: an application for Atlantic mackerel . ICES Journal of Marine Science , 66 : 445 – 454 . Google Scholar Crossref Search ADS WorldCat NEFSC (Northeast Fisheries Science Center) . 2012 . 54th Northeast Regional Stock Assessment Workshop (54th SAW) Assessment Report. Northeast Fisheries Science Center Reference Document 12-18. 600 pp. Ngo L. , Brand R. 1997 . Model selection in linear mixed effects models using SAS proc mixed. Proceedings of the 22nd Annual SAS Users Group International Conference, San Diego, CA, March 16–19, 1997 : 1335 – 1340 . Nunn A. D. , Tewson L. H., Cowx I. G. 2012 . The foraging ecology of larval and juvenile fishes . Reviews in Fish Biology and Fisheries , 22 : 377 – 408 . Google Scholar Crossref Search ADS WorldCat Overholtz W. J. , Link J. S. 2007 . Consumption impacts by marine mammals, fish, and seabirds on the Gulf of Maine-Georges Bank Atlantic herring complex during the years 1977–2002 . ICES Journal of Marine Science , 64 : 83 – 96 . Google Scholar Crossref Search ADS WorldCat Overholtz W. J. , Jacobson L. D., Link J. S. 2008 . An ecosystem approach for assessment advice and biological reference points for the Gulf of Maine-Georges Bank Atlantic herring complex . North American Journal of Fisheries Management , 28 : 247 – 257 . Google Scholar Crossref Search ADS WorldCat R Core Team . 2017 . R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (last accessed 6 March 2017). Read A. J. , Brownstein C. R. 2003 . Considering other consumers: fisheries, predators, and Atlantic herring in the Gulf of Maine . Conservation Ecology , 7 : 2 . Google Scholar Crossref Search ADS WorldCat Reid R. N. , Cargnelli L. M., Griesbach S. J., Packer D. B., Johnson D. L., Zettin C. A., Morse W. W., et al. 1999 . Essential fish habitat source document: Atlantic herring, Clupea harengus, life history and habitat characteristics. NOAA Technical Memorandum NMFS-NE-126. 48 pp. Reum J. C. P. , Essington T. E. 2008 . Seasonal variation in guild structure of the Puget Sound demersal fish community . Estuaries and Coasts , 31 : 790 – 801 . Google Scholar Crossref Search ADS WorldCat Scharf F. S. , Juanes F., Rountree R. A. 2000 . Predator size–prey size relationships of marine fish predators: interspecific variation and effects of ontogeny and body size on trophic-niche breadth . Marine Ecology Progress Series , 208 : 229 – 248 . Google Scholar Crossref Search ADS WorldCat Smith B. E. , Link J. S. 2010 . The trophic dynamics of 50 finfish and 2 squid species on the northeast US continental shelf. NOAA Technical Memorandum NMFS-NE-216. 640 pp. Stefánsson G. , Pálsson Ó. K. 1997 . Statistical evaluation and modelling of the stomach contents of Icelandic cod . Canadian Journal of Fisheries and Aquatic Sciences , 54 : 169 – 181 . Google Scholar Crossref Search ADS WorldCat Thomas A. C. , Pershing A. J., Friedland K. D., Nye J. A., Mills K. E., Alexander M. A., Record N. R., et al. 2017 . Seasonal trends and phenology shifts in sea surface temperature on the North American northeastern continental shelf . Elementa Science of Anthropocene , 5 : 48 . Google Scholar OpenURL Placeholder Text WorldCat Tyrrell M. C. , Link J. L., Moustahfid H., Overholtz W. J. 2008 . Evaluating the effect of predation mortality on forage species population dynamics in the Northeast US continental shelf ecosystem using multispecies virtual population analysis . ICES Journal of Marine Science , 65 : 1689 – 1700 . Google Scholar Crossref Search ADS WorldCat Van Kirk K. F. , Quinn T. J., Collie J. S. 2010 . A multispecies age-structured assessment model for the Gulf of Alaska . Canadian Journal of Fisheries and Aquatic Sciences , 67 : 1135 – 1148 . Google Scholar Crossref Search ADS WorldCat Van Kirk K. F. , Quinn T. J. II, Collie J. S., A’mar Z. T. 2015 . Assessing uncertainty in a multispecies age-structured assessment framework: the effects of data limitations and model assumptions . Natural Resource Modeling , 28 : 184 – 205 . Google Scholar Crossref Search ADS WorldCat Wood S. 2003 . Thin plate regression splines . Journal of the Royal Statistical Society B , 65 : 95 – 114 . Google Scholar Crossref Search ADS WorldCat Wood S. 2004 . Stable and efficient multiple smoothing parameter estimation for generalized additive models . Journal of the American Statistical Association , 99 : 673 – 686 . Google Scholar Crossref Search ADS WorldCat Wood S. 2006 . Generalized Additive Models: An Introduction with R . Chapman & Hall/CRC Press , Boca Raton, FL . 410 pp. Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Wood S. 2008 . Fast stable direct fitting and smoothness selection for generalized additive models . Journal of the Royal Statistical Society B , 70 : 495 – 518 . Google Scholar Crossref Search ADS WorldCat Wood S. , Scheipl F. 2014 . gamm4: Generalized additive mixed models using mgcv and lme4. R package version 0.2-3. https://CRAN.R-project.org/package=gamm4 (last accessed 23 August 2016). Published by International Council for the Exploration of the Sea 2018. This work is written by a US Government employee and is in the public domain in the US. Published by International Council for the Exploration of the Sea 2018.

Li, Futian; Beardall, John; Gao, Kunshan

doi: 10.1093/icesjms/fsx239pmid: N/A

Abstract Phytoplankton cells living in the surface waters of oceans are experiencing alterations in environmental conditions associated with global change. Given their importance in global primary productivity, it is of considerable concern to know how these organisms will perform physiologically under the changing levels of pH, temperatures, and nutrients predicted for future oceanic ecosystems. Here we show that the model diatom, Thalassiosira pseudonana, when grown at different temperatures (20 or 24 °C), pCO2 (400 or 1000 µatm), and nitrate concentrations (2.5 or 102.5 µmol l−1), displayed contrasting performance in its physiology. Elevated pCO2 (and hence seawater acidification) under the nitrate-limited conditions led to decreases in specific growth rate, cell size, pigment content, photochemical quantum yield of PSII, and photosynthetic carbon fixation. Furthermore, increasing the temperature exacerbated the negative effects of the seawater acidification associated with elevated pCO2 on specific growth rate and chlorophyll content under the N-limited conditions. These results imply that a reduced upward transport of nutrients due to enhanced stratification associated with ocean warming might act synergistically to reduce growth and carbon fixation by diatoms under progressive ocean acidification, with important ramifications for ocean productivity and the strength of the biological CO2 pump. Introduction Increasing emissions of anthropogenic carbon dioxide (CO2) result in global warming that leads to increases in sea surface temperatures (ocean warming). While continuous dissolution of CO2 into oceans remediates atmospheric CO2 increases and global warming, it also leads to ocean acidification (Gattuso et al., 2015). Average global ocean pH has already declined by 0.1 units since the Industrial Revolution and is predicted to further drop by up to 0.4 units by the end of this century (Gattuso et al., 2015). Both ocean warming and acidification affect the availability of nitrogen and other nutrients to phytoplankton living in the surface layers of the ocean. Intensified stratification of surface waters caused by ocean warming decreases upward transport of nutrients across the thermocline (Rost et al., 2008; Beardall et al., 2009; Gao et al., 2012a) leading to reduced nutrient supply. Moreover, ocean acidification could decrease nitrification and thus the supply of oxidized nitrogen in the surface ocean (Beman et al., 2011). These changes are expected to have far-reaching consequences for marine primary production and ecosystem sustainability and services (Gattuso et al., 2015; Boyd et al., 2016). Seawater acidification treatment has been shown, in a range of microalgal species, to cause down-regulation of the energy-costly CO2 concentrating mechanisms (CCMs) (Raven and Beardall, 2014 and references therein) and to reduce the intracellular dissolved inorganic carbon pool within phytoplankton cells (Liu et al., 2017). It also decreases intracellular pH (Suffrian et al., 2011) and increases the energy expenditure of the cells in maintaining cellular pH homeostasis (Taylor et al., 2012). While the physiological responses of phytoplankton to seawater acidification might vary between genera, species (Langer et al., 2006; King et al., 2015; Li et al., 2016), and even among strains (Langer et al., 2009), interactions with other environmental stressors can alter the responses to acidification (Li et al., 2012a; Beardall et al., 2014; Verspagen et al., 2014; Passow and Laws, 2015; Li et al., 2017a). In addition, the duration of growth under seawater acidification could also affect the physiological performance of phytoplankton cells due to acclimation and adaptation processes (Collins et al., 2014; Li et al., 2017b). Therefore, it is important to examine the effects of elevated CO2 under changing environmental conditions or under multiple stressors in both the short and the long term. Interactions between acidification and other factors, such as light intensity (Gao et al., 2012b; Li et al., 2014; Heiden et al., 2016), light fluctuation (Hoppe et al., 2015), solar UV radiation (Li et al., 2012a), temperature (Torstensson et al., 2012; Torstensson et al., 2013; Coello-Camba et al., 2014), and availability of iron (Sugie and Yoshimura, 2013, 2016; Segovia et al., 2017) and other nutrients (King et al., 2011; Sun et al., 2011; Li et al., 2012b) have been reported in diatoms. Diatoms are responsible for about 40% of oceanic primary production and account for over 50% of organic carbon burial in marine sediments (Falkowski et al., 2004). Hence, the responses of their growth and photosynthesis to changing marine environments are likely to influence the export of organic matter to deep water and the biogeochemical cycling of carbon, silicon, and other elements. For instance, diatoms dominate the open ocean microphytoplankton community and live cells have been found in the deep ocean (down to 4000 m), indicating their critical role in the biological CO2 pump (Agusti et al., 2015). On the other hand, N-limitation is usually exacerbated by higher half-saturation constants for nitrate uptake of diatoms compared to other algal classes (Eppley et al., 1969; Falkowski, 1975; Moore et al., 2002a). Indeed, about 50% of surface oceans are nitrogen-limited for diatoms, at least, during summer (Moore et al., 2002b). Given the decreased upward transport of nutrients, decreasing nitrification, and higher half-saturation constants for nitrate uptake, diatoms within the upper mixed layers will become especially exposed to increasing nutrient limitation under the influences of ocean warming and acidification. However, few studies have investigated combined effects among elevated CO2, temperature, and nutrient limitation (Taucher et al., 2015), in part because of the complexity of interactive effects (Boyd et al., 2016). In terms of CO2 and temperature interactions, growth rates of several diatom species freshly isolated from coastal New Zealand waters have been shown to be enhanced, depressed, or unaltered by elevated temperature and CO2, with temperature showing more prominent effects than CO2 (Tatters et al., 2013). In contrast, growth of Chaetoceros cf. wighamii was mainly affected by elevated CO2, rather than by the temperature levels tested (Araújo and Garcia, 2005). On the other hand, elevated levels of temperature and CO2 synergistically enhanced the growth of the Antarctic diatom Nitzschia lecointei (Torstensson et al., 2013), but depressed the primary production of Arctic phytoplankton (Coello-Camba et al., 2014), showing that phytoplankton may show different regional responses. In another study, however, seawater acidification treatment substantially enhanced the growth of diatoms in a phytoplankton community, whereas their growth was unaltered when CO2 and temperature were both elevated (Feng et al., 2009). Effects of elevated CO2 and associated seawater acidification on diatoms have been extensively studied under nitrogen-replete conditions (Gao and Campbell, 2014 and references therein). However, it has been suggested that the effects of acidification might differ under nutrient-limited compared to replete conditions (Taucher et al., 2015). For instance, the carbon to nitrogen ratios in Phaeodactylum tricornutum, Thalassiosira pseudonana, and Thalassiosira weissflogii were raised by acidification treatment only under N-limited conditions (Li et al., 2012b; Hong et al., 2017). Additionally, seawater acidification only increased photosynthetic carbon fixation and maximum relative electron transport of P. tricornutum when nitrogen availability was sufficient (Li et al., 2012b; Hong et al., 2017). There are also reports showing differential responses of diatoms to temperature changes grown at different concentrations of nutrients (Hagerthey et al., 2002; Doyle et al., 2005). While elevated temperature decreased the relative abundance of Nitzschia frustulum more conspicuously under low-nutrient conditions (Hagerthey et al., 2002), it increased the growth of two other diatom species greatly when nutrients were added (Doyle et al., 2005). As a much studied model diatom species, growth and photosynthesis of T. pseudonana have been shown to benefit from, or be unaltered by, seawater acidification (Sobrino et al., 2008; Trimborn et al., 2009; Crawfurd et al., 2011; Gao et al., 2012b; McCarthy et al., 2012; Yang and Gao, 2012; Shi et al., 2015; Hong et al., 2017). However, acidification showed negative effects on photosynthesis when T. pseudonana was cultured in a nitrate-limited chemostat (similar growth rates of ambient and elevated CO2 acclimated cells were controlled by the dilution rate and acidification effects were only studied under nitrate-limited conditions) (Hennon et al., 2014). Changes in temperature and the availability of nitrate may alter the response of diatoms to elevated CO2, as these changes would impact carbon and nitrogen metabolism and cellular energy budgets. We hypothesize that nitrate limitation may aggravate impacts of seawater acidification and superimposition of a warming treatment may further complicate the effects due to increased respiratory carbon loss under acidification and warming. In the present work, we tested this hypothesis by growing the diatom, T. pseudonana, under different levels of nitrate in combination with pCO2 and temperature treatments. Material and methods Culture conditions and experimental design The diatom T. pseudonana (CCMP 1335), originally isolated from coastal waters of Long Island (USA), was grown in 500 ml polycarbonate (PC) bottles maintained at 20 (LT) or 24 °C (HT) with a 12:12 h light–dark cycle. The temperatures were set according to the isolation temperature (20 °C), and the temperature rise of 4 °C was based on the RCP8.5 scenario that predicts the sea surface temperature will increase by this extent at the end of the 21st century (Bopp et al., 2013). In addition, the two temperature regimes are near or in the optimal temperature range of this strain (Boyd et al., 2013). The PC bottles were acid-cleaned and rinsed thoroughly with ultrapure water and autoclaved before being used for cell cultures. Photosynthetically active radiation (PAR) was set at 105 ± 6 μmol photons m−2 s−1 (measured by a US-SQS/WB spherical micro quantum sensor; Walz), with no differences among positions where bottles were placed. The culture media were prepared with autoclaved artificial seawater with nutrients added according to the Aquil medium recipe (Sunda et al., 2005), except for nitrate. The maximum cell concentration was controlled below 6 × 104 cells ml−1 by dilution with target-CO2 equilibrated media every 24 h in order to maintain constant seawater carbonate chemistry without aeration. All the cultures were grown in one growth chamber to prevent any potential influence of factors besides the ones under test from biasing the results. The nitrate concentration of the medium was set as 102.5 µmol l−1 for the high nitrogen (HN) treatment. It was set at 2.5 µmol l−1 for the low nitrogen (LN) treatment, a concentration that is limiting to the diatom’s growth (Li et al., 2012b; Hennon et al., 2014). The culture media were pre-aerated with low (∼400 µatm, LC) or high pCO2 (∼1000 µatm, HC) air for about 24 h to ensure equilibration with the target pCO2 levels. The CO2-enriched air was achieved with a CO2 plant growth chamber (HP1000G-D; Ruihua), in which the target pCO2 levels were obtained by mixing air and pure CO2 gas. Thus, eight treatments were achieved, and each treatment had three independent replicate cultures. All the culture bottles were placed randomly in the growth chamber to avoid artefacts related to inappropriate replication. Nitrate concentrations were determined by a nutrient autoanalyzer (AA3; Bran-Luebbe) in which nitrate was reduced to nitrite by cadmium. The lower detection limit of nitrate was 0.1 µmol l−1. Nitrate concentrations dropped from 117.1 (±0.9) to 108.3 (±1.1) µmol l−1 in the HN cultures and from 2.4 (±0.1) to 0 µmol l−1 (under the limit of detection) in the LN cultures over the 24 h period, respectively. Cultures were run for at least 15 generations before physiological parameters were measured during the photoperiod (5–7 h after the onset of light). Seawater carbonate chemistry The pHNBS was measured by a pH meter (Orion 2 STAR; Thermo Scientific) calibrated using standard National Bureau of Standards (NBS) buffers. Samples for total alkalinity (TA) measurement were poisoned by a saturated solution of mercuric chloride after filtering through cellulose acetate membranes (0.45 µm). TA was determined by Gran acidimetric titration on a 25-ml sample with a TA analyzer (AS-ALK1+; Apollo SciTech). Certified reference materials supplied by A. G. Dickson at the Scripps Institution of Oceanography were used to assure the accuracy of the TA measurement. Carbonate chemistry parameters were calculated based on the TA and pH values using the CO2SYS software, and are shown in Table 1. Table 1. Experimental treatments and carbonate chemistry parameters of culture media. Treatments . Temperature (°C) . Nitrate concentration (µmol l−1) . pCO2 (µatm) . pHNBS . TA (µmol kg−1) . DIC (µmol kg−1) . HCO3− (µmol kg−1) . CO2 (µmol kg−1) . LTHNLC 20 102.5 400 8.11 ± 0.03 2274 ± 20 2024 ± 6 1836 ± 8 15 ± 1 LTHNHC 20 102.5 1000 7.81 ± 0.03* 2272 ± 28 2155 ± 17* 2026 ± 13* 33 ± 2* LTLNLC 20 2.5 400 8.11 ± 0.02 2285 ± 20 2033 ± 8 1843 ± 1 15 ± 1 LTLNHC 20 2.5 1000 7.81 ± 0.02* 2298 ± 14 2181 ± 21* 2050 ± 22* 34 ± 2* HTHNLC 24 102.5 400 8.11 ± 0.03 2269 ± 32 1991 ± 33 1784 ± 35 14 ± 1 HTHNHC 24 102.5 1000 7.79 ± 0.01* 2273 ± 33 2143 ± 32* 2006 ± 30* 32 ± 1* HTLNLC 24 2.5 400 8.10 ± 0.02 2284 ± 15 2008 ± 10 1802 ± 13 14 ± 1 HTLNHC 24 2.5 1000 7.82 ± 0.02* 2308 ± 18 2165 ± 22* 2022 ± 22* 30 ± 2* Treatments . Temperature (°C) . Nitrate concentration (µmol l−1) . pCO2 (µatm) . pHNBS . TA (µmol kg−1) . DIC (µmol kg−1) . HCO3− (µmol kg−1) . CO2 (µmol kg−1) . LTHNLC 20 102.5 400 8.11 ± 0.03 2274 ± 20 2024 ± 6 1836 ± 8 15 ± 1 LTHNHC 20 102.5 1000 7.81 ± 0.03* 2272 ± 28 2155 ± 17* 2026 ± 13* 33 ± 2* LTLNLC 20 2.5 400 8.11 ± 0.02 2285 ± 20 2033 ± 8 1843 ± 1 15 ± 1 LTLNHC 20 2.5 1000 7.81 ± 0.02* 2298 ± 14 2181 ± 21* 2050 ± 22* 34 ± 2* HTHNLC 24 102.5 400 8.11 ± 0.03 2269 ± 32 1991 ± 33 1784 ± 35 14 ± 1 HTHNHC 24 102.5 1000 7.79 ± 0.01* 2273 ± 33 2143 ± 32* 2006 ± 30* 32 ± 1* HTLNLC 24 2.5 400 8.10 ± 0.02 2284 ± 15 2008 ± 10 1802 ± 13 14 ± 1 HTLNHC 24 2.5 1000 7.82 ± 0.02* 2308 ± 18 2165 ± 22* 2022 ± 22* 30 ± 2* Values of carbonate chemistry parameters are means ± SD of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). Open in new tab Table 1. Experimental treatments and carbonate chemistry parameters of culture media. Treatments . Temperature (°C) . Nitrate concentration (µmol l−1) . pCO2 (µatm) . pHNBS . TA (µmol kg−1) . DIC (µmol kg−1) . HCO3− (µmol kg−1) . CO2 (µmol kg−1) . LTHNLC 20 102.5 400 8.11 ± 0.03 2274 ± 20 2024 ± 6 1836 ± 8 15 ± 1 LTHNHC 20 102.5 1000 7.81 ± 0.03* 2272 ± 28 2155 ± 17* 2026 ± 13* 33 ± 2* LTLNLC 20 2.5 400 8.11 ± 0.02 2285 ± 20 2033 ± 8 1843 ± 1 15 ± 1 LTLNHC 20 2.5 1000 7.81 ± 0.02* 2298 ± 14 2181 ± 21* 2050 ± 22* 34 ± 2* HTHNLC 24 102.5 400 8.11 ± 0.03 2269 ± 32 1991 ± 33 1784 ± 35 14 ± 1 HTHNHC 24 102.5 1000 7.79 ± 0.01* 2273 ± 33 2143 ± 32* 2006 ± 30* 32 ± 1* HTLNLC 24 2.5 400 8.10 ± 0.02 2284 ± 15 2008 ± 10 1802 ± 13 14 ± 1 HTLNHC 24 2.5 1000 7.82 ± 0.02* 2308 ± 18 2165 ± 22* 2022 ± 22* 30 ± 2* Treatments . Temperature (°C) . Nitrate concentration (µmol l−1) . pCO2 (µatm) . pHNBS . TA (µmol kg−1) . DIC (µmol kg−1) . HCO3− (µmol kg−1) . CO2 (µmol kg−1) . LTHNLC 20 102.5 400 8.11 ± 0.03 2274 ± 20 2024 ± 6 1836 ± 8 15 ± 1 LTHNHC 20 102.5 1000 7.81 ± 0.03* 2272 ± 28 2155 ± 17* 2026 ± 13* 33 ± 2* LTLNLC 20 2.5 400 8.11 ± 0.02 2285 ± 20 2033 ± 8 1843 ± 1 15 ± 1 LTLNHC 20 2.5 1000 7.81 ± 0.02* 2298 ± 14 2181 ± 21* 2050 ± 22* 34 ± 2* HTHNLC 24 102.5 400 8.11 ± 0.03 2269 ± 32 1991 ± 33 1784 ± 35 14 ± 1 HTHNHC 24 102.5 1000 7.79 ± 0.01* 2273 ± 33 2143 ± 32* 2006 ± 30* 32 ± 1* HTLNLC 24 2.5 400 8.10 ± 0.02 2284 ± 15 2008 ± 10 1802 ± 13 14 ± 1 HTLNHC 24 2.5 1000 7.82 ± 0.02* 2308 ± 18 2165 ± 22* 2022 ± 22* 30 ± 2* Values of carbonate chemistry parameters are means ± SD of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). Open in new tab Table 3. Significance test results for temperature (T), nitrate (N), pCO2 (C), and their interactions based on the GLM. Parameter . T . N . C . T × N . T × C . N × C . T × N × C . . F . p . F . p . F . p . F . p . F . p . F . p . F . p . Specific growth rate 28.8 <0.001 406.3 <0.001 93.0 <0.001 85.5 <0.001 2.9 0.107 78.1 <0.001 10.2 0.006 Cell size 66.1 <0.001 166.2 <0.001 71.3 <0.001 8.0 0.012 4.4 0.052 53.0 <0.001 0.2 0.636 Chlorophyll a 0.3 0.559 432.6 <0.001 50.6 <0.001 82.3 <0.001 3.5 0.079 20.3 <0.001 0.01 0.907 BSi 3.0 0.100 0.5 0.492 0.4 0.504 0.01 0.898 0.4 0.520 7.2 0.016 2.8 0.109 POC 2.8 0.113 11.3 0.004 74.6 <0.001 6.3 0.023 9.7 0.007 77.1 <0.001 1.4 0.252 PON 58.5 <0.001 20.4 <0.001 85.4 <0.001 0.01 0.903 1.7 0.203 34.8 <0.001 5.1 0.037 C: N 70.1 <0.001 181.9 <0.001 0.1 0.688 32.0 <0.001 34.7 <0.001 13.9 0.002 23.7 <0.001 Si: C 5.2 0.035 0.1 0.706 94.8 <0.001 1.1 0.290 17.9 0.001 40.4 <0.001 6.4 0.022 Photosynthesis 7.4 0.015 780.6 <0.001 98.7 <0.001 61.2 <0.001 7.8 0.013 53.7 <0.001 10.6 0.005 Dark respiration 94.3 <0.001 73.1 <0.001 33.7 <0.001 31.8 <0.001 11.8 0.003 33.7 <0.001 16.2 0.001 Parameter . T . N . C . T × N . T × C . N × C . T × N × C . . F . p . F . p . F . p . F . p . F . p . F . p . F . p . Specific growth rate 28.8 <0.001 406.3 <0.001 93.0 <0.001 85.5 <0.001 2.9 0.107 78.1 <0.001 10.2 0.006 Cell size 66.1 <0.001 166.2 <0.001 71.3 <0.001 8.0 0.012 4.4 0.052 53.0 <0.001 0.2 0.636 Chlorophyll a 0.3 0.559 432.6 <0.001 50.6 <0.001 82.3 <0.001 3.5 0.079 20.3 <0.001 0.01 0.907 BSi 3.0 0.100 0.5 0.492 0.4 0.504 0.01 0.898 0.4 0.520 7.2 0.016 2.8 0.109 POC 2.8 0.113 11.3 0.004 74.6 <0.001 6.3 0.023 9.7 0.007 77.1 <0.001 1.4 0.252 PON 58.5 <0.001 20.4 <0.001 85.4 <0.001 0.01 0.903 1.7 0.203 34.8 <0.001 5.1 0.037 C: N 70.1 <0.001 181.9 <0.001 0.1 0.688 32.0 <0.001 34.7 <0.001 13.9 0.002 23.7 <0.001 Si: C 5.2 0.035 0.1 0.706 94.8 <0.001 1.1 0.290 17.9 0.001 40.4 <0.001 6.4 0.022 Photosynthesis 7.4 0.015 780.6 <0.001 98.7 <0.001 61.2 <0.001 7.8 0.013 53.7 <0.001 10.6 0.005 Dark respiration 94.3 <0.001 73.1 <0.001 33.7 <0.001 31.8 <0.001 11.8 0.003 33.7 <0.001 16.2 0.001 Bold and underlined values show where there are significant effects. Open in new tab Table 3. Significance test results for temperature (T), nitrate (N), pCO2 (C), and their interactions based on the GLM. Parameter . T . N . C . T × N . T × C . N × C . T × N × C . . F . p . F . p . F . p . F . p . F . p . F . p . F . p . Specific growth rate 28.8 <0.001 406.3 <0.001 93.0 <0.001 85.5 <0.001 2.9 0.107 78.1 <0.001 10.2 0.006 Cell size 66.1 <0.001 166.2 <0.001 71.3 <0.001 8.0 0.012 4.4 0.052 53.0 <0.001 0.2 0.636 Chlorophyll a 0.3 0.559 432.6 <0.001 50.6 <0.001 82.3 <0.001 3.5 0.079 20.3 <0.001 0.01 0.907 BSi 3.0 0.100 0.5 0.492 0.4 0.504 0.01 0.898 0.4 0.520 7.2 0.016 2.8 0.109 POC 2.8 0.113 11.3 0.004 74.6 <0.001 6.3 0.023 9.7 0.007 77.1 <0.001 1.4 0.252 PON 58.5 <0.001 20.4 <0.001 85.4 <0.001 0.01 0.903 1.7 0.203 34.8 <0.001 5.1 0.037 C: N 70.1 <0.001 181.9 <0.001 0.1 0.688 32.0 <0.001 34.7 <0.001 13.9 0.002 23.7 <0.001 Si: C 5.2 0.035 0.1 0.706 94.8 <0.001 1.1 0.290 17.9 0.001 40.4 <0.001 6.4 0.022 Photosynthesis 7.4 0.015 780.6 <0.001 98.7 <0.001 61.2 <0.001 7.8 0.013 53.7 <0.001 10.6 0.005 Dark respiration 94.3 <0.001 73.1 <0.001 33.7 <0.001 31.8 <0.001 11.8 0.003 33.7 <0.001 16.2 0.001 Parameter . T . N . C . T × N . T × C . N × C . T × N × C . . F . p . F . p . F . p . F . p . F . p . F . p . F . p . Specific growth rate 28.8 <0.001 406.3 <0.001 93.0 <0.001 85.5 <0.001 2.9 0.107 78.1 <0.001 10.2 0.006 Cell size 66.1 <0.001 166.2 <0.001 71.3 <0.001 8.0 0.012 4.4 0.052 53.0 <0.001 0.2 0.636 Chlorophyll a 0.3 0.559 432.6 <0.001 50.6 <0.001 82.3 <0.001 3.5 0.079 20.3 <0.001 0.01 0.907 BSi 3.0 0.100 0.5 0.492 0.4 0.504 0.01 0.898 0.4 0.520 7.2 0.016 2.8 0.109 POC 2.8 0.113 11.3 0.004 74.6 <0.001 6.3 0.023 9.7 0.007 77.1 <0.001 1.4 0.252 PON 58.5 <0.001 20.4 <0.001 85.4 <0.001 0.01 0.903 1.7 0.203 34.8 <0.001 5.1 0.037 C: N 70.1 <0.001 181.9 <0.001 0.1 0.688 32.0 <0.001 34.7 <0.001 13.9 0.002 23.7 <0.001 Si: C 5.2 0.035 0.1 0.706 94.8 <0.001 1.1 0.290 17.9 0.001 40.4 <0.001 6.4 0.022 Photosynthesis 7.4 0.015 780.6 <0.001 98.7 <0.001 61.2 <0.001 7.8 0.013 53.7 <0.001 10.6 0.005 Dark respiration 94.3 <0.001 73.1 <0.001 33.7 <0.001 31.8 <0.001 11.8 0.003 33.7 <0.001 16.2 0.001 Bold and underlined values show where there are significant effects. Open in new tab Specific growth rate, cell size, and chlorophyll a content Cell concentration and mean cell size were determined with 20 ml samples by a Coulter Particle Count and Size Analyzer (Z2; Beckman Coulter). Specific growth rate was calculated according to the equation: µ = (lnN1 − lnN0)/(t1 − t0), where N1 and N0 represent cell concentrations at t1 (before dilution) and t0 (initial or after dilution), (t1 − t0) = 24 h. Samples (150 ml) for chlorophyll measurement were filtered onto GF/F filters (25 mm; Whatman) and extracted in 5 ml of absolute methanol at 4 °C in darkness for 24 h. The supernatants obtained after centrifugation (5000 g for 10 min) were analysed by a UV-VIS Spectrophotometer (DU800; Beckman Coulter). The absorption values at 632, 665, and 750 nm were measured, and chlorophyll a concentrations were determined according to Ritchie (2006). Biogenic silica and particulate organic carbon and nitrogen Samples (120 ml) were gently filtered onto 1.2 µm PC filters (25 mm; Millipore) for determination of biogenic silica (BSi) according to Brzezinski and Nelson (1995). BSi on filters was digested by 4 ml of 0.2 mol l−1 NaOH for 45 min and neutralized with 1 ml of 1 mol l−1 HCl to terminate the extraction. The supernatants (1 ml) were then transferred to clean PE centrifuge plastic tubes (15 ml) and diluted with 4 ml of Milli-Q water. Ammonium molybdate (2 ml) and the reducing agent (3 ml) were added to the tubes and the absorption was measured at 810 nm by a UV-VIS Spectrophotometer (DU800; Beckman Coulter) within 2–3 h of extraction. Samples (100 ml) for measuring particulate organic carbon (POC) and nitrogen (PON) were harvested by gentle vacuum filtration (< 0.02 MPa) on pre-combusted (450 °C for 6 h) GF/F filters (25 mm; Whatman). HCl fumes were applied to remove any inorganic carbon on filters (12 h) before they were dried at 60 °C and analysed using a CHNS/O Analyzer (2400 SeriesII; PerkinElmer). Acetanilide of a known ratio of carbon to nitrogen and weight (range from 0.2 to 1 mg) was run every 6 samples to monitor instrument drift. Chlorophyll a fluorescence Samples were dark-adapted for 10 min before the maximum quantum yields of PSII (ΦPSII max) were measured with a multi-colour pulse amplitude modulated fluorometer (MULTI-COLOR-PAM; Walz). Preliminary experiments showed there was no difference between 10 and 15 min dark-adaptation times. Following dark-adaptation an actinic light of 94 μmol photons m−2 s−1, being similar to the growth light level, was applied for 3 min to determine the effective quantum yield of PSII (ΦPSII eff). The measuring beam and actinic light sources were blue and white light, respectively. The intensity of the saturating pulse was set at 4819 μmol photons m−2 s−1 for 0.8 s. ΦPSII max and ΦPSII eff were calculated according to the following equations: ΦPSII max  = (Fm − F0)/Fm for dark-adapted (10 min) samples; ΦPSII eff = (Fm′ − Ft)/Fm′ = ΔF/Fm′ for actinic light-adapted samples, where Fm and Fm′ indicate maximum chlorophyll fluorescence of dark- and light-adapted samples, respectively, F0 is the minimum chlorophyll fluorescence of dark-treated cells, and Ft is the steady state chlorophyll fluorescence of the light-exposed samples. Rapid light curves (RLCs) were measured to estimate relative maximum electron transport rate (rETRmax), apparent photon transfer efficiency (α), and light saturation point (Ik). Samples were incubated under growth conditions (105 μmol photons m−2 s−1; 20 and 24 °C for LT and HT treatments, respectively) for 10 min before RLC measurements. Eleven progressively increasing light intensities (60, 94, 128, 215, 368, 644, 798, 1149, 1599, 2120, and 2863 μmol photons m−2 s−1), each applied for 20 s, were used in the RLC measurements. Values of rETR at these light intensities were calculated according to: rETR = PAR × ΦPSII eff × 0.5, where PAR represents the photon flux density of actinic light (μmol photons m−2 s−1), ΦPSII eff is the effective photochemical quantum yield at each PAR level, and the factor 0.5 assumes that PSII receives half of all absorbed quanta. PAR and rETR data obtained from RLCs were fitted to the model: rETR = PAR/(a × PAR2 + b × PAR + c). The Ik, rETRmax and α were calculated from a, b and c according to Eilers and Peeters (1988). Photosynthesis and dark respiration Samples (20 ml) of cultures with a cell concentration range of 2–6 × 104 cells  ml−1 were inoculated with 5 µCi NaH14CO3 (ICN Radiochemicals) for determination of photosynthetic carbon fixation rates. They were incubated under growth conditions (105 μmol photons m−2 s−1, 20 or 24 °C) for 60 min. This relatively short-term incubation is usually taken to give an estimation of gross photosynthetic carbon fixation rate (Williams et al., 2002), though some other studies have indicated that N status might influence the balance between gross and net carbon fixation (Halsey et al., 2011, 2013). After the incubation, cells were gently filtered onto GF/F filters (25 mm; Whatman) and the filters were placed into scintillation vials. Filters were exposed to HCl fumes overnight, and dried at 50 °C for 6 h (Gao et al., 2007). Scintillation cocktail (5 ml) was added to the vials before assimilated radiocarbon was counted by a liquid scintillation counter (Tri-Carb 2800TR; PerkinElmer). Dark respiration was measured by a Clark-type oxygen electrode (Oxygraph; Hansatech). About 1 × 106 cells were harvested by gentle vacuum filtration (<0.02 MPa) onto cellulose acetate membranes (1 µm) and resuspended into 5 ml of 20 mmol l−1 Tris-buffered medium. Then they were injected into the oxygen electrode chamber at 20 or 24 °C controlled by a refrigerated circulating bath (GDH-0506; Shunma). ΦPSII max was measured to ensure physiological stability after the influence of filtration; no differences were found in ΦPSII max before and after the filtration. The pH values (8.10 and 7.80 for LC and HC conditions, respectively) of the Tris-buffered media were adjusted to the corresponding culture values by 1 mol l−1 HCl and NaOH. Dark respiration rates were derived from the linear portion of the slope of oxygen consumption vs. time (within about 10 min per measurement). Statistical analyses All data in the present study are reported as the mean ± SD (n = 3). Data were analysed by the general linear model (GLM) in SPSS statistics 18.0, with temperature, nitrate concentration, and pCO2 level as three factors in the model. Two levels were set in each factor: 20 and 24 °C for temperature; 2.5 and 102.5 µmol l−1for nitrate concentration; 400 and 1000 µatm for pCO2 level. Interactions among three factors were included in the model. The independent-samples t-test was applied to determine differences between two levels of a factor when p < 0.05 was found in the model. Results Specific growth rate and mean cell size While elevated pCO2 showed no significant impact on growth in HN-grown cells, it significantly decreased the specific growth rate of LN-grown cultures (31%, t-test, t = 4.1, df = 4, p = 0.014 for low temperature; 68%, t = 24.7, df = 4, p < 0.001 for high temperature; Figure 1a). Specific growth rates were 11–79% lower under N-limited conditions than replete conditions (t-test, t = 4.9, df = 4, p = 0.008 for low temperature plus ambient pCO2 treatment; t = 5.1, df = 4, p = 0.007 for low temperature plus elevated pCO2 treatment; t = 10.1, df = 4, p = 0.001 for high temperature plus ambient pCO2 treatment; t = 32.6, df = 4, p < 0.001 for high temperature plus elevated pCO2 treatment). Elevated temperature decreased cell growth under the reduced nitrate condition by 22 and 65% at the ambient and elevated pCO2 levels, respectively (t-test, t = 11.5, df = 4, p < 0.001 for ambient pCO2 level; t = 5.9, df = 4, p = 0.004 for elevated pCO2 level). A significant interaction was found between nitrate and pCO2 levels effects on specific growth rate (GLM, F1,16 = 78.1, p < 0.001; Table 3). Table 2. ΦPSII max and ΦPSII eff, α, rETRmax, and Ik (μmol photons m−2 s−1) of Thalassiosira pseudonana cells under different treatments. . . ΦPSII max . ΦPSII eff . α . rETRmax . Ik . LT (20 °C) HNLC 0.671 ± 0.008 0.582 ± 0.008 0.268 ± 0.005 119 ± 1 444 ± 13 HNHC 0.668 ± 0.003 0.575 ± 0.004 0.266 ± 0.002 117 ± 4 439 ± 15 LNLC 0.635 ± 0.016 0.549 ± 0.018 0.273 ± 0.010 97 ± 5 355 ± 32 LNHC 0.538 ± 0.036* 0.488 ± 0.028* 0.239 ± 0.021 77 ± 5* 322 ± 8 HT (24 °C) HNLC 0.678 ± 0.001 0.565 ± 0.005 0.277 ± 0.001 127 ± 1 458 ± 1 HNHC 0.672 ± 0.003* 0.557 ± 0.006 0.273 ± 0.002* 117 ± 4* 430 ± 15 LNLC 0.673 ± 0.004 0.593 ± 0.030 0.292 ± 0.002 109 ± 6 376 ± 21 LNHC 0.505 ± 0.038* 0.472 ± 0.012* 0.221 ± 0.008* 75 ± 9* 341 ± 53 . . ΦPSII max . ΦPSII eff . α . rETRmax . Ik . LT (20 °C) HNLC 0.671 ± 0.008 0.582 ± 0.008 0.268 ± 0.005 119 ± 1 444 ± 13 HNHC 0.668 ± 0.003 0.575 ± 0.004 0.266 ± 0.002 117 ± 4 439 ± 15 LNLC 0.635 ± 0.016 0.549 ± 0.018 0.273 ± 0.010 97 ± 5 355 ± 32 LNHC 0.538 ± 0.036* 0.488 ± 0.028* 0.239 ± 0.021 77 ± 5* 322 ± 8 HT (24 °C) HNLC 0.678 ± 0.001 0.565 ± 0.005 0.277 ± 0.001 127 ± 1 458 ± 1 HNHC 0.672 ± 0.003* 0.557 ± 0.006 0.273 ± 0.002* 117 ± 4* 430 ± 15 LNLC 0.673 ± 0.004 0.593 ± 0.030 0.292 ± 0.002 109 ± 6 376 ± 21 LNHC 0.505 ± 0.038* 0.472 ± 0.012* 0.221 ± 0.008* 75 ± 9* 341 ± 53 The data are means ± SD of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). HNLC, high nitrate low pCO2 treatment; HNHC, high nitrate high pCO2 treatment; LNLC, low nitrate low pCO2 treatment; LNHC, low nitrate high pCO2 treatment. Open in new tab Table 2. ΦPSII max and ΦPSII eff, α, rETRmax, and Ik (μmol photons m−2 s−1) of Thalassiosira pseudonana cells under different treatments. . . ΦPSII max . ΦPSII eff . α . rETRmax . Ik . LT (20 °C) HNLC 0.671 ± 0.008 0.582 ± 0.008 0.268 ± 0.005 119 ± 1 444 ± 13 HNHC 0.668 ± 0.003 0.575 ± 0.004 0.266 ± 0.002 117 ± 4 439 ± 15 LNLC 0.635 ± 0.016 0.549 ± 0.018 0.273 ± 0.010 97 ± 5 355 ± 32 LNHC 0.538 ± 0.036* 0.488 ± 0.028* 0.239 ± 0.021 77 ± 5* 322 ± 8 HT (24 °C) HNLC 0.678 ± 0.001 0.565 ± 0.005 0.277 ± 0.001 127 ± 1 458 ± 1 HNHC 0.672 ± 0.003* 0.557 ± 0.006 0.273 ± 0.002* 117 ± 4* 430 ± 15 LNLC 0.673 ± 0.004 0.593 ± 0.030 0.292 ± 0.002 109 ± 6 376 ± 21 LNHC 0.505 ± 0.038* 0.472 ± 0.012* 0.221 ± 0.008* 75 ± 9* 341 ± 53 . . ΦPSII max . ΦPSII eff . α . rETRmax . Ik . LT (20 °C) HNLC 0.671 ± 0.008 0.582 ± 0.008 0.268 ± 0.005 119 ± 1 444 ± 13 HNHC 0.668 ± 0.003 0.575 ± 0.004 0.266 ± 0.002 117 ± 4 439 ± 15 LNLC 0.635 ± 0.016 0.549 ± 0.018 0.273 ± 0.010 97 ± 5 355 ± 32 LNHC 0.538 ± 0.036* 0.488 ± 0.028* 0.239 ± 0.021 77 ± 5* 322 ± 8 HT (24 °C) HNLC 0.678 ± 0.001 0.565 ± 0.005 0.277 ± 0.001 127 ± 1 458 ± 1 HNHC 0.672 ± 0.003* 0.557 ± 0.006 0.273 ± 0.002* 117 ± 4* 430 ± 15 LNLC 0.673 ± 0.004 0.593 ± 0.030 0.292 ± 0.002 109 ± 6 376 ± 21 LNHC 0.505 ± 0.038* 0.472 ± 0.012* 0.221 ± 0.008* 75 ± 9* 341 ± 53 The data are means ± SD of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). HNLC, high nitrate low pCO2 treatment; HNHC, high nitrate high pCO2 treatment; LNLC, low nitrate low pCO2 treatment; LNHC, low nitrate high pCO2 treatment. Open in new tab Figure 1. Open in new tabDownload slide Specific growth rates (d−1) (a) and cell size (µm) (b) of Thalassiosira pseudonana acclimated to ambient (LC, filled bars) and elevated pCO2 (HC, open bars) at different temperature and nitrate levels. The data are mean ± SD values of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). Elevated pCO2 decreased the cell size under N-limited conditions (6%, t-test, t = 8.1, df = 4, p = 0.001 for low temperature; 5%, t = 4.8, df = 4, p = 0.008 for high temperature; Figure 1b), but had no significant effect on the HN-grown cells. Nitrate limitation decreased cell size by 3–7% compared to the corresponding HN treatments, with the exception of the low temperature plus ambient pCO2 treatment. Cells grown at the higher temperature had 1–5% smaller mean cell sizes than in the corresponding LT treatments (t-test, t = 4.3, df = 4, p = 0.012 for N-replete plus ambient pCO2 treatment; t = 3.5, df = 4, p = 0.024 for N-replete plus elevated pCO2 treatment; t = 5.0, df = 4, p = 0.007 for N-limited plus ambient pCO2 treatment; t = 3.6, df = 4, p = 0.021 for N-limited plus elevated pCO2 treatment). A significant interaction between nitrate and pCO2 levels on cell size was detected (GLM, F1,16 = 53.0, p < 0.001). Chlorophyll a, BSi, POC, and PON Differential responses of chlorophyll a content to elevated pCO2 were detected under N-limited compared to N-replete conditions: elevated pCO2 showed no significant effect under HN conditions but resulted in decreased chlorophyll a under LN conditions (56%, t-test, t = 10.3, df = 4, p < 0.001 for low temperature; 70%, t = 11.2, df = 4, p < 0.001 for high temperature; Figure 2a). LN treatments resulted in 22–89% lower chlorophyll a content compared to HN treatments (t-test, t = 4.3, df = 4, p = 0.012 for low temperature plus ambient pCO2 treatment; t = 4.9, df = 4, p = 0.008 for low temperature plus elevated pCO2 treatment; t = 24.4, df = 4, p < 0.001 for high temperature plus ambient pCO2 treatment; t = 28.9, df = 4, p < 0.001 for high temperature plus elevated pCO2 treatment), and elevated temperature increased the difference between LN and HN treatments. While elevated temperature decreased chlorophyll a content under LN conditions, it enhanced the content under HN conditions (t-test, t =−4.1, df = 4, p = 0.015 for N-replete plus ambient pCO2 treatment; t =−3.1, df = 4, p = 0.036 for N-replete plus elevated pCO2 treatment; t = 11.3, df = 4, p < 0.001 for N-limited plus ambient pCO2 treatment; t = 6.5, df = 4, p = 0.003 for N-limited plus elevated pCO2 treatment). A significant interaction between nitrate and pCO2 levels on chlorophyll a content was detected (GLM, F1,16 = 20.3, p < 0.001). No significant effects of the three factors on BSi content were found (Table 3), and cell quotas ranged from 0.09 to 0.13 pmol cell−1 (Figure 2b). Figure 2. Open in new tabDownload slide Cellular quotas of chlorophyll a (pg cell−1) (a), BSi (pmol cell−1) (b), POC (pmol cell−1) (c), and PON (pmol cell−1) (d) of Thalassiosira pseudonana acclimated to ambient (LC, filled bars) and elevated pCO2 (HC, open bars) at different temperature and nitrate levels. The data are mean ± SD values of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). Differential effects of elevated pCO2 on POC content under LN and HN conditions were detected: pCO2 had no significant impact under HN conditions, but increased POC content by 78–167% under LN conditions (t-test, t =−4.2, df = 4, p = 0.013 for low temperature; t =−29.2, df = 4, p < 0.001 for high temperature; Figure 2c). POC contents of LN cells were 24–28% lower than in the corresponding HN treatments at ambient pCO2 (t-test, t = 3.3, df = 4, p = 0.029 for low temperature; t = 4.5, df = 4, p = 0.010 for high temperature), but POC was enhanced by nitrate limitation at elevated pCO2. Temperature showed no significant effects on cellular POC content (GLM, F1,16 = 2.8, p = 0.113). Interactions between temperature or nitrate and pCO2 levels on POC content were detected (Table 3). Elevated pCO2 significantly enhanced PON contents by 23–131% (t-test, t =−3.4, df = 4, p = 0.026 for low temperature plus N-limited treatment; t =−5.3, df = 4, p = 0.006 for high temperature plus N-replete treatment; t =−20.0, df = 4, p < 0.001 for high temperature plus N-limited treatment; Figure 2d), with the exception of the low temperature plus N-replete treatment. LN-grown cells showed 41–48% lower PON content relative to HN-grown cells under ambient pCO2 conditions (t-test, t = 6.3, df = 4, p = 0.003 for low temperature; t = 12.1, df = 4, p < 0.001 for high temperature), but there was no difference between LN and HN cells when pCO2 was elevated. Elevated temperature decreased cellular PON content (t-test, t = 6.1, df = 4, p = 0.004 for N-replete plus ambient pCO2 treatment; t = 5.3, df = 4, p = 0.006 for N-replete plus elevated pCO2 treatment; t = 3.5, df = 4, p = 0.023 for N-limited plus elevated pCO2 treatment), with the exception of the N-limited plus ambient pCO2 treatment. A significant interaction between nitrate and pCO2 levels on PON content was detected (GLM, F1,16 = 34.8, p < 0.001). While no effects of pCO2 on C:N were detected under HN conditions, under LN conditions the ratio was decreased or raised by elevated pCO2 at low or high temperature, respectively (Figure 3a). LN-grown cells showed higher C:N relative to HN-grown cells (t-test, t =−7.1, df = 4, p = 0.002 for low temperature plus ambient pCO2 treatment; t = −5.6, df = 4, p = 0.005 for high temperature plus ambient pCO2 treatment; t =−13.8, df = 4, p < 0.001 for high temperature plus elevated pCO2 treatment), with the exception of the low temperature plus elevated pCO2 treatment. The higher temperature treatment resulted in 76% higher C:N for cells in the N-limited plus elevated pCO2 treatment (t-test, t =−9.3, df = 4, p = 0.001) but had no significant impact on the cells under other treatments. Interactions of the three factors on C:N were detected (Table 3). Figure 3. Open in new tabDownload slide Elemental ratios of POC and PON C:N (pmol:pmol) (a) and ratios of BSi and POC Si:C (pmol:pmol) (b) of Thalassiosira pseudonana acclimated to ambient (LC, filled bars) and elevated pCO2 (HC, open bars) at different temperature and nitrate levels. The data are mean ± SD values of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). Elevated pCO2 decreased the ratio of cellular BSi to POC (Si:C) by 18–67% (t-test, t = 3.9, df = 4, p = 0.018 for low temperature plus N-limited treatment; t = 3.9, df = 4, p = 0.017 for high temperature plus N-replete treatment; t = 11.6, df = 4, p < 0.001 for high temperature plus N-limited treatment; Figure 3b), with the exception of the low temperature plus N-replete treatment. Nitrate limitation increased the Si:C ratio of LC cells and reduced that of HC cells at the elevated temperature (t-test, t =−3.9, df = 4, p = 0.017 for LC cells; t = 8.8, df = 4, p = 0.001 for HC cells). The ratio was raised or decreased by elevated temperature at ambient or elevated pCO2 level under N-limited condition, respectively (t-test, t =−3.2, df = 4, p = 0.033 for ambient pCO2 level; t = 3.0, df = 4, p = 0.038 for elevated pCO2 level). Interactions between temperature or nitrate and pCO2 levels on Si:C were detected (Table 3). Chlorophyll a fluorescence Nitrate limitation decreased maximal (ΦPSII max) and effective (ΦPSII eff) photochemical efficiency of PSII (t-test, t = 3.3, df = 4, p = 0.028 for ΦPSII max in the low temperature plus ambient pCO2 treatment; t = 6.2, df = 4, p = 0.003 for ΦPSII max in the low temperature plus elevated pCO2 treatment; t = 7.5, df = 4, p = 0.002 for ΦPSII max in the high temperature plus ambient pCO2 treatment; t-test, t = 2.9, df = 4, p = 0.041 for ΦPSII eff in the low temperature plus ambient pCO2 treatment; t = 5.4, df = 4, p = 0.006 for ΦPSII eff in the low temperature plus elevated pCO2 treatment; t = 11.1, df = 4, p < 0.001 for ΦPSII eff in the high temperature plus ambient pCO2 treatment; Table 2), with the exception of the high temperature plus ambient pCO2 treatment. Under N-limited conditions, cells grown at elevated pCO2 always showed lower ΦPSII max and ΦPSII eff relative to LC cells (t-test, t = 4.2, df = 4, p = 0.013 for ΦPSII max at low temperature; t = 7.5, df = 4, p = 0.002 for ΦPSII max at high temperature; t-test, t = 3.2, df = 4, p = 0.033 for ΦPSII eff at low temperature; t = 6.5, df = 4, p = 0.003 for ΦPSII eff at high temperature). While elevated pCO2 did not show effects on the RLCs under HN conditions, it significantly depressed the rETR when nitrate was limiting, especially at high light intensities (Figure 4). Significantly negative effects of elevated pCO2 on rETRmax were detected under N-limited conditions (t-test, t = 4.5, df = 4, p = 0.011 for low temperature; t = 5.2, df = 4, p = 0.006 for high temperature). Nitrate limitation significantly decreased rETRmax by 14–36% (t-test, t = 6.9, df = 4, p = 0.002 for low temperature plus ambient pCO2 treatment; t = 10.7, df = 4, p < 0.001 for low temperature plus elevated pCO2 treatment; t = 4.8, df = 4, p = 0.009 for high temperature plus ambient pCO2 treatment; t = 7.0, df = 4, p = 0.002 for high temperature plus elevated pCO2 treatment), with a greater decrease at elevated pCO2 (Table 2). Elevated temperature increased rETRmax by 7% in the N-replete plus ambient pCO2 treatment (t-test, t =−10.7, df = 4, p < 0.001), but did not significantly impact the value of this parameter in other treatments. Elevated pCO2 decreased the apparent light harvesting efficiency (α) at elevated temperature, with a greater decrease under N-limitation (24%, t-test, t = 14.5, df = 4, p < 0.001). Elevated temperature significantly enhanced α (t-test, t =−3.0, df = 4, p = 0.039 for N-replete plus ambient pCO2 treatment; t =−4.0, df = 4, p = 0.016 for N-replete plus elevated pCO2 treatment; t =−3.4, df = 4, p = 0.028 for N-limited plus ambient pCO2 treatment), with the exception of the N-limited plus elevated pCO2 treatment. Elevated pCO2 or temperature showed no effects on the light saturation point (Ik). Ik was 18–27% lower under N-limited conditions compared to HN treatments (t-test, t = 4.4, df = 4, p = 0.011 for low temperature plus ambient pCO2 treatment; t = 11.8, df = 4, p < 0.001 for low temperature plus elevated pCO2 treatment; t = 6.9, df = 4, p = 0.002 for high temperature plus ambient pCO2 treatment; t = 2.8, df = 4, p = 0.049 for high temperature plus elevated pCO2 treatment). Figure 4. Open in new tabDownload slide RLCs determined by variations of rETR under a series of light intensities in Thalassiosira pseudonana cells at HN (a) and LN (b) levels. The data are means ± SD of triplicate cultures (n = 3). LTLC, low temperature low pCO2 treatment; LTHC, low temperature high pCO2 treatment; HTLC, high temperature low pCO2 treatment; HTHC, high temperature high pCO2 treatment. Photosynthetic C fixation and dark respiration The rate of photosynthetic carbon assimilation was decreased by elevated pCO2 in LN-grown cells (71%, t-test, t = 29.2, df = 4, p < 0.001 for low temperature; 69%, t = 5.9, df = 4, p = 0.004 for high temperature; Figure 5a), but it was unaltered in HN-grown cells. The C fixation rates were 22–89% lower under N-limited conditions compared to N-replete treatments, especially at elevated pCO2 (t-test, t = 10.4, df = 4, p < 0.001 for low temperature plus ambient pCO2 treatment; t = 19.2, df = 4, p < 0.001 for low temperature plus elevated pCO2 treatment; t = 16.3, df = 4, p < 0.001 for high temperature plus ambient pCO2 treatment; t = 13.4, df = 4, p < 0.001 for high temperature plus elevated pCO2 treatment). An effect of elevated temperature on the carbon fixation rate was only detected in cells at ambient pCO2: it reduced the rate when nitrate was limiting (t-test, t = 16.1, df = 4, p < 0.001), but enhanced the rate under the HN condition (t-test, t =−2.9, df = 4, p = 0.042). Interactions between the three factors were detected (Table 3). Figure 5. Open in new tabDownload slide Photosynthetic carbon fixation rates (pg C cell−1 h−1) (a) and dark respiration rates (fmol O2 cell−1 h−1) (b) of Thalassiosira pseudonana acclimated to ambient (LC, filled bars) and elevated pCO2 (HC, open bars) at different temperature and nitrate levels. The data are mean ± SD values of triplicate cultures (n = 3). Asterisks indicate significant (p < 0.05) differences between pCO2 treatments (t-test). While no effects of elevated pCO2 on respiration were detected in HN treatments, it significantly enhanced the rate under N-limited conditions (Figure 5b). Nitrate limitation increased respiration by 104–303% relative to corresponding HN treatments at elevated pCO2 level (t-test, t =−9.6, df = 4, p < 0.001 for low temperature; t =−13.1, df = 4, p < 0.001 for high temperature). Cells showed enhanced dark respiration rates at higher temperature (GLM, F1,16 = 94.3, p < 0.001), especially under the N-limited plus elevated pCO2 treatment where respiration was increased by 249% (t-test, t =−17.4, df = 4, p < 0.001). Interactions between the three factors were detected (Table 3). Discussion In the present work, T. pseudonana showed decreased specific growth rates, cell size, pigment content, photochemical quantum yield of PSII, and photosynthetic carbon fixation under multiple drivers (elevated levels of pCO2 and temperature and reduced availability of nitrate). Nitrate limitation appeared to act synergistically with elevated pCO2 and temperature to impact growth and photosynthesis of the diatom. The results imply that a reduction in upward transport of nutrients due to enhanced stratification as a consequence of sea surface warming might reduce growth and carbon fixation by diatoms as ocean acidification progresses. Effects of elevated temperature and its interaction with nitrate limitation Elevated temperature, in the present work, decreased the cell size of T. pseudonana, which has been suggested to be a general trend in diatoms (Montagnes and Franklin, 2001). Furthermore, higher temperature decreased cellular PON content in T. pseudonana, which may be partly correlated with smaller cell size, as indicated by positive relationship between cell size and PON content. The decreased content could also be caused by lowered activity (Gao et al., 2000) and reduced gene expression (Parker and Armbrust, 2005) of nitrate reductase (NR) in diatoms with warming treatments. Down-regulation of nitrogen metabolism by elevated temperature is also supported by observations of declining nitrate uptake rate with increasing temperature in a diatom-dominated phytoplankton community (Lomas and Glibert, 1999). Therefore, negatively affected nitrogen metabolism and enhanced respiration with rising temperature could lead to declining PON and shrinking cell size. This implies a possible slower sinking rate and lower export efficiency of diatoms with ocean warming. As the major component of photosynthetic architecture, nitrogen could impact the effects of elevated temperature on photosynthesis and growth due to the reduced contents of pigments (Li et al., 2012b) and PSII centres (Berges et al., 1996) and enhanced respiration (Li et al., 2012b) under N-limited conditions. While positive or neutral effects of elevated temperature on specific growth rate, chlorophyll a content and photosynthetic carbon fixation rate of T. pseudonana cells were detected under N-replete conditions, the warming treatment significantly decreased these parameters under N-limited conditions in the present study. Consistently, the positive effects of elevated temperature on net primary production and phytoplankton biomass were overcompensated by the negative effects of lower nutrient supply due to enhanced stratification associated with ocean warming (Behrenfeld et al., 2006; Lewandowska et al., 2014). Thus, warming could show differential effects on primary production of waters with distinct nutrient conditions. Effects of elevated CO2 under nitrate limitation Higher respiration observed in T. pseudonana cells in the N-limited plus elevated CO2 treatment might be attributed to enhanced glycolytic and tricarboxylic acid cycle pathways under nitrate limitation (Mock et al., 2008; Hockin et al., 2012) and elevated CO2 conditions (Jin et al., 2015). Enhanced mitochondrial respiration could theoretically lead to increased growth (Geider and Osborne, 1989), as it provides ATP and carbon skeletons for growth (Raven and Beardall, 2005). However, in T. pseudonana, the specific growth rate decreased in the nitrogen-limited cells but increased in cells grown under N-replete levels with increased mitochondrial respiration (Figure 6). In other words, negative and positive correlations of specific growth rate and respiration were evident under N-limited and replete conditions, respectively. The photosynthetic light reactions and mitochondrial respiration are two main processes generating ATP in photosynthetic organisms. When the light reactions were repressed under elevated CO2 and N-limited conditions, mitochondrial respiration could be enhanced to fulfil the cellular energy demand. Although the carbon utilized by mitochondrial respiration must have been initially fixed by photosynthesis, the proportion of fixed carbon for respiration might vary under different conditions. The higher respiration under elevated CO2 plus N-limited conditions did not, however, result in higher specific growth rate in T. pseudonana, which might indicate that the generated energy was allocated more to maintain intracellular homeostasis, rather than growth and biosynthesis. Enhanced (Wu et al., 2010; Yang and Gao, 2012) or unaltered (Trimborn et al., 2014) mitochondrial respiration under seawater acidification conditions have been reported previously in diatoms when nitrogen was replete. In the present work, mitochondrial respiration was substantially enhanced by acidification under N-limited conditions. However, a decreased oxygen uptake rate, determined by the 18 O method (Hennon et al., 2014), and reduced expression of the corresponding respiratory gene clusters (Hennon et al., 2015) were reported in T. pseudonana grown under seawater acidification (pHT = 7.71) conditions for about 15 generations in a N-limited chemostat. The lack of conformity of effects of acidification on respiration might be caused by differing levels of supplied light energy or light regimes. Continuous light exposure without a dark period, as used in Hennon et al. (2014), could have differential effects on mitochondrial respiration relative to a light–dark cycle regime, as shown in Skeletonema costatum (Gilstad et al., 1993). Moreover, the inclusion of photorespiration and the Mehler reaction in the oxygen uptake determination in Hennon et al. (2014) might also affect the observed changes of mitochondrial respiration to elevated CO2. Figure 6. Open in new tabDownload slide Relationships between dark respiration (fmol O2 cell−1 h−1) and specific growth rates (d−1) of Thalassiosira pseudonana cells in eight treatments (a) and the relationships at HN (b) and LN (c) levels. The data are means ± SD of triplicate cultures (n = 3). LTLC, low temperature low pCO2 treatment; LTHC, low temperature high pCO2 treatment; HTLC, high temperature low pCO2 treatment; HTHC, high temperature high pCO2 treatment. It is worth noting that the decreased chlorophyll content was not the only reason behind the depressed photosynthetic rate per cell at elevated pCO2 when nitrate was limiting. For instance, under low temperature plus N-limited condition, chlorophyll normalized photosynthesis rates were 0.26 ± 0.02 and 0.11 ± 0.02 µg C (µg chl a)−1 h−1 for cells at ambient and elevated pCO2, respectively. While differing effects of nitrogen limitation on the affinity of cells for inorganic carbon and CCMs were observed among species (Raven and Beardall, 2014), nitrogen limitation (Alipanah et al., 2015), and elevated CO2 (Nakajima et al., 2013; Hennon et al., 2015) could suppress the expression of genes encoding carbonic anhydrases and inorganic carbon transporters, which are essential for uptake and transport of bicarbonate, the predominant dissolved inorganic carbon (DIC) species in seawater, in T. pseudonana (Tsuji et al., 2017). Although elevated CO2 would partly compensate for the decreased bicarbonate uptake, these changes would significantly suppress the DIC uptake and assimilation, as shown by the lowest photosynthetic carbon fixation when nitrate limitation and seawater acidification coexisted (Figure 5a). A decreased number and percentage of active PSII centres of phytoplankton cells were observed under N-limited condition (Berges et al., 1996). Additionally, elevated CO2 was shown to increase the cost of maintaining functional PSII (McCarthy et al., 2012). However, the demand of nitrogen for repairing inactive PSII could not be fulfilled under N-limited conditions. Thus, lower effective quantum yield of PSII and apparent light harvesting efficiency were observed, which could also contribute to the lowest carbon fixation rate being found under N-limited and elevated CO2 conditions. Elevated CO2 and associated seawater chemistry changes usually enhance or do not significantly affect specific growth rate and photosynthesis of T. pseudonana (strain CCMP 1335) under nitrogen-replete conditions (Figure 7). However, in the present work, under nitrate limitation, the acidification treatment markedly impacted its physiological performance. As the two main electron sinks (Giordano and Raven, 2014), carbon and nitrogen assimilation processes compete for energy and reductant in photosynthetic organisms (Huppe and Turpin, 1994). Diatoms have evolved vacuoles to store nutrients (Falkowski et al., 2004), which enables cells to optimize carbon and nitrogen assimilation by reallocating energy and reductant when light is optimal and other resources are abundant. These characteristics could allow cells to maintain constant or higher growth and photosynthesis under seawater acidification. When nitrogen is limited, energy and reductant pools markedly decreased as photosynthetic capacity decreased. However, expressions of genes encoding nitrogen transport and metabolism (NR and NADPH-dependent nitrite reductase) were up-regulated under N-limited conditions, indicating the up-regulation of nitrogen metabolism (Alipanah et al., 2015). Elevated CO2 and associated seawater acidification, although saving some of the energy used for CCMs (Hopkinson et al., 2011), add an additional energy burden to cells to maintain intracellular pH homeostasis through adjusting cellular periplasmic redox activity and/or proton pumping (Taylor et al., 2012). Thus, elevated CO2 under N-limited conditions could impact the physiological performance of photosynthetic cells as clearly shown in the present study. Moreover, addition of a warming treatment exacerbated the negative effects of acidification on specific growth rate and chlorophyll content under the N-limited conditions (Figure 1a and 2a). The mechanism underpinning these changes might be that elevated temperature could increase the nitrogen to phosphate ratio of T. pseudonana (Toseland et al., 2013), which would increase cellular demand for nitrogen and intensify the nitrogen limitation on cells under the low N conditions. Figure 7. Open in new tabDownload slide Impact of seawater acidification on physiological parameters of Thalassiosira pseudonana (strain CCMP 1335) cells. Data are the fold difference of means of these parameters at elevated pCO2 level (HC) compared to means at ambient pCO2 level (LC), i.e. the ratios of HC to LC. Filled and open circles show the ratios under replete and limited nitrate conditions respectively. Interactions of elevated CO2 with other abiotic factors The effects of elevated CO2 on phytoplankton might be closely related to changes of other factors, as the organismal energy budget can be altered under a range of stress conditions (Wingler et al., 2000). Under optimal conditions, the effects of CO2 might be overshadowed by other factors, such as light, nutrient supply, and temperature (Boyd et al., 2010). However, the effects of elevated CO2 and associated seawater acidification tend to be more conspicuous when other factors are limiting or stressful. For instance, negative effects of acidification on growth of diatoms were detected under high levels of solar radiation (Gao et al., 2012b), low light, and low temperature (Passow and Laws, 2015) and in the presence of solar UV radiation (Li et al., 2017a). Recently, a depressed maximum quantum yield of PSII in T. pseudonana under acidification was found to occur only when nitrogen availability was reduced (in stationary phase) (Hong et al., 2017). Nevertheless, elevated CO2 and associated seawater acidification might also show more prominent effects on diatoms under nutrient-replete conditions relative to limited conditions, as documented in studies on T. weissflogii (Sugie and Yoshimura, 2016) and P. tricornutum (Li et al., 2012b). An increasing number of environmental factors could influence the effects of elevated CO2 on microalgae (Brennan and Collins, 2015), which is also shown in this study on the diatom T. pseudonana. Effects of the three factors on elemental ratios Changes in diatom elemental stoichiometry and macromolecular composition can impact predation by, and reproduction of, zooplankton (Elser et al., 2000; Wichard et al., 2007) and export of particulate organic matter to deep waters. In T. pseudonana, C:N was differentially impacted by elevated CO2 at low and high temperatures under nitrogen limitation (Figure 3), which resulted from variations in cellular POC and PON under elevated CO2 conditions at different temperatures. The higher C:N indicates a lower nutritional quality of phytoplankton as prey for higher trophic levels (Elser et al., 2000). The decreased Si:C at elevated CO2 under nitrogen limitation (Figure 3) was mainly due to increased POC. Decreased Si:C with seawater acidification has also been shown in other diatom species (Tatters et al., 2012; Mejia et al., 2013; Li et al., 2016). Interactions of acidification, N-limitation, and warming appear to give rise to reduced ratios of silicate per carbon. BSi contents were constant among treatments; thus, the ballasting effects by diatom frustules might not change. Nevertheless, the decreased Si:C would modify the primary production contributed by diatom communities and local carbon and silicon budgets in Si-limited waters (Mejia et al., 2013). Conclusions The present study emphasizes the importance of investigating effects of elevated CO2 under changes of other drivers. Until now, effects of seawater acidification on diatoms have been extensively studied under nitrogen-replete conditions (Gao and Campbell, 2014 and references therein). However, contrasting effects of acidification on T. pseudonana were detected under nitrate-limited and replete conditions (Figure 7), which indicates that elevated CO2 could show distinct effects on phytoplankton living in waters with different nutritional conditions. Moreover, the present study highlights the critical role of nitrogen availability in influencing the effects of seawater acidification and elevated sea surface temperatures on growth and photosynthesis in diatoms. Positive or neutral effects of acidification and warming on growth and photosynthesis of diatoms under nutrient replete conditions (Montagnes and Franklin, 2001; Kroeker et al., 2013) might be reversed to negative impacts when cells are nutrient-limited. T. pseudonana cells showed the lowest specific growth rate and photosynthetic carbon fixation rate under the combined conditions of elevated temperature, N-limitation, and seawater acidification, which is the scenario predicted for future oceanic ecosystems (Boyd et al., 2015). Thus, the negative effects of ocean warming on net primary production and phytoplankton biomass of low- and mid-latitude oceans (Behrenfeld et al., 2006; Boyce et al., 2010) could be further exacerbated under future ocean conditions. As the indispensable base and component of marine food webs, diatoms might be negatively impacted by changes of oceanic environmental factors. Hence, the base of marine food webs and the strength of the biological CO2 pump could be impacted severely by future CO2-induced seawater acidification and elevated temperature in a way that is also dependent on the nutritional conditions of local waters. However, it should be noticed that the responses of diatoms to seawater acidification might depend on the timescale over which they are exposed as longer term exposure to changed conditions leads to different physiological responses (Li et al., 2017b), and therefore caution should be exercised in directly extrapolating the results obtained from relatively short-term studies to the long-term ocean acidification process. Data availability Data in the present study are available at https://doi.pangaea.de/10.1594/PANGAEA.880576. Acknowledgements The authors would like to thank the two anonymous reviewers and the editor for their insightful comments on the manuscript. This study was supported by National Natural Science Foundation of China (41430967, 41720104005, 41721005) and Joint project of National Natural Science Foundation of China and Shandong province (No. U1606404). JB’s visit to Xiamen was supported by Xiamen University. References Agusti S. , González-Gordillo J. I., Vaqué D., Estrada M., Cerezo M. I., Salazar G., Gasol J. M. et al. 2015 . Ubiquitous healthy diatoms in the deep sea confirm deep carbon injection by the biological pump . Nature Communications 6 : 7608. Google Scholar Crossref Search ADS PubMed WorldCat Alipanah L. , Rohloff J., Winge P., Bones A. M., Brembu T. 2015 . Whole-cell response to nitrogen deprivation in the diatom Phaeodactylum tricornutum . Journal of Experimental Botany 66 : 6281 – 6296 . Google Scholar Crossref Search ADS PubMed WorldCat Araújo S. C. , Garcia V. M. T. 2005 . Growth and biochemical composition of the diatom Chaetoceros cf. wighamii brightwell under different temperature, salinity and carbon dioxide levels. I. Protein, carbohydrates and lipids . Aquaculture 246 : 405 – 412 . Google Scholar Crossref Search ADS WorldCat Beardall J. , Stojkovic S., Larsen S. 2009 . Living in a high CO2 world: impacts of global climate change on marine phytoplankton . Plant Ecology & Diversity 2 : 191 – 205 . Google Scholar Crossref Search ADS WorldCat Beardall J. , Stojkovic S., Gao K. 2014 . Interactive effects of nutrient supply and other environmental factors on the sensitivity of marine primary producers to ultraviolet radiation: implications for the impacts of global change . Aquatic Biology 22 : 5 – 23 . Google Scholar Crossref Search ADS WorldCat Behrenfeld M. J. , O’Malley R. T., Siegel D. A., McClain C. R., Sarmiento J. L., Feldman G. C., Milligan A. J. et al. 2006 . Climate-driven trends in contemporary ocean productivity . Nature 444 : 752 – 755 . Google Scholar Crossref Search ADS PubMed WorldCat Beman J. M. , Chow C.-E., King A. L., Feng Y., Fuhrman J. A., Andersson A., Bates N. R. et al. 2011 . Global declines in oceanic nitrification rates as a consequence of ocean acidification . Proceedings of the National Academy of Sciences of the United States of America 108 : 208 – 213 . Google Scholar Crossref Search ADS PubMed WorldCat Berges J. A. , Charlebois D. O., Mauzerall D. C., Falkowski P. G. 1996 . Differential effects of nitrogen limitation on photosynthetic efficiency of photosystems I and II in microalgae . Plant Physiology 110 : 689 – 696 . Google Scholar Crossref Search ADS PubMed WorldCat Bopp L. , Resplandy L., Orr J. C., Doney S. C., Dunne J. P., Gehlen M., Halloran P. et al. 2013 . Multiple stressors of ocean ecosystems in the 21st century: projections with CMIP5 models . Biogeosciences 10 : 6225 – 6245 . Google Scholar Crossref Search ADS WorldCat Boyce D. G. , Lewis M. R., Worm B. 2010 . Global phytoplankton decline over the past century . Nature 466 : 591 – 596 . Google Scholar Crossref Search ADS PubMed WorldCat Boyd P. W. , Strzepek R., Fu F., Hutchins D. A. 2010 . Environmental control of open-ocean phytoplankton groups: now and in the future . Limnology and Oceanography 55 : 1353 – 1376 . Google Scholar Crossref Search ADS WorldCat Boyd P. W. , Rynearson T. A., Armstrong E. A., Fu F., Hayashi K., Hu Z., Hutchins D. A. et al. 2013 . Marine phytoplankton temperature versus growth responses from polar to tropical waters – outcome of a scientific community-wide study . PLoS One 8 : e63091. Google Scholar Crossref Search ADS PubMed WorldCat Boyd P. W. , Lennartz S. T., Glover D. M., Doney S. C. 2015 . Biological ramifications of climate-change-mediated oceanic multi-stressors . Nature Climate Change 5 : 71 – 79 . Google Scholar Crossref Search ADS WorldCat Boyd P. W. , Cornwall C. E., Davison A., Doney S. C., Fourquez M., Hurd C. L., Lima I. D. et al. 2016 . Biological responses to environmental heterogeneity under future ocean conditions . Global Change Biology 22 : 2633 – 2650 . Google Scholar Crossref Search ADS PubMed WorldCat Brennan G. , Collins S. 2015 . Growth responses of a green alga to multiple environmental drivers . Nature Climate Change 5 : 892 – 897 . Google Scholar Crossref Search ADS WorldCat Brzezinski M. A. , Nelson D. M. 1995 . The annual silica cycle in the Sargasso Sea near Bermuda . Deep Sea Research Part I: Oceanographic Research Papers 42 : 1215 – 1237 . Google Scholar Crossref Search ADS WorldCat Coello-Camba A. , Agustí S., Holding J., Arrieta J. M., Duarte C. M. 2014 . Interactive effect of temperature and CO2 increase in Arctic phytoplankton . Frontiers in Marine Science 1 : 49 . Google Scholar Crossref Search ADS WorldCat Collins S. , Rost B., Rynearson T. A. 2014 . Evolutionary potential of marine phytoplankton under ocean acidification . Evolutionary Applications 7 : 140 – 155 . Google Scholar Crossref Search ADS PubMed WorldCat Crawfurd K. J. , Raven J. A., Wheeler G. L., Baxter E. J., Joint I. 2011 . The response of Thalassiosira pseudonana to long-term exposure to increased CO2 and decreased pH . PLoS One 6 : e26695. Google Scholar Crossref Search ADS PubMed WorldCat Doyle S. A. , Saros J. E., Williamson C. E. 2005 . Interactive effects of temperature and nutrient limitation on the response of alpine phytoplankton growth to ultraviolet radiation . Limnology and Oceanography 50 : 1362 – 1367 . Google Scholar Crossref Search ADS WorldCat Eilers P. , Peeters J. 1988 . A model for the relationship between light intensity and the rate of photosynthesis in phytoplankton . Ecological Modelling 42 : 199 – 215 . Google Scholar Crossref Search ADS WorldCat Elser J. J. , Fagan W. F., Denno R. F., Dobberfuhl D. R., Folarin A., Huberty A., Interlandi S. et al. 2000 . Nutritional constraints in terrestrial and freshwater food webs . Nature 408 : 578 – 580 . Google Scholar Crossref Search ADS PubMed WorldCat Eppley R. W. , Rogers J. N., McCarthy J. J. 1969 . Half-saturation constants for uptake of nitrate and ammonium by marine phytoplankton . Limnology and Oceanography 14 : 912 – 920 . Google Scholar Crossref Search ADS WorldCat Falkowski P. G. 1975 . Nitrate uptake in marine phytoplankton: comparison of half-saturation constants from seven species . Limnology and Oceanography 20 : 412 – 417 . Google Scholar Crossref Search ADS WorldCat Falkowski P. G. , Katz M. E., Knoll A. H., Quigg A., Raven J. A., Schofield O., Taylor F. 2004 . The evolution of modern eukaryotic phytoplankton . Science 305 : 354 – 360 . Google Scholar Crossref Search ADS PubMed WorldCat Feng Y. , Hare C. E., Leblanc K., Rose J. M., Zhang Y., DiTullio G. R., Lee P. et al. 2009 . Effects of increased pCO2 and temperature on the North Atlantic Spring Bloom: I. The phytoplankton community and biogeochemical response . Marine Ecology Progress Series 388 : 13 – 25 . Google Scholar Crossref Search ADS WorldCat Gao Y. , Smith G. J., Alberte R. S. 2000 . Temperature dependence of nitrate reductase activity in marine phytoplankton: biochemical analysis and ecological implications . Journal of Phycology 36 : 304 – 313 . Google Scholar Crossref Search ADS WorldCat Gao K. , Wu Y., Li G., Wu H., Villafane V. E., Helbling E. W. 2007 . Solar UV radiation drives CO2 fixation in marine phytoplankton: a double-edged sword . Plant Physiology 144 : 54 – 59 . Google Scholar Crossref Search ADS PubMed WorldCat Gao K. , Helbling E. W., Häder D. P., Hutchins D. A. 2012a . Responses of marine primary producers to interactions between ocean acidification, solar radiation, and warming . Marine Ecology Progress Series 470 : 167 – 189 . Google Scholar Crossref Search ADS WorldCat Gao K. , Xu J., Gao G., Li Y., Hutchins D. A., Huang B., Wang L. et al. 2012b . Rising CO2 and increased light exposure synergistically reduce marine primary productivity . Nature Climate Change 2 : 519 – 523 . Google Scholar Crossref Search ADS WorldCat Gao K. , Campbell D. A. 2014 . Photophysiological responses of marine diatoms to elevated CO2 and decreased pH: a review . Functional Plant Biology 41 : 449 – 459 . Google Scholar Crossref Search ADS WorldCat Gattuso J.-P. , Magnan A., Bille R., Cheung W. W. L., Howes E. L., Joos F., Allemand D. et al. 2015 . Contrasting futures for ocean and society from different anthropogenic CO2 emissions scenarios . Science 349 : aac4722 . Google Scholar Crossref Search ADS PubMed WorldCat Geider R. J. , Osborne B. A. 1989 . Respiration and microalgal growth: a review of the quantitative relationship between dark respiration and growth . New Phytologist 112 : 327 – 341 . Google Scholar Crossref Search ADS WorldCat Gilstad M. , Johnsen G., Sakshaug E. 1993 . Photosynthetic parameters, pigment composition and respiration rates of the marine diatom Skeletonema costatum grown in continuous light and a 12: 12 h light-dark cycle . Journal of Plankton Research 15 : 939 – 951 . Google Scholar Crossref Search ADS WorldCat Giordano M. , Raven J. A. 2014 . Nitrogen and sulfur assimilation in plants and algae . Aquatic Botany 118 : 45 – 61 . Google Scholar Crossref Search ADS WorldCat Hagerthey S. E. , Defew E. C., Paterson D. M. 2002 . Influence of Corophium volutator and Hydrobia ulvae on intertidal benthic diatom assemblages under different nutrient and temperature regimes . Marine Ecology Progress Series 245 : 47 – 59 . Google Scholar Crossref Search ADS WorldCat Halsey K. H. , Milligan A. J., Behrenfeld M. J. 2011 . Linking time-dependent carbon-fixation efficiencies in Dunaliella tertiolecta (Chlorophyceae) to underlying metabolic pathways . Journal of Phycology 47 : 66 – 76 . Google Scholar Crossref Search ADS PubMed WorldCat Halsey K. H. , O'Malley R. T., Graff J. R., Milligan A. J., Behrenfeld M. J. 2013 . A common partitioning strategy for photosynthetic products in evolutionarily distinct phytoplankton species . New Phytologist 198 : 1030 – 1038 . Google Scholar Crossref Search ADS PubMed WorldCat Heiden J. P. , Bischof K., Trimborn S. 2016 . Light intensity modulates the response of two Antarctic diatom species to ocean acidification . Frontiers in Marine Science 3 : 260. Google Scholar Crossref Search ADS WorldCat Hennon G. M. M. , Quay P., Morales R. L., Swanson L. M., Virginia Armbrust E. 2014 . Acclimation conditions modify physiological response of the diatom Thalassiosira pseudonana to elevated CO2 concentrations in a nitrate-limited chemostat . Journal of Phycology 50 : 243 – 253 . Google Scholar Crossref Search ADS PubMed WorldCat Hennon G. M. M. , Ashworth J., Groussman R. D., Berthiaume C., Morales R. L., Baliga N. S., Orellana M. V. et al. 2015 . Diatom acclimation to elevated CO2 via cAMP signalling and coordinated gene expression . Nature Climate Change 5 : 761 – 765 . Google Scholar Crossref Search ADS WorldCat Hockin N. L. , Mock T., Mulholland F., Kopriva S., Malin G. 2012 . The response of diatom central carbon metabolism to nitrogen starvation is different from that of green algae and higher plants . Plant Physiology 158 : 299 – 312 . Google Scholar Crossref Search ADS PubMed WorldCat Hong H. , Li D., Lin W., Li W., Shi D. 2017 . Nitrogen nutritional condition affects the response of energy metabolism in diatoms to elevated carbon dioxide . Marine Ecology Progress Series 567 : 41 – 56 . Google Scholar Crossref Search ADS WorldCat Hopkinson B. M. , Dupont C. L., Allen A. E., Morel F. M. 2011 . Efficiency of the CO2-concentrating mechanism of diatoms . Proceedings of the National Academy of Sciences of the United States of America 108 : 3830 – 3837 . Google Scholar Crossref Search ADS PubMed WorldCat Hoppe C. J. , Holtz L. M., Trimborn S., Rost B. 2015 . Ocean acidification decreases the light-use efficiency in an Antarctic diatom under dynamic but not constant light . New Phytologist 207 : 159 – 171 . Google Scholar Crossref Search ADS PubMed WorldCat Huppe H. , Turpin D. 1994 . Integration of carbon and nitrogen metabolism in plant and algal cells . Annual Review of Plant Biology 45 : 577 – 607 . Google Scholar Crossref Search ADS WorldCat Jin P. , Wang T., Liu N., Dupont S., Beardall J., Boyd P. W., Riebesell U. et al. 2015 . Ocean acidification increases the accumulation of toxic phenolic compounds across trophic levels . Nature Communications 6 : 8714. Google Scholar Crossref Search ADS PubMed WorldCat King A. L. , Sanudo-Wilhelmy S. A., Leblanc K., Hutchins D. A., Fu F. 2011 . CO2 and vitamin B12 interactions determine bioactive trace metal requirements of a subarctic Pacific diatom . The ISME Jornal 5 : 1388 – 1396 . Google Scholar Crossref Search ADS WorldCat King A. L. , Jenkins B. D., Wallace J. R., Liu Y., Wikfors G. H., Milke L. M., Meseck S. L. 2015 . Effects of CO2 on growth rate, C: N: P, and fatty acid composition of seven marine phytoplankton species . Marine Ecology Progress Series 537 : 59 – 69 . Google Scholar Crossref Search ADS WorldCat Kroeker K. J. , Kordas R. L., Crim R., Hendriks I. E., Ramajo L., Singh G. S., Duarte C. M. et al. 2013 . Impacts of ocean acidification on marine organisms: quantifying sensitivities and interaction with warming . Global Change Biology 19 : 1884 – 1896 . Google Scholar Crossref Search ADS PubMed WorldCat Langer G. , Geisen M., Baumann K. H., Kläs J., Riebesell U., Thoms S., Young J. R. 2006 . Species-specific responses of calcifying algae to changing seawater carbonate chemistry . Geochemistry, Geophysics, Geosystems 7 : Q09006. Google Scholar Crossref Search ADS WorldCat Langer G. , Nehrke G., Probert I., Ly J., Ziveri P. 2009 . Strain-specific responses of Emiliania huxleyi to changing seawater carbonate chemistry . Biogeosciences 6 : 2637 – 2646 . Google Scholar Crossref Search ADS WorldCat Lewandowska A. M. , Boyce D. G., Hofmann M., Matthiessen B., Sommer U., Worm B. 2014 . Effects of sea surface warming on marine plankton . Ecology Letters 17 : 614 – 623 . Google Scholar Crossref Search ADS PubMed WorldCat Li F. , Wu Y., Hutchins D. A., Fu F., Gao K. 2016 . Physiological responses of coastal and oceanic diatoms to diurnal fluctuations in seawater carbonate chemistry under two CO2 concentrations . Biogeosciences 13 : 6247 – 6259 . Google Scholar Crossref Search ADS WorldCat Li F. , Beardall J., Collins S., Gao K. 2017b . Decreased photosynthesis and growth with reduced respiration in the model diatom Phaeodactylum tricornutum grown under elevated CO2 over 1800 generations . Global Change Biology 23 : 127 – 137 . Google Scholar Crossref Search ADS WorldCat Li W. , Gao K., Beardall J. 2012b . Interactive effects of ocean acidification and nitrogen-limitation on the diatom Phaeodactylum tricornutum . PLoS One 7 : e51590 . Google Scholar Crossref Search ADS WorldCat Li W. , Yang Y., Li Z., Xu J., Gao K. 2017a . Effects of seawater acidification on the growth rates of the diatom Thalassiosira (Conticribra) weissflogii under different nutrient, light, and UV radiation regimes . Journal of Applied Phycology 29 : 133 – 142 . Google Scholar Crossref Search ADS WorldCat Li Y. , Gao K., Villafañe V. E., Helbling E. W. 2012a . Ocean acidification mediates photosynthetic response to UV radiation and temperature increase in the diatom Phaeodactylum tricornutum . Biogeosciences 9 : 3931 – 3942 . Google Scholar Crossref Search ADS WorldCat Li Y. , Xu J., Gao K. 2014 . Light-modulated responses of growth and photosynthetic performance to ocean acidification in the model diatom Phaeodactylum tricornutum . PLoS One 9 : e96173. Google Scholar Crossref Search ADS PubMed WorldCat Liu N. , Beardall J., Gao K. 2017 . Elevated CO2 and associated seawater chemistry do not benefit a model diatom grown with increased availability of light . Aquatic Microbial Ecology 79 : 137 – 147 . Google Scholar Crossref Search ADS WorldCat Lomas M. W. , Glibert P. M. 1999 . Temperature regulation of nitrate uptake: a novel hypothesis about nitrate uptake and reduction in cool-water diatoms . Limnology and Oceanography 44 : 556 – 572 . Google Scholar Crossref Search ADS WorldCat McCarthy A. , Rogers S. P., Duffy S. J., Campbell D. A. 2012 . Elevated carbon dioxide differentially alters the photophysiology of Thalassiosira pseudonana (Bacillariophyceae) and Emiliania huxleyi (Haptophyta) . Journal of Phycology 48 : 635 – 646 . Google Scholar Crossref Search ADS PubMed WorldCat Mejia L. M. , Isensee K., Méndez-Vicente A., Pisonero J., Shimizu N., González C., Monteleone B. 2013 . B content and Si/C ratios from cultured diatoms (Thalassiosira pseudonana and Thalassiosira weissflogii): relationship to seawater pH and diatom carbon acquisition . Geochimica Et Cosmochimica Acta 123 : 322 – 337 . Google Scholar Crossref Search ADS WorldCat Mock T. , Samanta M. P., Iverson V., Berthiaume C., Robison M., Holtermann K., Durkin C. et al. 2008 . Whole-genome expression profiling of the marine diatom Thalassiosira pseudonana identifies genes involved in silicon bioprocesses . Proceedings of the National Academy of Sciences of the United States of America 105 : 1579 – 1584 . Google Scholar Crossref Search ADS PubMed WorldCat Montagnes D. J. , Franklin M. 2001 . Effect of temperature on diatom volume, growth rate, and carbon and nitrogen content: reconsidering some paradigms . Limnology and Oceanography 46 : 2008 – 2018 . Google Scholar Crossref Search ADS WorldCat Moore J. K. , Doney S. C., Kleypas J. A., Glover D. M., Fung I. Y. 2002a . An intermediate complexity marine ecosystem model for the global domain . Deep Sea Research Part II: Topical Studies in Oceanography 49 : 403 – 462 . Google Scholar Crossref Search ADS WorldCat Moore J. K. , Doney S. C., Glover D. M., Fung I. Y. 2002b . Iron cycling and nutrient-limitation patterns in surface waters of the World Ocean . Deep Sea Research Part II: Topical Studies in Oceanography 49 : 463 – 507 . Google Scholar Crossref Search ADS WorldCat Nakajima K. , Tanaka A., Matsuda Y. 2013 . SLC4 family transporters in a marine diatom directly pump bicarbonate from seawater . Proceedings of the National Academy of Sciences of the United States of America 110 : 1767 – 1772 . Google Scholar Crossref Search ADS PubMed WorldCat Parker M. S. , Armbrust E. V. 2005 . Synergistic effects of light, temperature, and nitrogen source on transcription of genes for carbon and nitrogen metabolism in the centric diatom Thalassiosira pseudonana (Bacillariophyceae) . Journal of Phycology 41 : 1142 – 1153 . Google Scholar Crossref Search ADS WorldCat Passow U. , Laws E. A. 2015 . Ocean acidification as one of multiple stressors: growth response of Thalassiosira weissflogii (diatom) under temperature and light stress . Marine Ecology Progress Series 541 : 75 – 90 . Google Scholar Crossref Search ADS WorldCat Raven J. A. , Beardall J. 2005 . Respiration in aquatic photolithotrophs. In Respiration in Aquatic Ecosystems , pp. 36–46. Ed. by del Giorgio P. A., Williams P. J. le B.. Oxford University Press , New York . 315 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Raven J. A. , Beardall J. 2014 . CO2 concentrating mechanisms and environmental change . Aquatic Botany 118 : 24 – 37 . Google Scholar Crossref Search ADS WorldCat Ritchie R. J. 2006 . Consistent sets of spectrophotometric chlorophyll equations for acetone, methanol and ethanol solvents . Photosynthesis Research 89 : 27 – 41 . Google Scholar Crossref Search ADS PubMed WorldCat Rost B. , Zondervan I., Wolf-Gladrow D. 2008 . Sensitivity of phytoplankton to future changes in ocean carbonate chemistry: current knowledge, contradictions and research directions . Marine Ecology Progress Series 373 : 227 – 237 . Google Scholar Crossref Search ADS WorldCat Segovia M. , Lorenzo M. R., Maldonado M. T., Larsen A., Berger S. A., Tsagaraki T. M., Lázaro F. J. et al. 2017 . Iron availability modulates the effects of future CO2 levels within the marine planktonic food web . Marine Ecology Progress Series 565 : 17 – 33 . Google Scholar Crossref Search ADS WorldCat Shi D. , Li W., Hopkinson B. M., Hong H., Li D., Kao S.-J., Lin W. 2015 . Interactive effects of light, nitrogen source, and carbon dioxide on energy metabolism in the diatom Thalassiosira pseudonana . Limnology and Oceanography 60 : 1805 – 1822 . Google Scholar Crossref Search ADS WorldCat Sobrino C. , Ward M. L., Neale P. J. 2008 . Acclimation to elevated carbon dioxide and ultraviolet radiation in the diatom Thalassiosira pseudonana: effects on growth, photosynthesis, and spectral sensitivity of photoinhibition . Limnology and Oceanography 53 : 494 – 505 . Google Scholar Crossref Search ADS WorldCat Suffrian K. , Schulz K. G., Gutowska M. A., Riebesell U., Bleich M. 2011 . Cellular pH measurements in Emiliania huxleyi reveal pronounced membrane proton permeability . New Phytologist 190 : 595 – 608 . Google Scholar Crossref Search ADS PubMed WorldCat Sugie K. , Yoshimura T. 2013 . Effects of pCO2 and iron on the elemental composition and cell geometry of the marine diatom Pseudo‐nitzschia pseudodelicatissima (Bacillariophyceae) . Journal of Phycology 49 : 475 – 488 . Google Scholar Crossref Search ADS PubMed WorldCat Sugie K. , Yoshimura T. 2016 . Effects of high CO2 levels on the ecophysiology of the diatom Thalassiosira weissflogii differ depending on the iron nutritional status . ICES Journal of Marine Science 73 : 680 – 692 . Google Scholar Crossref Search ADS WorldCat Sun J. , Hutchins D. A., Feng Y., Seubert E. L., Caron D. A., Fu F. 2011 . Effects of changing pCO2 and phosphate availability on domoic acid production and physiology of the marine harmful bloom diatom Pseudo-nitzschia multiseries . Limnology and Oceanography 56 : 829 – 840 . Google Scholar Crossref Search ADS WorldCat Sunda W. G. , Price N. M., Morel F. M. 2005 . Trace metal ion buffers and their use in culture studies. In Algal Culturing Techniques , pp. 35–63. Ed. by Aderson R. A.. Elsevier Academic Press , Burlington . 578 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Tatters A. O. , Fu F.-X., Hutchins D. A., Neilan B. 2012 . High CO2 and silicate limitation synergistically increase the toxicity of Pseudo-nitzschia fraudulenta . PLoS One 7 : e32116. Google Scholar Crossref Search ADS PubMed WorldCat Tatters A. O. , Roleda M. Y., Schnetzer A., Fu F., Hurd C. L., Boyd P. W., Caron D. A. et al. 2013 . Short- and long-term conditioning of a temperate marine diatom community to acidification and warming . Philosophical Transactions of the Royal Society B: Biological Sciences 368 : 20120437. Google Scholar Crossref Search ADS WorldCat Taucher J. , Jones J., James A., Brzezinski M. A., Carlson C. A., Riebesell U., Passow U. 2015 . Combined effects of CO2 and temperature on carbon uptake and partitioning by the marine diatoms Thalassiosira weissflogii and Dactyliosolen fragilissimus . Limnology and Oceanography 60 : 901 – 919 . Google Scholar Crossref Search ADS WorldCat Taylor A. R. , Brownlee C., Wheeler G. L. 2012 . Proton channels in algae: reasons to be excited . Trends in Plant Science 17 : 675 – 684 . Google Scholar Crossref Search ADS PubMed WorldCat Torstensson A. , Chierici M., Wulff A. 2012 . The influence of increased temperature and carbon dioxide levels on the benthic/sea ice diatom Navicula directa . Polar Biology 35 : 205 – 214 . Google Scholar Crossref Search ADS WorldCat Torstensson A. , Hedblom M., Andersson J., Andersson M. X., Wulff A. 2013 . Synergism between elevated pCO2 and temperature on the Antarctic sea ice diatom Nitzschia lecointei . Biogeosciences 10 : 6391 – 6401 . Google Scholar Crossref Search ADS WorldCat Toseland A. , Daines S. J., Clark J. R., Kirkham A., Strauss J., Uhlig C., Lenton T. M. et al. 2013 . The impact of temperature on marine phytoplankton resource allocation and metabolism . Nature Climate Change 3 : 979 – 984 . Google Scholar Crossref Search ADS WorldCat Trimborn S. , Wolf-Gladrow D., Richter K.-U., Rost B. 2009 . The effect of pCO2 on carbon acquisition and intracellular assimilation in four marine diatoms . Journal of Experimental Marine Biology and Ecology 376 : 26 – 36 . Google Scholar Crossref Search ADS WorldCat Trimborn S. , Thoms S., Petrou K., Kranz S. A., Rost B. 2014 . Photophysiological responses of Southern Ocean phytoplankton to changes in CO2 concentrations: short-term versus acclimation effects . Journal of Experimental Marine Biology and Ecology 451 : 44 – 54 . Google Scholar Crossref Search ADS WorldCat Tsuji Y. , Mahardika A., Matsuda Y. 2017 . Evolutionarily distinct strategies for the acquisition of inorganic carbon from seawater in marine diatoms . Journal of Experimental Botany , 68 : 3949 – 3958 . Google Scholar Crossref Search ADS PubMed WorldCat Verspagen J. M. H. , Van de Waal D. B., Finke J. F., Visser P. M., Huisman J. 2014 . Contrasting effects of rising CO2 on primary production and ecological stoichiometry at different nutrient levels . Ecology Letters 17 : 951 – 960 . Google Scholar Crossref Search ADS PubMed WorldCat Wichard T. , Gerecht A., Boersma M., Poulet S. A., Wiltshire K., Pohnert G. 2007 . Lipid and fatty acid composition of diatoms revisited: rapid wound-activated change of food quality parameters influences herbivorous copepod reproductive success . ChemBioChem 8 : 1146 – 1153 . Google Scholar Crossref Search ADS PubMed WorldCat Williams P. J. , Le B., Thomas D. N., Reynolds C. S. 2002 . Phytoplankton Productivity: Carbon Assimilation in Marine and Freshwater Ecology . Wiley-Blackwell , Oxford . 400 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wingler A. , Lea P. J., Quick W. P., Leegood R. C. 2000 . Photorespiration: metabolic pathways and their role in stress protection . Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 355 : 1517 – 1529 . Google Scholar Crossref Search ADS WorldCat Wu Y. , Gao K., Riebesell U. 2010 . CO2-induced seawater acidification affects physiological performance of the marine diatom Phaeodactylum tricornutum . Biogeosciences 7 : 2915 – 2923 . Google Scholar Crossref Search ADS WorldCat Yang G. , Gao K. 2012 . Physiological responses of the marine diatom Thalassiosira pseudonana to increased pCO2 and seawater acidity . Marine Environmental Research 79 : 142 – 151 . Google Scholar Crossref Search ADS PubMed WorldCat © International Council for the Exploration of the Sea 2018. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © International Council for the Exploration of the Sea 2018. All rights reserved. For Permissions, please email: [email protected]