UNIQUE ROLE OF JELLYFISH IN THE PLANKTON ECOSYSTEM 1 REVEALED USING A GLOBAL OCEAN BIOGEOCHEMICAL 2 MODEL 3

Abstract. Jellyfish are increasingly recognised as important components of the marine ecosystem, yet their specific role is poorly defined compared to that of other zooplankton groups. This paper presents the first global ocean biogeochemical model that includes an explicit representation of jellyfish, and uses the model to gain insight into the influence of jellyfish on the plankton community. The PlankTOM11 model groups organisms into Plankton Functional Types (PFT). The jellyfish PFT is parameterised here based on our synthesis of observations on jellyfish growth, grazing, respiration and mortality rates as functions of temperature and on jellyfish biomass. The distribution of jellyfish is unique compared to that of other PFTs in the model. The jellyfish global biomass of 0.13 PgC is within the observational range, and comparable to the biomass of other zooplankton and phytoplankton PFTs. The introduction of jellyfish in the model has a large direct influence on the crustacean macrozooplankton PFT, and influences indirectly the rest of the plankton ecosystem through trophic cascades. The zooplankton community in PlankTOM11 is highly sensitive to the jellyfish mortality rate, with jellyfish increasingly dominating the zooplankton community as its mortality diminishes. Overall the results suggest that jellyfish play an important and unique role in regulating marine plankton ecosystems, which has been neglected so far. 


The introduction of jellyfish in the model has a large direct influence on the crustacean 29 macrozooplankton PFT, and influences indirectly the rest of the plankton ecosystem through trophic 30 cascades. The zooplankton community in PlankTOM11 is highly sensitive to the jellyfish mortality 31 rate, with jellyfish increasingly dominating the zooplankton community as its mortality diminishes. 32 Overall the results suggest that jellyfish play an important and unique role in regulating marine 33 plankton ecosystems, which has been neglected so far. Gelatinous zooplankton are increasingly recognised as influential organisms in the marine 37 environment, not just for the disruptions they can cause to coastal economies (fisheries, aquaculture, 38 beach closures and power plants etc.; Purcell et al., 2007), but also as important consumers of 39 plankton (Lucas and Dawson, 2014), a food source for many marine species (Lamb et al., 2017) and 40 as key components in marine biogeochemical cycles (Crum et al., 2014;Lebrato et al., 2012). The 41 term gelatinous zooplankton can encompass a wide range of organisms across three phyla: Tunicata 42 (salps), Ctenophora (comb-jellies), and Cnidaria (true jellyfish). This study focuses on Cnidaria 43 (including Hydrozoa, Cubozoa and Scyphozoa), which contribute 92% of the total global biomass of 44 gelatinous zooplankton . The other gelatinous zooplankton groups, Tunicata and 45 Ctenophora, are excluded from this study because there are far fewer data available on their biomass 46 and vital rates than for Cnidaria, and they only contribute a combined global biomass of 8% of total 47 gelatinous zooplankton . Cnidaria are both independent enough from other 48 gelatinous zooplankton, and cohesive enough to be represented as a single Plankton Functional Type 49 (PFT) for global modelling (Le Quéré et al., 2005). For the rest of this paper pelagic Cnidaria are 50 referred to as jellyfish. 51 Jellyfish exhibit a radially symmetrical body plan and are characterised by a bell-shaped body 52 (medusae). Swimming is achieved by muscular, "pulsing" contractions and animals have one opening 53 for both feeding and excretion. Most scyphozoans and cubozoans, and many hydrozoans, follow a 54 meroplanktonic life cycle. A sessile (generally) benthic polyps buds off planktonic ephyrae asexually. 55 These, in turn, grow into medusae that reproduce sexually to generate planula larvae, which then 56 settle and transform into polyps. Within this general life cycle, there is large reproductive and life 57 cycle variety, including some holoplanktonic species that skip the benthic polyp stage as well as 58 holobenthic species that skip the pelagic phase, and much plasticity (Boero et al., 2008;Lucas and 59 Dawson, 2014). 60 Jellyfish are significant consumers of plankton, feeding mostly on zooplankton using tentacles and/or 61 oral arms containing stinging cells called nematocysts (Lucas and Dawson, 2014). The large body size 62 to carbon content ratio of jellyfish creates a low maintenance, large feeding structure, which, because 63 they do not use sight to capture prey, allow them to efficiently clear plankton throughout 24 hours 64 (Acuña et al., 2011;Lucas and Dawson, 2014). Jellyfish are connected to lower trophic levels, with 65 the ability to influence the plankton ecosystem structure and thus the larger marine ecosystem through 66 trophic cascades (Pitt et al., 2007; West et al., 2009). Jellyfish have the ability to rapidly form 67 large high-density aggregations known as blooms that can temporarily dominate local ecosystems 68 (Graham et al., 2001;Hamner and Dawson, 2009). Jellyfish contribute to the biogeochemical cycle 69 https://doi.org/10.5194/bg-2020-136 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License.
For growth through grazing, % " # ! is the grazing rate by zooplankton ! on food source & and is 135 the growth efficiency. For loss through grazing, # ! # " is the grazing of other zooplankton on ! . For 136 basal respiration, *°# ! is the respiration rate at 0°C, is temperature, # ! is the temperature 137 dependence of respiration ( )* = )* ). Mortality is the closure term of the model, and is mostly due 138 to predation by higher trophic levels than are represented by the model. *°# ! is the mortality rate at 139 0°C, # ! is the temperature dependence of the mortality ( )* = )* ) and # ! is the half saturation 140 constant for mortality. ∑ 0 is the sum of all PFTs, excluding bacteria, and is used as a proxy for the 141 biomass of predators not explicitly included in the model. More details on each term are provided 142 below. 143 144

PFT Growth 145
Growth rate is the trait that most distinguishes PFTs in models (Buitenhuis et al., 2006(Buitenhuis et al., , 2013a. 146 Jellyfish growth rates were compiled as a function of temperature from the literature. In previous 147 published versions of the PlankTOM model, growth as a function of temperature ( , ) was fitted with 148 two parameters: 149 where * is the growth at 0°C, )* is the temperature dependence of growth derived from 151 observations, and is the temperature (Le Quéré et al., 2016). Jellyfish growth rate is poorly captured 152 by an exponential fit to temperature. To better capture the observations, the growth calculation has 153 now been updated with a three-parameter growth rate, which produces a bell-shaped curve centred 154 around an optimal growth rate at a given temperature ( Fig. 2 and Table 2). The three-parameter fit is 155 suitable for the global modelling of plankton because it can represent an exponential increase if the 156 data support this (Schoemann et al., 2005). The growth rate as a function of temperature ( , ) is now 157 defined by; the optimal temperature ( 12$ ), maximum growth rate ( 345 ) at 12$ , and the temperature 158 The available observations measure growth rate, but the model requires specification of the grazing 161 rate (Eq. 1). Growth of zooplankton and grazing ( , ) are related through the gross growth efficiency 162 The food web, and thus the trophic level of PFTs is determined through grazing preferences. The 169 relative preference of jellyfish zooplankton for the other PFTs was determined through a literature 170 search (Colin et al., 2005;Flynn and Gibbons, 2007;Malej et al., 2007;Purcell, 1992Purcell, , 1997Purcell, , 2003171 Stoecker et al., 1987;Uye and Shimauchi, 2005). The dominant food source was mesozooplankton 172 (specifically copepods), followed by proto-zooplankton and then macrozooplankton (Table 3). There 173 is little evidence in the literature for jellyfish actively consuming autotrophs. One of the few pieces of 174 evidence is a gut content analysis where 'unidentified protists… some chlorophyll bearing' were 175 found (Colin et al., 2005). Another is a study by Boero et al. (2007) which showed that very small 176 medusae such as Obelia will consume bacteria, but not necessarily actively. The ephyrae stage of 177 scyphozoans are likely to have a higher clearance rate of autotrophs, due to their smaller size, but this 178 will have a minimal effect on the overall preferences and the biomass consumed. Table 3 shows the 179 relative preference of jellyfish for its prey assigned in the model, along with the preferences of the 180 other zooplankton PFTs. The preference ratios are weighted using the global carbon biomass for each 181 type, calculated from the MAREDAT database, following the methodology used for the other PFTs 182 Previous analysis of respiration rates of jellyfish found that temperature manipulation experiments 186 with Q10 values of >3 were flawed because the temperature was changed too rapidly (Purcell, 2009;187 Purcell et al., 2010). In a natural environment, jellyfish gradually acclimate to temperature changes 188 which has a smaller effect on their respiration rates. Purcell et al. (2010) instead collated values from 189 experiments that measured respiration at ambient temperatures, providing a range of temperature data 190 across different studies. They found that Q10 for respiration was 1.67 for Aurelia species (Purcell,  Where RR is the respiration rate, BM is the body mass, and T and R T are the observed temperature and 203 associated respiration rate. The parameters values were then calculated using * = 4 , and )* = 204 ( ? ) )* , where e is the exponential function. The resulting fit to data is shown in Fig. 3. The parameter 205 values for respiration used in the model are given in Table 4. Macrozooplankton respiration values are 206 also given in Fig. 3 and Table 4, to provide a comparison to another zooplankton PFT of the most 207 similar size available. 208 209

Jellyfish PFT Mortality 210
There is limited data on mortality rates for jellyfish and to use mortality data from the literature on is in a steady state where mortality equals recruitment, reproduction is constant and that mortality is 214 independent of age (Moriarty, 2009). All models with zooplankton mortality rates follow these 215 assumptions. In reality the mortality of a zooplankton population is highly variable. Steady states are 216 balanced over a long period (if a population remains viable), reproduction is restricted to certain times 217 of year and the early stages of life cycles are many times more vulnerable to mortality. Despite these 218 assumptions, with the limited data on mortality rates, the larger uncertainty lies with the data rather 219 than the assumptions (Moriarty, 2009). The half saturation constant for mortality ( # ! in Eq. 1) is set 220 to 20 µmol C L -1 the same as other zooplankton types, due to the lack of PFT specific data. In the 221 small amount of data available and suitable for use in the model (16 data points from two studies) 222 mortality ranged from 0.006 -0.026 per day (Acevedo et al., 2013;Malej and Malej, 1992). Applying 223 the exponential fit to this data gave a mortality rate at 0°C ( *°# ! in Eq. 1) of 0.018 per day. Sensitivity 224 tests were carried out from this mortality rate due to low confidence in the value. 225 Results from a subset of the sensitivity tests are shown in Fig. 4. The model was found to best 226 represent a range of observations when jellyfish mortality was increased to 0.12 per day. The fit to 227 mortality for the data (µ0 = 0.018) and the adjusted mortality (µ0 = 0.12) is shown in Fig. 3 (Table 3), to account for this additional grazing the mortality term for 252 macrozooplankton and the respiration term for mesozooplankton were reduced compared to model 253 versions where no jellyfish are present (Table 5). Respiration is reduced in place of mortality for 254 mesozooplankton as their mortality term had already been reduced to zero to account for predation by 255 identical to PlankTOM11 except for the top predator mortality term for meso-and macrozooplankton, 283 which were returned to pre-jellyfish values, to account for the lack of predation by jellyfish. 284 Macrozooplankton mortality was then tuned from this value to account for the change to the growth 285 calculation ( Table 5). The second additional simulation is carried out to test the addition of an 11 th 286 PFT in comparison to the addition of jellyfish as the 11 th PFT. This is done by parameterising the 287 jellyfish PFT identically to the macrozooplankton PFT, so that there are 11 PFTs active, with two 288 macrozooplankton. This simulation is called PlankTOM10.5. Otherwise, these simulations were 289 identical to PlankTOM11.  Jellyfish biomass in PlankTOM11 is within the range but towards the lower end of observations at 328 0.13 PgC, with jellyfish accounting for 16% of the total zooplankton biomass (Table 6). When the 329 https://doi.org/10.5194/bg-2020-136 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. modelled biomass was tuned to match the higher observed biomass by adjusting the mortality rate, 330 jellyfish dominate the entire ecosystem significantly reducing levels of the other zooplankton and 331 increasing chlorophyll above observations for the Northern and Southern Hemispheres ( Fig. 4 and 332 representative of reality. The maximum biomass in the southern hemisphere is mostly around coastal 340 areas i.e. South America and southern Australia (Fig. 6). This is expected from reports and papers on 341   (Table 8). The spatial patchiness is 356 somewhat replicated in PlankTOM11, although with a smaller variation (Fig. 7). PlankTOM11 357 replicates the mean seasonal shape and biomass of jellyfish with a small peak over the summer 358 followed by a large peak in September in the observations and in October in PlankTOM11 (Fig. 7). 359 Overall, PlankTOM11 replicates the mean but underestimates the maximum biomass and temporal 360 patchiness of the observations (Fig. 7 and Table 8). concentrations around the equator (Fig. 8). PlankTOM11 also reproduces higher chlorophyll 367 concentrations in the Northern Hemisphere than the Southern (Fig. 9), and higher concentrations in 368 the southern Atlantic than the southern Pacific Ocean (Fig. 8) concentrations increasing in summer compared to the winter for each hemisphere (Fig. 8). 372 To assess the effect of adding jellyfish to PlankTOM, two additional simulations were conducted: 373 PlankTOM10 where jellyfish growth is set to zero and PlankTOM10.5 where all jellyfish parameters 374 are set equal to macrozooplankton parameters (Sect. 2.1.6). The two simulations show similar spatial 375 patterns of surface chlorophyll to PlankTOM11, but different concentration levels. PlankTOM11 376 closely replicates the chlorophyll ratio between the north and south with a ratio of 2.12, compared to 377 the observed ratio of 2.16 (Fig. 9). PlankTOM10 and PlankTOM10.5 underestimate the observed ratio 378 with ratios of 1.57 and 1.96 respectively (Fig. 9). Adding an 11 th PFT improves the chlorophyll ratio, 379 however, the regional chlorophyll concentrations for PlankTOM10.5 are a poorer match to the 380 observations than PlankTOM11, especially in the north (Fig. 9). PlankTOM10 overestimates the 381 observed chlorophyll concentration in the south (0.22 and 0.18 respectively; Fig. 9). All three 382 simulations underestimate chlorophyll concentration in the tropics compared to observations (Fig. 9).

PFTs. 386
PlankTOM11 underestimates primary production by 10 PgC y -1 , export production and N2 fixation are 387 within the observational range, and CaCO3 export is slightly overestimated (Table 6). 388 In PlankTOM11 each PFT shows unique spatial distribution in carbon biomass (Fig. 5). The total 389 biomass of phytoplankton is within the range of observations, but the partitioning of this biomass 390 between phytoplankton types differs from observations (Table 6). PlankTOM11 is dominated by 391 mixed-phytoplankton and coccolithophores, together making up 47% of the total phytoplankton 392 biomass. Diatoms and Phaeocystis are the next most abundant and fall within the observed range, 393 followed by Picophytoplankton with around half the observed biomass (Table 6). The observations 394 are dominated by picophytoplankton, followed by Phaeocystis and Diatoms (Table 6). The modelled 395 mixed-phytoplankton is likely taking up the ecosystem niche of picophytoplankton. Coccolithophores 396 https://doi.org/10.5194/bg-2020-136 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. are overestimated by a factor of 10 and may also be filling the ecosystem niche of picophytoplankton 397 in the model (Table 6). 398 399

ROLE OF JELLYFISH IN THE PLANKTON ECOSYSTEM 400 401
Other than jellyfish, macrozooplankton exhibit the largest change in biomass between the three 402 simulations, followed by mesozooplankton (Fig. 10). This is despite the higher preference of jellyfish 403 grazing on mesozooplankton (ratio of 10) than on macrozooplankton (ratio of 5; Table 3), suggesting 404 that the interaction between macrozooplankton and jellyfish is dominated by competition for 405 resources rather than by their mutual predation. Jellyfish and macrozooplankton both preferentially 406 graze on mesozooplankton and protozooplankton. The greatest difference in PFT biomass between 407 simulations occurs in latitudes higher than 30º (Fig. 10). In the tropics, jellyfish have a low impact on 408 the ecosystem due to their low biomass in this region ( Fig. 6 and Fig. 10). 409 The seasonality of the PFTs in each simulation is shown in Fig. 11 for 30-70º north and south, as the 410 regions with the greatest differences between simulations (Fig. 10). In PlankTOM10 411 macrozooplankton represent the highest trophic level. The addition of another PFT at the same or at a 412 higher trophic level (PlankTOM10.5 and PlankTOM11 respectively) greatly reduces the biomass of 413 the macrozooplankton, through a combination of competition and low-level predation ( Fig. 10 and  414   Fig. 11). The addition of jellyfish changes the zooplankton with the highest biomass from 415 macrozooplankton to protozooplankton, in both the north and south (Fig. 11). However, the addition 416 of jellyfish has a small impact on the biomass of protozooplankton (Fig. 11), despite the high prey 417 preference of jellyfish for protozooplankton. The small impact of jellyfish on protozooplankton 418 biomass may be due to trophic cascade effects where jellyfish reduce the biomass of 419 macrozooplankton, which reduces the predation pressure of macrozooplankton on protozooplankton, 420 whilst jellyfish provide an additional predation pressure on protozooplankton. The decrease in 421 predation by macrozooplankton may be compensated for by the increase in predation by jellyfish. 422 In PlankTOM11 there is a clear distinction between the biomass in the north and south, with higher 423 biomass for each PFT in the north compared to the south ( Fig. 10 and Fig. 11). Plankton types have 424 higher concentrations in the respective hemisphere's summer, and a double peak in phytoplankton in 425 the north (Fig. 10 and Fig. 11). PlankTOM10 also has a higher biomass of each PFT in the north 426 compared to the south, but the difference is smaller than that in PlankTOM11 ( Fig. 10 and Fig. 11). 427 The key difference between the two models is the biomass of macrozooplankton. In PlankTOM10 428 macrozooplankton are the dominant zooplankton, especially in late summer and autumn where their 429 biomass matches and even exceeds the biomass of phytoplankton in the region (Fig. 11). In 430 https://doi.org/10.5194/bg-2020-136 Preprint. Discussion started: 30 April 2020 c Author(s) 2020. CC BY 4.0 License. PlankTOM11 neither macrozooplankton, nor any other zooplankton, come close to matching the 431 biomass of phytoplankton. The largest direct influence of jellyfish in these regions is its control on 432 macrozooplankton biomass. 433 In PlankTOM11 in the north, phytoplankton display a double peak in seasonal biomass, with a 434 smaller peak in April of 2.4 µmol C L -1 , followed by a larger peak in July of 3.2 µmol C L -1 (Fig. 11).

440
Model results suggest high competition between macrozooplankton (crustaceans) and jellyfish. The 441 growth rate of jellyfish is higher than that of macrozooplankton for the majority of the ocean (where 442 the temperature is less than 25°C) but the mortality of jellyfish is also significantly higher than 443 macrozooplankton, again for the majority of the ocean. In situations where jellyfish mortality is 444 reduced (but still higher than macrozooplankton mortality), jellyfish outcompete macrozooplankton 445 for grazing. Because jellyfish also prey directly on macrozooplankton, the biomass of The high patchiness of jellyfish in the observations is partly but not fully captured in PlankTOM11 452 ( Fig. 7 and Table 7 Jellyfish in PlankTOM11 are parameterised using data largely from temperate species, because this is 490 the majority of the data available. This may explain some of the prevalence of jellyfish in 491 PlankTOM11 at mid-to high-latitudes and the lower biomass in the tropics. Experimental rate data 492 for a wider range of jellyfish species from a wider range of latitudes is required to address this bias. 493 Another limitation of jellyfish representation in the model is the lack of body size representation. 494 Most biological activity is from small individuals, while most of the biomass is from large 495 individuals. The size distribution of body mass in jellyfish is particularly wide compared to other 496 PFTs (Table 1) to the model (PlankTOM10 to PlankTOM10.5 to PlankTOM11), especially the North/South ratio. 500 The three simulations have identical physical environments, with the influence of jellyfish as the only 501 alteration, so any differences between the three can be attributed to the ecosystem structure. Jellyfish 502 are the highest trophic level represented in PlankTOM11, with preference for meso-, followed by 503 proto-, and then macrozooplankton. However, the largest influence of jellyfish is on the 504 macrozooplankton, because the grazing pressure on mesozooplankton from macrozooplankton is 505 reduced, and the grazing on protozooplankton by macro-and mesozooplankton is reduced, while the 506 grazing pressure from jellyfish on both meso-and protozooplankton is increased. The combined 507 changes to macrozooplankton and jellyfish grazing pressure counteract to reduce the overall change in 508 grazing pressure. The top down trophic cascade from jellyfish on the other zooplankton also changes 509 some of the grazing pressures on the phytoplankton, which translates into regional and seasonal 510 effects on chlorophyll. The replication of global mean jellyfish biomass, 0.13 PgC, is within the observational range, and in 517 the region with the highest density of observations PlankTOM11 closely replicates the mean and 518 seasonal jellyfish biomass. There is a deficit of data on jellyfish carbon biomass observations and 519 physiological rates. Monitoring and data collection efforts have increased over recent years; we 520 recommend a further increase especially focussing in less-surveyed regions and on non-temperate 521

species. 522
Jellyfish exert control over the other zooplankton, with the greatest influence on macrozooplankton. 523 Through trophic cascades jellyfish also influence the phytoplankton. PlankTOM11 is a successful first 524 step in the inclusion of jellyfish in global ocean biogeochemical modelling. The model raises 525 interesting questions about the sensitivity of the zooplankton community to changes in jellyfish 526 mortality and calls for an investigation in interactions between macrozooplankton and jellyfish. 527    Table 1 for 533 PFT definitions). The arrows represent the grazing fluxes by protozooplankton (orange), 534 mesozooplankton (red), macrozooplankton (blue) and jellyfish zooplankton (purple). Only fluxes with 535 relative preferences above 0.1 are shown (see Table 3).    to the data is shown in black, using the parameter values from Table 2 and Table 4. Growth rates are 548 the same as shown in Fig. 2, on a different scale. For jellyfish mortality the thin dashed line is the fit 549 to data and the solid line is the adjusted fit (Table 4).   for PlankTOM11 is shown by the black filled circle and the fit to the data simulation is shown by the 556 grey filled circle; global mean PFT biomass (µg C L -1 ) for 0-200m depth (top -middle), regional 557 mean surface chlorophyll concentration (µg chl L -1 ; bottom). For the regional mean chlorophyll the 558 observations are calculated from SeaWiFS. All data are averaged for 1985-2015, and between 30º and 559 55º latitude in both hemispheres: 140-240ºE in the north and 140-290ºE in the south (see Fig. 8).

560
Phyto is the sum of all the phytoplankton PFTs.         Table A1. Global mean values for rates and biomass from observations with the associated references. In parenthesis is the percentage share of the plankton type of the total Phytoplankton or Zooplankton biomass.

Observations
Reference for the data