Modelling risk areas in the North Sea for blooms of the invasive comb jelly

Recent records of the invasive ctenophore Mnemiopis leidyi A. Agassiz, 1865 in the North Sea are a cause for concern due to the detrimental effects this invader has had on fish stocks in the Black and Caspian seas. The North Sea is a major fishing ground and has spawning and nursery areas for many important fish species. These may be affected by competition and predation from Mnemiopsis leidyi, so it is important to determine whether the species, having been introduced, is likely to become established and produce blooms. This study applies temperature, salinity, and food constraints to data from the GETM-ERSEM-BFM model to evaluate the suitability of the North Sea for survival and reproduction of this invasive species. Large parts of the North Sea were found to be suitable for Mnemiopsis leidyi reproduction in summer months, although in most areas the suitable time window would not allow completion of more than two life cycles. The highest risk areas were in southern coastal and estuarine regions and in the Skagerrak and Kattegat, due to a combination of high temperatures and high food concentrations. Importantly, food was found to limit winter survival and so may restrict the overwintering population. Continued monitoring of this species, especially in areas predicted to be at a high risk, will be essential to determine whether it is likely to become a problem in the North Sea.


Background and aims
Mnemiopsis leidyi A. Agassiz, 1865, referred to hereafter as M. leidyi, is commonly known as the sea walnut or warty comb jelly.It is a planktivorous ctenophore native to temperate and sub-tropical waters along the East coast of the Americas from 20˚N to 46˚S (Lehtiniemi et al. 2012;Boersma et al. 2007;Purcell et al. 2001).It was first recorded outside native areas in 1980, when it appeared in the Black Sea (Schaber et al. 2011), probably introduced via ballast water (Fuentes et al. 2010).It spread both within the Black Sea and to other seas in the Mediterranean basin (Fuentes et al. 2010;Schaber et al. 2011) and has had dramatic effects on the native ecosystems of the Black and Caspian seas (Shiganova et al. 2003).The subsequent collapse of fish stocks (in particular anchovies and sprat (Shiganova et al. 2003)) in these seas has been attributed both to M. leidyi predation on fish eggs and larvae, and competition with zooplanktivorous fish (Shiganova et al. 2003).Another contributing factor may have been the already depleted status of the fish stocks, which released food resources for M. leidyi that may not otherwise have been available (Boersma et al. 2007;Richardson et al. 2009;Hamer et al. 2011).The events in these seas make this invasive species a cause for concern in other seas where it has been introduced (Hamer et al. 2011), especially in the North Sea, which has important fish stocks and spawning grounds (Ellis et al. 2011) and also shares the depleted state of fish stocks of the Black Sea, especially in southern areas (Boersma et al. 2007;Fuentes et al. 2010).
In November 2005, M. leidyi was first recorded in the Baltic (Olveira 2007) where it is now believed to be widespread (Javidpour et al. 2009).The first records of its occurrence in the North Sea were in summer 2006 (Faasse and Bayha 2006;Boersma et al. 2007;Tendal et al. 2007;van Ginderdeuren et al. 2012), though it is likely it was present but undetected prior to this  (Faasse and Bayha 2006).M. leidyi has now spread along French, Belgian, and Dutch coasts and it was found in high numbers in some of the Zeeland estuaries during a recent survey, as part of the Interreg MEMO (Mnemiopsis Ecology and Modelling: Observation of an invasive comb jelly in the North Sea') project, in October 2012.Genetic analysis has shown that the Baltic and North Sea populations are distinct from the Black Sea ones and were introduced directly from native areas, not via the Black Sea (Reusch et al. 2010;Bolte et al. 2013).
In addition to the potential effects on North Sea fish stocks, M. leidyi could have other detrimental effects.As with blooms of other gelatinous species, one of the problems posed is that when high densities are reached they can clog fishing nets, making them hard to lift out of the water: M. leidyi blooms have been reported to cause this problem in Israel in 2009 (Fuentes et al. 2010).Gelatinous species can also block power plant water intake pipes (Richardson et al. 2009): in Scotland in 2010; Torness power plant had to be closed to clear jellyfish that had blocked the pipes (BBC 2011), and this has also been a problem in France (Valo 2012).
Due to these potential detrimental effects on the North Sea ecosystem and economy, it is important to establish the risk posed by M. leidyi in this area.Sampling carried out in the North Sea (Figure 2) gives an idea of the species' current distribution, but sampling is limited in spatial and temporal scope, and as a new invader this species may not have yet reached all environmentally suitable areas (for example, the UK coast).As such, we use the best available knowledge of the species' environmental requirements in combination with modelled temperature, salinity, and food fields to assess the suitability of the North Sea for this invasive species, in an attempt to predict whether and where it is likely to become established.

Environmental requirements
M. leidyi is a hermaphrodite ctenophore capable of self-fertilisation (Purcell et al. 2001).It is tolerant of a wide range of environmental conditions, which, in conjunction with its generalist diet (Purcell et al. 2001;Fuentes et al. 2010) and fast feeding, growth and reproductive rates (Purcell et al. 2001;Jaspers et al. 2011) contributes to its success as an invader (Salihoglu et al. 2011).
In its native range, M. leidyi is found between 2-32˚C and salinity 2-38 PSU (Purcell et al. 2001), and in invaded waters from 2-31˚C and salinity 3-39 PSU (Purcell et al. 2001, E. Antajan, Ifremer, Laboratoire Environnement Ressources de Boulogne-sur-Mer, France, pers. comm.).Temperature and salinity thus are not usually constraints on survival, as most seas fall within this range.However, its distribution is limited by winter temperatures in some places (Purcell et al. 2001;Boersma et al. 2007): low winter temperatures are thought to prevent it from overwintering in the Sea of Asov (Purcell et al. 2001).M. leidyi can also tolerate hypoxia, which gives it an advantage over less tolerant fish species in eutrophic environments (Purcell et al. 2001;Richardson et al. 2009;Kolesar et al. 2010).
M. leidyi reproduction requires more specific conditions: M. leidyi is thought to reproduce only where temperatures exceed 12˚C (Lehtiniemi et al. 2012) and salinity exceeds ~10 PSU (Lehtiniemi et al. 2012;Jaspers et al. 2011).Thus, blooms could be limited by temperature and salinity.In addition to sufficient food for adults (3 mg C m -3 ), sufficient densities of larval food groups (approximately 40 mg C m -3 of microplankton (Sullivan and Gifford 2007)) are necessary for the population to grow, and inadequate food can halt development of tentaculate into lobate life stages (Salihoglu et al. 2011).Salihoglu et al. (2011) found a strong influence of temperature and food concentrations on population growth rates, and the highest growth rates can only be found where neither condition is limiting (Purcell et al. 2001): i.e. 25 mg C m -3 of adult food, 90 mg C m -3 of larval food, and temperatures of 20˚C (Salihoglu et al. 2011;Lehtiniemi et al. 2012).Table 1 summarises the known constraints on M. leidyi survival and reproduction.
The M. leidyi life cycle is estimated to take between 16 and 40 days from egg to adult (Salihoglu et al. 2011), due to the sensitivity of metabolism and growth to temperature (Kremer 1994;Purcell et al. 2001).At 15˚C it is estimated to take 40 days, compared to 16 days at 30˚C (Salihoglu et al. 2011).

Methods
The GIS programs GRASS 6.4.2 and ArcMap 10.1 were used to assess 6 nautical mile grid cells of the North Sea to determine their suitability for M. leidyi survival and reproduction, using the literature-derived constraints for temperature, salinity and food discussed in the introduction (see Table 1).
For the years of the model hindcast (running from 1958-2011) for which M. leidyi has been recorded in the North Sea (2005)(2006)(2007)(2008)(2009)(2010)(2011), daily data were extracted from GETM-ERSEM-BFM for the variables temperature, salinity, and biomass of the main food groups of M. leidyi larvae and adults (in mg C m -3 ).The adult food groups extracted were omnivorous mesozooplankton and carnivorous mesozooplankton (0.2-20 mm), and the larval food groups were microzooplankton and microphytoplankton (diatoms, flagellates and dinoflagellates, < 0.2 mm) (Sullivan and Gifford 2004;Salihoglu et al. 2011).Daily concentrations for adult food types (AF) were summed and multiplied by a factor of 10 to account for underestimation of the model, found during validation for three locations in the North Sea (see van der Molen et al. 2013): 10 where B denotes biomass, OZ indicates omnivorous mesozooplankton, and CZ is carnivorous mesozooplankton.Daily concentrations for larval food (LF) groups (microplankton) were also summed: with MZ and MP indicating microzooplankton and microphytoplankton, respectively.

Survival and reproduction suitability layers
For each day in each year, binary daily layers were created by selecting cells with temperature, salinity, and mesozooplankton concentrations greater than the respective survival thresholds, using a spatial query.The daily layers were then summed across each year to find the number of days, N s , suitable for survival of M. leidyi in the year: with TS d (t) a step function representing the temperature suitability of day d for survival with value 0 or 1 below or above the threshold t = 2˚C, respectively.Similarly, Ss d (t) is a step function representing the salinity suitability of day d for survival with value 0 or 1 below or above the threshold t = 4.5 PSU, and As d (t) is a step function representing the adult food suitability of day d for survival with value 0 or 1 below or above the threshold t = 3 mg C m -3 .
The same approach was used to find the number of days per year, N r , suitable for reproduction: with TS d (t) a step function representing the temperature suitability of day d for reproduction with value 0 or 1 below or above the threshold t = 12˚C, respectively.Similarly, Ss d (t) is a step function representing the salinity suitability of day d for reproduction with value 0 or 1 below or above the threshold t = 10 PSU, and As d (t) is a step function representing the adult food suitability of day d for reproduction with value 0 or 1 below or above the threshold t = 3 mg C m -3 , and Ls d (t) is a step function representing the larval food suitability of day d for reproduction with value 0 or 1 below or above the threshold t = 40 mg C m -3 .
To give a better idea of the risk of M. leidyi blooms, the reproduction map was classified according to the number of life cycles that could be completed while the conditions were suitable for reproduction, using a life cycle length of 40 days, as at 15˚C (Salihoglu et al. 2011).

Reproductive suitability with temperature and food indices
To incorporate the increase in egg production and M. leidyi biomass with temperature (exponential), and food (linear), found both in the field and in the laboratory (Salihoglu et al. 2011;Purcell et al. 2001;Lehtiniemi et al. 2012), daily temperature and food values were reassigned to a reproductive suitability index of 0-10 (0 = unsuitable, 1 = fulfils minimum thresholds, to 10 = non-limiting).Food values were assigned using a linear function and temperature using a power function.The suitability indices for temperature and food and the binary index for salinity (below 10 = 0, above 10 = 1) were then multiplied for

10/
with LF d the larval food (microzooplankton and microphytoplankton) biomass of day d, LF opt the non-limiting larval food requirement (90 mg C m -3 ), LF min the minimum food requirement of larvae (40 mg C m -3 ), and γ a constant equalling 8 that is used to scale the index to 1-10.
The above processes were repeated for each year from 2005-2011, and the average and standard deviations calculated (Table 2).All maps show the average for the time series unless otherwise stated.

Sensitivity analysis
To account for uncertainty in the input food fields from ERSEM-BFM, the model was rerun with a range of correction factors for larval (1, 2, 5, and 10) and adult (1, 5, 10, and 15) food.These factors were chosen to reflect the potential bias for mesozooplankton obtained with a 1D version of the model (van der Molen et al. 2013).Adults also feed on macrozooplankton (copepods, fish larvae) (Kolesar et al. 2010) which are not included in ERSEM-BFM but can be expected to have spatial distributions similar to mesozooplankton, and could, for the current purpose, be represented by an increased mesozooplankton biomass.Observations of microzooplankton were not available for validation, but we expect a smaller bias as these species are closer in trophic level to phytoplankton in the ecosystem model, chlorophyll is better represented by the model, and microplankton concentrations are significantly higher than macrozooplankton concentrations.
Additionally, to account for the imperfect knowledge of the parameters constraining the species' survival and reproduction, the survival and reproduction thresholds for temperature, salinity, and food were varied to encompass the range of values found in the literature.One parameter was changed while holding others constant, and the deviation from the reference run calculated for key statistics.Tables 3 and 4 show the values used for the sensitivity analysis.

Species records
Records of where M. leidyi has been found from 2005 to present were obtained from the literature (Tendal et al. 2007;Faasse and Bayha 2006;Tulp 2006;Boersma et al. 2007;Riisgard et al. 2007;Antajan et al. 2010;Soenen et al. 2010;Riisgard et al. 2012;Van Ginderdeuren et al. 2012;Bolte et al. 2013;Haraldsson et al. 2013;Kellnreitner et al. 2013), surveys done at Cefas and as part of the MEMO project, and are displayed for each season in Figure 2A-D.Absences were included where available.These records, along with the sampling method and source, are presented in Table 1S in Supplementary material.

Survival
Most areas were found to be suitable for M. leidyi survival for between 200 and 300 days per year, with, on average, a maximum of 350 suitable days per year (Figure 3, Table 2).The Norwegian Trench, Skagerrak and the East Anglian plume were the most suitable areas, and Western Danish and German Waters had the lowest suitability.Temperature and salinity did not restrict M. leidyi survival in the North Sea: the whole area was above 2˚C and 4.5 PSU year round.Some areas did fall below mesozooplankton biomass of 3 mg C m -3 in the winter, which limited the suitability for M. leidyi survival.

Reproduction
The number of days per year suitable for M. leidyi reproduction ranged from 0 to a maximum of 212 (in 2011), with a mean of 55 (Table 2).Coastal and nearshore areas, especially the Skagerrak, Kattegat and Norwegian Trench, were suitable for longest period (Figure 4A).River mouths, such as the Thames estuary and the Rhine delta, were also suitable for a large proportion of the year.Larval food was the main limiting factor and the spatial pattern for larval food suitability was very similar to the overall pattern (Figure 4A and 4D).
On average, 57% of the study area was suitable for M. leidyi reproduction for longer than 40  days per year (the life cycle length at 15˚C) (Figure 4A).In most areas only one or two 40day cycles could have been completed, although some areas could host up to four 40-day life cycles (Figure 5).

Reproductive suitability
Figures 6 and 7 show the average suitability for M. leidyi reproduction for 2005-2011; taking into account the relationships between food and reproduction, and temperature and reproduction.
Reproductive suitability was highest in the Kattegat and along southern coasts; the Jutland Bank and German Bight were also suitable.Temperature restricted reproductive suitability at a broad scale, with areas in the South and West of the North Sea being more suitable for reproduction in terms of temperature (Figure 6B), and the coasts of Northern England and Scotland having low suitability for reproduction although sufficient food was predicted to be available (Figures 6C and 6D).However, as with the survival and minimum reproduction maps (Figures 3 and 4), larval food (Figure 6D) was the more restrictive constraint on suitability for reproduction and matched most closely the pattern in overall suitability.The suitability varied interannually, with 2006 being the most suitable year, but the patterns were consistent between years.Interannual variability was low in North-western and central areas and highest in the Thames estuary and along the continental coast (with the exception of the Netherlands) (Figure 6A).
Reproductive suitability peaked in mid to late summer in most areas (Figure 7).Coastal Dutch, German and Danish waters had early peaks and the central North Sea had the latest peak -in early autumn.

Sensitivity analysis
The model was insensitive to changes in the salinity and temperature survival thresholds but was affected by changes in adult food requirements (see Table 3).Decreasing this threshold to 1 mg C m -3 increased the average number days per year suitable for survival to 300, and the maximum to 365, but did not affect reproduction.
Lowering the temperature reproductive threshold had little effect, but increasing it to 16˚C reduced the area suitable for reproduction by 34%.Varying the larval food requirement by 5 mg C m -3 had a small effect on the suitability for reproduction.
Using an adult food correction factor of 15 resulted in some areas being suitable for survival year-round, with an average of 255 suitable days per year (Table 4).Changing the adult food correction factor had no impact on reproductive suitability as it was non-limiting.Increasing the larval food input by factors of 2, 5, and 10 increased the average number of days per year suitable for reproduction by 36, 68 and 73 respectively.Larval food was still the most limiting factor, although the role of temperature became more important when the correction factor was increased.

Survival
Our results suggest that winter conditions in the North Sea are not ideal for M. leidyi survival, as the number of suitable days per year is predicted to be below the survival thresholds for, on average, at least 15 days and generally more (Table 2, Figure 3).However, it is known that M. leidyi can survive North Sea winters as it has been found in consecutive years and during winter sampling (van Ginderdeuren et al. 2012;E. Antajan, personal communication), and it is unlikely that it is introduced afresh every year.Olveira (2007) found that M. leidyi can survive without food for at least 17 days, during which time specimens decreased in length by 43% on average.This may be long enough to enable M. leidyi to survive in restricted areas of the North Sea, such as the Rhine estuary, which, in several years, is suitable for the highest number of days, and does have winter records of M. leidyi (Figure 2).
The zooplankton concentration survival threshold was set at 3 mg C m -3 because it is the lowest level at which M. leidyi has been observed in the field (where this was also measured) (Kremer 1994;Purcell et al. 2001;Fuentes et al. 2010).So far, no experiments have investigated the survival of M. leidyi at very low food concentrations, but slowed metabolism at lower winter temperatures means food requirements are likely to be lower (Purcell et al. 2001) and, if M. leidyi could survive at very low food concentrations (e.g. 1 mg C m -3 , as used in the sensitivity analysis) for longer periods, the suitable overwintering area would be increased.It is also important to consider that winter food concentrations may be underestimated by the biogeochemical model; where the correction factor was increased this resulted in an increased number of days suitable for survival.There may also be additional food sources (i.e.fish larvae) available to M. leidyi that are not included in the biogeochemical model.This is supported by a parallel study of winter survival based on field data (David C, Vaz S, Loots C, Antajan E, van der Molen J, Travers-Trolet M, Alfred Wegener Institute, Bremerhaven, Germany, unpublished data), which found positive correlations with larval fish abundance in the North Sea.
Despite the limitations, our results suggest that winter conditions are not ideal for M. leidyi survival.This could help to explain why M. leidyi populations, and the condition of specimens found, have been so variable from year to year (Javidpour et al. 2009;Fuentes et al. 2010;Antajan, personal communication): a small change in the food availability in winter could have a large impact on overwintering populations.

Reproduction
Our results show that M. leidyi could reproduce in much of the North Sea during the summer months.Whether the conditions are suitable for long enough for blooms to occur is less certain.Although M. leidyi populations can increase rapidly under favourable conditions (Lehtiniemi et al. 2012), with an assumed life cycle length of 40 days (as at 15ºC) only one or two life cycles could be completed in the suitable time window in most areas, suggesting that large-scale blooms are unlikely except in the North Sea west of the Skagerrak.However, these results do not take into account the scaling of reproduction with temperature and food, so blooms may occur in areas with highly suitable food and temperature conditions, such as those highlighted in Figure 6A.
Existing records of M. leidyi M. leidyi has so far been found mostly in coastal waters and estuaries (Figure 2), which is consistent with the predictions of this study (Figure 6A).We know of some more central records from the winter IBTS (International Bottom Trawl Survey, E. Antajan, unpublished data), which do not match our prediction for winter survival, but these individuals are likely to have been advected by currents from more suitable overwintering areas, and are likely to only survive for a brief time until lack of food limits survival.
There are some regions that are predicted here to be suitable yet have no records of M. leidyi.For some areas -e.g. on the French coast of the Channel, to the authors' knowledge, there are no published records of surveys that could have detected the presence of M. leidyi.Other areas have been sampled, but sampling was typically sparse in space (tens of km between samples, limited survey areas) and time (one-off cruises, or annual cruises), and may have missed patches, low numbers or infrequent presence.Investigations using a particle tracking model suggest that low input from areas with M. leidyi presence may explain its absence in UK waters (David et al., unpublished data).However, even for areas where transport by currents is unlikely, ballast water transport could be an additional introduction pathway, and it is only with continued monitoring that we will discover when M. leidyi is present.

Overlap with spawning grounds and fish larvae
The North Sea has important stocks of Atlantic cod, Gadus morhua Linnaeus, 1758, and spawning is widespread across the area (Ellis et al. 2011).Cod larvae are found in high abundance along the Dutch and German coasts and in Norwegian offshore waters (Edwards et al. 2011), but this is earlier in the year than the time of peak M. leidyi bloom risk (Figure 7).However, coastal and estuarine areas are used as nursery grounds by cod (Ellis et al. 2011), so M. leidyi may pose a threat to cod through competition with larvae or small juveniles for food (Jaspers et al. 2011).
Atlantic herring, Clupea harengus Linnaeus, 1758, spawn in spring and autumn in coastal areas, including the Greater Thames Estuary and the English Channel (Ellis et al. 2011), and these do overlap with areas of high risk for M. leidyi reproduction, although it is the autumn spawners that are most likely to overlap temporally.There is also an overlap with spawning grounds for European plaice, Pleuronectes platessa Linnaeus, 1758, and whiting, Merlangius merlangus (Linnaeus, 1758) (Ellis et al. 2011), and with areas of high clupeid larval abundance on the UK coast and in the southern North Sea (Edwards et al. 2011).Dietary overlap with larval clupeids could also result in competition, though this is likely to be low due to low temporal overlap (Kellnreitner et al. 2013) Lesser sandeels, Ammodytes marinus Raitt, 1934, are one of the most abundant mid-trophic species in the North Sea ecosystem, playing a key role both as zooplanktivores (e.g.van der Kooij et al. 2008;van Deurs et al. 2009) and as the main prey for a wide range of fish and seabird species (e.g.Engelhard et al. 2008;Wanless et al. 2008).Sandeels have also been targeted by the largest single species fisheries in the North Sea and their population has declined in recent years, which is thought to be related to sandeel fisheries and possibly the impacts of climate change on their prey (Wanless et al. 2008;ICES 2012a).M. leidyi could have an important competitive impact on this vital zooplanktivore, particularly when sandeel populations are reduced in size (Boersma et al. 2007).

Limitations and further study
The main limitation of this study is that it is based on model data, so any uncertainty in the output of GETM-ERSEM-BFM is reflected in our results and limits confidence in them.However, using model output has the advantage that it is possible to get a broad spatial data coverage at a consistent resolution, which is not possible using empirical data.In addition, conditions, such as riverine input (e.g.Lenhart et al. 2010), may be manipulated to observe how conditions may influence outcomes.
A 1D version of the model used here (GOTM-ERSEM) has been validated for three locations in the North Sea based on data from Smart Buoys, bottom landers, and cruises (van der Molen et al. 2013).Temperature was predicted well, but omnivorous mesozooplankton biomass was underestimated by approximately 90% on average.To account for this, the biomass of both mesozooplankton groups was multiplied by 10 for the reference run of the model.Sensitivity analysis using correction factors of 5 and 15 had no impact on suitability for reproduction but did affect the suitability for survival; so, if ERSEM-BFM under-predicts mesozooplankton biomass by more than 90%, the suitability for survival could be increased.
Microplankton biomass has yet to be validated, so it is not known how well this is predicted by GETM-ERSEM-BFM, though underestimation is likely.One problem with carrying out such validations is the limited availability of observations within the domain of the model, and the high uncertainty associated with them.The sensitivity analysis shows that underestimation of microplankton biomass would have an effect on the results; applying correction factors of 2, 5 and 10 increases the suitability, though larval food is still the most limiting factor to reproduction.This indicates that the results presented here should be taken as a conservative estimate of the suitability for M. leidyi reproduction.
Further validation and subsequent improvement of the ecosystem model is necessary to improve the accuracy of the zooplankton biomass output and increase confidence in the results of studies such as this that use the output.In addition to validation of the model, one way to better predict M. leidyi distribution would be to include it in the ecosystem model (ERSEM-BFM) either directly or indirectly through a particle tracking individual behaviour model.A stage-structured population model by Salihoglu et al. (2011) already exists and could be used as a starting point for this purpose.
The risk maps produced here accounted for the scaling of reproduction with temperature and food, but not salinity, which may also affect reproductive output (Lehtiniemi et al. 2012;Jaspers et al. 2011).Salinity was omitted (except the condition of being greater than 10) because we found only two published experiments investigating the effects of salinity on M. leidyi reproduction, which found contrasting results: either a linear increase with increasing salinity (Jaspers et al. 2011) or a dome-shaped relationship (Lehtiniemi et al. 2012).Salinity is also less likely to affect reproduction in the North Sea (compared to the Baltic) as, with the exception of estuaries, it is high (approximately 34.5) and varies very little.However, if the results of Lehtiniemi et al. (2011) hold true then it may be that M. leidyi can only reproduce in estuaries in the North Sea, which would place a major constraint on blooms.A recent study by Haraldsson et al (2013) found that 80% of M. leidyi were found in the range of 22-29, which further suggests that the higher salinity of the central North Sea may not be suitable for reproduction.Further study into the effects of salinity on reproductive output would be necessary to establish whether this is the case.Additionally, further study into the ability of M. leidyi to tolerate low food levels is necessary, as the ability to withstand starvation or low food conditions appears to be important to winter survival.

M. leidyi predators
For many invasive species, including M. leidyi, one of the reasons for their success as invaders is a lack of natural predators in invaded regions (Purcell et al. 2001).In native areas M. leidyi is fed upon by the ctenophore Beroe ovata Bruguière, 1979 in addition to two jellyfish species and two fish species (Fuentes et al. 2010).In the Black Sea, M. leidyi populations have declined and the ecosystem recovered since B. ovata appeared (Shiganova et al. 2003), and in the Aegean Sea, where B. ovata has been present since the beginning of the M. leidyi invasion, M. leidyi has never become abundant (Shiganova et al. 2004;Fuentes et al. 2010).B. ovata is not yet known to be established in the North Sea, although it has been recorded in Denmark (Shiganova et al. 2014), but there are other ctenophore predators of this genus present, such as Beroe gracilis Künne, 1939.This species has been shown to eat small M. leidyi in laboratory studies, so may exert a predation pressure on M. leidyi (Hosia et al. 2011).Additionally, cod is thought to predate upon M. leidyi, as M. leidyi has been found in cod stomachs (Schaber et al. 2011).The presence of potential predators in the North Sea may limit the impact of M. leidyi on ecosystem structure and functioning (Boersma et al. 2007).However, the population status of these potential predators must be taken into account, especially in the case of cod, which is depleted in the North Sea (ICES, 2012b).

Climate change
Although conditions are already suitable for M. leidyi in the North Sea, warmer conditions due to climate change would likely enhance survival and so the size of the overwintering population, and increase reproductive output in summer.However, it is also important to consider that increased temperatures cause an increase in metabolic rates of M. leidyi which will increase its food requirements (Purcell et al. 2001).
Climate change may also affect the interactions of M. leidyi with its zooplankton prey by altering the phenology of blooms of both predator and prey, affecting the temporal overlap.In the North West Atlantic, the temporal match between M. leidyi and its copepod prey Acartia tonsa Dana 1849 has increased due to altered phenology (Sullivan et al. 2001;Costello et al. 2006), which is hypothesised to be due to warming of winter refugia (Costello et al. 2006).

Management
Managing introduced species such as M. leidyi in the marine environment poses a complex challenge, due to the ease of transport between seas by vectors such as shipping, and the difficulties of managing species in an environment with little in the way of physical barriers to prevent spread (Molnar et al. 2008).Since M. leidyi is already present in the North Sea, it is likely only a matter of time before M. leidyi reaches all suitable areas of the North Sea, whether by natural dispersal or in ballast water.
There are some preventative management measures that may be effective in reducing the risk of M. leidyi blooms, by reducing the underlying drivers.Nutrient runoff from land greatly affects both the abundance and composition of the marine plankton community, favouring the phytoplankton types consumed by jellyfish rather than those consumed by fish, and increasing the abundance of phytoplankton and zooplankton (Richardson et al. 2009).Since these species are the prey of M. leidyi, the amount of nutrient runoff could have a large impact on its ability to bloom.By reducing artificial nutrient input from rivers (Richardson et al. 2009), M. leidyi food concentrations would be limited (Salihoglu et al. 2011), which would limit M. leidyi population size (Purcell et al. 2001).
Another indirect driver is overfishing (Gucu 2002;Richardson et al. 2009); the better the population status of potential fish competitors (such as sandeels) the less food is available for M. leidyi, making it less likely to become established (Purcell et al. 2001).Also, as mentioned previously, healthy cod populations may have a detrimental effect on M. leidyi due to predation (Schaber et al. 2011).
Even if it is possible to limit M. leidyi blooms by reducing eutrophication and overfishing, they may still occur, so other management strategies may be necessary to mitigate the effects of the blooms.For example, for power plants in areas with a high risk of M. leidyi blooms, it could be worthwhile to adopt techniques to avoid jellyfish clogging intake pipes, such as bubblers to float jellyfish away (Lo 1991), and to monitor the surrounding waters to provide earlier warning of a bloom.For areas where high risk regions for M. leidyi blooms overlap with fish spawning grounds, it may be necessary to take the extra mortality of eggs and larvae due to predation by M. leidyi into account in determining how heavily the stock can be fished.
Figure 2 shows the existing records for where M. leidyi has so far been found in the North Sea, but this is by no means a definitive map of where M. leidyi presently occurs (or has occurred in previous years), as sampling effort has been very patchy in time and space.More monitoring with better spatial coverage is essential to understand where M. leidyi is currently established in the North Sea, and to study interactions with environmental and biotic variables such as those used in this study (Schaber et al. 2011).The results of this study could be used to prioritise areas and times of year for monitoring for this species.

Figure 1 .
Figure 1.Map of the study area.The blue area represents the model domain of GETM-ERSEM-BFM.

Figure 2 .
Figure 2. Records of Mnemiopsis leidyi in the North Sea from 2005-2012 during spring (A: Mar-May), summer (B: Jun-Aug), autumn (C: Sep-Nov), and winter (D: Dec-Feb).Closed triangles represent presences and open triangles represent absences.For details and sources see Table1Sin Supplementary material.
each day and the daily layers summed to obtain overall suitability for the year.whereall indices are zero below the minimum threshold and 10 above the optimum (nonlimiting) threshold, and: / with T d the temperature of day d (where T d is between 12 and 20˚C), and α a constant equalling 1.02410 12 that is used to scale the index to 1-10.10/withAi d the adult food (mesozooplankton) biomass of day d, AF opt the non-limiting adult food requirement (25 mg C m -3 ), AF min the minimum adult food requirement for reproduction (3 mg C m -3 ), and β a constant equalling 1.36, used to scale the index to 1-10.

Figure 3 .
Figure 3. Map showing the average (for 2005-2011) number of days per year suitable for survival of Mnemiopsis leidyi as determined by adult food, temperature and salinity.Inset shows standard deviation for the same period.

Figure 5 .
Figure 5. Map showing the average (for 2005-2008) number of reproductive cycles possible within the time window when conditions are suitable for Mnemiopsis leidyi reproduction, assuming a life cycle length of 40 days, as at 15˚C.

Figure 6 .
Figure 6.Map showing the average (for 2005-2011) suitability for Mnemiopsis leidyi reproduction (A).This is based on suitability indices for temperature (B) and food (C (adult) and D (larval)), and a salinity threshold of 10.Inset (E) shows standard deviation for the same period.

Figure 7 .
Figure 7. Map showing the month with the maximum suitability for Mnemiopsis leidyi reproduction (mode for 2005-2011).

Table 2 .
Survival and reproduction results.

Table 3 .
Sensitivity analysis results for species environmental requirements.Reference runs are in bold.

Table 4 .
Sensitivity analysis results for food input fields.Reference runs are in bold.