The necessity of a holistic approach when managing marine mammal–fisheries interactions: Environment and fisheries impact are stronger than seal predation

Seal populations are recovering in many regions around the world and, consequently, they are increasingly interacting with fisheries. We used an Ecopath with Ecosim model for the offshore Central Baltic Sea to investigate the interactions between the changes in fish stocks and grey seal (Halichoerus grypus) population under different fishing and environmental scenarios for the twenty-first century. The assumed climate, eutrophication and cod (Gadus morhua) fisheries scenarios modified seal predation impacts on fish. Fish biomass and catches are more affected by fishing mortality and the environment than by seal predation. Our results highlight that the impacts of the increasing seal population on lower trophic levels are complex; thus, we emphasize the need to consider a range of possible ecosystem contexts when evaluating potential impacts of top predators. Finally, we suggest that an increasing seal population is not likely to hinder the preservation of the main Baltic fish stocks. Electronic supplementary material The online version of this article (10.1007/s13280-018-1131-y) contains supplementary material, which is available to authorized users.

Departing from the existing Baltic Proper Ecopath model built to describe the Baltic open sea area food-web dynamics between 1974-2006Tomczak et al. 2013), we divided the seal group into juvenile and adult groups using the 'multi-stanza' representation (Gales et al. 2003;Walters et al. 2010) and defined new Ecopath parameters and diet composition for these two groups. The split was necessary as juveniles and adults of grey seal have different energetic requirements (Innes et al. 1987;Hammill and Stenson 2000) and their diets differ in the prey size and proportion (Lundström et al. 2007;Lundström et al. 2010;Hanson et al. 2017). We derived population biomass estimations from the number of counted seals during four decades  of surveys obtained from the Baltic Marine Environment Protection Commission (HELCOM) and from the Swedish Museum of Natural History (SMNH). We assumed that the number of counted seals in the monitoring reported by HELCOM was 85% of the actual population size (Hiby et al. 2007) and that the distribution during monitoring (grey seal molting period, i.e. May-June) was representative for the entire year. It was assumed that one third of the counted grey seal individuals in the Baltic Sea corresponds to juveniles and the rest to adults (Hårding et al. 2013). The average weight of a grey seal juvenile as considered for the model (i.e. < 60 months old) is 100 kg (Innes et al. 1987), while the weight range of a female adult is 100-180 kg and of a male adult is 180-300 kg (Perrin et al. 2009); thus, we considered the average weight of an adult in the model to be 200 kg. We included new data on the diet and biological parameters corresponding to grey seals from the North Atlantic (data from the Baltic seals were preferentially used, when available; Table S1).
The Electivity (selection index) describes a predator's preference for prey. It scales from -1 to 1; where -1 indicates total avoidance of a prey; 0 indicates that a prey is taken in proportion to its abundance in the ecosystem; and 1 indicates total preference for a prey. The electivity index as displayed in Fig. S1 is the standardized forage ratio of Chesson (1983).
The ecotrophic efficiency (EE) is a dimensionless parameter representing the fraction of the production that is used in the system, i.e. either passed up the food web, used for biomass accumulation, migration or export. Ecotrophic efficiency varies between 0 and 1 and can be expected to approach 1 for groups with considerable predation pressure.

Appendix S2. Ecosim parameters.
The 'Group info' parameters for the seal groups (Table S3) were adjusted so the feeding effects of the seals were maximized. Seals do not need to limit the time spent feeding to avoid predation in the Baltic, thus we gave 'maximum relative feeding time' values higher than the default value of 2 (meaning than that feeding time may double when prey became more scarce), i.e. 5 for the juvenile seal stanza and 10 for the adult seal stanza. As usually recommended for marine mammals, we set the 'feeding time adjustment rate' to 0.5, meaning that seals will adjust their foraging time to some extent based on food availability, to reach 'normal' individual consumption rates, i.e. as under conditions when the Ecopath model was estimated. QBmax/QBo values represent the upper limit of individual intake rates and are by default set to 1000 in Ecosim models (i.e., predators are not limited by handling times and can increase their intake greatly when prey densities increase). Here this value was set at 500, although preliminary tests did not show any differences in biomass projected with QBmax/QBo values between 100-1000, which is not surprising as this parameter only becomes relevant at a particular combination of very low predator density, very high prey density and high prey vulnerability. We set the 'predator effect on feeding time' parameter for seals to 0 (that is, no effect). Grey seals are considered opportunistic and thus they are thought to feed predominantly on the more available prey species (Murie and Lavigne 1992;Mark Jessopp et al. 2013).
In Ecosim, the temporal changes in the flow rates between available and unavailable prey behavioral states are calculated based on the "foraging arena" concept and are regulated by the predator-and prey-specific vulnerability parameters. Ecosim users can specify vulnerability multipliers (k ij ) which determine how much maximum predation mortality rate of prey i can increase with an increasing biomass of predator j. When vulnerability multipliers are high (k ij ≫ 2), the predators can increase their consumption of prey when their biomass increases, while at low values (k ij = 1) total prey consumption rates stay close to their Ecopath rates even when predator biomass increases. The former situation is usually referred to as 'top-down control' of the trophic interaction in EwE models and typically occurs when the prey has no refuge from predation and/or the predator is far from carrying capacity in the year which the Ecopath model describes and thus, has a large potential to increase its predation pressure on prey (Christensen and Walters 2004). Both criteria are fulfilled in the case of grey seals and their fish prey. Thus, we specified a value of k ij = 100 for this parameter for both juvenile and adult seals.

Appendix S3. Mixed Trophic Impact (MTI).
To estimate the direct and indirect effects (positive or negative) that changes of the biomass of one functional group might have on another, Ecopath uses the Network Analysis routine called Mixed Trophic Impact (MTI). It synthesizes all the effects that an infinitesimal change in the biomass of a group will have on the biomass of other groups in a system (Ulanowicz and Puccia 1990) and gives an ecosystem overview of the trophic interactions.
It is calculated from the difference between the proportion that group i contributes to the diet of group j, and the proportion that group i takes from the production of group j (Christensen and Walters 2004 where DC ij is the diet composition term expressing how much j contributes to the diet of i, and FC j,i is a host composition term giving the proportion of the predation on j that is due to i as a predator (or fishery).

Appendix S4. Assumptions and limitations of the model.
The main assumption in Ecopath and Ecosim models is that the modeled functional groups are related to the main components of the food-web. However, this structure is only a simplification of the ecosystem and not all food-web components are included in the model and thus these models present certain limitations. For instance, Plagányi and Butterworth (2004) discussed the implications of the foraging arena hypothesis. Furthermore, Aydin (2004) explored the implications of the "fixed growth efficiency" function in Ecosim represented by the P/Q rates, which implies that the model does not adequately account for changes in population energetics as a population's size and structure changes due to heavy fishing or release of fishing pressure. Also, as Coll et al. (2009) mentioned, Ecosim is dependent on the mass-balance (equilibrium) assumption of Ecopath. This infers the risk of encountering errors in the interpretation of the Ecosim results, when these are extrapolated far beyond the equilibrium for which Ecopath data are available (Mackinson et al. 1997;Walters et al. 1997). Moreover, Ecopath and Ecosim models present limitations when interpreting the outcomes for highly migratory species or stocks patchiness (see Christensen and Walters (2004)), such as herring and sprat. Thus, when interpreting the simulations, we carefully considered the assumptions and limitations underlying Ecopath and Ecosim models. Assumptions and limitations of the model used for this study in particular are discussed in detail in Tomczak et al. (2012). Moreover, Niiranen et al. (2012) conducted a study to specifically assess the uncertainties and sensitivity of the BaltProWeb model, in which our model is based. In their study, Niiranen et al. (2012) also assessed the potential uncertainty under different future conditions with the same fishery and climate scenarios as in our model.       Figure S1. Electivity index showing seal's preference for the modelled prey. It scales from -1 to 1; where -1 indicates total avoidance of a prey; 0 indicates that a prey is taken in proportion to its abundance in the ecosystem; and 1 indicates total preference for a prey. The electivity values are highlighted using a blue color scale. The electivity index displayed is the standardized forage ration of Chesson (1983). Reverse hypoxic area anomaly Primary production anomaly Phytoplankton production per biomass anomaly Reverse hypoxic area anomaly C Figure S2. Forcing functions on fishing mortality (F) for adult herring and sprat (A), August and spring water temperature anomalies and salinity anomaly (B), primary production and reversed hypoxic area anomalies (C), and cod reproductive volume and herring recruitment biomass anomalies (D). See Table 1 for reference and description of the forcing functions. Seal abundance Figure S4. Combinations of seal abundance (seal0, seal, seal50), cod fishing mortality (Fcod0.3, Fcod1) and environmental (Env0, Env1, Env2) scenarios for future  projections. See Table 2 for reference and description of each scenario. Figure S5. Pre-balance diagnostics (PREBAL) of the Baltic Proper Ecopath model plotting biomass estimates (t/km -2 ) on a log scale with functional groups ordered by trophic level from highest to lowest. Biomass spanned five orders of magnitude. PREBAL described by Link (2010) are a way to judge the quality of Ecopath models. Pre-bal criteria state that biomass estimates and P/B and Q/B ratios should increase with decreasing trophic level. According to Link (2010), consumption by a taxon should be more than production by that taxon, as it occurs in our case. Therefore, P should not exceed Q (P/Q < 1). Figure S6. Biomass (A), consumption by seals (B) and catches (C) of adult cod, herring and sprat under environment scenario combinations of Env0, Env1 and Env2, seal0, seal1 and seal100 seal biomass scenarios and cod fishing mortality scenarios Fcod = 0.3 (orange) and Fcod = 1 (blue).  Table S4 (upper panel), and MTI values for the year 2094 if seal consumed exclusively herring, sprat and cod within the studied system (i.e. redistributing all the 'import' diet proportionally into the modelled prey, that is Import diet = 0).