An overview of the marine food web in Icelandic waters using Ecopath with Ecosim
Joana P.C. Ribeiro, Bjarki Þ. Elvarsson, Erla Sturludóttir, Gunnar Stefánsson
AAn overview of the marine food web in Icelandic watersusing Ecopath with Ecosim
Joana P.C. Ribeiro a, ∗ , Bjarki Þ. Elvarsson b , Erla Sturludóttir a , GunnarStefánsson a a Science Institute, University of Iceland, Dunhagi 7, 107 Reykjavík, Iceland b Marine and Freshwater Research Institute, Skúlagata 4, 101 Reykjavík, Iceland
Abstract
Fishing activities have broad impacts that affect, although not exclusively, thetargeted stocks. These impacts affect predators and prey of the harvestedspecies, as well as the whole ecosystem it inhabits. Ecosystem models canbe used to study the interactions that occur within a system, including thosebetween different organisms and those between fisheries and targeted species.Trophic web models like Ecopath with Ecosim (EwE) can handle fishing fleets asa top predator, with top-down impact on harvested organisms. The aim of thisstudy was to better understand the Icelandic marine ecosystem and the interac-tions within. This was done by constructing an EwE model of Icelandic waters.The model was run from 1984 to 2013 and was fitted to time series of biomassestimates, landings data and mean annual temperature. The final model waschosen by selecting the model with the lowest Akaike information criterion. Askill assessment was performed using the Pearson’s correlation coefficient, thecoefficient of determination, the modelling efficiency and the reliability index toevaluate the model performance. The model performed satisfactorily when sim-ulating previously estimated biomass and known landings. Most of the groupswith time series were estimated to have top-down control over their prey. Theseare harvested species with direct and/or indirect links to lower trophic levels andfuture fishing policies should take this into account. This model could be usedas a tool to investigate how such policies could impact the marine ecosystem inIcelandic waters.
Keywords:
Ecosystem dynamics, Trophic web, Ecopath with Ecosim,Icelandic waters
1. Introduction
Fisheries affect targeted stocks, as well as the whole ecosystem. Fisheriescan have negative effects due to maximizing stock profitability. These include ∗ Corresponding author
Email address: [email protected] (Joana P.C. Ribeiro) a r X i v : . [ q - b i o . P E ] F e b amage to the habitat and impacts to predators and prey of the fished species(Pikitch et al., 2004). The need for fisheries management arose from overex-ploitation of fish stocks. Since they were first developed, different measures offisheries management have been employed across systems and their effectivenesshas been previously studied (Stefansson and Rosenberg, 2005). Fishing policies,or lack thereof, have not always prevented overexploitation (Cook et al., 1997).Fisheries management strategies have had a tendency to focus on a singlespecies and not on wider, more realistic multi species scenarios (Hanna, 1999;Pikitch et al., 2004). Ecosystems harbour different species within them, whichinteract with each other and with their environment. The species in a givensystem are connected through trophic links and other types of ecological in-teractions. Ecosystem models can be used to gain better understanding of theunderlying dynamics within a system. Trophic links in particular can be studiedusing food web models, also referred to as trophic models. When a system isharvested, the interaction between the targeted species and the fishing fleet isintroduced in the system. In this case, fishing fleets can be handled by food webmodels as a top predator which interact with targeted stocks through top-downtrophic control.A widely used food web model is Ecopath with Ecosim (EwE; Christensenand Pauly, 1992; Walters et al., 1997; Pauly et al., 2000; Walters et al., 2000).Typical uses of EwE include answering ecological questions, quantifying trophicflows and studying the food-web structure and investigating potential impactsof fisheries on ecosystems. EwE has also been used as a fisheries managementscenarios evaluation tool (for a complete overview on EwE capacities and limi-tations see Christensen and Walters, 2004).Two EwE models for Icelandic waters had previously been published (Buchary,E.A.; Mendy, A.N. and Buchary, E.A., in Guénette et al., 2001; Mendy, 1998).The earlier model (Buchary, E.A. and Mendy, A.N. in Guénette et al., 2001;Mendy, 1998) modelled the trophic web in Icelandic waters in the year 1997 anddid not include Ecosim. The latter model (Buchary E.A. in Guénette et al.,2001) was a reconstruction of the food web in Icelandic waters in 1950 andincluded a dynamic component that simulated the trophic dynamics between1950 and 1997. These models did not cover more recent dynamics nor werethey evaluated using a skill assessment. Other work on trophic interactions inIcelandic waters include specific predator-prey interactions (Stefansson et al.,1998), isotope analyses of organisms in the Icelandic sea (Petursdottir et al.,2012) and end to end modelling of the marine ecosystem (Sturludottir et al.,2018).The marine environment around Iceland has been exploited for centuries. Inthe 20 th century, fisheries were the main source of income in Iceland (Popescuand Poulsen, 2012). Economically important species include demersal fish suchas cod ( Gadus morhua ), saithe (
Pollachius virens ), haddock (
Melanogrammusaeglefinus ), redfish (
Sebastes spp. ) and Greenland halibut (
Rinhardtius hip-poglossoides ), as well as pelagic species such as herring (
Clupea harengus ),capelin (
Mallotus villosus ) and, in later years, blue whiting (
Micromesistiuspoutassou ; Astthorsson et al., 2007). 2he aim of this study was to gain better understanding of the ecosystem inIcelandic waters by using EwE to model the trophic interactions between speciesin the ecosystem, as well as the interactions between fisheries and targetedspecies. The aim was also to evaluate the performance of the model with a skillassessment.
2. Methods
Iceland is an island located in the North Atlantic ocean, at the junctionof the Middle Atlantic Ruidge and the Greenland-Scotland ridge (Astthorssonet al., 2007). Its exclusive economic zone (EEZ) spans an area of 758 000 km (Popescu and Poulsen, 2012) (Figure 1). This area includes the continental shelfaround Iceland. Figure 1: Icelandic EEZ with Iceland marked in black and highlighted continental shelf.
The Icelandic waters harbour organisms ranging from phytoplankton to ma-rine mammals, as well as sea birds. Some of these are differently distributedalong the ecosystem in Icelandic waters, due to different water salinity and tem-perature between the southern and western areas and the northern and easternparts (Astthorsson et al., 2007). The southern and western areas have warmand saline Atlantic water, while the water in northern and eastern areas is a mixof Atlantic, Arctic and Polar water (Malmberg and Jonsson, 1997; Jonsson and3aldimarsson, 2005; Astthorsson et al., 2007). The different habitats resultingfrom this difference in water composition is exploited by some species, includ-ing commercially important fish stocks. These spawn in southern and westernwaters and feed off northern and western waters (Astthorsson et al., 1994, 2007).
To study the marine ecosystem of Icelandic waters, EwE (Christensen andPauly, 1992; Walters et al., 1997; Pauly et al., 2000; Walters et al., 2000) waschosen as a tool to give a simple overview of the entire system. EwE is composedof two modelling stages, Ecopath and Ecosim. Ecopath is the mass-balance partof EwE and its master equation is as follows: B i (cid:18) PB (cid:19) i = (cid:88) j B j (cid:18) QB (cid:19) j DC ij + Y i + E i + BA i + B i (cid:18) PB (cid:19) i (1 − EE i ) (1)Here, B i is the biomass of functional group i , ( P/B ) i is the production tobiomass ratio, equivalent to total mortality ( Z ) in closed systems and corre-sponding to the sum of natural mortality ( M ) and fishing mortality ( F ), B j is the biomass of predator j , ( Q/B ) j is the consumption to biomass ratio of j , DC ij is the proportion of i found in the stomach of j , Y i is the yield, E i isthe net migration, BA i is the biomass accumulation and EE i is the ecotrophicefficiency, which corresponds to the production of i explained by the model. F is calculated in Ecopath as the ratio Y /B . The Icelandic Ecopath model was comprised of 35 functional groups, ofwhich six contained two stanzas each. The most relevant species of phyto-plankton, zooplankton, invertebrates, fish, marine mammals and sea birds wereincluded in the 35 functional groups used in the EwE model here presented (Ta-ble 1). Species were chosen as multi-stanza and/or individual functional groupsdepending on their economical importance and/or their role on the ecosystemand available biomass estimates and/or landings data. The aggregation of dif-ferent species in a single functional group was based on species type and habitatuse (Table A.1).Commercial species were targeted by a single fishing fleet, without discrim-ination regarding fishing gear. 4 able 1: List of the functional groups for the Icelandic waters EwE model.Functional groupsSeabirds Herring* (0-3) Commercial demersalMinke whale Herring* (4+) Other demersalBaleen whales Redfish* (0-7) CephalopodTooth whales Redfish* (8+) MolluscPinniped Greenland halibut* (0-5) LobsterGreenland shark Greenland halibut* (6+) ShrimpDogshark Capelin BenthosSkate rays Blue whiting JellyfishCod* (0-3) Mackerel KrillCod* (4+) Sandeel Crustacean zooplanktonSaithe* (0-3) Large pelagic Small zooplanktonSaithe* (4+) Small pelagic PhytoplanktonHaddock* (0-2) Flatfish DetritusHaddock* (3+) Other codfish*Multi-stanza group
Ecopath balances the model by solving a series of linear equations (Eq. 1).For each equation there must be an unknown value. The four basic values inEcopath are B , P/B , Q/B and EE . Ideally, B , P/B and
Q/B should be inputvalues. However, if estimates of B are lacking, the EE can be used as an inputvalue to estimate B . In multi-stanza groups, B , P/B and
Q/B have to beinserted for the leading stanza, whereas only
P/B needs to be inserted for allstanzas. Even though
P/B and
Q/B can be estimated, it is highly recommendedthat both are part of the input. The assumption of a closed system was madewhen modelling the marine ecosystem in Iceland. In a closed system, there isno BA nor E , thus these values were set to 0.Ecopath was set up for the year 1984 and the input biomass and landingscorresponded to estimates for this year or for the next year with available data.A comprehensive list of the literature used can be found in Table 2.5 able 2: Literature sources of biomass and Q/B estimates and of landings.Functional group Literature usedSeabirds Lilliendahl and Solmundsson (1997)Dommasnes et al. (2001)Umhverfisstofnun (2016)Minke whale; Sigurjónsson and Víkingsson (1997)Baleen whales and Borchers et al. (2009)Tooth whales Pike et al. (2009)Víkingsson et al. (2009)Institute (2014)Pinniped Gunnarsson et al. (1998)Dommasnes et al. (2001)All fish groups Institute (2014)Froese and Pauly (2016)Greenland shark (
Somniosus microcephalus ; biomass) MacNeil et al. (2012)Cod (
Q/B ) Magnússon and Pálsson (1989)Mollusc; jellyfish; benthos;Small zooplankton and crustacean zooplanktonPhytoplankton Kleisner and Hoornaert (2016)
No biomass was found for functional groups dogshark, skate rays, sandeel,large pelagic, small pelagic, flatfish, other demersal, lobster, shrimp, benthos,jellyfish and both zooplankton groups. Since these were not harvested in Icelandin 1984, it was not possible to estimate biomass and as such the EE was usedinstead. The EE of dogshark and skate rays was set to 0.500. For all remaininggroups, EE was set to 0.950. The choice of EE values was made based onrecommendations in the literature (Heymans et al., 2016). Stomach content data of most fish groups was provided by the MFRI (Insti-tute, 2014) and then transformed into diet composition using R (R Core Team,2016), to be later used in Ecopath. Diet composition of all other groups, as wellas of small pelagic fish, other demersal fish and invertebrates was set accordingto the literature (Table 3).
Table 3: Literature sources of diet composition.Functional group Literature usedSeabirds Lilliendahl and Solmundsson (1997)Minke whale Vikingsson et al. (2014)Baleen whales Sigurjónsson and Víkingsson (1997)Tooth whales Canning et al. (2008)Sigurjónsson and Víkingsson (1997)Víkingsson et al. (2003)Pinnipeds Hauksson and Bogason (1997)Large pelagic fish Hutchings (2002)Small pelagic fish Tyler and Pearcy (1975)Other demersal fish Hutchings (2002)All inverebrate groups Gunnarsson et al. (1998) .2.4. Model balancing After initialising Ecopath for the first time, two groups had
EE > . Thesewere functional group herring 0-3 ( EE = 1 . ) and capelin ( EE = 2 . ). Inorder to balance the model, P/B of the herring 0-3 group was raised from 0.490to 0.570 and the EE was set to 0.950 for capelin. The latter decision was basedon the hypothesis that capelin biomass has been underestimated in assessments(Magnússon and Pálsson, 1989). According to a study by Magnússon and Páls-son (1989), assessments of capelin biomass have to be approximately doubledto explain consumption by predators. Ecosim is the dynamic part of the EwE modelling suite (Walters et al.,1997, 2000; Christensen and Walters, 2004). Ecopath estimates are used asinitial values in Ecosim, which runs a simulation for a user-defined period. TheEcosim model presented here had a time span of 30 years. Ecosim uses a systemof differential equations to give a time series of simulated biomass and landings: dB i dt = g i (cid:88) j Q ji − (cid:88) j Q ij + I i − ( M i + F i + e i ) B i (2)Here, dB i /dt represents the rate of change in biomass of functional group i ,described in terms of its biomass, B i , g i is net growth efficiency ( P/Q ), Q ji isthe consumption of j by i , Q ij is the predation by j on i , M i is the mortalitynot explained by the model, F i is the fishing mortality rate, e i is the emigrationrate and I i is the immigration rate.Consumption in Ecosim is calculated based on the foraging arena theory(Walters et al., 1997; Ahrens et al., 2012), in which the biomass of the prey issplit into vulnerable and non-vulnerable pools. Consumption is computed as: Q ij = v ij · a ij · B i · B j · T i · T j · S ij · M ij /D j v ij + v ij · T i · M ij + a ij · M ij B j · S j · T j /D j (3)Here, v ij is the vulnerability of prey i to predator j , a ij is the rate of effectivesearch of i by j , T i is the relative feeding time of i , T j is the relative feedingtime of j , S ij is the representation of seasonal or long-term forcing effects, M ij represents mediation forcing effects and D j is the impact of handling time as alimit to consumption. Ecosim benefits from reference temporal data used to calibrate the model.This reference data includes biomass and/or landings. In addition, biomass,landings, F and environmental variables can be forced to drive the model. Tocalibrate the model, a time series of both forced and reference values was used(Table 4). 7 able 4: Description of the time-series used to calibrate Ecosim. F - Fishing mortality (forced);FB - Forced biomass; RB - Reference biomass; RC - Reference landings; EV - Environmentalvariables. Type Group PeriodF Cod 4+; Saithe 4+; Haddock 3+; 1984-2013Herring 4+; Redfish 8+; Capelin;Blue whiting; Flatfish; Other codfish;Mackerel; 2005-2013Commercial demersal; 1994-2013FB Mackerel; 1984-2013RB Cod 4+; Saithe 4+; Haddock 3+; 1984-2013Herring 4+; Redfish 8+; Blue whiting;RC Cod 4+; Saithe 4+; Haddock 3+; 1984-2013Herring 4+; Redfish 8+; Greenland halibut 6+;Capelin; Flatfish; Blue whiting;Mackerel; 2005-2013Other codfish; Commercial demmersal. 1984-2013EV Temperature 1984-2013 Biomass forcing was used for mackerel. Numbers of mackerel have been in-creasing in Icelandic waters since around 2007 (Astthorsson et al., 2012). Initialbiomass of mackerel was set to low values (0.200 t.km − ), as new functionalgroups cannot be introduced in Ecosim. The invasion by mackerel was thensimulated by forcing its biomass from 2005 on. This was the chosen method tohandle the increasing numbers of mackerel in the ecosystem as a method to sim-ulate species invasion in EwE. Average annual temperature was used as forcingfunction for PP. The time series was provided the MFRI (Institute, 2014).The model was fitted to the time series by conducting anomaly and vulnera-bility searches. The former reduced the total sum of squares (SSQ) by adjustingthe forcing function on PP, while the latter reduced SSQ by adjusting vulnerabil-ity parameters. In EwE, the anomaly search uses spline regression to search fortime series values of annual relative PP that could explain productivity shiftsimpacting biomass across the ecosystem (Christensen et al., 2008). Anomalysearches using 5, 10, 15 and 20 spline points were conducted and model selec-tion was done using the Akaike method (Akaike, 1974). The Akaike methodfor model selection uses the Akaike information criterion (AIC) as a measure ofmodel how well the model fits to the data considering the number of parametersand the lower the AIC, the better the model (Akaike, 1974). A skill assessment of the model was done using three measures suggestedby Stow et al. (2009). These were the Pearson’s correlation coefficient ( r ), thesquared Pearson’s correlation coefficient or determination coefficient (R ), themodelling efficiency coefficient (MEF; eq. 4) and the reliability index (RI; eq.8). The model efficiency coefficient and the reliability index are described as: M EF = (cid:0)(cid:80) ni =1 ( O i − ¯ O ) − (cid:80) ni =1 ( P i − O i ) (cid:1)(cid:80) ni =1 ( O i − ¯ O ) (4)And: RI = exp (cid:118)(cid:117)(cid:117)(cid:116) n n (cid:88) i =1 (cid:18) log O i P i (cid:19) (5)Here, n is the number of observations, O i is the ith of observation, ¯ O is theaverage of the observations and P i is the ith prediction.The correlation coefficient, r , can range from -1 to 1. An r value of 0indicates no correlation, whereas values of 1 indicate that the simulated andreference values variate with an equal trend. Negative r values indicate inversevariation between simulated and reference values. MEF can vary between −∞ and 1. Negative values of MEF indicate that the model has a worse fit to thedata than an average through the time series; M EF = 0 indicates that themodel has an equal fit to an average through the time series; and
M EF = 1 indicates that the model perfectly fits to the time series. Models with negativeMEF can still be useful, if r is positive. Negative MEF values could be dueto differences in magnitude between simulated and observed values, despitethe simulations following the same trend as the observations. Finally, RI is ameasure of the difference in magnitude between simulated and reference values.It varies between 1 to + ∞ and a value of 1.5 would indicate that the modelsimulates the reference values with a 50% difference in magnitude.
3. Results and Discussion
The estimated biomass for capelin in 1984 was 4 681 408 tonnes, . timeshigher than MFRI estimates for that year (Institute, 2014). These results areconsistent with a previous study by Magnússon and Pálsson (1989), in which theauthors hypothesised that capelin biomass was being grossly underestimated inthe assessments and should be approximately doubled in order to explain theamount of capelin that is being consumed by predators.The combined predation mortality of cod on capelin accounted for of M of capelin ( M = 0 . ; total M = 1 . ). Blue whiting had the second highesteffect on capelin M , with its predation mortality on capelin accounting for of M ( M = 0 . ). 9 .2. Dynamic model Overall, the Ecosim model performed in a satisfactory manner when repli-cating previously known dynamics (Table 6 and Figures 2 and 3). The anomalysearch using 10 spline points resulted in the fit with the lowest AIC Table 5.The anomaly search on the final model was thus done using 10 spline points.When fitting the model to the time series of biomass and landings by estimatingthe vulnerability parameters, the original fit (
SSQ = 84 . ) was improved by4% ( SSQ = 81 . ). After the addition of temperature as a forcing function ofPP to the time series of biomass, F and landings, the original fit was improvedby 16% ( SSQ = 71 . ). Considering these results, temperature was included inthe final model as a forcing function of PP. Table 5: Summary of SSQ and AIC using different number of spline points.Spline points SSQ AIC0 75.52 -880.45 75.01 -944.510 71.04 -961.415 72.17 -942.420 68.53 -957.5
The best fit was achieved when annual temperature was included in themodel, suggesting that fisheries alone cannot explain the Icelandic marine ecosys-tem. Temperature is considered to be one of the main determining factors inmarine fish distribution (Sunday et al., 2012; Genner, 2016). Cod has beenshown to change its habitat use in the southern North sea according to surfacetemperature (Freitas et al., 2016). Considering the fundamental role tempera-ture plays in fish ecology, it is not surprising that the addition of temperatureto the model resulted in a better fit. 10 able 6: Skill assessment per group, where r is the Pearson’s correlation coefficient, R isthe determination coefficient, MEF is the modelling efficiency and RI is the reliability index.Reference values for a good fit: . ≤ r ≤ ; . ≤ R ≤ ; < MEF ≤ ; ≤ RI ≤ . .Functional group Biomass Catchesr R MEF RI r R MEF RICod 4+ 0.605 0.366 -1.079 1.291 0.810 0.656 0.172 1.291Saithe 4+ 0.750 0.563 0.599 1.242 0.844 0.712 0.479 1.242Haddock 3+ 0.768 0.590 0.635 1.241 0.623 0.388 0.471 1.241Herring 4+ 0.323 0.104 -0.223 1.398 0.577 0.333 0.047 1.398Redfish 8+ 0.567 0.321 0.147 1.230 0.796 0.634 0.279 1.230Greenland halibut 6+ 0.188 0.035 -0.354 1.909Capelin 0.763 0.583 -0.419 2.006Blue whiting 0.535 0.286 -0.012 1.909 0.929 0.863 0.424 1.755Mackerel 1.000 1.000 1.000 1.000Flatfish 0.778 0.605 0.391 1.238Other codfish 0.754 0.568 0.506 1.251Comercial demmersal 0.687 0.472 -0.444 1.318
The skill assessment had overall satisfactory results ( . ≤ r ≤ ; . ≤ R ≤ ; < M EF ≤ ; ≤ RI ≤ . ; Table 6). There was a general tendencyfrom the model to replicate the landings better than the biomass. With theexception of group haddock 3+, fitted biomass had higher R values than fittedlandings (Table 6). These results are unsurprising, as the landings were actualobservations, while the biomass used were previous estimates by the MFRI(Institute, 2014).There were three groups (cod 4+, herring 4+ and blue whiting) for whichthe model had a worse fit for biomass than a straight line through the averageof the reference values ( M EF < ; Table 6). The model performed poorly whenpredicting the landings of Greenland halibut 6+, capelin and commercial dem-ersal fish as well ( M EF < ; Table 6). The MEF is sensitive to the differencebetween the simulated and the reference values (Eq. 4) and can be negativeeven when the R is within reasonably good range ( . ≤ R ≤ ), as was thecase of capelin landings (Table 6). The model underestimated capelin landingswith a difference in magnitude of approximately 100% ( RI = 2 . ; Table 6),which could account for the low MEF value.The model was calibrated using biomass estimates from single-species assess-ments, as well as F calculated using those estimates. This method was chosen, asthere were no biomass estimates from multi-species assessments available andbecause of how Ecopath internally computes F as Y /B . The use of biomassestimated by single species assessments in a multi-species context could havecontributed to the poor skill assessments results observed for herring 4+ andredfish 8+ biomass, as well as for Greenland halibut 6+ landings.11 igure 2: Biomass estimates by year for all groups and reference biomass time-series. Esti-mates by Ecosim are represented by a solid line and time-series reference data is representedby dots. Biomass is thousands of tonnes. Notice the different scales.
With the exception of mackerel, all groups had
RI > . , which indicatesthat the simulated biomass and landings varied at least 20% in magnitude inrelation to the reference biomass and landings (Table 6). However, none of thesedifferences in magnitude were constant through the simulations, indicating thatthere were no systematic errors (Figures 2 and 3).The skill assessment of mackerel caches shows optimal results for all mea-sures (Table 6). These can be explained by the use of forced biomass for thisgroup. Using forced biomass for mackerel in EwE can have consequences forthe modelled trophic dynamics, as it could lead to over-fitting. Over-fittingthe model can, in turn, lead to unnatural simulations of the groups with directtrophic links to this group. However, the use of fisheries and temperature wasnot enough to satisfactorily simulate the invasion by mackerel.12 igure 3: Landings estimates by year for all groups and reference landings time-series. Esti-mates by Ecosim are represented by a solid line and time-series reference data is representedby dots. Catches are in thousands of tonnes. Notice the different scales. Ecosim uses a vulnerability scale of 1 to + ∞ , where v = 2 represents mixed-trophic control, v < represents bottom-up control and v > , representstop-down control (Christensen et al., 2008). The higher the vulnerability, thestronger top-down effect a predator has on its prey, this being relevant for valuesup to v = 100 .Of the groups with vulnerability parameters freed from default, only cod 4+had v < (Table 7), suggesting that this group is affected by the abundanceand/or availability of their prey, as contrast to acting as a controlling forceover them. The groups capelin, blue whiting and commercial demersal had v = 1 . , v = 2 . and v = 2 . , respectively (Table 7), which suggests thatthese groups have mixed trophic control. The remaining groups had v > ,suggesting top-down control from these groups over their prey (Table 7). Onlyhigh TL predators ( T L > ; Figure 4) had their vulnerability parameters freedfrom default (Table 7), as these were the only groups with time-series of biomassand/or landings available, and these usually are linked to their prey throughtop-down control. As such, it is not possible to speculate about which type oftrophic control dominates the Icelandic marine ecosystem from this model.13 igure 4: Flow diagram for the Ecopath model of Icelandic waters in 1984 showing the dis-tribution of the functional groups by trophic level. Larger nodes represent bigger stock sizes.The trophic level scale can be found on the left of the diagram. The vulnerability search was done by predator. When freeing vulnerabilitiesfrom default in this manner, it is assumed that how vulnerable a prey is to theirpredator depends mostly on the predator’s foraging behaviour (Christensen andPauly, 1992). Conducting the vulnerability search by predator will thus lead toone vulnerability value for every predator-prey interaction of a given predator(Christensen and Pauly, 1992). Assuming that a predator’s foraging behaviourremains unchanged independently of the prey sought might not always depictreality. An example of this is cod, which was estimated to be affected by bottom-up control ( v < ; Table 7) by this model. However, a previous study hasshown that cod has top-down control over shrimp, but is affected by bottom-upcontrol by capelin (Stefansson et al., 1998). The vulnerability search methodused prevents individual predator/prey pairs from having a unique vulnerabilityvalue. The vulnerability search was done in this manner as it was the only optionavailable to restrict it to the groups with time series. Table 7: Estimated vulnerabilities ( v )for predator.Predator v Functional group v Cod 4+ 1.186 Capelin 1.979Saithe 4+ > Blue whiting 2.156Haddock 3+ > Mackerel > Herring 4+ > Flatfish > Redfish 8+ > Other codfish > Greenland halibut 6+ > Commercial demersal 2.000 .3. Comparisson with other models There were two previously published Ecopath models describing the Ice-landic marine ecosystem in 1950 and 1997 (Buchary, E.A.; Mendy, A.N andBuchary, E.A., respectively, in Guénette et al., 2001; Mendy, 1998). The modelfor 1950 included an Ecosim simulation, ran from 1950 to 1997.In the 1950 study, the author constructed the model to be comparable tothe Ecopath model for 1997 (Buchary, E.A. in Guénette et al., 2001). The1950 model Ecopath was an attempt to reconstruct the ecosystem for thatyear (Buchary, E.A. in Guénette et al., 2001; Mendy, 1998), by estimating thebiomass for most functional groups. The author then used a time series of Fand reference biomass (calculated as Y/F) of cod and herring to fit the model(Buchary, E.A. in Guénette et al., 2001). No vulnerability search was done inthis study and all vulnerabilities were set to 0.3 (Buchary, E.A. in Guénetteet al., 2001). There were differences between the input biomass and landingsof the 1997 model (Mendy, A.N and Buchary, E.A. in Guénette et al., 2001;Mendy, 1998) and the simulations by the present model for the same year formost groups. It is reasonable to think that these discrepancies were due to dif-ferent modelling assumptions such as modelled area and species density, as thedifferences are also found between the input data of the older model and thedata sources for the current model.Both the 1950 and 1997 models used 25 functional groups each (Buchary,E.A.; Mendy, A.N and Buchary, E.A., respectively, in Guénette et al., 2001;Mendy, 1998) in contrast to the 35 functional groups of the here presentedmodel. The current model was more complete than the 1950 and 1997 models,as it included more detailed functional groups, a more complete time series thanthe one used in the 1950 model and because of it being fitted to the time seriesthrough vulnerability and anomaly searches. Furthermore, a skill assessmentwas carried out in the current model, which was not the case for the previousmodels.
It was estimated by this model that commercial fishes like haddock, redfish,Greenland halibut, flatfish and other codfish have top-down control over theirprey. These being predators implies that their more or less heavy removal fromthe system could have serious implications for the ecosystem dynamics andshould be taken into account by fisheries management authorities. Furthermore,mid trophic level species like herring and mackerel were also estimated to havetop-down control over their prey. These species have a direct trophic link tolower trophic level organisms and their exploitation should surely take this intoaccount.The use of EE to estimate biomass in Ecopath, might raise concerns, asit is inversely related to biomass. An alternative to using EE could be usingspecies density from other EwE models in neighbouring areas. However, usingthis method would imply assuming that either the marine ecosystem in Ice-land would have the same physical, hydrological and biological dynamics as the15eighbouring areas or that species density is independent of these. This wouldbe an assumption with potentially serious consequences, as the distribution ofspecies in marine systems is thought to be linked to biotic and abiotic factors(Genner, 2016).This model provides insight on the marine trophic web in Iceland and itcould be used as a tool to investigate how different fishing policies could impactthe trophic web dynamics in Icelandic waters.
4. Acknowledgements
This study has received funding from the European Union’s Seventh Frame-work Programme for research, technological development and demonstrationunder grant agreement no. 613571 for the project MareFrame and from the Eu-ropean Commission’s Horizon 2020 Research and Innovation Programme underGrant Agreement No. 634495 for the project Science, Technology, and SocietyInitiative to minimize Unwanted Catches in European Fisheries (Minouw), aswell as from the Icelandic Research Fund (Rannis, No. 152039051).Furthermore, the authors would like to acknowledge the MFRI for providingthe data that made the work presented here possible, Maciej T. Tomczak, in pro-viding useful comments on an early version of the paper and Villy Christiansenfor giving helpful input on the time series fitting.
ReferencesReferences
Ahrens, R. N. M., Walters, C. J., Christensen, V., 2012. Foraging arena theory.Fish and Fisheries 13 (1), 41–59.URL
Akaike, H., 1974. New look at statistical-model identification. Ieee Transactionson Automatic Control AC19 (6), 716–723.URL
Astthorsson, O. S., Gislason, A., Gudmundsdottir, A., 1994. Distribution, abun-dance and length of pelagic juvenile cod in icelandic waters in relation to en-vironmental conditions. In: ICES Marine Science Symposia. Vol. 198. Copen-hagen, Denmark: International Council for the Exploration of the Sea, 1991-,pp. 529–541.Astthorsson, O. S., Gislason, A., Jonsson, S., 2007. Climate variability and theicelandic marine ecosystem. Deep Sea Research Part II: Topical Studies inOceanography 54 (23), 2456–2477.Astthorsson, O. S., Valdimarsson, H., Gudmundsdottir, A., Óskarsson, G. J.,2012. Climate-related variations in the occurrence and distribution of mack-erel (scomber scombrus) in icelandic waters. ICES Journal of Marine Science:Journal du Conseil 69 (7), 1289–1297.16orchers, D. L., Pike, D. G., Gunnlaugsson, T., Víkingsson, G., 2009. Minkewhale abundance estimation from the nass 1987 and 2001 aerial cue–countingsurveys taking appropriate account of distance estimation errors. NAMMCOScientific Publications 7, 95–110.Canning, S. J., Santos, M. B., Reid, R. J., Evans, P. G. H., Sabin, R. C., Bai-ley, N., Pierce, G. J., 2008. Seasonal distribution of white-beaked dolphins(lagenorhynchus albirostris) in uk waters with new information on diet andhabitat use. Journal of the Marine Biological Association of the United King-dom 88 (6), 1159–1166.URL
Christensen, V., Pauly, D., 1992. Ecopath-ii - a software for balancing steady-state ecosystem models and calculating network characteristics. EcologicalModelling 61 (3-4), 169–185.URL
Christensen, V., Walters, C., Pauly, D., Forrest, R., 2008. Ecopath with ecosimversion 6 user guide. Lenfest Ocean Futures Project, 235.Christensen, V., Walters, C. J., 2004. Ecopath with ecosim: methods, capabili-ties and limitations. Ecological Modelling 172 (2-4), 109–139.URL
Cook, R. M., Sinclair, A., Stefansson, G., 1997. Potential collapse of north seacod stocks. Nature 385 (6616), 521–522.URL
Dommasnes, A., Christensen, V., Ellertsen, B., Kvamme, C., Melle, W., Nottes-tad, L., Torstein, P., Tjelmeland, S., Zeller, D., 2001. An ecopath model forthe norwegian sea and barents sea. Fish. Cent. Res. Rep. 9 (4), 213–240.Freitas, C., Olsen, E. M., Knutsen, H., Albretsen, J., Moland, E., 2016.Temperature-associated habitat selection in a cold-water marine fish. Journalof Animal Ecology 85 (3), 628–637.URL
Froese, R., Pauly, D., 2016. Fishbase.URL
Genner, M. J., 2016. Staying out of the heat: how habitat use is determined bylocal temperature. Journal of Animal Ecology 85 (3), 611–613.URL
Guénette, S., Christensen, V., Pauly, D., 2001. Fisheries impacts on north at-lantic ecosystems: Models and analyses.URL https://open.library.ubc.ca/cIRcle/collections/52383/items/1.0348145
Hauksson, E., Bogason, V., 1997. Comparative feeding of grey (halichoerusgrypus) and common seals (phoca vitulina) in coastal waters of iceland, witha note on the diet of hooded (cystophora cristata) and harp seals (phocagroenlandica). J. Northwest Atl. Fish. Sci 22, 125–135.Heymans, J. J., Coll, M., Link, J. S., Mackinson, S., Steenbeek, J., Walters, C.,Christensen, V., 2016. Best practice in ecopath with ecosim food-web modelsfor ecosystem-based management. Ecological Modelling 331, 173–184.URL
Hutchings, J. A., 2002. Ecology and biodiversity of commercially unexploitedmarine fishes in the northwest atlantic. Final Report, Dalhousie University,Halifax, Nova Scotia.Institute, M. F. R., 2014. State of marine stocks in icelandic waters 2013/2014- prospects for the quota year 2014/2015.Jonsson, S., Valdimarsson, H., 2005. Recent developments in oceanographicresearch in icelandic waters. In: Developments in Quaternary Sciences. Vol. 5.Elsevier, pp. 79–92.Kleisner, K., Hoornaert, C., 2016. Primary production.URL
Lilliendahl, K., Solmundsson, J., 1997. An estimate of summer food consump-tion of six seabird species in iceland. Ices Journal of Marine Science 54 (4),624–630.URL
MacNeil, M. A., McMeans, B. C., Hussey, N. E., Vecsei, P., Svavarsson, J.,Kovacs, K. M., Lydersen, C., Treble, M. A., Skomal, G. B., Ramsey, M.,Fisk, A. T., 2012. Biology of the greenland shark somniosus microcephalus.Journal of Fish Biology 80 (5), 991–1018.URL
Magnússon, K. G., Pálsson, l. K., 1989. Trophic ecological relationships of ice-landic cod. In: ICES Rapp. Proc.-Verb. Vol. 188. pp. 206–224.Malmberg, S., Jonsson, S., 1997. Timing of deep convection in the greenlandand iceland seas. Ices Journal of Marine Science 54 (3), 300–309.URL
Petursdottir, H., Falk-Petersen, S., Gislason, A., 2012. Trophic interactions ofmeso- and macrozooplankton and fish in the iceland sea as evaluated by fattyacid and stable isotope analysis. Ices Journal of Marine Science 69 (7), 1277–1288.Pike, D. G., Paxton, C. G., Gunnlaugsson, T., Víkingsson, G., 2009. Trends inthe distribution and abundance of cetaceans from aerial surveys in icelandiccoastal waters, 1986-2001. NAMMCO Scientific Publications 7, 117–142.Pikitch, E. K., Santora, C., Babcock, E. A., Bakun, A., Bonfil, R., Conover,D. O., Dayton, P., Doukakis, P., Fluharty, D., Heneman, B., Houde, E. D.,Link, J., Livingston, P. A., Mangel, M., McAllister, M. K., Pope, J., Sains-bury, K. J., 2004. Ecosystem-based fishery management. Science 305 (5682),346–347.URL
Popescu, I., Poulsen, K., 2012. Icelandic fisheries: a review. Policy DepartmentB: Structural and Cohesion Policies Fisheries Note, European Parliament.R Core Team, 2016. R: A Language and Environment for Statistical Computing.R Foundation for Statistical Computing, Vienna, Austria.URL
Sigurjónsson, J., Víkingsson, G. A., 1997. Seasonal abundance of and estimatedfood consumption by cetaceans in icelandic and adjacent waters. Journal ofNorthwest Atlantic Fishery Science 22, 271–287.Stefansson, G., Rosenberg, A. A., 2005. Combining control measures for moreeffective management of fisheries under uncertainty: quotas, effort limitationand protected areas. Philosophical Transactions of the Royal Society of Lon-don B: Biological Sciences 360 (1453), 133–146.Stefansson, G., Skuladottir, U., Steinarsson, B. A. E., 1998. Aspects of theecology of a boreal system. Ices Journal of Marine Science 55, 859–862.Stow, C. A., Jolliff, J., McGillicuddy, D. J., Doney, S. C., Allen, J. I., Friedrichs,M. A., Rose, K. A., Wallhead, P., 2009. Skill assessment for coupled biolog-ical/physical models of marine systems. Journal of Marine Systems 76 (1),4–15. 19turludottir, E., Desjardins, C., Elvarsson, B., Fulton, E. A., Gorton, R., Loge-mann, K., Stefansson, G., 2018. End-to-end model of icelandic waters usingthe atlantis framework: Exploring system dynamics and model reliability.Fisheries Research 207, 9–24.Sunday, J. M., Bates, A. E., Dulvy, N. K., 2012. Thermal tolerance and theglobal redistribution of animals. Nature Climate Change 2 (9), 686–690.URL
Tyler, H. R., Pearcy, W. G., 1975. Feeding habits of 3 species of lanternfishes(family myctophidae) off oregon, usa. Marine Biology 32 (1), 7–11.URL
Umhverfisstofnun, 2016. Veiditolur.URL http://ust.is/?PageId=2ab56cba-246d-11e4-93bd-00505695691b
Vikingsson, G. A., Elvarsson, B. P., Olafsdottir, D., Sigurjonsson, J., Chosson,V., Galan, A., 2014. Recent changes in the diet composition of common minkewhales (balaenoptera acutorostrata) in icelandic waters. a consequence of cli-mate change? Marine Biology Research 10 (2), 138–152.URL
Víkingsson, G., Pike, D. G., Desportes, G., Øien, N., Gunnlaugsson, T., Bloch,D., 2009. Distribution and abundance of fin whales (balaenoptera physalus) inthe northeast and central atlantic as inferred from the north atlantic sightingssurveys 1987–2001. NAMMCO Sci. Publ 7, 49–72.Víkingsson, G. A., Ólafsdóttir, D., Sigurjónsson, J., 2003. Geographical, andseasonal variation in the diet of harbour porpoises (phocoena phocoena) inicelandic coastal waters. NAMMCO Scientific Publications 5, 243–270.Walters, C., Christensen, V., Pauly, D., 1997. Structuring dynamic models ofexploited ecosystems from trophic mass-balance assessments. Reviews in FishBiology and Fisheries 7 (2), 139–172.URL
Walters, C., Pauly, D., Christensen, V., Kitchell, J. F., 2000. Representing den-sity dependent consequences of life history strategies in aquatic ecosystems:Ecosim ii. Ecosystems 3 (1), 70–83.URL
Alca tordaFratercula arcticaFulmarus GlacialisRissa tridactylaUria aalgeUria lomvia
Marine Mammals Minke whale
Balaenoptera acutorostrata
Baleen whales
Balaenoptera borealisBalaenoptera musculusBalaenoptera physalusMegaptera novaeangliae
Tooth whales
Globicephala melasHyperoodon ampullatusLagenorhyncus acutusLagenorhyncus albirostrisOrcinus orcaPhocoena phocoenaPhyseter macrocephalus
Pinniped
Halichoerus grypusPhoca vitulina
Fish Greenland shark
Somniosus microcephalus
Dogshark
Apristurus aphyodesApristurus laurussoniiCentroscyllium fabriciiEtmopterus princepsSqualus acanthias
Skate rays
Raja sp.
Cod*
Gadus morhua
Saithe*
Pollachius virens
Haddock*
Melanogrammus aeglefinus
Herring*
Clupea harengus
Redfish*
Sebastes norvegicusSebastes mentellaSebastes viviparus
Greenland halibut*
Reinhardtius hippoglossoides
Capelin
Mallotus villosus
Blue whiting
Micromesistius poutassou
Mackerel
Scomber scombrus
Sandeel
Ammodytes sp.
Large pelagic
Magnisudis atlanticaSerrivomer beanii
Small pelagic MyctophidaeParalepididaeFlatfish
Glyptocephalus cynoglossusHippoglossoides platessoidesHippoglossus hippoglossusLimanda limandaMicrostomus kittPleuronectes platessa
Other codfish
Brosme brosmeMerlangius merlangus ype Functional Group Species Molva dipterygiaMolva molva
Commercial demersal
Anarhicas sp.Cyclopterus lumpusLophius piscatorius
Other demersal CottidaeLumpenidaeMacroridaeZoarcinaeMolluscs Cephalopod OctopodaTeuthidaSepiidaMollusc
Arctica islandica
BivalviaCrustaceans Lobster Nephropidae
Nephrops norvegicus
Shrimp Caridea
Pandalus sp.
Benthos Benthos AnomuraBrachyuraPolichaetaCnidarians Jellyfish ScyphozoaPlankton Krill EuphasiidaeCrustacean zooplankton AmphipodaCopepodaOstracodaSmall zooplankton ChaetognathaCiliophoraPhytoplankton AlgaeDetritus*Multi-stanza group able A.2: Input data for the balanced Ecopath model for Icelandic waters. Biomass andlandings are given in thousands of tonnes.Functional group Biomass P/B Q/B EE LandingsSeabirds 2.831 0.177 38.00 0.191Minke whale 81.65 0.042 6.340 0.985Baleen whale 189.5 0.060 4.000 5.449Tooth whale 26.84 0.063 15.00Pinniped 1.538 0.141 15.00 0.078Greenland shark 17.50 0.043 0.500 0.051Dogshark 0.320 4.834 0.500Skate rays 0.225 1.750 0.500 2.000CodCod 0-3 289.4 0.500 4.120 8.694Cod 4+ 914.0 0.520 1.850 282.0SaitheSaithe 0-3 116.9 0.500 9.687 0.088Saithe 4+ 287.0 0.520 4.760 63.00HaddockHaddock 0-2 119.9 0.900 9.794 0.872Haddock 3+ 147.8 0.900 4.960 48.00HerringHerring 0-3 953.2 0.570 10.82 0.723Herring 4+ 387.3 0.950 7.345 49.58RedfishRedfish 0-7 837.9 0.220 4.684Redfish 8+ 883.6 0.380 2.800 109.0Greenland halibutGreenland halibut 0-5 110.9 0.300 2.557Greenland halibut 6+ 220.0 0.340 1.400 30.21Capelin 1.262 4.800 0.950 865.0Blue whiting 1159 0.780 9.060 7.077Mackerel 151.6 0.390 4.400Sandeel 0.640 6.700 0.950Large pelagic 0.555 5.700 0.950Small pelagic 0.785 9.800 0.950Flatfish 0.326 3.420 0.950 13.57Other codfish 0.577 2.300 0.950 9.856Commercial demersal 431.7 0.712 1.400 24.67Other demersal 0.265 3.100 0.950Cephalopod 1995 2.440 12.00Mollusc 1.500 5.000 0.950 15.58Lobster 0.354 5.850 0.950 2.459Shrimp 1.250 5.000 0.950 24.42Benthos 1.500 9.750 0.950Jellyfish 3.000 10.00 0.950Krill 2.500 15.00 0.950Crustacean Zooplankton 4.000 15.00 0.950Small Zooplankton 13.00 25.00 0.950Phytoplankton 12151 117.6Detritus 0.027 able A.3: Output for the Ecopath model for Icelandic waters. Biomass is given inthousands of tonnes.Functional group Trophic level Biomass EE Q/BSeabirds 4.160 2.831 0.379 0.005Minke whale 4.261 81.65 0.287 0.007Baleen whale 3.333 189.5 0.479 0.015Tooth whale 4.666 26.84 0.000 0.004Pinniped 4.934 1.538 0.360 0.009Greenland shark 4.778 17.50 0.068 0.086Dogshark 4.142 13.05 0.500 0.066Skate rays 3.897 220.30 0.500 0.129CodCod 0-3 4.036 289.4 0.752 0.121Cod 4+ 4.179 914.0 0.817 0.281SaitheSaithe 0-3 3.913 116.9 0.927 0.052Saithe 4+ 4.021 287.0 0.950 0.109HaddockHaddock 0-2 3.657 119.9 0.631 0.092Haddock 3+ 3.760 147.8 0.975 0.181HerringHerring 0-3 3.223 953.2 0.913 0.053Herring 4+ 3.250 387.3 0.948 0.129RedfishRedfish 0-7 3.481 837.9 0.577 0.047Redfish 8+ 3.668 883.6 0.648 0.136Greenland halibutGreenland halibut 0-5 4.160 110.9 0.242 0.117Greenland halibut 6+ 4.370 220.0 0.404 0.243Capelin 3.250 4681 0.950 0.263Blue whiting 3.523 1159 0.220 0.086Mackerel 3.248 151.6 0.776 0.089Sandeel 3.250 3030 0.950 0.096Large pelagic 3.250 68.76 0.950 0.097Small pelagic 3.250 2193 0.950 0.080Flatfish 4.118 279.1 0.950 0.095Other codfish 4.235 470.2 0.950 0.251Commercial demersal 3.779 431.7 0.386 0.508Other demersal 3.416 2861 0.950 0.085Cephalopod 3.151 1995 0.420 0.203Mollusc 2.000 481.1 0.950 0.300Lobster 3.359 1784 0.950 0.060Shrimp 2.853 4217 0.950 0.250Benthos 2.248 38230 0.950 0.154Jellyfish 3.150 22.48 0.950 0.300Krill 2.250 39521 0.950 0.167Crustacean Zooplankton 2.164 14741 0.950 0.267Small Zooplankton 2.000 16707 0.950 0.520Phytoplankton 1.000 12151 0.656Detritus 1.000 20466 0.471able A.3: Output for the Ecopath model for Icelandic waters. Biomass is given inthousands of tonnes.Functional group Trophic level Biomass EE Q/BSeabirds 4.160 2.831 0.379 0.005Minke whale 4.261 81.65 0.287 0.007Baleen whale 3.333 189.5 0.479 0.015Tooth whale 4.666 26.84 0.000 0.004Pinniped 4.934 1.538 0.360 0.009Greenland shark 4.778 17.50 0.068 0.086Dogshark 4.142 13.05 0.500 0.066Skate rays 3.897 220.30 0.500 0.129CodCod 0-3 4.036 289.4 0.752 0.121Cod 4+ 4.179 914.0 0.817 0.281SaitheSaithe 0-3 3.913 116.9 0.927 0.052Saithe 4+ 4.021 287.0 0.950 0.109HaddockHaddock 0-2 3.657 119.9 0.631 0.092Haddock 3+ 3.760 147.8 0.975 0.181HerringHerring 0-3 3.223 953.2 0.913 0.053Herring 4+ 3.250 387.3 0.948 0.129RedfishRedfish 0-7 3.481 837.9 0.577 0.047Redfish 8+ 3.668 883.6 0.648 0.136Greenland halibutGreenland halibut 0-5 4.160 110.9 0.242 0.117Greenland halibut 6+ 4.370 220.0 0.404 0.243Capelin 3.250 4681 0.950 0.263Blue whiting 3.523 1159 0.220 0.086Mackerel 3.248 151.6 0.776 0.089Sandeel 3.250 3030 0.950 0.096Large pelagic 3.250 68.76 0.950 0.097Small pelagic 3.250 2193 0.950 0.080Flatfish 4.118 279.1 0.950 0.095Other codfish 4.235 470.2 0.950 0.251Commercial demersal 3.779 431.7 0.386 0.508Other demersal 3.416 2861 0.950 0.085Cephalopod 3.151 1995 0.420 0.203Mollusc 2.000 481.1 0.950 0.300Lobster 3.359 1784 0.950 0.060Shrimp 2.853 4217 0.950 0.250Benthos 2.248 38230 0.950 0.154Jellyfish 3.150 22.48 0.950 0.300Krill 2.250 39521 0.950 0.167Crustacean Zooplankton 2.164 14741 0.950 0.267Small Zooplankton 2.000 16707 0.950 0.520Phytoplankton 1.000 12151 0.656Detritus 1.000 20466 0.471